[ { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0304_00/00000.jpg", "scene0304_00/00040.jpg", "scene0304_00/00080.jpg", "scene0304_00/00120.jpg", "scene0304_00/00160.jpg" ], "turn_id": 1, "scan_id": "scene0304_00", "origin_question": "Remember, have you seen any coffee maker(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any coffee maker(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0304_00/00200.jpg", "scene0304_00/00240.jpg", "scene0304_00/00280.jpg", "scene0304_00/00320.jpg", "scene0304_00/00360.jpg" ], "turn_id": 2, "scan_id": "scene0304_00", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the red backpack with black straps and handles\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the red backpack with black straps and handles\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0304_00/00400.jpg", "scene0304_00/00440.jpg", "scene0304_00/00480.jpg", "scene0304_00/00520.jpg", "scene0304_00/00560.jpg" ], "turn_id": 3, "scan_id": "scene0304_00", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "2.4", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0304_00/00600.jpg", "scene0304_00/00640.jpg", "scene0304_00/00680.jpg", "scene0304_00/00720.jpg", "scene0304_00/00760.jpg" ], "turn_id": 4, "scan_id": "scene0304_00", "origin_question": "Which two objects are on the same side of you now? \"the medium-sized rectangular gray door\", \"the red backpack with black straps and handles\", \"the medium-sized rectangular gray plastic bin\". ", "option": [ "the medium-sized rectangular gray door, the red backpack with black straps and handles", "the medium-sized rectangular gray door, the medium-sized rectangular gray plastic bin", "the red backpack with black straps and handles, the medium-sized rectangular gray plastic bin" ], "answer": "the red backpack with black straps and handles, the medium-sized rectangular gray plastic bin", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the medium-sized rectangular gray door\", \"the red backpack with black straps and handles\", \"the medium-sized rectangular gray plastic bin\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the medium-sized rectangular gray door, the red backpack with black straps and handles', 'the medium-sized rectangular gray door, the medium-sized rectangular gray plastic bin', 'the red backpack with black straps and handles, the medium-sized rectangular gray plastic bin'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0304_00/00800.jpg", "scene0304_00/00840.jpg", "scene0304_00/00880.jpg", "scene0304_00/00920.jpg", "scene0304_00/00960.jpg" ], "turn_id": 5, "scan_id": "scene0304_00", "origin_question": "What will be the first-time appearance order of the following categories: backpack, bin, pillar?", "option": [ "backpack, pillar, bin", "pillar, bin, backpack", "bin, backpack, pillar", "pillar, backpack, bin" ], "answer": "pillar, backpack, bin", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: backpack, bin, pillar? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['backpack, pillar, bin', 'pillar, bin, backpack', 'bin, backpack, pillar', 'pillar, backpack, bin'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0304_00/01000.jpg", "scene0304_00/01040.jpg", "scene0304_00/01080.jpg", "scene0304_00/01120.jpg", "scene0304_00/01160.jpg" ], "turn_id": 6, "scan_id": "scene0304_00", "origin_question": "Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the tall cylindrical wooden pillar, light brown\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the tall cylindrical wooden pillar, light brown\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0304_00/01200.jpg", "scene0304_00/01240.jpg", "scene0304_00/01280.jpg", "scene0304_00/01320.jpg", "scene0304_00/01360.jpg" ], "turn_id": 7, "scan_id": "scene0304_00", "origin_question": "In which turn were you closest to \"the small, white and blue metal cabinet\"? ", "option": [], "answer": "5", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the small, white and blue metal cabinet\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0304_00/01400.jpg", "scene0304_00/01440.jpg", "scene0304_00/01480.jpg", "scene0304_00/01520.jpg", "scene0304_00/01560.jpg" ], "turn_id": 8, "scan_id": "scene0304_00", "origin_question": "Is the medium-sized rectangular gray door to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the medium-sized rectangular gray door to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0585_01/00000.jpg", "scene0585_01/00040.jpg", "scene0585_01/00080.jpg", "scene0585_01/00120.jpg", "scene0585_01/00160.jpg" ], "turn_id": 1, "scan_id": "scene0585_01", "origin_question": "Remember, have you seen any stool(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any stool(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0585_01/00200.jpg", "scene0585_01/00240.jpg", "scene0585_01/00280.jpg", "scene0585_01/00320.jpg", "scene0585_01/00360.jpg" ], "turn_id": 2, "scan_id": "scene0585_01", "origin_question": "In which turn did you see both \"the large rectangular beige board for posting notes\" and \"the cylindrical creamy yellow plastic stool with textures\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the large rectangular beige board for posting notes\" and \"the cylindrical creamy yellow plastic stool with textures\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0585_01/00400.jpg", "scene0585_01/00440.jpg", "scene0585_01/00480.jpg", "scene0585_01/00520.jpg", "scene0585_01/00560.jpg" ], "turn_id": 3, "scan_id": "scene0585_01", "origin_question": "Which direction is the rectangular white sheet with red printed text in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the rectangular white sheet with red printed text in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "scene0585_01/00600.jpg", "scene0585_01/00640.jpg", "scene0585_01/00680.jpg", "scene0585_01/00720.jpg", "scene0585_01/00760.jpg" ], "turn_id": 4, "scan_id": "scene0585_01", "origin_question": "Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Right", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0585_01/02100.jpg", "scene0585_01/02140.jpg", "scene0585_01/02180.jpg", "scene0585_01/02220.jpg", "scene0585_01/02260.jpg" ], "turn_id": 5, "scan_id": "scene0585_01", "origin_question": "Please recall \"the slim, black fabric rack with multiple compartments\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.44", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the slim, black fabric rack with multiple compartments\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0585_01/02300.jpg", "scene0585_01/02340.jpg", "scene0585_01/02380.jpg", "scene0585_01/02420.jpg", "scene0585_01/02460.jpg" ], "turn_id": 6, "scan_id": "scene0585_01", "origin_question": "Please recall \"the large rectangular beige board for posting notes\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.88", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the large rectangular beige board for posting notes\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0585_01/02500.jpg", "scene0585_01/02540.jpg", "scene0585_01/02580.jpg", "scene0585_01/02620.jpg", "scene0585_01/02660.jpg" ], "turn_id": 7, "scan_id": "scene0585_01", "origin_question": "Which direction is the rectangular white sheet with red printed text in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the rectangular white sheet with red printed text in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0585_01/02700.jpg", "scene0585_01/02740.jpg", "scene0585_01/02780.jpg", "scene0585_01/02820.jpg", "scene0585_01/02860.jpg" ], "turn_id": 8, "scan_id": "scene0585_01", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the blue fabric-covered chair with a metal frame\"; \"the large rectangular beige board for posting notes\"; \"the mailbox\" ", "option": [ "the blue fabric-covered chair with a metal frame", "the large rectangular beige board for posting notes", "the mailbox" ], "answer": "the blue fabric-covered chair with a metal frame", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the blue fabric-covered chair with a metal frame\"; \"the large rectangular beige board for posting notes\"; \"the mailbox\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the blue fabric-covered chair with a metal frame', 'the large rectangular beige board for posting notes', 'the mailbox'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0413_00/00000.jpg", "scene0413_00/00040.jpg", "scene0413_00/00080.jpg", "scene0413_00/00120.jpg", "scene0413_00/00160.jpg" ], "turn_id": 1, "scan_id": "scene0413_00", "origin_question": "Remember, have you seen any toilet(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any toilet(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0413_00/00200.jpg", "scene0413_00/00240.jpg", "scene0413_00/00280.jpg", "scene0413_00/00320.jpg", "scene0413_00/00360.jpg" ], "turn_id": 2, "scan_id": "scene0413_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0413_00/00400.jpg", "scene0413_00/00440.jpg", "scene0413_00/00480.jpg", "scene0413_00/00520.jpg", "scene0413_00/00560.jpg" ], "turn_id": 3, "scan_id": "scene0413_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0413_00/00600.jpg", "scene0413_00/00640.jpg", "scene0413_00/00680.jpg", "scene0413_00/00720.jpg", "scene0413_00/00760.jpg" ], "turn_id": 4, "scan_id": "scene0413_00", "origin_question": "When did you first discover \"the rectangular light-colored wooden vanity with a granite countertop\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the rectangular light-colored wooden vanity with a granite countertop\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0413_00/00800.jpg", "scene0413_00/00840.jpg", "scene0413_00/00880.jpg", "scene0413_00/00920.jpg", "scene0413_00/00960.jpg" ], "turn_id": 5, "scan_id": "scene0413_00", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "1.65", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0413_00/01000.jpg", "scene0413_00/01040.jpg", "scene0413_00/01080.jpg", "scene0413_00/01120.jpg", "scene0413_00/01160.jpg" ], "turn_id": 6, "scan_id": "scene0413_00", "origin_question": "Compared to your position at the end of the 5 turn, are you now closer or farther away from \"the standard white oval toilet\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 5 turn, are you now closer or farther away from \"the standard white oval toilet\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0413_00/01200.jpg", "scene0413_00/01240.jpg", "scene0413_00/01280.jpg", "scene0413_00/01320.jpg", "scene0413_00/01360.jpg" ], "turn_id": 7, "scan_id": "scene0413_00", "origin_question": "Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the sink\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the sink\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0413_00/01400.jpg", "scene0413_00/01440.jpg", "scene0413_00/01480.jpg", "scene0413_00/01520.jpg", "scene0413_00/01560.jpg" ], "turn_id": 8, "scan_id": "scene0413_00", "origin_question": "What will be the first-time appearance order of the following categories: door, sink, toilet?", "option": [ "sink, toilet, door", "door, toilet, sink", "toilet, sink, door", "toilet, door, sink" ], "answer": "sink, toilet, door", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: door, sink, toilet? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['sink, toilet, door', 'door, toilet, sink', 'toilet, sink, door', 'toilet, door, sink'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0413_00/01600.jpg", "scene0413_00/01640.jpg", "scene0413_00/01680.jpg", "scene0413_00/01720.jpg", "scene0413_00/01760.jpg" ], "turn_id": 9, "scan_id": "scene0413_00", "origin_question": "Which direction is the sink in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the sink in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "1mp3d_0040_region19/128b6ca35e664c84a457de8d856c8444_i1_3.jpg", "1mp3d_0040_region19/128b6ca35e664c84a457de8d856c8444_i2_3.jpg", "1mp3d_0040_region19/128b6ca35e664c84a457de8d856c8444_i1_1.jpg", "1mp3d_0040_region19/128b6ca35e664c84a457de8d856c8444_i2_1.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0040_region19", "origin_question": "Remember, how many plant(s) have you seen so far? ", "option": [], "answer": "2", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many plant(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "1mp3d_0040_region19/ff9beb62a14240cf8addbee231fc3582_i1_4.jpg", "1mp3d_0040_region19/ff9beb62a14240cf8addbee231fc3582_i2_4.jpg", "1mp3d_0040_region19/6fe29c72b80b449492666e34067f9008_i1_4.jpg", "1mp3d_0040_region19/6fe29c72b80b449492666e34067f9008_i2_4.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0040_region19", "origin_question": "Is the brown swivel chair with cushioned seat to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the brown swivel chair with cushioned seat to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0040_region19/6fe29c72b80b449492666e34067f9008_i1_5.jpg", "1mp3d_0040_region19/6fe29c72b80b449492666e34067f9008_i2_5.jpg", "1mp3d_0040_region19/ff9beb62a14240cf8addbee231fc3582_i1_3.jpg", "1mp3d_0040_region19/ff9beb62a14240cf8addbee231fc3582_i2_3.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0040_region19", "origin_question": "", "option": [], "answer": "", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "1mp3d_0040_region19/128b6ca35e664c84a457de8d856c8444_i2_5.jpg", "1mp3d_0040_region19/128b6ca35e664c84a457de8d856c8444_i1_4.jpg", "1mp3d_0040_region19/128b6ca35e664c84a457de8d856c8444_i2_4.jpg", "1mp3d_0040_region19/ff9beb62a14240cf8addbee231fc3582_i1_5.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0040_region19", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the medium-sized wooden table with black rods\"; \"the compact white inkjet printer with a smooth top\"; \"the rectangular dark brown wooden window\" ", "option": [ "the medium-sized wooden table with black rods", "the compact white inkjet printer with a smooth top", "the rectangular dark brown wooden window" ], "answer": "the rectangular dark brown wooden window", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the medium-sized wooden table with black rods\"; \"the compact white inkjet printer with a smooth top\"; \"the rectangular dark brown wooden window\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the medium-sized wooden table with black rods', 'the compact white inkjet printer with a smooth top', 'the rectangular dark brown wooden window'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "1mp3d_0040_region19/ff9beb62a14240cf8addbee231fc3582_i2_5.jpg", "1mp3d_0040_region19/ff9beb62a14240cf8addbee231fc3582_i1_0.jpg", "1mp3d_0040_region19/ff9beb62a14240cf8addbee231fc3582_i2_0.jpg", "1mp3d_0040_region19/ff9beb62a14240cf8addbee231fc3582_i1_1.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0040_region19", "origin_question": "Which two objects are on the same side of you now? \"the moderate-sized rectangular dark brown wooden desk\", \"the brown swivel chair with cushioned seat\", \"the medium-sized wooden table with black rods\". ", "option": [ "the moderate-sized rectangular dark brown wooden desk, the brown swivel chair with cushioned seat", "the moderate-sized rectangular dark brown wooden desk, the medium-sized wooden table with black rods", "the brown swivel chair with cushioned seat, the medium-sized wooden table with black rods" ], "answer": "the moderate-sized rectangular dark brown wooden desk, the brown swivel chair with cushioned seat", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the moderate-sized rectangular dark brown wooden desk\", \"the brown swivel chair with cushioned seat\", \"the medium-sized wooden table with black rods\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the moderate-sized rectangular dark brown wooden desk, the brown swivel chair with cushioned seat', 'the moderate-sized rectangular dark brown wooden desk, the medium-sized wooden table with black rods', 'the brown swivel chair with cushioned seat, the medium-sized wooden table with black rods'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0067_region7/effea130a5984e53bd2bef2bbe4ba4c1_i1_5.jpg", "1mp3d_0067_region7/effea130a5984e53bd2bef2bbe4ba4c1_i2_5.jpg", "1mp3d_0067_region7/fbc5ba2686d24a2d86cc1ac0575325ae_i1_2.jpg", "1mp3d_0067_region7/fbc5ba2686d24a2d86cc1ac0575325ae_i2_2.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0067_region7", "origin_question": "Remember, have you seen any kitchen island(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any kitchen island(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0067_region7/57b84179a9a24ff49697d9349a21bc60_i1_1.jpg", "1mp3d_0067_region7/57b84179a9a24ff49697d9349a21bc60_i2_1.jpg", "1mp3d_0067_region7/57b84179a9a24ff49697d9349a21bc60_i1_0.jpg", "1mp3d_0067_region7/57b84179a9a24ff49697d9349a21bc60_i2_0.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0067_region7", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the rectangular brown wooden window frame\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the rectangular brown wooden window frame\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "1mp3d_0067_region7/57b84179a9a24ff49697d9349a21bc60_i1_5.jpg", "1mp3d_0067_region7/57b84179a9a24ff49697d9349a21bc60_i2_5.jpg", "1mp3d_0067_region7/fbc5ba2686d24a2d86cc1ac0575325ae_i1_5.jpg", "1mp3d_0067_region7/fbc5ba2686d24a2d86cc1ac0575325ae_i2_5.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0067_region7", "origin_question": "What will be the first-time appearance order of the following categories: cabinet, kitchen island, sink?", "option": [ "sink, kitchen island, cabinet", "kitchen island, sink, cabinet", "sink, cabinet, kitchen island", "cabinet, kitchen island, sink" ], "answer": "cabinet, kitchen island, sink", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: cabinet, kitchen island, sink? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['sink, kitchen island, cabinet', 'kitchen island, sink, cabinet', 'sink, cabinet, kitchen island', 'cabinet, kitchen island, sink'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "1mp3d_0067_region7/fbc5ba2686d24a2d86cc1ac0575325ae_i1_4.jpg", "1mp3d_0067_region7/fbc5ba2686d24a2d86cc1ac0575325ae_i2_4.jpg", "1mp3d_0067_region7/fbc5ba2686d24a2d86cc1ac0575325ae_i1_3.jpg", "1mp3d_0067_region7/fbc5ba2686d24a2d86cc1ac0575325ae_i2_3.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0067_region7", "origin_question": "Is the shiny silver rectangular stainless steel dishwasher to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the shiny silver rectangular stainless steel dishwasher to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "1mp3d_0067_region7/8025576c59c04913bc7df70a8737ecee_i1_3.jpg", "1mp3d_0067_region7/8025576c59c04913bc7df70a8737ecee_i2_3.jpg", "1mp3d_0067_region7/96b25d845ff447cc989dfd87c88ff8b8_i1_3.jpg", "1mp3d_0067_region7/96b25d845ff447cc989dfd87c88ff8b8_i2_3.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0067_region7", "origin_question": "Please recall \"the black, medium-sized leaf-shaped bowl\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "3.23", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the black, medium-sized leaf-shaped bowl\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "1mp3d_0067_region7/dd66f2f7f4a743aa8f07c053c3636c67_i1_3.jpg", "1mp3d_0067_region7/dd66f2f7f4a743aa8f07c053c3636c67_i2_3.jpg", "1mp3d_0067_region7/a8c2dfb0f8c849ba83c20c7f7d6a00dc_i2_0.jpg", "1mp3d_0067_region7/a8c2dfb0f8c849ba83c20c7f7d6a00dc_i1_5.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0067_region7", "origin_question": "Among these three objects(\"the rectangular dark granite kitchen island\",\"the wide rectangular shiny dark gray countertop\",\"the range hood\" ), which one is the farthest from you?", "option": [ "the rectangular dark granite kitchen island", "the wide rectangular shiny dark gray countertop", "the range hood" ], "answer": "the range hood", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects(\"the rectangular dark granite kitchen island\",\"the wide rectangular shiny dark gray countertop\",\"the range hood\" ), which one is the farthest from you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular dark granite kitchen island', 'the wide rectangular shiny dark gray countertop', 'the range hood'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0067_region7/a8c2dfb0f8c849ba83c20c7f7d6a00dc_i2_5.jpg", "1mp3d_0067_region7/a8c2dfb0f8c849ba83c20c7f7d6a00dc_i1_4.jpg", "1mp3d_0067_region7/a8c2dfb0f8c849ba83c20c7f7d6a00dc_i2_4.jpg", "1mp3d_0067_region7/dd66f2f7f4a743aa8f07c053c3636c67_i1_0.jpg" ], "turn_id": 7, "scan_id": "1mp3d_0067_region7", "origin_question": "", "option": [], "answer": "", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "1mp3d_0067_region7/dd66f2f7f4a743aa8f07c053c3636c67_i2_0.jpg", "1mp3d_0067_region7/96b25d845ff447cc989dfd87c88ff8b8_i1_5.jpg", "1mp3d_0067_region7/96b25d845ff447cc989dfd87c88ff8b8_i2_5.jpg", "1mp3d_0067_region7/8025576c59c04913bc7df70a8737ecee_i1_4.jpg" ], "turn_id": 8, "scan_id": "1mp3d_0067_region7", "origin_question": "Which two objects are on the same side of you now? \"the wide rectangular shiny dark gray countertop\", \"the stove\", \"the rectangular picture in shades of brown and beige\". ", "option": [ "the wide rectangular shiny dark gray countertop, the stove", "the wide rectangular shiny dark gray countertop, the rectangular picture in shades of brown and beige", "the stove, the rectangular picture in shades of brown and beige" ], "answer": "the wide rectangular shiny dark gray countertop, the stove", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the wide rectangular shiny dark gray countertop\", \"the stove\", \"the rectangular picture in shades of brown and beige\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the wide rectangular shiny dark gray countertop, the stove', 'the wide rectangular shiny dark gray countertop, the rectangular picture in shades of brown and beige', 'the stove, the rectangular picture in shades of brown and beige'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0392_00/00000.jpg", "scene0392_00/00040.jpg", "scene0392_00/00080.jpg", "scene0392_00/00120.jpg", "scene0392_00/00160.jpg" ], "turn_id": 1, "scan_id": "scene0392_00", "origin_question": "Remember, have you seen any mirror(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any mirror(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0392_00/00200.jpg", "scene0392_00/00240.jpg", "scene0392_00/00280.jpg", "scene0392_00/00320.jpg", "scene0392_00/00360.jpg" ], "turn_id": 2, "scan_id": "scene0392_00", "origin_question": "At the end of which turn were both of \"the black chair with a mesh back\" and \"the bed\" on your right side?", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the black chair with a mesh back\" and \"the bed\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0392_00/00400.jpg", "scene0392_00/00440.jpg", "scene0392_00/00480.jpg", "scene0392_00/00520.jpg", "scene0392_00/00560.jpg" ], "turn_id": 3, "scan_id": "scene0392_00", "origin_question": "In which turn were you closest to \"the bed\"? ", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the bed\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0392_00/00600.jpg", "scene0392_00/00640.jpg", "scene0392_00/00680.jpg", "scene0392_00/00720.jpg", "scene0392_00/00760.jpg" ], "turn_id": 4, "scan_id": "scene0392_00", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the white rectangular kitchen counter made of laminated material\"; \"the dark-colored jacket with visible stitching\"; \"the bed\" ", "option": [ "the white rectangular kitchen counter made of laminated material", "the dark-colored jacket with visible stitching", "the bed" ], "answer": "the white rectangular kitchen counter made of laminated material", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the white rectangular kitchen counter made of laminated material\"; \"the dark-colored jacket with visible stitching\"; \"the bed\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the white rectangular kitchen counter made of laminated material', 'the dark-colored jacket with visible stitching', 'the bed'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0392_00/00800.jpg", "scene0392_00/00840.jpg", "scene0392_00/00880.jpg", "scene0392_00/00920.jpg", "scene0392_00/00960.jpg" ], "turn_id": 5, "scan_id": "scene0392_00", "origin_question": "Please recall \"the bed\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.86", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the bed\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0392_00/01000.jpg", "scene0392_00/01040.jpg", "scene0392_00/01080.jpg", "scene0392_00/01120.jpg", "scene0392_00/01160.jpg" ], "turn_id": 6, "scan_id": "scene0392_00", "origin_question": "Is the black chair with a mesh back to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the black chair with a mesh back to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "scene0392_00/01200.jpg", "scene0392_00/01240.jpg", "scene0392_00/01280.jpg", "scene0392_00/01320.jpg", "scene0392_00/01360.jpg" ], "turn_id": 7, "scan_id": "scene0392_00", "origin_question": "Among these three objects(\"the dark-colored jacket with visible stitching\",\"the black chair with a mesh back\",\"the small rectangular white range hood made of metal and plastic\" ), which one is the closest to you?", "option": [ "the dark-colored jacket with visible stitching", "the black chair with a mesh back", "the small rectangular white range hood made of metal and plastic" ], "answer": "the small rectangular white range hood made of metal and plastic", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the dark-colored jacket with visible stitching\",\"the black chair with a mesh back\",\"the small rectangular white range hood made of metal and plastic\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the dark-colored jacket with visible stitching', 'the black chair with a mesh back', 'the small rectangular white range hood made of metal and plastic'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0392_00/01400.jpg", "scene0392_00/01440.jpg", "scene0392_00/01480.jpg", "scene0392_00/01520.jpg", "scene0392_00/01560.jpg" ], "turn_id": 8, "scan_id": "scene0392_00", "origin_question": "Is the rectangular silver stainless steel sink to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the rectangular silver stainless steel sink to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0392_00/01600.jpg", "scene0392_00/01640.jpg", "scene0392_00/01680.jpg", "scene0392_00/01720.jpg", "scene0392_00/01760.jpg" ], "turn_id": 9, "scan_id": "scene0392_00", "origin_question": "In which turn did you see both \"the standard-sized white refrigerator with stainless steel features\" and \"the light brown rectangular wooden table with four legs\" simultaneously? ", "option": [], "answer": "7", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the standard-sized white refrigerator with stainless steel features\" and \"the light brown rectangular wooden table with four legs\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "scene0392_00/01800.jpg", "scene0392_00/01840.jpg", "scene0392_00/01880.jpg", "scene0392_00/01920.jpg", "scene0392_00/01960.jpg" ], "turn_id": 10, "scan_id": "scene0392_00", "origin_question": "Assuming the direction you are facing at the end of the turn 7 is forward, did you move a certain distance forward or backward from that position?", "option": [ "Forward", "Backward" ], "answer": "Forward", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 7 is forward, did you move a certain distance forward or backward from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Forward', 'Backward'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0550_00/00000.jpg", "scene0550_00/00040.jpg", "scene0550_00/00080.jpg", "scene0550_00/00120.jpg", "scene0550_00/00160.jpg" ], "turn_id": 1, "scan_id": "scene0550_00", "origin_question": "Remember, how many hook(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many hook(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0550_00/00200.jpg", "scene0550_00/00240.jpg", "scene0550_00/00280.jpg", "scene0550_00/00320.jpg", "scene0550_00/00360.jpg" ], "turn_id": 2, "scan_id": "scene0550_00", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the deep brown rectangular wooden door\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the deep brown rectangular wooden door\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0550_00/00400.jpg", "scene0550_00/00440.jpg", "scene0550_00/00480.jpg", "scene0550_00/00520.jpg", "scene0550_00/00560.jpg" ], "turn_id": 3, "scan_id": "scene0550_00", "origin_question": "Which two objects are on the same side of you now? \"the white rectangular tiled shower stall\", \"the standard white toilet with a black lid\", \"the rectangular white porcelain vanity on a white countertop\". ", "option": [ "the white rectangular tiled shower stall, the standard white toilet with a black lid", "the white rectangular tiled shower stall, the rectangular white porcelain vanity on a white countertop", "the standard white toilet with a black lid, the rectangular white porcelain vanity on a white countertop" ], "answer": "the white rectangular tiled shower stall, the standard white toilet with a black lid", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the white rectangular tiled shower stall\", \"the standard white toilet with a black lid\", \"the rectangular white porcelain vanity on a white countertop\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the white rectangular tiled shower stall, the standard white toilet with a black lid', 'the white rectangular tiled shower stall, the rectangular white porcelain vanity on a white countertop', 'the standard white toilet with a black lid, the rectangular white porcelain vanity on a white countertop'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0550_00/00600.jpg", "scene0550_00/00640.jpg", "scene0550_00/00680.jpg", "scene0550_00/00720.jpg", "scene0550_00/00760.jpg" ], "turn_id": 4, "scan_id": "scene0550_00", "origin_question": "Is the rectangular white porcelain vanity on a white countertop to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the rectangular white porcelain vanity on a white countertop to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0550_00/00800.jpg", "scene0550_00/00840.jpg", "scene0550_00/00880.jpg", "scene0550_00/00920.jpg", "scene0550_00/00960.jpg" ], "turn_id": 5, "scan_id": "scene0550_00", "origin_question": "Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0550_00/01000.jpg", "scene0550_00/01040.jpg", "scene0550_00/01080.jpg", "scene0550_00/01120.jpg", "scene0550_00/01160.jpg" ], "turn_id": 6, "scan_id": "scene0550_00", "origin_question": "By the end of the 4 turn, were you getting closer to or farther away from these two objects: X: \"the window\", Y: \"the rectangular white porcelain vanity on a white countertop\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 4 turn, were you getting closer to or farther away from these two objects: X: \"the window\", Y: \"the rectangular white porcelain vanity on a white countertop\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0550_00/01200.jpg", "scene0550_00/01240.jpg", "scene0550_00/01280.jpg", "scene0550_00/01320.jpg", "scene0550_00/01360.jpg" ], "turn_id": 7, "scan_id": "scene0550_00", "origin_question": "Which direction is the white rectangular tiled shower stall in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the white rectangular tiled shower stall in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0550_00/01400.jpg", "scene0550_00/01440.jpg", "scene0550_00/01480.jpg", "scene0550_00/01520.jpg", "scene0550_00/01560.jpg" ], "turn_id": 8, "scan_id": "scene0550_00", "origin_question": "Is the white rectangular tiled shower stall to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the white rectangular tiled shower stall to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0550_00/01600.jpg", "scene0550_00/01640.jpg", "scene0550_00/01680.jpg", "scene0550_00/01720.jpg", "scene0550_00/01760.jpg" ], "turn_id": 9, "scan_id": "scene0550_00", "origin_question": "What will be the first-time appearance order of the following categories: bin, mirror, stopcock?", "option": [ "bin, stopcock, mirror", "mirror, bin, stopcock", "mirror, stopcock, bin", "bin, mirror, stopcock" ], "answer": "bin, stopcock, mirror", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bin, mirror, stopcock? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['bin, stopcock, mirror', 'mirror, bin, stopcock', 'mirror, stopcock, bin', 'bin, mirror, stopcock'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0550_00/01800.jpg", "scene0550_00/01840.jpg", "scene0550_00/01880.jpg", "scene0550_00/01920.jpg", "scene0550_00/01960.jpg" ], "turn_id": 10, "scan_id": "scene0550_00", "origin_question": "Which direction is the white rectangular tiled shower stall in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the white rectangular tiled shower stall in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0591_01/00000.jpg", "scene0591_01/00040.jpg", "scene0591_01/00080.jpg", "scene0591_01/00120.jpg", "scene0591_01/00160.jpg" ], "turn_id": 1, "scan_id": "scene0591_01", "origin_question": "Remember, how many bin(s) have you seen so far? ", "option": [], "answer": "2", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many bin(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0591_01/00200.jpg", "scene0591_01/00240.jpg", "scene0591_01/00280.jpg", "scene0591_01/00320.jpg", "scene0591_01/00360.jpg" ], "turn_id": 2, "scan_id": "scene0591_01", "origin_question": "In which turn did you see both \"the black single-seater couch with a soft material\" and \"the window\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the black single-seater couch with a soft material\" and \"the window\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0591_01/00400.jpg", "scene0591_01/00440.jpg", "scene0591_01/00480.jpg", "scene0591_01/00520.jpg", "scene0591_01/00560.jpg" ], "turn_id": 3, "scan_id": "scene0591_01", "origin_question": "", "option": [], "answer": "", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0591_01/00600.jpg", "scene0591_01/00640.jpg", "scene0591_01/00680.jpg", "scene0591_01/00720.jpg", "scene0591_01/00760.jpg" ], "turn_id": 4, "scan_id": "scene0591_01", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the sleek oval-shaped black mouse\"; \"the large rectangular blackboard with a white background\"; \"the window\" ", "option": [ "the sleek oval-shaped black mouse", "the large rectangular blackboard with a white background", "the window" ], "answer": "the large rectangular blackboard with a white background", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the sleek oval-shaped black mouse\"; \"the large rectangular blackboard with a white background\"; \"the window\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the sleek oval-shaped black mouse', 'the large rectangular blackboard with a white background', 'the window'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0591_01/00800.jpg", "scene0591_01/00840.jpg", "scene0591_01/00880.jpg", "scene0591_01/00920.jpg", "scene0591_01/00960.jpg" ], "turn_id": 5, "scan_id": "scene0591_01", "origin_question": "Which direction is the pair of black headphones with soft cushioning in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the pair of black headphones with soft cushioning in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0591_01/01000.jpg", "scene0591_01/01040.jpg", "scene0591_01/01080.jpg", "scene0591_01/01120.jpg", "scene0591_01/01160.jpg" ], "turn_id": 6, "scan_id": "scene0591_01", "origin_question": "Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "97.24", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0591_01/01200.jpg", "scene0591_01/01240.jpg", "scene0591_01/01280.jpg", "scene0591_01/01320.jpg", "scene0591_01/01360.jpg" ], "turn_id": 7, "scan_id": "scene0591_01", "origin_question": "When did you first discover \"the large rectangular blackboard with a white background\"? (index of the turn)", "option": [], "answer": "4", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the large rectangular blackboard with a white background\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0591_01/01400.jpg", "scene0591_01/01440.jpg", "scene0591_01/01480.jpg", "scene0591_01/01520.jpg", "scene0591_01/01560.jpg" ], "turn_id": 8, "scan_id": "scene0591_01", "origin_question": "Which direction is the large rectangular blackboard with a white background in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the large rectangular blackboard with a white background in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0044_region38/f74898f691214cd784c6af1ce5b53f6a_i1_0.jpg", "1mp3d_0044_region38/f74898f691214cd784c6af1ce5b53f6a_i2_0.jpg", "1mp3d_0044_region38/711406aa1d4f4b798532cf0c77cce3d2_i1_3.jpg", "1mp3d_0044_region38/711406aa1d4f4b798532cf0c77cce3d2_i2_3.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0044_region38", "origin_question": "Remember, have you seen any picture(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any picture(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0044_region38/89c7f892dc2e412a953ab16548c541b4_i1_2.jpg", "1mp3d_0044_region38/89c7f892dc2e412a953ab16548c541b4_i2_2.jpg", "1mp3d_0044_region38/9922aafc4ca442db9ff9aac3b0fd5b34_i1_2.jpg", "1mp3d_0044_region38/9922aafc4ca442db9ff9aac3b0fd5b34_i2_2.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0044_region38", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "1mp3d_0044_region38/d4fba379bcee4c53bf62282ec8aded53_i1_5.jpg", "1mp3d_0044_region38/d4fba379bcee4c53bf62282ec8aded53_i2_5.jpg", "1mp3d_0044_region38/9922aafc4ca442db9ff9aac3b0fd5b34_i1_1.jpg", "1mp3d_0044_region38/9922aafc4ca442db9ff9aac3b0fd5b34_i2_1.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0044_region38", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "1mp3d_0044_region38/89c7f892dc2e412a953ab16548c541b4_i1_0.jpg", "1mp3d_0044_region38/89c7f892dc2e412a953ab16548c541b4_i2_0.jpg", "1mp3d_0044_region38/711406aa1d4f4b798532cf0c77cce3d2_i1_1.jpg", "1mp3d_0044_region38/711406aa1d4f4b798532cf0c77cce3d2_i2_1.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0044_region38", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "4.06", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "1mp3d_0044_region38/89c7f892dc2e412a953ab16548c541b4_i1_4.jpg", "1mp3d_0044_region38/89c7f892dc2e412a953ab16548c541b4_i2_4.jpg", "1mp3d_0044_region38/9922aafc4ca442db9ff9aac3b0fd5b34_i1_3.jpg", "1mp3d_0044_region38/9922aafc4ca442db9ff9aac3b0fd5b34_i2_3.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0044_region38", "origin_question": "What will be the first-time appearance order of the following categories: door, picture, tv?", "option": [ "tv, door, picture", "picture, door, tv", "door, picture, tv", "picture, tv, door" ], "answer": "picture, door, tv", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: door, picture, tv? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['tv, door, picture', 'picture, door, tv', 'door, picture, tv', 'picture, tv, door'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "1mp3d_0044_region38/98adaae2e097459da98a6bfc3cf42cc4_i1_3.jpg", "1mp3d_0044_region38/98adaae2e097459da98a6bfc3cf42cc4_i2_3.jpg", "1mp3d_0044_region38/98adaae2e097459da98a6bfc3cf42cc4_i1_2.jpg", "1mp3d_0044_region38/98adaae2e097459da98a6bfc3cf42cc4_i2_2.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0044_region38", "origin_question": "Please recall \"the small triangular wooden rack in brown\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "4.05", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the small triangular wooden rack in brown\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0351_01/00000.jpg", "scene0351_01/00030.jpg", "scene0351_01/00060.jpg", "scene0351_01/00090.jpg" ], "turn_id": 1, "scan_id": "scene0351_01", "origin_question": "Remember, have you seen any cart(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any cart(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0351_01/00120.jpg", "scene0351_01/00150.jpg", "scene0351_01/00180.jpg", "scene0351_01/00210.jpg" ], "turn_id": 2, "scan_id": "scene0351_01", "origin_question": "Remember, how many humidifier(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many humidifier(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0351_01/00240.jpg", "scene0351_01/00270.jpg", "scene0351_01/00300.jpg", "scene0351_01/00330.jpg" ], "turn_id": 3, "scan_id": "scene0351_01", "origin_question": "What will be the first-time appearance order of the following categories: chair, door, plant?", "option": [ "chair, door, plant", "chair, plant, door", "plant, door, chair", "plant, chair, door" ], "answer": "chair, plant, door", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: chair, door, plant? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['chair, door, plant', 'chair, plant, door', 'plant, door, chair', 'plant, chair, door'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0351_01/00360.jpg", "scene0351_01/00390.jpg", "scene0351_01/00420.jpg", "scene0351_01/00450.jpg" ], "turn_id": 4, "scan_id": "scene0351_01", "origin_question": "Which two objects are on the same side of you now? \"the rectangular blackboard with a white plastic background\", \"the round potted plant with green and white colors\", \"the pen\". ", "option": [ "the rectangular blackboard with a white plastic background, the round potted plant with green and white colors", "the rectangular blackboard with a white plastic background, the pen", "the round potted plant with green and white colors, the pen" ], "answer": "the round potted plant with green and white colors, the pen", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the rectangular blackboard with a white plastic background\", \"the round potted plant with green and white colors\", \"the pen\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular blackboard with a white plastic background, the round potted plant with green and white colors', 'the rectangular blackboard with a white plastic background, the pen', 'the round potted plant with green and white colors, the pen'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0351_01/00480.jpg", "scene0351_01/00510.jpg", "scene0351_01/00540.jpg", "scene0351_01/00570.jpg" ], "turn_id": 5, "scan_id": "scene0351_01", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the pen\" from your position?", "option": [], "answer": "36.94", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the pen\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0351_01/00600.jpg", "scene0351_01/00630.jpg", "scene0351_01/00660.jpg", "scene0351_01/00690.jpg" ], "turn_id": 6, "scan_id": "scene0351_01", "origin_question": "Which two objects are on the same side of you now? \"the round potted plant with green and white colors\", \"the rectangular brown board in excellent condition\", \"the white corded telephone with a rectangular base\". ", "option": [ "the round potted plant with green and white colors, the rectangular brown board in excellent condition", "the round potted plant with green and white colors, the white corded telephone with a rectangular base", "the rectangular brown board in excellent condition, the white corded telephone with a rectangular base" ], "answer": "the rectangular brown board in excellent condition, the white corded telephone with a rectangular base", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the round potted plant with green and white colors\", \"the rectangular brown board in excellent condition\", \"the white corded telephone with a rectangular base\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the round potted plant with green and white colors, the rectangular brown board in excellent condition', 'the round potted plant with green and white colors, the white corded telephone with a rectangular base', 'the rectangular brown board in excellent condition, the white corded telephone with a rectangular base'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0351_01/00720.jpg", "scene0351_01/00750.jpg", "scene0351_01/00780.jpg", "scene0351_01/00810.jpg" ], "turn_id": 7, "scan_id": "scene0351_01", "origin_question": "Please recall \"the round potted plant with green and white colors\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.72", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the round potted plant with green and white colors\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0029_region0/8aaa453a204c4b91b61bff24567196a5_i1_0.jpg", "1mp3d_0029_region0/8aaa453a204c4b91b61bff24567196a5_i2_0.jpg", "1mp3d_0029_region0/652313c43e064574a464d97f67c52007_i1_0.jpg", "1mp3d_0029_region0/652313c43e064574a464d97f67c52007_i2_0.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0029_region0", "origin_question": "Remember, have you seen any door(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any door(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 3 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "1mp3d_0029_region0/8aaa453a204c4b91b61bff24567196a5_i1_3.jpg", "1mp3d_0029_region0/8aaa453a204c4b91b61bff24567196a5_i2_3.jpg", "1mp3d_0029_region0/652313c43e064574a464d97f67c52007_i1_1.jpg", "1mp3d_0029_region0/652313c43e064574a464d97f67c52007_i2_1.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0029_region0", "origin_question": "When did you first discover \"the door\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the door\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 3 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0029_region0/8aaa453a204c4b91b61bff24567196a5_i1_2.jpg", "1mp3d_0029_region0/8aaa453a204c4b91b61bff24567196a5_i2_2.jpg", "1mp3d_0029_region0/9b72664399a34e4f9dbe470571c73187_i1_3.jpg", "1mp3d_0029_region0/9b72664399a34e4f9dbe470571c73187_i2_3.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0029_region0", "origin_question": "", "option": [], "answer": "", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 3 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0247_00/00000.jpg", "scene0247_00/00040.jpg", "scene0247_00/00080.jpg", "scene0247_00/00120.jpg", "scene0247_00/00160.jpg" ], "turn_id": 1, "scan_id": "scene0247_00", "origin_question": "Remember, have you seen any cabinet(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any cabinet(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0247_00/00200.jpg", "scene0247_00/00240.jpg", "scene0247_00/00280.jpg", "scene0247_00/00320.jpg", "scene0247_00/00360.jpg" ], "turn_id": 2, "scan_id": "scene0247_00", "origin_question": "When did you first discover \"the sleek stainless steel double-door refrigerator\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the sleek stainless steel double-door refrigerator\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0247_00/00400.jpg", "scene0247_00/00440.jpg", "scene0247_00/00480.jpg", "scene0247_00/00520.jpg", "scene0247_00/00560.jpg" ], "turn_id": 3, "scan_id": "scene0247_00", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the sleek stainless steel double-door refrigerator\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the sleek stainless steel double-door refrigerator\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0247_00/00600.jpg", "scene0247_00/00640.jpg", "scene0247_00/00680.jpg", "scene0247_00/00720.jpg", "scene0247_00/00760.jpg" ], "turn_id": 4, "scan_id": "scene0247_00", "origin_question": "At the end of which turn were both of \"the rectangular stainless steel sink with a shiny surface\" and \"the sleek stainless steel double-door refrigerator\" on your left side?", "option": [], "answer": "4", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the rectangular stainless steel sink with a shiny surface\" and \"the sleek stainless steel double-door refrigerator\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0247_00/00800.jpg", "scene0247_00/00840.jpg", "scene0247_00/00880.jpg", "scene0247_00/00920.jpg", "scene0247_00/00960.jpg" ], "turn_id": 5, "scan_id": "scene0247_00", "origin_question": "Is the sleek stainless steel double-door refrigerator to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the sleek stainless steel double-door refrigerator to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0247_00/01000.jpg", "scene0247_00/01040.jpg", "scene0247_00/01080.jpg", "scene0247_00/01120.jpg", "scene0247_00/01160.jpg" ], "turn_id": 6, "scan_id": "scene0247_00", "origin_question": "In which turn did you see both \"the large rectangular glass window with a black frame\" and \"the small, rectangular green plastic bin\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the large rectangular glass window with a black frame\" and \"the small, rectangular green plastic bin\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0247_00/01200.jpg", "scene0247_00/01240.jpg", "scene0247_00/01280.jpg", "scene0247_00/01320.jpg", "scene0247_00/01360.jpg" ], "turn_id": 7, "scan_id": "scene0247_00", "origin_question": "Please recall \"the sleek stainless-steel dishwasher\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "3.83", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the sleek stainless-steel dishwasher\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0247_00/01400.jpg", "scene0247_00/01440.jpg", "scene0247_00/01480.jpg", "scene0247_00/01520.jpg", "scene0247_00/01560.jpg" ], "turn_id": 8, "scan_id": "scene0247_00", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "4.26", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0247_00/01600.jpg", "scene0247_00/01640.jpg", "scene0247_00/01680.jpg", "scene0247_00/01720.jpg", "scene0247_00/01760.jpg" ], "turn_id": 9, "scan_id": "scene0247_00", "origin_question": "Which direction is the sleek stainless steel double-door refrigerator in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the sleek stainless steel double-door refrigerator in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0247_00/01800.jpg", "scene0247_00/01840.jpg", "scene0247_00/01880.jpg", "scene0247_00/01920.jpg", "scene0247_00/01960.jpg" ], "turn_id": 10, "scan_id": "scene0247_00", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the stainless steel teapot with a black handle and lid\" from your position?", "option": [], "answer": "171.82", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the stainless steel teapot with a black handle and lid\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0047_region21/8f518222854346d1b805bb33a3f4784a_i1_1.jpg", "1mp3d_0047_region21/8f518222854346d1b805bb33a3f4784a_i2_1.jpg", "1mp3d_0047_region21/5a79bc7472c74af1a6a74fc3a6f0dea0_i1_2.jpg", "1mp3d_0047_region21/5a79bc7472c74af1a6a74fc3a6f0dea0_i2_2.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0047_region21", "origin_question": "Remember, have you seen any cabinet(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any cabinet(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "1mp3d_0047_region21/5a79bc7472c74af1a6a74fc3a6f0dea0_i1_3.jpg", "1mp3d_0047_region21/5a79bc7472c74af1a6a74fc3a6f0dea0_i2_3.jpg", "1mp3d_0047_region21/7fa2c356661f48639ace29ba71805d1a_i1_3.jpg", "1mp3d_0047_region21/7fa2c356661f48639ace29ba71805d1a_i2_3.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0047_region21", "origin_question": "In which turn were you closest to \"the rectangular flat-screen TV with a metallic finish\"? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the rectangular flat-screen TV with a metallic finish\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "1mp3d_0047_region21/db60b270524f4038841419f83473d08d_i1_4.jpg", "1mp3d_0047_region21/db60b270524f4038841419f83473d08d_i2_4.jpg", "1mp3d_0047_region21/d8cf4d9317e0460db924bc6a6e4f14b8_i1_4.jpg", "1mp3d_0047_region21/d8cf4d9317e0460db924bc6a6e4f14b8_i2_4.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0047_region21", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "1.48", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0047_region21/db60b270524f4038841419f83473d08d_i1_3.jpg", "1mp3d_0047_region21/db60b270524f4038841419f83473d08d_i2_3.jpg", "1mp3d_0047_region21/7fa2c356661f48639ace29ba71805d1a_i1_2.jpg", "1mp3d_0047_region21/7fa2c356661f48639ace29ba71805d1a_i2_2.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0047_region21", "origin_question": "Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the rectangular flat-screen TV with a metallic finish\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the rectangular flat-screen TV with a metallic finish\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0047_region21/f416e46dfb5c483f9d958d1c119238d0_i1_1.jpg", "1mp3d_0047_region21/f416e46dfb5c483f9d958d1c119238d0_i2_1.jpg", "1mp3d_0047_region21/206999c99e3b4e748b4b8f23145eed99_i1_1.jpg", "1mp3d_0047_region21/206999c99e3b4e748b4b8f23145eed99_i2_1.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0047_region21", "origin_question": "", "option": [], "answer": "", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0047_region21/f416e46dfb5c483f9d958d1c119238d0_i1_2.jpg", "1mp3d_0047_region21/f416e46dfb5c483f9d958d1c119238d0_i2_2.jpg", "1mp3d_0047_region21/f416e46dfb5c483f9d958d1c119238d0_i1_3.jpg", "1mp3d_0047_region21/f416e46dfb5c483f9d958d1c119238d0_i2_3.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0047_region21", "origin_question": "Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the rectangular flat-screen TV with a metallic finish\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the rectangular flat-screen TV with a metallic finish\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0030_region2/333115d876ef49909460d9c3e1a4e0e9_i1_0.jpg", "1mp3d_0030_region2/333115d876ef49909460d9c3e1a4e0e9_i2_0.jpg", "1mp3d_0030_region2/8310b328af0a4096be09f78e30f2c53e_i2_5.jpg", "1mp3d_0030_region2/8310b328af0a4096be09f78e30f2c53e_i1_4.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0030_region2", "origin_question": "Remember, have you seen any chair(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any chair(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "1mp3d_0030_region2/8310b328af0a4096be09f78e30f2c53e_i2_4.jpg", "1mp3d_0030_region2/333115d876ef49909460d9c3e1a4e0e9_i1_5.jpg", "1mp3d_0030_region2/333115d876ef49909460d9c3e1a4e0e9_i2_5.jpg", "1mp3d_0030_region2/f5cf1375de4d46b38328dda8ebcd6ce0_i1_5.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0030_region2", "origin_question": "In which turn were you farthest from \"the bed\"? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the bed\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "1mp3d_0030_region2/f5cf1375de4d46b38328dda8ebcd6ce0_i2_5.jpg", "1mp3d_0030_region2/f5cf1375de4d46b38328dda8ebcd6ce0_i2_3.jpg", "1mp3d_0030_region2/f5cf1375de4d46b38328dda8ebcd6ce0_i1_4.jpg", "1mp3d_0030_region2/f5cf1375de4d46b38328dda8ebcd6ce0_i2_4.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0030_region2", "origin_question": "When did you first discover \"the classic wooden chair with a patterned cushion\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the classic wooden chair with a patterned cushion\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "1mp3d_0030_region2/333115d876ef49909460d9c3e1a4e0e9_i1_4.jpg", "1mp3d_0030_region2/333115d876ef49909460d9c3e1a4e0e9_i2_4.jpg", "1mp3d_0030_region2/333115d876ef49909460d9c3e1a4e0e9_i1_3.jpg", "1mp3d_0030_region2/333115d876ef49909460d9c3e1a4e0e9_i2_3.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0030_region2", "origin_question": "When did you first discover \"the rectangular window with a white frame\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the rectangular window with a white frame\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "1mp3d_0030_region2/333115d876ef49909460d9c3e1a4e0e9_i1_2.jpg", "1mp3d_0030_region2/333115d876ef49909460d9c3e1a4e0e9_i2_2.jpg", "1mp3d_0030_region2/8310b328af0a4096be09f78e30f2c53e_i1_2.jpg", "1mp3d_0030_region2/8310b328af0a4096be09f78e30f2c53e_i2_2.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0030_region2", "origin_question": "What will be the first-time appearance order of the following categories: bed, door, mirror?", "option": [ "bed, mirror, door", "door, bed, mirror", "mirror, door, bed", "bed, door, mirror" ], "answer": "bed, door, mirror", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bed, door, mirror? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['bed, mirror, door', 'door, bed, mirror', 'mirror, door, bed', 'bed, door, mirror'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0047_region17/8f518222854346d1b805bb33a3f4784a_i1_4.jpg", "1mp3d_0047_region17/8f518222854346d1b805bb33a3f4784a_i2_4.jpg", "1mp3d_0047_region17/d5f67d76e52b44aaa5e6e4b21090d86b_i1_5.jpg", "1mp3d_0047_region17/d5f67d76e52b44aaa5e6e4b21090d86b_i2_5.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0047_region17", "origin_question": "Remember, have you seen any remote control(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any remote control(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 3 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0047_region17/d5f67d76e52b44aaa5e6e4b21090d86b_i1_1.jpg", "1mp3d_0047_region17/d5f67d76e52b44aaa5e6e4b21090d86b_i2_1.jpg", "1mp3d_0047_region17/8f518222854346d1b805bb33a3f4784a_i1_3.jpg", "1mp3d_0047_region17/8f518222854346d1b805bb33a3f4784a_i2_3.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0047_region17", "origin_question": "Remember, have you seen any table(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any table(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 3 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "1mp3d_0047_region17/8f518222854346d1b805bb33a3f4784a_i1_2.jpg", "1mp3d_0047_region17/8f518222854346d1b805bb33a3f4784a_i2_2.jpg", "1mp3d_0047_region17/5a79bc7472c74af1a6a74fc3a6f0dea0_i1_5.jpg", "1mp3d_0047_region17/5a79bc7472c74af1a6a74fc3a6f0dea0_i2_5.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0047_region17", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "76.87", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 3 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0072_region10/615acadbd02d4570b997e6021233996a_i1_2.jpg", "1mp3d_0072_region10/615acadbd02d4570b997e6021233996a_i2_2.jpg", "1mp3d_0072_region10/f74428d06fa34b79bcfccb38f448e2c8_i1_1.jpg", "1mp3d_0072_region10/f74428d06fa34b79bcfccb38f448e2c8_i2_1.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0072_region10", "origin_question": "Remember, have you seen any beam(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any beam(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "1mp3d_0072_region10/5f70dea238aa4c8eb85954cd768c55fc_i1_1.jpg", "1mp3d_0072_region10/5f70dea238aa4c8eb85954cd768c55fc_i2_1.jpg", "1mp3d_0072_region10/9b66262361814acda74fb9444039622c_i1_1.jpg", "1mp3d_0072_region10/9b66262361814acda74fb9444039622c_i2_1.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0072_region10", "origin_question": "When did you first discover \"the grass\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the grass\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "1mp3d_0072_region10/9b66262361814acda74fb9444039622c_i1_0.jpg", "1mp3d_0072_region10/9b66262361814acda74fb9444039622c_i2_0.jpg", "1mp3d_0072_region10/5f70dea238aa4c8eb85954cd768c55fc_i1_0.jpg", "1mp3d_0072_region10/5f70dea238aa4c8eb85954cd768c55fc_i2_0.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0072_region10", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "1mp3d_0072_region10/f74428d06fa34b79bcfccb38f448e2c8_i1_5.jpg", "1mp3d_0072_region10/f74428d06fa34b79bcfccb38f448e2c8_i2_5.jpg", "1mp3d_0072_region10/f74428d06fa34b79bcfccb38f448e2c8_i1_0.jpg", "1mp3d_0072_region10/f74428d06fa34b79bcfccb38f448e2c8_i2_0.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0072_region10", "origin_question": "What will be the first-time appearance order of the following categories: door, grass, pool?", "option": [ "pool, grass, door", "grass, door, pool", "door, grass, pool", "door, pool, grass" ], "answer": "grass, door, pool", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: door, grass, pool? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['pool, grass, door', 'grass, door, pool', 'door, grass, pool', 'door, pool, grass'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0072_region10/615acadbd02d4570b997e6021233996a_i1_1.jpg", "1mp3d_0072_region10/615acadbd02d4570b997e6021233996a_i2_1.jpg", "1mp3d_0072_region10/615acadbd02d4570b997e6021233996a_i1_0.jpg", "1mp3d_0072_region10/615acadbd02d4570b997e6021233996a_i2_0.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0072_region10", "origin_question": "Remember, have you seen any monitor(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any monitor(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "1mp3d_0072_region10/54eb4ad83ed34c308336e83d3d20799c_i1_3.jpg", "1mp3d_0072_region10/54eb4ad83ed34c308336e83d3d20799c_i2_3.jpg", "1mp3d_0072_region10/54eb4ad83ed34c308336e83d3d20799c_i1_2.jpg", "1mp3d_0072_region10/54eb4ad83ed34c308336e83d3d20799c_i2_2.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0072_region10", "origin_question": "What will be the first-time appearance order of the following categories: chair, plant, pool?", "option": [ "chair, plant, pool", "chair, pool, plant", "plant, pool, chair", "plant, chair, pool" ], "answer": "plant, chair, pool", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: chair, plant, pool? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['chair, plant, pool', 'chair, pool, plant', 'plant, pool, chair', 'plant, chair, pool'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "1mp3d_0072_region10/d1d21a951d804f5189c5ab3249b63b57_i1_2.jpg", "1mp3d_0072_region10/d1d21a951d804f5189c5ab3249b63b57_i2_2.jpg", "1mp3d_0072_region10/fa1f71b06e894ff6a4e2173012ac7a3b_i1_3.jpg", "1mp3d_0072_region10/fa1f71b06e894ff6a4e2173012ac7a3b_i2_3.jpg" ], "turn_id": 7, "scan_id": "1mp3d_0072_region10", "origin_question": "When did you first discover \"the large rectangular sliding glass door with a black frame\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the large rectangular sliding glass door with a black frame\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0474_05/00000.jpg", "scene0474_05/00040.jpg", "scene0474_05/00080.jpg", "scene0474_05/00120.jpg", "scene0474_05/00160.jpg" ], "turn_id": 1, "scan_id": "scene0474_05", "origin_question": "Remember, have you seen any door(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any door(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0474_05/00200.jpg", "scene0474_05/00240.jpg", "scene0474_05/00320.jpg", "scene0474_05/00360.jpg", "scene0474_05/00400.jpg" ], "turn_id": 2, "scan_id": "scene0474_05", "origin_question": "When did you first discover \"the rectangular, smooth matte blackboard with a metal frame\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the rectangular, smooth matte blackboard with a metal frame\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0474_05/00440.jpg", "scene0474_05/00480.jpg", "scene0474_05/01610.jpg", "scene0474_05/01650.jpg", "scene0474_05/01690.jpg" ], "turn_id": 3, "scan_id": "scene0474_05", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0474_05/01730.jpg", "scene0474_05/01770.jpg", "scene0474_05/01810.jpg", "scene0474_05/01850.jpg", "scene0474_05/01890.jpg" ], "turn_id": 4, "scan_id": "scene0474_05", "origin_question": "At the end of which turn were both of \"the rectangular light brown wooden shelf\" and \"the rectangular, regular white sheet of paper\" on your left side?", "option": [], "answer": "2", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the rectangular light brown wooden shelf\" and \"the rectangular, regular white sheet of paper\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0474_05/01930.jpg", "scene0474_05/01970.jpg", "scene0474_05/02010.jpg", "scene0474_05/02050.jpg", "scene0474_05/02130.jpg" ], "turn_id": 5, "scan_id": "scene0474_05", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the black metal cabinet with three drawers\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the black metal cabinet with three drawers\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0474_05/02430.jpg", "scene0474_05/02470.jpg", "scene0474_05/02510.jpg", "scene0474_05/02550.jpg", "scene0474_05/02600.jpg" ], "turn_id": 6, "scan_id": "scene0474_05", "origin_question": "At the end of which turn were both of \"the rectangular, smooth matte blackboard with a metal frame\" and \"the rectangular white plastic keyboard\" on your left side?", "option": [], "answer": "6", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the rectangular, smooth matte blackboard with a metal frame\" and \"the rectangular white plastic keyboard\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0474_05/02640.jpg", "scene0474_05/02680.jpg", "scene0474_05/02720.jpg", "scene0474_05/02760.jpg", "scene0474_05/02800.jpg" ], "turn_id": 7, "scan_id": "scene0474_05", "origin_question": "Please recall \"the rectangular, regular white sheet of paper\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.38", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the rectangular, regular white sheet of paper\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0474_05/02840.jpg", "scene0474_05/02880.jpg", "scene0474_05/02920.jpg", "scene0474_05/02960.jpg", "scene0474_05/03000.jpg" ], "turn_id": 8, "scan_id": "scene0474_05", "origin_question": "When did you first discover \"the light brown wooden door\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the light brown wooden door\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0678_02/00000.jpg", "scene0678_02/00040.jpg", "scene0678_02/00080.jpg", "scene0678_02/00120.jpg", "scene0678_02/00160.jpg" ], "turn_id": 1, "scan_id": "scene0678_02", "origin_question": "Remember, how many bag(s) have you seen so far? ", "option": [], "answer": "4", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many bag(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0678_02/00200.jpg", "scene0678_02/00240.jpg", "scene0678_02/00280.jpg", "scene0678_02/00320.jpg", "scene0678_02/00360.jpg" ], "turn_id": 2, "scan_id": "scene0678_02", "origin_question": "Please recall \"the small rectangular white plastic basket\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.96", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the small rectangular white plastic basket\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0678_02/00400.jpg", "scene0678_02/00440.jpg", "scene0678_02/00480.jpg", "scene0678_02/00520.jpg", "scene0678_02/00560.jpg" ], "turn_id": 3, "scan_id": "scene0678_02", "origin_question": "Is the white oscillating fan with a rounded shape to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the white oscillating fan with a rounded shape to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "scene0678_02/00600.jpg", "scene0678_02/00640.jpg", "scene0678_02/00680.jpg", "scene0678_02/00720.jpg", "scene0678_02/00760.jpg" ], "turn_id": 4, "scan_id": "scene0678_02", "origin_question": "Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Right", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0678_02/00800.jpg", "scene0678_02/00840.jpg", "scene0678_02/00880.jpg", "scene0678_02/00920.jpg", "scene0678_02/00960.jpg" ], "turn_id": 5, "scan_id": "scene0678_02", "origin_question": "Which direction is the white oscillating fan with a rounded shape in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the white oscillating fan with a rounded shape in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0678_02/01000.jpg", "scene0678_02/01040.jpg", "scene0678_02/01080.jpg", "scene0678_02/01120.jpg", "scene0678_02/01160.jpg" ], "turn_id": 6, "scan_id": "scene0678_02", "origin_question": "Which direction is the large stainless-steel sink with a smooth surface in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the large stainless-steel sink with a smooth surface in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "scene0678_02/01200.jpg", "scene0678_02/01240.jpg", "scene0678_02/01280.jpg", "scene0678_02/01320.jpg", "scene0678_02/01360.jpg" ], "turn_id": 7, "scan_id": "scene0678_02", "origin_question": "Among these three objects(\"the red cylindrical mesh hamper\",\"the large stainless-steel sink with a smooth surface\",\"the rectangular metal fire extinguisher with a white frame\" ), which one is the closest to you?", "option": [ "the red cylindrical mesh hamper", "the large stainless-steel sink with a smooth surface", "the rectangular metal fire extinguisher with a white frame" ], "answer": "the rectangular metal fire extinguisher with a white frame", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the red cylindrical mesh hamper\",\"the large stainless-steel sink with a smooth surface\",\"the rectangular metal fire extinguisher with a white frame\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the red cylindrical mesh hamper', 'the large stainless-steel sink with a smooth surface', 'the rectangular metal fire extinguisher with a white frame'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0678_02/01400.jpg", "scene0678_02/01440.jpg", "scene0678_02/01480.jpg", "scene0678_02/01520.jpg", "scene0678_02/01560.jpg" ], "turn_id": 8, "scan_id": "scene0678_02", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the folded white cotton towel\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the folded white cotton towel\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0678_02/01600.jpg", "scene0678_02/01640.jpg", "scene0678_02/01680.jpg", "scene0678_02/01720.jpg", "scene0678_02/01760.jpg" ], "turn_id": 9, "scan_id": "scene0678_02", "origin_question": "Compared to your position at the end of the 5 turn, are you now closer or farther away from \"the large rectangular green stone table\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 5 turn, are you now closer or farther away from \"the large rectangular green stone table\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0678_02/01800.jpg", "scene0678_02/01840.jpg", "scene0678_02/01880.jpg", "scene0678_02/01920.jpg", "scene0678_02/01960.jpg" ], "turn_id": 10, "scan_id": "scene0678_02", "origin_question": "Is the rectangular metal fire extinguisher with a white frame to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the rectangular metal fire extinguisher with a white frame to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0578_01/00000.jpg", "scene0578_01/00040.jpg", "scene0578_01/00080.jpg", "scene0578_01/00120.jpg" ], "turn_id": 1, "scan_id": "scene0578_01", "origin_question": "Remember, how many blackboard(s) have you seen so far? ", "option": [], "answer": "2", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many blackboard(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0578_01/00160.jpg", "scene0578_01/00200.jpg", "scene0578_01/00240.jpg", "scene0578_01/00280.jpg" ], "turn_id": 2, "scan_id": "scene0578_01", "origin_question": "In which turn were you farthest from \"the window\"? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the window\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0578_01/00320.jpg", "scene0578_01/00360.jpg", "scene0578_01/00400.jpg", "scene0578_01/00440.jpg" ], "turn_id": 3, "scan_id": "scene0578_01", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the black metal monitor with a silver base\" from your position?", "option": [], "answer": "141.38", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the black metal monitor with a silver base\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0578_01/00480.jpg", "scene0578_01/00520.jpg", "scene0578_01/00560.jpg", "scene0578_01/00600.jpg" ], "turn_id": 4, "scan_id": "scene0578_01", "origin_question": "At the end of which turn were both of \"the small rectangular white keyboard\" and \"the blue backpack with a white emblem and pockets\" on your left side?", "option": [], "answer": "3", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the small rectangular white keyboard\" and \"the blue backpack with a white emblem and pockets\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0578_01/00640.jpg", "scene0578_01/00680.jpg", "scene0578_01/00720.jpg", "scene0578_01/00760.jpg" ], "turn_id": 5, "scan_id": "scene0578_01", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the black metal monitor with a silver base\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the black metal monitor with a silver base\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0578_01/00800.jpg", "scene0578_01/00850.jpg", "scene0578_01/00890.jpg", "scene0578_01/00930.jpg" ], "turn_id": 6, "scan_id": "scene0578_01", "origin_question": "Which two objects are on the same side of you now? \"the white plastic computer mouse with a simple design\", \"the blue backpack with a white emblem and pockets\", \"the rectangular light brown wooden door with a glass window\". ", "option": [ "the white plastic computer mouse with a simple design, the blue backpack with a white emblem and pockets", "the white plastic computer mouse with a simple design, the rectangular light brown wooden door with a glass window", "the blue backpack with a white emblem and pockets, the rectangular light brown wooden door with a glass window" ], "answer": "the blue backpack with a white emblem and pockets, the rectangular light brown wooden door with a glass window", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the white plastic computer mouse with a simple design\", \"the blue backpack with a white emblem and pockets\", \"the rectangular light brown wooden door with a glass window\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the white plastic computer mouse with a simple design, the blue backpack with a white emblem and pockets', 'the white plastic computer mouse with a simple design, the rectangular light brown wooden door with a glass window', 'the blue backpack with a white emblem and pockets, the rectangular light brown wooden door with a glass window'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0578_01/00970.jpg", "scene0578_01/01010.jpg", "scene0578_01/01050.jpg", "scene0578_01/01090.jpg" ], "turn_id": 7, "scan_id": "scene0578_01", "origin_question": "By the end of the 4 turn, were you getting closer to or farther away from these two objects: X: \"the black metal monitor with a silver base\", Y: \"the cylindrical black plastic bin\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to X and farther away from Y", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 4 turn, were you getting closer to or farther away from these two objects: X: \"the black metal monitor with a silver base\", Y: \"the cylindrical black plastic bin\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0029_region9/fdab6422162e49db822a37178ab70481_i1_0.jpg", "1mp3d_0029_region9/fdab6422162e49db822a37178ab70481_i2_0.jpg", "1mp3d_0029_region9/7d9664bc035641d995b523a19650d178_i1_1.jpg", "1mp3d_0029_region9/7d9664bc035641d995b523a19650d178_i2_1.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0029_region9", "origin_question": "Remember, have you seen any door(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any door(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "1mp3d_0029_region9/7d9664bc035641d995b523a19650d178_i1_4.jpg", "1mp3d_0029_region9/7d9664bc035641d995b523a19650d178_i2_4.jpg", "1mp3d_0029_region9/7d9664bc035641d995b523a19650d178_i1_3.jpg", "1mp3d_0029_region9/7d9664bc035641d995b523a19650d178_i2_3.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0029_region9", "origin_question": "Is the round white metal light fixture to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the round white metal light fixture to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0029_region9/fdab6422162e49db822a37178ab70481_i1_5.jpg", "1mp3d_0029_region9/fdab6422162e49db822a37178ab70481_i2_5.jpg", "1mp3d_0029_region9/99b1210b63c94f9184a9f06032a2ea4a_i1_4.jpg", "1mp3d_0029_region9/99b1210b63c94f9184a9f06032a2ea4a_i2_4.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0029_region9", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the smooth white rectangular wooden door with panels\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the smooth white rectangular wooden door with panels\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0029_region9/99b1210b63c94f9184a9f06032a2ea4a_i1_5.jpg", "1mp3d_0029_region9/99b1210b63c94f9184a9f06032a2ea4a_i2_5.jpg", "1mp3d_0029_region9/fdab6422162e49db822a37178ab70481_i1_1.jpg", "1mp3d_0029_region9/fdab6422162e49db822a37178ab70481_i2_1.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0029_region9", "origin_question": "", "option": [], "answer": "", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0029_region9/fdab6422162e49db822a37178ab70481_i1_2.jpg", "1mp3d_0029_region9/fdab6422162e49db822a37178ab70481_i2_2.jpg", "1mp3d_0029_region9/fdab6422162e49db822a37178ab70481_i1_3.jpg", "1mp3d_0029_region9/fdab6422162e49db822a37178ab70481_i2_3.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0029_region9", "origin_question": "Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the smooth white rectangular wooden door with panels\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the smooth white rectangular wooden door with panels\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0029_region9/99b1210b63c94f9184a9f06032a2ea4a_i1_2.jpg", "1mp3d_0029_region9/99b1210b63c94f9184a9f06032a2ea4a_i2_2.jpg", "1mp3d_0029_region9/99b1210b63c94f9184a9f06032a2ea4a_i1_1.jpg", "1mp3d_0029_region9/99b1210b63c94f9184a9f06032a2ea4a_i2_1.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0029_region9", "origin_question": "", "option": [], "answer": "", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "1mp3d_0044_region3/6e2a63a762af433ba81e3eb6dfe0c73f_i1_2.jpg", "1mp3d_0044_region3/6e2a63a762af433ba81e3eb6dfe0c73f_i2_2.jpg", "1mp3d_0044_region3/fe8cd80fdfd14aefb885e48408384db6_i1_2.jpg", "1mp3d_0044_region3/fe8cd80fdfd14aefb885e48408384db6_i2_2.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0044_region3", "origin_question": "Remember, how many piano(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many piano(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0044_region3/e004abe1362144c08cbbf2ff4dd08eff_i1_2.jpg", "1mp3d_0044_region3/e004abe1362144c08cbbf2ff4dd08eff_i2_2.jpg", "1mp3d_0044_region3/fe8cd80fdfd14aefb885e48408384db6_i1_1.jpg", "1mp3d_0044_region3/fe8cd80fdfd14aefb885e48408384db6_i2_1.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0044_region3", "origin_question": "Remember, have you seen any clothes(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any clothes(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "1mp3d_0044_region3/fe8cd80fdfd14aefb885e48408384db6_i1_0.jpg", "1mp3d_0044_region3/fe8cd80fdfd14aefb885e48408384db6_i2_0.jpg", "1mp3d_0044_region3/e004abe1362144c08cbbf2ff4dd08eff_i1_5.jpg", "1mp3d_0044_region3/e004abe1362144c08cbbf2ff4dd08eff_i2_5.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0044_region3", "origin_question": "How far is your current position from where you were at the end of the turn 1? (in meters)", "option": [], "answer": "1.12", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 1? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "1mp3d_0044_region3/e004abe1362144c08cbbf2ff4dd08eff_i1_4.jpg", "1mp3d_0044_region3/e004abe1362144c08cbbf2ff4dd08eff_i2_4.jpg", "1mp3d_0044_region3/003ac44fc17f43f8b2bfb5054ea65d6d_i1_3.jpg", "1mp3d_0044_region3/003ac44fc17f43f8b2bfb5054ea65d6d_i2_3.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0044_region3", "origin_question": "Assuming the direction you are facing at the end of the turn 3 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Right", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 3 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "1mp3d_0044_region3/ff647edc697b4c90b3290d9330d677db_i1_2.jpg", "1mp3d_0044_region3/ff647edc697b4c90b3290d9330d677db_i2_2.jpg", "1mp3d_0044_region3/fccf31238735441097bb9b462aac34f7_i1_1.jpg", "1mp3d_0044_region3/fccf31238735441097bb9b462aac34f7_i2_1.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0044_region3", "origin_question": "Using your orientation at the end of turn 4 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "81.9", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 4 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0044_region3/fccf31238735441097bb9b462aac34f7_i1_0.jpg", "1mp3d_0044_region3/fccf31238735441097bb9b462aac34f7_i2_0.jpg", "1mp3d_0044_region3/ff647edc697b4c90b3290d9330d677db_i1_3.jpg", "1mp3d_0044_region3/ff647edc697b4c90b3290d9330d677db_i2_3.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0044_region3", "origin_question": "", "option": [], "answer": "", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "1mp3d_0057_region4/6a289d9b67cd43f4b59aab2037af221b_i1_5.jpg", "1mp3d_0057_region4/6a289d9b67cd43f4b59aab2037af221b_i2_5.jpg", "1mp3d_0057_region4/50e5eb51e92446a78ccfb851123e801d_i1_5.jpg", "1mp3d_0057_region4/50e5eb51e92446a78ccfb851123e801d_i2_5.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0057_region4", "origin_question": "Remember, how many chandelier(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many chandelier(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "1mp3d_0057_region4/cae5bc03a39a4790aa93f8ce3a55da06_i1_5.jpg", "1mp3d_0057_region4/cae5bc03a39a4790aa93f8ce3a55da06_i2_5.jpg", "1mp3d_0057_region4/d0a970634a7c498d93a69eddf46f3f3f_i1_4.jpg", "1mp3d_0057_region4/d0a970634a7c498d93a69eddf46f3f3f_i2_4.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0057_region4", "origin_question": "In which turn were you farthest from \"the rectangular flat-screen TV with a sleek black border\"? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the rectangular flat-screen TV with a sleek black border\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0057_region4/728392f2d97740438342b1a3b917f971_i1_3.jpg", "1mp3d_0057_region4/728392f2d97740438342b1a3b917f971_i2_3.jpg", "1mp3d_0057_region4/c57bb84a32be4c9a8f3a0661ce990ac8_i1_1.jpg", "1mp3d_0057_region4/c57bb84a32be4c9a8f3a0661ce990ac8_i2_1.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0057_region4", "origin_question": "", "option": [], "answer": "", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "1mp3d_0057_region4/728392f2d97740438342b1a3b917f971_i1_4.jpg", "1mp3d_0057_region4/728392f2d97740438342b1a3b917f971_i2_4.jpg", "1mp3d_0057_region4/d0a970634a7c498d93a69eddf46f3f3f_i1_5.jpg", "1mp3d_0057_region4/d0a970634a7c498d93a69eddf46f3f3f_i2_5.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0057_region4", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "1mp3d_0057_region4/7623ec113f9e464fa9c2974c8841896b_i1_4.jpg", "1mp3d_0057_region4/7623ec113f9e464fa9c2974c8841896b_i2_4.jpg", "1mp3d_0057_region4/faa32c91d8e341c38b633e5a3acac2eb_i1_5.jpg", "1mp3d_0057_region4/faa32c91d8e341c38b633e5a3acac2eb_i2_5.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0057_region4", "origin_question": "When did you first discover \"the medium-sized rectangular wooden door in brown\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the medium-sized rectangular wooden door in brown\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "1mp3d_0057_region4/f3c0b6536ff04efaae476fd547aea850_i1_2.jpg", "1mp3d_0057_region4/ff14bec47e17474dbdbfeccfe40e5d39_i1_2.jpg", "1mp3d_0057_region4/ff14bec47e17474dbdbfeccfe40e5d39_i1_1.jpg", "1mp3d_0057_region4/f3c0b6536ff04efaae476fd547aea850_i1_1.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0057_region4", "origin_question": "Assuming the direction you are facing at the end of the turn 4 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Left", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 4 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "1mp3d_0057_region4/f3c0b6536ff04efaae476fd547aea850_i2_1.jpg", "1mp3d_0057_region4/faa32c91d8e341c38b633e5a3acac2eb_i1_4.jpg", "1mp3d_0057_region4/faa32c91d8e341c38b633e5a3acac2eb_i2_4.jpg", "1mp3d_0057_region4/aeccd4a96d1e4d819bc5f59bff1d156d_i1_5.jpg" ], "turn_id": 7, "scan_id": "1mp3d_0057_region4", "origin_question": "Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "51.36", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "1mp3d_0057_region4/aeccd4a96d1e4d819bc5f59bff1d156d_i2_5.jpg", "1mp3d_0057_region4/f7bfe15839b94a548dec94d2875b9cc3_i1_1.jpg", "1mp3d_0057_region4/f7bfe15839b94a548dec94d2875b9cc3_i2_1.jpg", "1mp3d_0057_region4/aeccd4a96d1e4d819bc5f59bff1d156d_i1_0.jpg" ], "turn_id": 8, "scan_id": "1mp3d_0057_region4", "origin_question": "Assuming the direction you are facing at the end of the turn 5 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Left", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 5 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0685_02/00000.jpg", "scene0685_02/00040.jpg", "scene0685_02/00080.jpg", "scene0685_02/00120.jpg", "scene0685_02/00160.jpg" ], "turn_id": 1, "scan_id": "scene0685_02", "origin_question": "Remember, have you seen any picture(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any picture(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0685_02/00200.jpg", "scene0685_02/00240.jpg", "scene0685_02/00280.jpg", "scene0685_02/00320.jpg", "scene0685_02/00360.jpg" ], "turn_id": 2, "scan_id": "scene0685_02", "origin_question": "At the end of which turn were both of \"the sturdy green rectangular plastic bin\" and \"the rectangular black metal table with a reflective surface\" on your left side?", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the sturdy green rectangular plastic bin\" and \"the rectangular black metal table with a reflective surface\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0685_02/00400.jpg", "scene0685_02/00440.jpg", "scene0685_02/00480.jpg", "scene0685_02/00520.jpg", "scene0685_02/00560.jpg" ], "turn_id": 3, "scan_id": "scene0685_02", "origin_question": "Is the sturdy green rectangular plastic bin to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the sturdy green rectangular plastic bin to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0685_02/00600.jpg", "scene0685_02/00640.jpg", "scene0685_02/00680.jpg", "scene0685_02/00720.jpg", "scene0685_02/00760.jpg" ], "turn_id": 4, "scan_id": "scene0685_02", "origin_question": "In which turn were you farthest from \"the rectangular, chrome-plated stainless steel shelf\"? ", "option": [], "answer": "2", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the rectangular, chrome-plated stainless steel shelf\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0685_02/00800.jpg", "scene0685_02/00840.jpg", "scene0685_02/00880.jpg", "scene0685_02/00920.jpg", "scene0685_02/00960.jpg" ], "turn_id": 5, "scan_id": "scene0685_02", "origin_question": "Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0685_02/01000.jpg", "scene0685_02/01040.jpg", "scene0685_02/01080.jpg", "scene0685_02/01120.jpg", "scene0685_02/01160.jpg" ], "turn_id": 6, "scan_id": "scene0685_02", "origin_question": "What will be the first-time appearance order of the following categories: bin, paper, shelf?", "option": [ "paper, bin, shelf", "bin, paper, shelf", "paper, shelf, bin", "bin, shelf, paper" ], "answer": "bin, shelf, paper", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bin, paper, shelf? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['paper, bin, shelf', 'bin, paper, shelf', 'paper, shelf, bin', 'bin, shelf, paper'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0685_02/01200.jpg", "scene0685_02/01240.jpg", "scene0685_02/01280.jpg", "scene0685_02/01320.jpg", "scene0685_02/01360.jpg" ], "turn_id": 7, "scan_id": "scene0685_02", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the large, flat, rectangular board\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the large, flat, rectangular board\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0685_02/01400.jpg", "scene0685_02/01440.jpg", "scene0685_02/01480.jpg", "scene0685_02/01520.jpg", "scene0685_02/01560.jpg" ], "turn_id": 8, "scan_id": "scene0685_02", "origin_question": "Please recall \"the cylindrical pink bottle with a red nozzle\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.05", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the cylindrical pink bottle with a red nozzle\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0685_02/01600.jpg", "scene0685_02/01640.jpg", "scene0685_02/01680.jpg", "scene0685_02/01720.jpg", "scene0685_02/01760.jpg" ], "turn_id": 9, "scan_id": "scene0685_02", "origin_question": "In which turn did you see both \"the rectangular, deep brown wooden door\" and \"the cylindrical pink bottle with a red nozzle\" simultaneously? ", "option": [], "answer": "7", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the rectangular, deep brown wooden door\" and \"the cylindrical pink bottle with a red nozzle\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0685_02/01800.jpg", "scene0685_02/01840.jpg", "scene0685_02/01880.jpg", "scene0685_02/01920.jpg", "scene0685_02/01960.jpg" ], "turn_id": 10, "scan_id": "scene0685_02", "origin_question": "What will be the first-time appearance order of the following categories: bottle, picture, shelf?", "option": [ "bottle, shelf, picture", "bottle, picture, shelf", "shelf, bottle, picture", "picture, shelf, bottle" ], "answer": "picture, shelf, bottle", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bottle, picture, shelf? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['bottle, shelf, picture', 'bottle, picture, shelf', 'shelf, bottle, picture', 'picture, shelf, bottle'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0703_01/00000.jpg", "scene0703_01/00030.jpg", "scene0703_01/00060.jpg", "scene0703_01/00090.jpg" ], "turn_id": 1, "scan_id": "scene0703_01", "origin_question": "Remember, have you seen any door(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any door(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0703_01/00120.jpg", "scene0703_01/00150.jpg", "scene0703_01/00180.jpg", "scene0703_01/00210.jpg" ], "turn_id": 2, "scan_id": "scene0703_01", "origin_question": "At the end of which turn were both of \"the pink rectangular sheet of thin paper\" and \"the small white rectangular plastic switch\" on your left side?", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the pink rectangular sheet of thin paper\" and \"the small white rectangular plastic switch\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0703_01/00240.jpg", "scene0703_01/00270.jpg", "scene0703_01/00300.jpg", "scene0703_01/00330.jpg" ], "turn_id": 3, "scan_id": "scene0703_01", "origin_question": "Please recall \"the small white rectangular plastic switch\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.46", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the small white rectangular plastic switch\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0703_01/00360.jpg", "scene0703_01/00390.jpg", "scene0703_01/00420.jpg", "scene0703_01/00450.jpg" ], "turn_id": 4, "scan_id": "scene0703_01", "origin_question": "Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "102.69", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0703_01/00480.jpg", "scene0703_01/00510.jpg", "scene0703_01/00540.jpg", "scene0703_01/00570.jpg" ], "turn_id": 5, "scan_id": "scene0703_01", "origin_question": "When did you first discover \"the pink rectangular sheet of thin paper\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the pink rectangular sheet of thin paper\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0703_01/00600.jpg", "scene0703_01/00630.jpg", "scene0703_01/00660.jpg", "scene0703_01/00690.jpg" ], "turn_id": 6, "scan_id": "scene0703_01", "origin_question": "When did you first discover \"the small white rectangular plastic switch\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the small white rectangular plastic switch\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0040_region52/22371c52a50b48419ce5042d631eb1a6_i1_3.jpg", "1mp3d_0040_region52/22371c52a50b48419ce5042d631eb1a6_i2_3.jpg", "1mp3d_0040_region52/c9e3b241fc8c4bad8b8d5d9e3192ae0a_i1_3.jpg", "1mp3d_0040_region52/c9e3b241fc8c4bad8b8d5d9e3192ae0a_i2_3.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0040_region52", "origin_question": "Remember, have you seen any stairs(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any stairs(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0040_region52/22371c52a50b48419ce5042d631eb1a6_i1_2.jpg", "1mp3d_0040_region52/22371c52a50b48419ce5042d631eb1a6_i2_2.jpg", "1mp3d_0040_region52/22371c52a50b48419ce5042d631eb1a6_i1_1.jpg", "1mp3d_0040_region52/22371c52a50b48419ce5042d631eb1a6_i2_1.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0040_region52", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0040_region52/22371c52a50b48419ce5042d631eb1a6_i1_0.jpg", "1mp3d_0040_region52/22371c52a50b48419ce5042d631eb1a6_i2_0.jpg", "1mp3d_0040_region52/c9e3b241fc8c4bad8b8d5d9e3192ae0a_i1_0.jpg", "1mp3d_0040_region52/c9e3b241fc8c4bad8b8d5d9e3192ae0a_i2_0.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0040_region52", "origin_question": "", "option": [], "answer": "", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "1mp3d_0040_region52/c9e3b241fc8c4bad8b8d5d9e3192ae0a_i1_5.jpg", "1mp3d_0040_region52/c9e3b241fc8c4bad8b8d5d9e3192ae0a_i2_5.jpg", "1mp3d_0040_region52/e4ee9fc1de3f44779759a399b5c0ed7b_i1_5.jpg", "1mp3d_0040_region52/e4ee9fc1de3f44779759a399b5c0ed7b_i2_5.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0040_region52", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "1mp3d_0040_region52/2d03c8e5efeb4f729b4e14bc84a6fc63_i1_5.jpg", "1mp3d_0040_region52/2d03c8e5efeb4f729b4e14bc84a6fc63_i2_5.jpg", "1mp3d_0040_region52/b5dc366ccba3406db2e04b8e1fd2270e_i1_5.jpg", "1mp3d_0040_region52/b5dc366ccba3406db2e04b8e1fd2270e_i2_5.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0040_region52", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "2.55", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0040_region52/2b9575c049fc4539967fffe928f95c2f_i1_0.jpg", "1mp3d_0040_region52/2b9575c049fc4539967fffe928f95c2f_i2_0.jpg", "1mp3d_0040_region52/2b9575c049fc4539967fffe928f95c2f_i1_4.jpg", "1mp3d_0040_region52/2b9575c049fc4539967fffe928f95c2f_i2_4.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0040_region52", "origin_question": "", "option": [], "answer": "", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "1mp3d_0040_region52/2b9575c049fc4539967fffe928f95c2f_i1_5.jpg", "1mp3d_0040_region52/2b9575c049fc4539967fffe928f95c2f_i2_5.jpg", "1mp3d_0040_region52/b5dc366ccba3406db2e04b8e1fd2270e_i1_0.jpg", "1mp3d_0040_region52/b5dc366ccba3406db2e04b8e1fd2270e_i2_0.jpg" ], "turn_id": 7, "scan_id": "1mp3d_0040_region52", "origin_question": "Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "83.39", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "1mp3d_0040_region52/b5dc366ccba3406db2e04b8e1fd2270e_i1_4.jpg", "1mp3d_0040_region52/b5dc366ccba3406db2e04b8e1fd2270e_i2_4.jpg", "1mp3d_0040_region52/2d03c8e5efeb4f729b4e14bc84a6fc63_i1_0.jpg", "1mp3d_0040_region52/2d03c8e5efeb4f729b4e14bc84a6fc63_i2_0.jpg" ], "turn_id": 8, "scan_id": "1mp3d_0040_region52", "origin_question": "Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0067_region22/fd7ac20630e24aff97bb73f292d0da12_i1_5.jpg", "1mp3d_0067_region22/fd7ac20630e24aff97bb73f292d0da12_i2_5.jpg", "1mp3d_0067_region22/f4939bf6f00a4864832a358f1ea8394e_i1_1.jpg", "1mp3d_0067_region22/f4939bf6f00a4864832a358f1ea8394e_i2_1.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0067_region22", "origin_question": "Remember, have you seen any picture(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any picture(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "1mp3d_0067_region22/fd7ac20630e24aff97bb73f292d0da12_i1_4.jpg", "1mp3d_0067_region22/fd7ac20630e24aff97bb73f292d0da12_i2_4.jpg", "1mp3d_0067_region22/fd7ac20630e24aff97bb73f292d0da12_i1_3.jpg", "1mp3d_0067_region22/fd7ac20630e24aff97bb73f292d0da12_i2_3.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0067_region22", "origin_question": "In which turn were you farthest from \"the white ceramic toilet with a round bowl\"? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the white ceramic toilet with a round bowl\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "1mp3d_0067_region22/f4939bf6f00a4864832a358f1ea8394e_i1_0.jpg", "1mp3d_0067_region22/f4939bf6f00a4864832a358f1ea8394e_i2_0.jpg", "1mp3d_0067_region22/41e8be0d514945b1a74adb3308597196_i1_5.jpg", "1mp3d_0067_region22/41e8be0d514945b1a74adb3308597196_i2_5.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0067_region22", "origin_question": "In which turn did you see both \"the medium-sized rectangular picture with soft tones\" and \"the standard-sized white cylindrical cup made of glass\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the medium-sized rectangular picture with soft tones\" and \"the standard-sized white cylindrical cup made of glass\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "1mp3d_0067_region22/f4939bf6f00a4864832a358f1ea8394e_i1_4.jpg", "1mp3d_0067_region22/f4939bf6f00a4864832a358f1ea8394e_i2_4.jpg", "1mp3d_0067_region22/41e8be0d514945b1a74adb3308597196_i2_4.jpg", "1mp3d_0067_region22/e8527069c00e4d929f5eece0d766ebd8_i1_4.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0067_region22", "origin_question": "What will be the first-time appearance order of the following categories: door, toilet, window?", "option": [ "toilet, window, door", "door, window, toilet", "door, toilet, window", "toilet, door, window" ], "answer": "toilet, window, door", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: door, toilet, window? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['toilet, window, door', 'door, window, toilet', 'door, toilet, window', 'toilet, door, window'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "1mp3d_0067_region22/e8527069c00e4d929f5eece0d766ebd8_i2_4.jpg", "1mp3d_0067_region22/e8527069c00e4d929f5eece0d766ebd8_i1_5.jpg", "1mp3d_0067_region22/e8527069c00e4d929f5eece0d766ebd8_i2_5.jpg", "1mp3d_0067_region22/41e8be0d514945b1a74adb3308597196_i1_2.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0067_region22", "origin_question": "What will be the first-time appearance order of the following categories: cup, door, window?", "option": [ "window, door, cup", "cup, door, window", "cup, window, door", "window, cup, door" ], "answer": "cup, window, door", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: cup, door, window? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['window, door, cup', 'cup, door, window', 'cup, window, door', 'window, cup, door'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0044_region22/628a2cb0bb6d483cbc726c8aacd90bf4_i1_2.jpg", "1mp3d_0044_region22/628a2cb0bb6d483cbc726c8aacd90bf4_i2_2.jpg", "1mp3d_0044_region22/96ce4a25cc054509b9aff5cf697b716c_i1_1.jpg", "1mp3d_0044_region22/96ce4a25cc054509b9aff5cf697b716c_i2_1.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0044_region22", "origin_question": "Remember, have you seen any cushion(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any cushion(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "1mp3d_0044_region22/628a2cb0bb6d483cbc726c8aacd90bf4_i1_1.jpg", "1mp3d_0044_region22/628a2cb0bb6d483cbc726c8aacd90bf4_i2_1.jpg", "1mp3d_0044_region22/f66c39dfa41346939c8253266b130559_i1_2.jpg", "1mp3d_0044_region22/f66c39dfa41346939c8253266b130559_i2_2.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0044_region22", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the large, round black metal chandelier\"; \"the framed vineyard picture with a dark decorative frame\"; \"the sturdy, dark wooden or metal banister with a curved top\" ", "option": [ "the large, round black metal chandelier", "the framed vineyard picture with a dark decorative frame", "the sturdy, dark wooden or metal banister with a curved top" ], "answer": "the framed vineyard picture with a dark decorative frame", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the large, round black metal chandelier\"; \"the framed vineyard picture with a dark decorative frame\"; \"the sturdy, dark wooden or metal banister with a curved top\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the large, round black metal chandelier', 'the framed vineyard picture with a dark decorative frame', 'the sturdy, dark wooden or metal banister with a curved top'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "1mp3d_0044_region22/95906009592d4f8eadc55e48507593be_i1_3.jpg", "1mp3d_0044_region22/95906009592d4f8eadc55e48507593be_i2_3.jpg", "1mp3d_0044_region22/54412fe5b92e49239cb028c9234aaa0c_i1_5.jpg", "1mp3d_0044_region22/54412fe5b92e49239cb028c9234aaa0c_i2_5.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0044_region22", "origin_question": "How far is your current position from where you were at the end of the turn 1? (in meters)", "option": [], "answer": "2.77", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 1? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "1mp3d_0044_region22/81fbf84311b34c679917c3712d163618_i1_3.jpg", "1mp3d_0044_region22/81fbf84311b34c679917c3712d163618_i2_3.jpg", "1mp3d_0044_region22/81fbf84311b34c679917c3712d163618_i1_2.jpg", "1mp3d_0044_region22/81fbf84311b34c679917c3712d163618_i2_2.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0044_region22", "origin_question": "In which turn did you see both \"the large, round black metal chandelier\" and \"the framed vineyard picture with a dark decorative frame\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the large, round black metal chandelier\" and \"the framed vineyard picture with a dark decorative frame\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "1mp3d_0044_region22/54412fe5b92e49239cb028c9234aaa0c_i1_4.jpg", "1mp3d_0044_region22/54412fe5b92e49239cb028c9234aaa0c_i2_4.jpg", "1mp3d_0044_region22/54412fe5b92e49239cb028c9234aaa0c_i1_3.jpg", "1mp3d_0044_region22/54412fe5b92e49239cb028c9234aaa0c_i2_3.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0044_region22", "origin_question": "By the end of the 4 turn, were you getting closer to or farther away from these two objects: X: \"the lush green leafy plant in a dark-colored pot\", Y: \"the framed vineyard picture with a dark decorative frame\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "farther away from both X and Y", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 4 turn, were you getting closer to or farther away from these two objects: X: \"the lush green leafy plant in a dark-colored pot\", Y: \"the framed vineyard picture with a dark decorative frame\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "1mp3d_0044_region22/95906009592d4f8eadc55e48507593be_i1_2.jpg", "1mp3d_0044_region22/95906009592d4f8eadc55e48507593be_i2_2.jpg", "1mp3d_0044_region22/f66c39dfa41346939c8253266b130559_i2_0.jpg", "1mp3d_0044_region22/f66c39dfa41346939c8253266b130559_i1_1.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0044_region22", "origin_question": "At the end of which turn were both of \"the lush green leafy plant in a dark-colored pot\" and \"the framed vineyard picture with a dark decorative frame\" on your right side?", "option": [], "answer": "5", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the lush green leafy plant in a dark-colored pot\" and \"the framed vineyard picture with a dark decorative frame\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0084_region12/7d708a5b80ee45979870bae83b2bdd44_i1_0.jpg", "1mp3d_0084_region12/7d708a5b80ee45979870bae83b2bdd44_i1_4.jpg", "1mp3d_0084_region12/7d708a5b80ee45979870bae83b2bdd44_i2_4.jpg", "1mp3d_0084_region12/7d708a5b80ee45979870bae83b2bdd44_i2_5.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0084_region12", "origin_question": "Remember, have you seen any switch(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any switch(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 2 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0084_region12/b2af08ee382d4746b235a45d0a7032bb_i1_2.jpg", "1mp3d_0084_region12/b2af08ee382d4746b235a45d0a7032bb_i2_2.jpg", "1mp3d_0084_region12/b2af08ee382d4746b235a45d0a7032bb_i1_1.jpg", "1mp3d_0084_region12/b2af08ee382d4746b235a45d0a7032bb_i2_1.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0084_region12", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the small rectangular white plastic switch\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the small rectangular white plastic switch\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 2 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0593_01/00000.jpg", "scene0593_01/00040.jpg", "scene0593_01/00080.jpg", "scene0593_01/00120.jpg" ], "turn_id": 1, "scan_id": "scene0593_01", "origin_question": "Remember, have you seen any backpack(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any backpack(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0593_01/00160.jpg", "scene0593_01/00200.jpg", "scene0593_01/00240.jpg", "scene0593_01/00280.jpg" ], "turn_id": 2, "scan_id": "scene0593_01", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the couch\"; \"the medium-sized red and black backpack with a strap\"; \"the table\" ", "option": [ "the couch", "the medium-sized red and black backpack with a strap", "the table" ], "answer": "the table", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the couch\"; \"the medium-sized red and black backpack with a strap\"; \"the table\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the couch', 'the medium-sized red and black backpack with a strap', 'the table'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0593_01/00320.jpg", "scene0593_01/00360.jpg", "scene0593_01/00400.jpg", "scene0593_01/00440.jpg" ], "turn_id": 3, "scan_id": "scene0593_01", "origin_question": "Is the table to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the table to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0593_01/00480.jpg", "scene0593_01/00520.jpg", "scene0593_01/00560.jpg", "scene0593_01/00600.jpg" ], "turn_id": 4, "scan_id": "scene0593_01", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the medium-sized red and black backpack with a strap\" from your position?", "option": [], "answer": "137.26", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the medium-sized red and black backpack with a strap\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0593_01/00640.jpg", "scene0593_01/00680.jpg", "scene0593_01/00720.jpg", "scene0593_01/00760.jpg" ], "turn_id": 5, "scan_id": "scene0593_01", "origin_question": "Which two objects are on the same side of you now? \"the door\", \"the medium-sized red and black backpack with a strap\", \"the paper cutter\". ", "option": [ "the door, the medium-sized red and black backpack with a strap", "the door, the paper cutter", "the medium-sized red and black backpack with a strap, the paper cutter" ], "answer": "the door, the medium-sized red and black backpack with a strap", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the door\", \"the medium-sized red and black backpack with a strap\", \"the paper cutter\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the door, the medium-sized red and black backpack with a strap', 'the door, the paper cutter', 'the medium-sized red and black backpack with a strap, the paper cutter'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0593_01/00800.jpg", "scene0593_01/00840.jpg", "scene0593_01/00880.jpg", "scene0593_01/00920.jpg" ], "turn_id": 6, "scan_id": "scene0593_01", "origin_question": "Which two objects are on the same side of you now? \"the medium-sized red and black backpack with a strap\", \"the white standing fan with three blades\", \"the paper cutter\". ", "option": [ "the medium-sized red and black backpack with a strap, the white standing fan with three blades", "the medium-sized red and black backpack with a strap, the paper cutter", "the white standing fan with three blades, the paper cutter" ], "answer": "the medium-sized red and black backpack with a strap, the white standing fan with three blades", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the medium-sized red and black backpack with a strap\", \"the white standing fan with three blades\", \"the paper cutter\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the medium-sized red and black backpack with a strap, the white standing fan with three blades', 'the medium-sized red and black backpack with a strap, the paper cutter', 'the white standing fan with three blades, the paper cutter'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0593_01/00960.jpg", "scene0593_01/01000.jpg", "scene0593_01/01040.jpg", "scene0593_01/01080.jpg" ], "turn_id": 7, "scan_id": "scene0593_01", "origin_question": "Remember, how many window(s) have you seen so far? ", "option": [], "answer": "2", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many window(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0072_region4/b893c0ba92f64e9e88698fc4225fad6a_i1_0.jpg", "1mp3d_0072_region4/b893c0ba92f64e9e88698fc4225fad6a_i2_0.jpg", "1mp3d_0072_region4/a3dfeebf01d24ad19c009d3527fcb6cc_i1_0.jpg", "1mp3d_0072_region4/a3dfeebf01d24ad19c009d3527fcb6cc_i2_0.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0072_region4", "origin_question": "Remember, have you seen any door(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any door(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "1mp3d_0072_region4/fa1f71b06e894ff6a4e2173012ac7a3b_i1_5.jpg", "1mp3d_0072_region4/fa1f71b06e894ff6a4e2173012ac7a3b_i2_5.jpg", "1mp3d_0072_region4/d2a62238e6a8472381a0111b06f35f72_i1_5.jpg", "1mp3d_0072_region4/d2a62238e6a8472381a0111b06f35f72_i2_5.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0072_region4", "origin_question": "In which turn did you see both \"the rectangular light made of golden metal\" and \"the large rectangular wooden door, deep brown in color\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the rectangular light made of golden metal\" and \"the large rectangular wooden door, deep brown in color\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "1mp3d_0072_region4/d2a62238e6a8472381a0111b06f35f72_i1_0.jpg", "1mp3d_0072_region4/d2a62238e6a8472381a0111b06f35f72_i2_0.jpg", "1mp3d_0072_region4/fa1f71b06e894ff6a4e2173012ac7a3b_i1_0.jpg", "1mp3d_0072_region4/fa1f71b06e894ff6a4e2173012ac7a3b_i2_0.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0072_region4", "origin_question": "Please recall \"the large rectangular wooden door, deep brown in color\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "5.14", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the large rectangular wooden door, deep brown in color\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "1mp3d_0072_region4/a3dfeebf01d24ad19c009d3527fcb6cc_i1_1.jpg", "1mp3d_0072_region4/a3dfeebf01d24ad19c009d3527fcb6cc_i2_1.jpg", "1mp3d_0072_region4/b893c0ba92f64e9e88698fc4225fad6a_i1_2.jpg", "1mp3d_0072_region4/b893c0ba92f64e9e88698fc4225fad6a_i2_2.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0072_region4", "origin_question": "Which direction is the large, lush plant with green, spotted leaves in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the large, lush plant with green, spotted leaves in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0072_region4/501671cb342e44198f8867b37af987b2_i2_5.jpg", "1mp3d_0072_region4/5791bf29c55c4664b970a4d1adcddf06_i1_5.jpg", "1mp3d_0072_region4/5791bf29c55c4664b970a4d1adcddf06_i2_5.jpg", "1mp3d_0072_region4/501671cb342e44198f8867b37af987b2_i1_4.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0072_region4", "origin_question": "Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the large, lush plant with green, spotted leaves\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the large, lush plant with green, spotted leaves\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "1mp3d_0072_region4/501671cb342e44198f8867b37af987b2_i2_4.jpg", "1mp3d_0072_region4/5791bf29c55c4664b970a4d1adcddf06_i1_4.jpg", "1mp3d_0072_region4/5791bf29c55c4664b970a4d1adcddf06_i2_4.jpg", "1mp3d_0072_region4/591d52de25f64bbbb3bc6365669c5a0f_i1_5.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0072_region4", "origin_question": "Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "1mp3d_0072_region4/591d52de25f64bbbb3bc6365669c5a0f_i2_5.jpg", "1mp3d_0072_region4/b8b37a15b628463d9cfacb168332912b_i1_5.jpg", "1mp3d_0072_region4/b8b37a15b628463d9cfacb168332912b_i2_5.jpg", "1mp3d_0072_region4/884d01546f7d45ab977658ff323da87b_i1_5.jpg" ], "turn_id": 7, "scan_id": "1mp3d_0072_region4", "origin_question": "In which turn were you farthest from \"the large rectangular pale yellow wooden beam\"? ", "option": [], "answer": "7", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the large rectangular pale yellow wooden beam\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "1mp3d_0072_region4/884d01546f7d45ab977658ff323da87b_i2_5.jpg", "1mp3d_0072_region4/3eed2cb0dead4d9eb9bd7030ed437dcf_i1_5.jpg", "1mp3d_0072_region4/3eed2cb0dead4d9eb9bd7030ed437dcf_i2_5.jpg", "1mp3d_0072_region4/236a4dea3b6e415baa0606e642d48784_i1_5.jpg" ], "turn_id": 8, "scan_id": "1mp3d_0072_region4", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "60.27", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0426_02/00000.jpg", "scene0426_02/00040.jpg", "scene0426_02/00080.jpg", "scene0426_02/00120.jpg" ], "turn_id": 1, "scan_id": "scene0426_02", "origin_question": "Remember, have you seen any ladder(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any ladder(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0426_02/00160.jpg", "scene0426_02/00200.jpg", "scene0426_02/00240.jpg", "scene0426_02/00280.jpg" ], "turn_id": 2, "scan_id": "scene0426_02", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0426_02/00320.jpg", "scene0426_02/00360.jpg", "scene0426_02/00400.jpg", "scene0426_02/00440.jpg" ], "turn_id": 3, "scan_id": "scene0426_02", "origin_question": "What will be the first-time appearance order of the following categories: light, paper, tv?", "option": [ "paper, tv, light", "paper, light, tv", "light, tv, paper", "tv, light, paper" ], "answer": "tv, light, paper", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: light, paper, tv? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['paper, tv, light', 'paper, light, tv', 'light, tv, paper', 'tv, light, paper'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0426_02/00480.jpg", "scene0426_02/00520.jpg", "scene0426_02/00560.jpg", "scene0426_02/00600.jpg" ], "turn_id": 4, "scan_id": "scene0426_02", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the grey rectangular leather ladder with white text\" from your position?", "option": [], "answer": "177.36", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the grey rectangular leather ladder with white text\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0426_02/00640.jpg", "scene0426_02/00680.jpg", "scene0426_02/00720.jpg", "scene0426_02/00760.jpg" ], "turn_id": 5, "scan_id": "scene0426_02", "origin_question": "In which turn did you see both \"the case\" and \"the light rectangular wooden dresser with six drawers\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the case\" and \"the light rectangular wooden dresser with six drawers\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0426_02/00800.jpg", "scene0426_02/00840.jpg", "scene0426_02/00880.jpg", "scene0426_02/00920.jpg" ], "turn_id": 6, "scan_id": "scene0426_02", "origin_question": "Is the grey rectangular leather ladder with white text to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the grey rectangular leather ladder with white text to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0426_02/00960.jpg", "scene0426_02/01000.jpg", "scene0426_02/01040.jpg", "scene0426_02/01080.jpg" ], "turn_id": 7, "scan_id": "scene0426_02", "origin_question": "What will be the first-time appearance order of the following categories: ladder, paper, window?", "option": [ "window, paper, ladder", "ladder, paper, window", "ladder, window, paper", "paper, window, ladder" ], "answer": "ladder, paper, window", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: ladder, paper, window? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['window, paper, ladder', 'ladder, paper, window', 'ladder, window, paper', 'paper, window, ladder'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0426_02/01120.jpg", "scene0426_02/01160.jpg", "scene0426_02/01200.jpg", "scene0426_02/01240.jpg" ], "turn_id": 8, "scan_id": "scene0426_02", "origin_question": "What will be the first-time appearance order of the following categories: chair, remote control, window?", "option": [ "chair, remote control, window", "chair, window, remote control", "remote control, chair, window", "window, remote control, chair" ], "answer": "chair, window, remote control", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: chair, remote control, window? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['chair, remote control, window', 'chair, window, remote control', 'remote control, chair, window', 'window, remote control, chair'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0426_02/01280.jpg", "scene0426_02/01320.jpg", "scene0426_02/01360.jpg", "scene0426_02/01400.jpg" ], "turn_id": 9, "scan_id": "scene0426_02", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the medium-sized pink velvety clothes\" from your position?", "option": [], "answer": "115.88", "user_message": "For the 9 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the medium-sized pink velvety clothes\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0656_01/00000.jpg", "scene0656_01/00040.jpg", "scene0656_01/00080.jpg", "scene0656_01/00120.jpg", "scene0656_01/00160.jpg" ], "turn_id": 1, "scan_id": "scene0656_01", "origin_question": "Remember, have you seen any stand(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any stand(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0656_01/00200.jpg", "scene0656_01/00240.jpg", "scene0656_01/00280.jpg", "scene0656_01/00320.jpg", "scene0656_01/00360.jpg" ], "turn_id": 2, "scan_id": "scene0656_01", "origin_question": "Which direction is the small, square metal and plastic clock in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the small, square metal and plastic clock in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0656_01/00400.jpg", "scene0656_01/00440.jpg", "scene0656_01/00480.jpg", "scene0656_01/00520.jpg", "scene0656_01/00560.jpg" ], "turn_id": 3, "scan_id": "scene0656_01", "origin_question": "Which direction is the red cylindrical plastic bin in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the red cylindrical plastic bin in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0656_01/00600.jpg", "scene0656_01/00640.jpg", "scene0656_01/00680.jpg", "scene0656_01/00720.jpg", "scene0656_01/00760.jpg" ], "turn_id": 4, "scan_id": "scene0656_01", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the standard wooden chair with a dark cushioned seat\" from your position?", "option": [], "answer": "128.75", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the standard wooden chair with a dark cushioned seat\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0656_01/00800.jpg", "scene0656_01/00840.jpg", "scene0656_01/00880.jpg", "scene0656_01/00920.jpg", "scene0656_01/00960.jpg" ], "turn_id": 5, "scan_id": "scene0656_01", "origin_question": "", "option": [], "answer": "", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0656_01/01000.jpg", "scene0656_01/01040.jpg", "scene0656_01/01080.jpg", "scene0656_01/01120.jpg", "scene0656_01/01160.jpg" ], "turn_id": 6, "scan_id": "scene0656_01", "origin_question": "In which turn were you closest to \"the light yellow rectangular wooden three-tiered shelf\"? ", "option": [], "answer": "5", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the light yellow rectangular wooden three-tiered shelf\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0656_01/01200.jpg", "scene0656_01/01240.jpg", "scene0656_01/01280.jpg", "scene0656_01/01320.jpg", "scene0656_01/01360.jpg" ], "turn_id": 7, "scan_id": "scene0656_01", "origin_question": "What will be the first-time appearance order of the following categories: curtain, shelf, vanity?", "option": [ "shelf, curtain, vanity", "curtain, shelf, vanity", "curtain, vanity, shelf", "vanity, shelf, curtain" ], "answer": "curtain, shelf, vanity", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: curtain, shelf, vanity? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['shelf, curtain, vanity', 'curtain, shelf, vanity', 'curtain, vanity, shelf', 'vanity, shelf, curtain'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0656_01/01400.jpg", "scene0656_01/01440.jpg", "scene0656_01/01480.jpg", "scene0656_01/01520.jpg", "scene0656_01/01560.jpg" ], "turn_id": 8, "scan_id": "scene0656_01", "origin_question": "Which direction is the standard wooden chair with a dark cushioned seat in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the standard wooden chair with a dark cushioned seat in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0656_01/01600.jpg", "scene0656_01/01640.jpg", "scene0656_01/01680.jpg", "scene0656_01/01720.jpg", "scene0656_01/01760.jpg" ], "turn_id": 9, "scan_id": "scene0656_01", "origin_question": "What will be the first-time appearance order of the following categories: bag, pillow, table?", "option": [ "bag, pillow, table", "bag, table, pillow", "pillow, table, bag", "pillow, bag, table" ], "answer": "pillow, table, bag", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bag, pillow, table? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['bag, pillow, table', 'bag, table, pillow', 'pillow, table, bag', 'pillow, bag, table'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0656_01/01800.jpg", "scene0656_01/01840.jpg", "scene0656_01/01880.jpg", "scene0656_01/01920.jpg", "scene0656_01/01960.jpg" ], "turn_id": 10, "scan_id": "scene0656_01", "origin_question": "Is the light yellow rectangular wooden three-tiered shelf to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the light yellow rectangular wooden three-tiered shelf to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0169_00/00000.jpg", "scene0169_00/00040.jpg", "scene0169_00/00080.jpg", "scene0169_00/00120.jpg", "scene0169_00/00160.jpg" ], "turn_id": 1, "scan_id": "scene0169_00", "origin_question": "Remember, have you seen any paper(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any paper(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0169_00/00200.jpg", "scene0169_00/00240.jpg", "scene0169_00/00280.jpg", "scene0169_00/00320.jpg", "scene0169_00/00360.jpg" ], "turn_id": 2, "scan_id": "scene0169_00", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the red backpack with black zippers and handle\"; \"the blackboard\"; \"the light brown rectangular wooden cabinet with a door\" ", "option": [ "the red backpack with black zippers and handle", "the blackboard", "the light brown rectangular wooden cabinet with a door" ], "answer": "the red backpack with black zippers and handle", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the red backpack with black zippers and handle\"; \"the blackboard\"; \"the light brown rectangular wooden cabinet with a door\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the red backpack with black zippers and handle', 'the blackboard', 'the light brown rectangular wooden cabinet with a door'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0169_00/00400.jpg", "scene0169_00/00440.jpg", "scene0169_00/00480.jpg", "scene0169_00/00520.jpg", "scene0169_00/00560.jpg" ], "turn_id": 3, "scan_id": "scene0169_00", "origin_question": "In which turn were you farthest from \"the vibrant landscape picture with a light wooden frame\"? ", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the vibrant landscape picture with a light wooden frame\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0169_00/00600.jpg", "scene0169_00/00640.jpg", "scene0169_00/00680.jpg", "scene0169_00/00720.jpg", "scene0169_00/00760.jpg" ], "turn_id": 4, "scan_id": "scene0169_00", "origin_question": "How far is your current position from where you were at the end of the turn 1? (in meters)", "option": [], "answer": "4.48", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 1? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0169_00/00800.jpg", "scene0169_00/00840.jpg", "scene0169_00/00880.jpg", "scene0169_00/00920.jpg", "scene0169_00/00960.jpg" ], "turn_id": 5, "scan_id": "scene0169_00", "origin_question": "What will be the first-time appearance order of the following categories: backpack, blackboard, paper?", "option": [ "blackboard, backpack, paper", "paper, backpack, blackboard", "blackboard, paper, backpack", "paper, blackboard, backpack" ], "answer": "paper, blackboard, backpack", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: backpack, blackboard, paper? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['blackboard, backpack, paper', 'paper, backpack, blackboard', 'blackboard, paper, backpack', 'paper, blackboard, backpack'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0169_00/01000.jpg", "scene0169_00/01040.jpg", "scene0169_00/01080.jpg", "scene0169_00/01120.jpg", "scene0169_00/01160.jpg" ], "turn_id": 6, "scan_id": "scene0169_00", "origin_question": "Using your orientation at the end of turn 5 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "22.98", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 5 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0169_00/01200.jpg", "scene0169_00/01240.jpg", "scene0169_00/01280.jpg", "scene0169_00/01320.jpg", "scene0169_00/01360.jpg" ], "turn_id": 7, "scan_id": "scene0169_00", "origin_question": "Which direction is the red backpack with black zippers and handle in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the red backpack with black zippers and handle in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0169_00/01400.jpg", "scene0169_00/01440.jpg", "scene0169_00/01480.jpg", "scene0169_00/01520.jpg", "scene0169_00/01560.jpg" ], "turn_id": 8, "scan_id": "scene0169_00", "origin_question": "What will be the first-time appearance order of the following categories: bin, blackboard, paper?", "option": [ "paper, bin, blackboard", "blackboard, bin, paper", "paper, blackboard, bin", "bin, paper, blackboard" ], "answer": "paper, blackboard, bin", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bin, blackboard, paper? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['paper, bin, blackboard', 'blackboard, bin, paper', 'paper, blackboard, bin', 'bin, paper, blackboard'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0169_00/01600.jpg", "scene0169_00/01640.jpg", "scene0169_00/01680.jpg", "scene0169_00/01720.jpg", "scene0169_00/01760.jpg" ], "turn_id": 9, "scan_id": "scene0169_00", "origin_question": "At the end of which turn were both of \"the vibrant landscape picture with a light wooden frame\" and \"the bin\" on your left side?", "option": [], "answer": "1", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the vibrant landscape picture with a light wooden frame\" and \"the bin\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0169_00/01800.jpg", "scene0169_00/01840.jpg", "scene0169_00/01880.jpg", "scene0169_00/01920.jpg", "scene0169_00/01960.jpg" ], "turn_id": 10, "scan_id": "scene0169_00", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the red backpack with black zippers and handle\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the red backpack with black zippers and handle\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0114_00/00000.jpg", "scene0114_00/00040.jpg", "scene0114_00/00080.jpg", "scene0114_00/00120.jpg", "scene0114_00/00160.jpg" ], "turn_id": 1, "scan_id": "scene0114_00", "origin_question": "Remember, have you seen any window(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any window(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0114_00/00200.jpg", "scene0114_00/00240.jpg", "scene0114_00/00280.jpg", "scene0114_00/00320.jpg", "scene0114_00/00360.jpg" ], "turn_id": 2, "scan_id": "scene0114_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0114_00/00400.jpg", "scene0114_00/00440.jpg", "scene0114_00/00480.jpg", "scene0114_00/00520.jpg", "scene0114_00/00560.jpg" ], "turn_id": 3, "scan_id": "scene0114_00", "origin_question": "Remember, have you seen any fan(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any fan(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0114_00/00600.jpg", "scene0114_00/00640.jpg", "scene0114_00/00680.jpg", "scene0114_00/00720.jpg", "scene0114_00/00760.jpg" ], "turn_id": 4, "scan_id": "scene0114_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0114_00/00800.jpg", "scene0114_00/00840.jpg", "scene0114_00/00880.jpg", "scene0114_00/00920.jpg", "scene0114_00/00960.jpg" ], "turn_id": 5, "scan_id": "scene0114_00", "origin_question": "How far is your current position from where you were at the end of the turn 3? (in meters)", "option": [], "answer": "1.58", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 3? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0114_00/01000.jpg", "scene0114_00/01040.jpg", "scene0114_00/01080.jpg", "scene0114_00/01120.jpg", "scene0114_00/01160.jpg" ], "turn_id": 6, "scan_id": "scene0114_00", "origin_question": "Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the standard moderate-sized blackboard made of painted wood\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the standard moderate-sized blackboard made of painted wood\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0114_00/01200.jpg", "scene0114_00/01240.jpg", "scene0114_00/01280.jpg", "scene0114_00/01320.jpg", "scene0114_00/01360.jpg" ], "turn_id": 7, "scan_id": "scene0114_00", "origin_question": "Which direction is the standard moderate-sized blackboard made of painted wood in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the standard moderate-sized blackboard made of painted wood in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0114_00/01400.jpg", "scene0114_00/01440.jpg", "scene0114_00/01480.jpg", "scene0114_00/01520.jpg", "scene0114_00/01560.jpg" ], "turn_id": 8, "scan_id": "scene0114_00", "origin_question": "Which direction is the bin in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the bin in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0114_00/01600.jpg", "scene0114_00/01640.jpg", "scene0114_00/01680.jpg", "scene0114_00/01720.jpg", "scene0114_00/01760.jpg" ], "turn_id": 9, "scan_id": "scene0114_00", "origin_question": "Is the white wall-mounted storage rack with multiple shelves to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the white wall-mounted storage rack with multiple shelves to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0078_region10/1e6190c84052401d8a2435eed54b82aa_i1_4.jpg", "1mp3d_0078_region10/1e6190c84052401d8a2435eed54b82aa_i2_4.jpg", "1mp3d_0078_region10/28a8a22473e342f7be7ed0f33e4272e3_i1_0.jpg", "1mp3d_0078_region10/28a8a22473e342f7be7ed0f33e4272e3_i2_0.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0078_region10", "origin_question": "Remember, have you seen any adhesive tape(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any adhesive tape(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "1mp3d_0078_region10/89b25402eae54a36a9b9d2b9889ab632_i1_0.jpg", "1mp3d_0078_region10/89b25402eae54a36a9b9d2b9889ab632_i2_0.jpg", "1mp3d_0078_region10/a860e5bf7afe44179bcdc838db421802_i1_4.jpg", "1mp3d_0078_region10/a860e5bf7afe44179bcdc838db421802_i2_4.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0078_region10", "origin_question": "At the end of which turn were both of \"the gray and yellow three-seater couch\" and \"the natural light brown wooden double door\" on your right side?", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the gray and yellow three-seater couch\" and \"the natural light brown wooden double door\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "1mp3d_0078_region10/89b25402eae54a36a9b9d2b9889ab632_i1_5.jpg", "1mp3d_0078_region10/89b25402eae54a36a9b9d2b9889ab632_i2_5.jpg", "1mp3d_0078_region10/28a8a22473e342f7be7ed0f33e4272e3_i1_1.jpg", "1mp3d_0078_region10/28a8a22473e342f7be7ed0f33e4272e3_i2_1.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0078_region10", "origin_question": "At the end of which turn were both of \"the large arched windowpane with a wooden frame\" and \"the natural light brown wooden double door\" on your right side?", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the large arched windowpane with a wooden frame\" and \"the natural light brown wooden double door\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "1mp3d_0078_region10/d375a9c7bc4748a2b36ee572d56aa9e2_i1_2.jpg", "1mp3d_0078_region10/d375a9c7bc4748a2b36ee572d56aa9e2_i2_2.jpg", "1mp3d_0078_region10/fa30e6859ca249a78879975b0b659bfb_i1_1.jpg", "1mp3d_0078_region10/fa30e6859ca249a78879975b0b659bfb_i2_1.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0078_region10", "origin_question": "By the end of the 3 turn, were you getting closer to or farther away from these two objects: X: \"the large arched windowpane with a wooden frame\", Y: \"the rectangular canvas with a vibrant green background\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 3 turn, were you getting closer to or farther away from these two objects: X: \"the large arched windowpane with a wooden frame\", Y: \"the rectangular canvas with a vibrant green background\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "1mp3d_0078_region10/fa30e6859ca249a78879975b0b659bfb_i1_0.jpg", "1mp3d_0078_region10/fa30e6859ca249a78879975b0b659bfb_i2_0.jpg", "1mp3d_0078_region10/b14d29bea4b547d5923b3a09323b443d_i1_2.jpg", "1mp3d_0078_region10/b14d29bea4b547d5923b3a09323b443d_i2_2.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0078_region10", "origin_question": "By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the large arched windowpane with a wooden frame\", Y: \"the gray and yellow three-seater couch\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the large arched windowpane with a wooden frame\", Y: \"the gray and yellow three-seater couch\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "1mp3d_0078_region10/45f892a4837c49028f47cb6a341f65df_i1_5.jpg", "1mp3d_0078_region10/45f892a4837c49028f47cb6a341f65df_i2_5.jpg", "1mp3d_0078_region10/45f892a4837c49028f47cb6a341f65df_i1_4.jpg", "1mp3d_0078_region10/45f892a4837c49028f47cb6a341f65df_i2_4.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0078_region10", "origin_question": "Using your orientation at the end of turn 5 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "62.89", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 5 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "1mp3d_0078_region10/44055efcd82c49daae1cb67e4fc79ebe_i1_2.jpg", "1mp3d_0078_region10/44055efcd82c49daae1cb67e4fc79ebe_i2_2.jpg", "1mp3d_0078_region10/ee1a56ee68544ff9a09fbb3044bf1cb3_i1_3.jpg", "1mp3d_0078_region10/ee1a56ee68544ff9a09fbb3044bf1cb3_i2_3.jpg" ], "turn_id": 7, "scan_id": "1mp3d_0078_region10", "origin_question": "In which turn were you closest to \"the natural light brown wooden double door\"? ", "option": [], "answer": "6", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the natural light brown wooden double door\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0078_region10/44055efcd82c49daae1cb67e4fc79ebe_i1_3.jpg", "1mp3d_0078_region10/44055efcd82c49daae1cb67e4fc79ebe_i2_3.jpg", "1mp3d_0078_region10/ee1a56ee68544ff9a09fbb3044bf1cb3_i1_2.jpg", "1mp3d_0078_region10/ee1a56ee68544ff9a09fbb3044bf1cb3_i2_2.jpg" ], "turn_id": 8, "scan_id": "1mp3d_0078_region10", "origin_question": "", "option": [], "answer": "", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0131_00/00000.jpg", "scene0131_00/00040.jpg", "scene0131_00/00080.jpg", "scene0131_00/00120.jpg" ], "turn_id": 1, "scan_id": "scene0131_00", "origin_question": "Remember, have you seen any stapler(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any stapler(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0131_00/00160.jpg", "scene0131_00/00200.jpg", "scene0131_00/00240.jpg", "scene0131_00/00280.jpg" ], "turn_id": 2, "scan_id": "scene0131_00", "origin_question": "In which turn were you closest to \"the computer\"? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the computer\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0131_00/00320.jpg", "scene0131_00/00360.jpg", "scene0131_00/00400.jpg", "scene0131_00/00440.jpg" ], "turn_id": 3, "scan_id": "scene0131_00", "origin_question": "At the end of which turn were both of \"the fan\" and \"the black plastic keyboard with a standard layout\" on your left side?", "option": [], "answer": "2", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the fan\" and \"the black plastic keyboard with a standard layout\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0131_00/00480.jpg", "scene0131_00/00520.jpg", "scene0131_00/00560.jpg", "scene0131_00/00600.jpg" ], "turn_id": 4, "scan_id": "scene0131_00", "origin_question": "Which two objects are on the same side of you now? \"the desk\", \"the rectangular plastic switch with a black border\", \"the black plastic keyboard with a standard layout\". ", "option": [ "the desk, the rectangular plastic switch with a black border", "the desk, the black plastic keyboard with a standard layout", "the rectangular plastic switch with a black border, the black plastic keyboard with a standard layout" ], "answer": "the desk, the rectangular plastic switch with a black border", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the desk\", \"the rectangular plastic switch with a black border\", \"the black plastic keyboard with a standard layout\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the desk, the rectangular plastic switch with a black border', 'the desk, the black plastic keyboard with a standard layout', 'the rectangular plastic switch with a black border, the black plastic keyboard with a standard layout'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0131_00/00640.jpg", "scene0131_00/00680.jpg", "scene0131_00/00720.jpg", "scene0131_00/00760.jpg" ], "turn_id": 5, "scan_id": "scene0131_00", "origin_question": "Which direction is the white ceramic cup with a cylindrical shape in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the white ceramic cup with a cylindrical shape in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0131_00/00800.jpg", "scene0131_00/00840.jpg", "scene0131_00/00880.jpg", "scene0131_00/00920.jpg" ], "turn_id": 6, "scan_id": "scene0131_00", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0078_region1/acdaf6966e6d48c789599b0059c54f3c_i1_0.jpg", "1mp3d_0078_region1/acdaf6966e6d48c789599b0059c54f3c_i2_0.jpg", "1mp3d_0078_region1/90f5e6e359b949fdaabf8ac3ee50a82d_i1_5.jpg", "1mp3d_0078_region1/90f5e6e359b949fdaabf8ac3ee50a82d_i2_5.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0078_region1", "origin_question": "Remember, have you seen any water heater(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any water heater(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0078_region1/75e8fbb7bb864c19bf8672b303c5b6a7_i1_0.jpg", "1mp3d_0078_region1/75e8fbb7bb864c19bf8672b303c5b6a7_i2_0.jpg", "1mp3d_0078_region1/a4457a8d962f4a93b82214b4ad5ef9ef_i1_0.jpg", "1mp3d_0078_region1/a4457a8d962f4a93b82214b4ad5ef9ef_i2_0.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0078_region1", "origin_question": "Remember, have you seen any beam(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any beam(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "1mp3d_0078_region1/a4457a8d962f4a93b82214b4ad5ef9ef_i1_1.jpg", "1mp3d_0078_region1/75e8fbb7bb864c19bf8672b303c5b6a7_i1_1.jpg", "1mp3d_0078_region1/75e8fbb7bb864c19bf8672b303c5b6a7_i2_1.jpg", "1mp3d_0078_region1/90f5e6e359b949fdaabf8ac3ee50a82d_i1_4.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0078_region1", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the light wooden beam with a straight, sturdy design\"; \"the light-colored rectangular wooden door with a sloping top\"; \"the standard blackboard with a white surface\" ", "option": [ "the light wooden beam with a straight, sturdy design", "the light-colored rectangular wooden door with a sloping top", "the standard blackboard with a white surface" ], "answer": "the light-colored rectangular wooden door with a sloping top", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the light wooden beam with a straight, sturdy design\"; \"the light-colored rectangular wooden door with a sloping top\"; \"the standard blackboard with a white surface\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the light wooden beam with a straight, sturdy design', 'the light-colored rectangular wooden door with a sloping top', 'the standard blackboard with a white surface'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "1mp3d_0078_region1/90f5e6e359b949fdaabf8ac3ee50a82d_i2_4.jpg", "1mp3d_0078_region1/acdaf6966e6d48c789599b0059c54f3c_i1_5.jpg", "1mp3d_0078_region1/acdaf6966e6d48c789599b0059c54f3c_i2_5.jpg", "1mp3d_0078_region1/625c0d1148e54581b8ae5ace722ac094_i1_2.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0078_region1", "origin_question": "By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the small pink plastic chair with backrest and armrests\", Y: \"the light wooden beam with a straight, sturdy design\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to Y and farther away from X", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the small pink plastic chair with backrest and armrests\", Y: \"the light wooden beam with a straight, sturdy design\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0078_region1/625c0d1148e54581b8ae5ace722ac094_i2_2.jpg", "1mp3d_0078_region1/2b926a9b858a4c118d903b3899fa5b59_i1_2.jpg", "1mp3d_0078_region1/2b926a9b858a4c118d903b3899fa5b59_i2_2.jpg", "1mp3d_0078_region1/2b926a9b858a4c118d903b3899fa5b59_i1_1.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0078_region1", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the small pink plastic chair with backrest and armrests\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the small pink plastic chair with backrest and armrests\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "1mp3d_0078_region1/2b926a9b858a4c118d903b3899fa5b59_i2_1.jpg", "1mp3d_0078_region1/0fb3575b26954645b116ecd661dac744_i1_0.jpg", "1mp3d_0078_region1/0fb3575b26954645b116ecd661dac744_i2_0.jpg", "1mp3d_0078_region1/1685d78305d3419fb8cbabc690b77ac3_i1_4.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0078_region1", "origin_question": "When did you first discover \"the light-colored rectangular wooden door with a sloping top\"? (index of the turn)", "option": [], "answer": "3", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the light-colored rectangular wooden door with a sloping top\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "1mp3d_0078_region1/1685d78305d3419fb8cbabc690b77ac3_i2_4.jpg", "1mp3d_0078_region1/1685d78305d3419fb8cbabc690b77ac3_i1_3.jpg", "1mp3d_0078_region1/1685d78305d3419fb8cbabc690b77ac3_i2_3.jpg", "1mp3d_0078_region1/ef12b1ecc3be4693b327e30468697a9b_i1_4.jpg" ], "turn_id": 7, "scan_id": "1mp3d_0078_region1", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the large rectangular beige bag made of thick fabric\" from your position?", "option": [], "answer": "168.81", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the large rectangular beige bag made of thick fabric\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "1mp3d_0078_region1/ef12b1ecc3be4693b327e30468697a9b_i2_4.jpg", "1mp3d_0078_region1/efcb9e4c410841ca8665f6cab94e943b_i1_4.jpg", "1mp3d_0078_region1/efcb9e4c410841ca8665f6cab94e943b_i2_4.jpg", "1mp3d_0078_region1/7e23cbe5e28047d49ab4641f530c9f00_i1_4.jpg" ], "turn_id": 8, "scan_id": "1mp3d_0078_region1", "origin_question": "Using your orientation at the end of turn 7 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 7 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0040_region48/46959f49c0b9487daf8ee51d602d3a4f_i1_4.jpg", "1mp3d_0040_region48/46959f49c0b9487daf8ee51d602d3a4f_i2_4.jpg", "1mp3d_0040_region48/6f5e70ae957d4a95af44068aa4946744_i1_5.jpg", "1mp3d_0040_region48/6f5e70ae957d4a95af44068aa4946744_i2_5.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0040_region48", "origin_question": "Remember, have you seen any fireplace(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any fireplace(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "1mp3d_0040_region48/35fbae92aa284ba19652ccb31f4cff9c_i1_0.jpg", "1mp3d_0040_region48/35fbae92aa284ba19652ccb31f4cff9c_i2_0.jpg", "1mp3d_0040_region48/35fbae92aa284ba19652ccb31f4cff9c_i1_1.jpg", "1mp3d_0040_region48/35fbae92aa284ba19652ccb31f4cff9c_i2_1.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0040_region48", "origin_question": "Remember, how many commode(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many commode(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "1mp3d_0040_region48/13539678a339494dbe249d3e8137778d_i1_3.jpg", "1mp3d_0040_region48/13539678a339494dbe249d3e8137778d_i2_3.jpg", "1mp3d_0040_region48/13539678a339494dbe249d3e8137778d_i1_2.jpg", "1mp3d_0040_region48/13539678a339494dbe249d3e8137778d_i2_2.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0040_region48", "origin_question": "At the end of which turn were both of \"the circular glossy black pedestal table\" and \"the rectangular window with a white frame\" on your left side?", "option": [], "answer": "2", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the circular glossy black pedestal table\" and \"the rectangular window with a white frame\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "1mp3d_0040_region48/f9b338cf59ea46cd810548040b85ca3b_i1_3.jpg", "1mp3d_0040_region48/f9b338cf59ea46cd810548040b85ca3b_i2_3.jpg", "1mp3d_0040_region48/5530eac6cc8540a8ba22d50487a5dd2f_i1_2.jpg", "1mp3d_0040_region48/5530eac6cc8540a8ba22d50487a5dd2f_i2_2.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0040_region48", "origin_question": "In which turn were you closest to \"the circular glossy black pedestal table\"? ", "option": [], "answer": "1", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the circular glossy black pedestal table\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "1mp3d_0040_region48/89930128220c45b8bb04c3d3990b202c_i1_5.jpg", "1mp3d_0040_region48/89930128220c45b8bb04c3d3990b202c_i2_5.jpg", "1mp3d_0040_region48/5530eac6cc8540a8ba22d50487a5dd2f_i1_5.jpg", "1mp3d_0040_region48/5530eac6cc8540a8ba22d50487a5dd2f_i2_5.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0040_region48", "origin_question": "Please recall \"the rectangular window with a white frame\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "7.47", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the rectangular window with a white frame\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "1mp3d_0040_region48/5530eac6cc8540a8ba22d50487a5dd2f_i1_0.jpg", "1mp3d_0040_region48/5530eac6cc8540a8ba22d50487a5dd2f_i2_0.jpg", "1mp3d_0040_region48/5530eac6cc8540a8ba22d50487a5dd2f_i1_1.jpg", "1mp3d_0040_region48/5530eac6cc8540a8ba22d50487a5dd2f_i2_1.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0040_region48", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "1mp3d_0040_region48/f9b338cf59ea46cd810548040b85ca3b_i1_5.jpg", "1mp3d_0040_region48/f9b338cf59ea46cd810548040b85ca3b_i2_5.jpg", "1mp3d_0040_region48/f9b338cf59ea46cd810548040b85ca3b_i1_4.jpg", "1mp3d_0040_region48/f9b338cf59ea46cd810548040b85ca3b_i2_4.jpg" ], "turn_id": 7, "scan_id": "1mp3d_0040_region48", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "87.41", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "1mp3d_0040_region48/fb5d8954758f4bb990c8d78d2a5d63f2_i1_4.jpg", "1mp3d_0040_region48/fb5d8954758f4bb990c8d78d2a5d63f2_i2_4.jpg", "1mp3d_0040_region48/227f460a7ab143ebb0716888379db76f_i1_4.jpg", "1mp3d_0040_region48/227f460a7ab143ebb0716888379db76f_i2_4.jpg" ], "turn_id": 8, "scan_id": "1mp3d_0040_region48", "origin_question": "Is the rectangular black television of standard size to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the rectangular black television of standard size to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "1mp3d_0040_region48/fb5d8954758f4bb990c8d78d2a5d63f2_i1_5.jpg", "1mp3d_0040_region48/fb5d8954758f4bb990c8d78d2a5d63f2_i2_5.jpg", "1mp3d_0040_region48/9fd63460efcf4292a96791b9531b6f94_i1_5.jpg", "1mp3d_0040_region48/9fd63460efcf4292a96791b9531b6f94_i2_5.jpg" ], "turn_id": 9, "scan_id": "1mp3d_0040_region48", "origin_question": "Which direction is the large, round beige velvet ottoman in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 9 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the large, round beige velvet ottoman in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0362_01/00000.jpg", "scene0362_01/00040.jpg", "scene0362_01/00080.jpg", "scene0362_01/00120.jpg" ], "turn_id": 1, "scan_id": "scene0362_01", "origin_question": "Remember, have you seen any bowl(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any bowl(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0362_01/00160.jpg", "scene0362_01/00200.jpg", "scene0362_01/00240.jpg", "scene0362_01/00280.jpg" ], "turn_id": 2, "scan_id": "scene0362_01", "origin_question": "When did you first discover \"the horizontal beige radiator with protrusions for heat dissipation\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the horizontal beige radiator with protrusions for heat dissipation\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0362_01/00320.jpg", "scene0362_01/00360.jpg", "scene0362_01/00400.jpg", "scene0362_01/00440.jpg" ], "turn_id": 3, "scan_id": "scene0362_01", "origin_question": "At the end of which turn were both of \"the medium-sized lamp with a conical neutral lampshade\" and \"then abstract picture with red, yellow, and white shapes\" on your right side?", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the medium-sized lamp with a conical neutral lampshade\" and \"then abstract picture with red, yellow, and white shapes\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0362_01/00480.jpg", "scene0362_01/00520.jpg", "scene0362_01/00560.jpg", "scene0362_01/00600.jpg" ], "turn_id": 4, "scan_id": "scene0362_01", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the horizontal beige radiator with protrusions for heat dissipation\"; \"the white rectangular sheet of printed paper\"; \"the rectangular two-seater couch with deep brown fabric\" ", "option": [ "the horizontal beige radiator with protrusions for heat dissipation", "the white rectangular sheet of printed paper", "the rectangular two-seater couch with deep brown fabric" ], "answer": "the white rectangular sheet of printed paper", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the horizontal beige radiator with protrusions for heat dissipation\"; \"the white rectangular sheet of printed paper\"; \"the rectangular two-seater couch with deep brown fabric\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the horizontal beige radiator with protrusions for heat dissipation', 'the white rectangular sheet of printed paper', 'the rectangular two-seater couch with deep brown fabric'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0362_01/00640.jpg", "scene0362_01/00680.jpg", "scene0362_01/00720.jpg", "scene0362_01/00760.jpg" ], "turn_id": 5, "scan_id": "scene0362_01", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular two-seater couch with deep brown fabric\", Y: \"the sleek black flat-panel TV with a modern design\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to X and farther away from Y", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular two-seater couch with deep brown fabric\", Y: \"the sleek black flat-panel TV with a modern design\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0362_01/00800.jpg", "scene0362_01/00840.jpg", "scene0362_01/00880.jpg", "scene0362_01/00920.jpg" ], "turn_id": 6, "scan_id": "scene0362_01", "origin_question": "Which direction is the rectangular two-seater couch with deep brown fabric in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the rectangular two-seater couch with deep brown fabric in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0362_01/00960.jpg", "scene0362_01/01000.jpg", "scene0362_01/01040.jpg", "scene0362_01/01080.jpg" ], "turn_id": 7, "scan_id": "scene0362_01", "origin_question": "Is the medium-sized lamp with a conical neutral lampshade to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the medium-sized lamp with a conical neutral lampshade to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0362_01/01120.jpg", "scene0362_01/01160.jpg", "scene0362_01/01200.jpg", "scene0362_01/01240.jpg" ], "turn_id": 8, "scan_id": "scene0362_01", "origin_question": "By the end of the 4 turn, were you getting closer to or farther away from these two objects: X: \"then abstract picture with red, yellow, and white shapes\", Y: \"the modern office chair with a black mesh backrest\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to Y and farther away from X", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 4 turn, were you getting closer to or farther away from these two objects: X: \"then abstract picture with red, yellow, and white shapes\", Y: \"the modern office chair with a black mesh backrest\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "1mp3d_0029_region31/f3e7de357f254d15bb79bc62ef0c803e_i1_3.jpg", "1mp3d_0029_region31/f3e7de357f254d15bb79bc62ef0c803e_i2_3.jpg", "1mp3d_0029_region31/50df19b5012e4451870d16320747ff9b_i1_4.jpg", "1mp3d_0029_region31/50df19b5012e4451870d16320747ff9b_i2_4.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0029_region31", "origin_question": "Remember, how many window(s) have you seen so far? ", "option": [], "answer": "2", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many window(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 3 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "1mp3d_0029_region31/50df19b5012e4451870d16320747ff9b_i1_5.jpg", "1mp3d_0029_region31/50df19b5012e4451870d16320747ff9b_i2_5.jpg", "1mp3d_0029_region31/50df19b5012e4451870d16320747ff9b_i1_3.jpg", "1mp3d_0029_region31/50df19b5012e4451870d16320747ff9b_i2_3.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0029_region31", "origin_question": "In which turn did you see both \"the thick, blue and green woven blanket with tassels\" and \"the large vertical canvas picture of a basketball player\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the thick, blue and green woven blanket with tassels\" and \"the large vertical canvas picture of a basketball player\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 3 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "1mp3d_0029_region31/50df19b5012e4451870d16320747ff9b_i1_2.jpg", "1mp3d_0029_region31/50df19b5012e4451870d16320747ff9b_i2_2.jpg", "1mp3d_0029_region31/f3e7de357f254d15bb79bc62ef0c803e_i1_2.jpg", "1mp3d_0029_region31/f3e7de357f254d15bb79bc62ef0c803e_i2_2.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0029_region31", "origin_question": "At the end of which turn were both of \"the large vertical canvas picture of a basketball player\" and \"the white wooden vertically-acting door\" on your left side?", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the large vertical canvas picture of a basketball player\" and \"the white wooden vertically-acting door\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 3 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0274_00/00000.jpg", "scene0274_00/00040.jpg", "scene0274_00/00080.jpg", "scene0274_00/00120.jpg", "scene0274_00/00160.jpg" ], "turn_id": 1, "scan_id": "scene0274_00", "origin_question": "Remember, have you seen any trophy(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any trophy(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0274_00/00200.jpg", "scene0274_00/00240.jpg", "scene0274_00/00280.jpg", "scene0274_00/00320.jpg", "scene0274_00/00360.jpg" ], "turn_id": 2, "scan_id": "scene0274_00", "origin_question": "When did you first discover \"the green cylindrical plastic bin with a black lid\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the green cylindrical plastic bin with a black lid\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0274_00/00400.jpg", "scene0274_00/00440.jpg", "scene0274_00/00480.jpg", "scene0274_00/00520.jpg", "scene0274_00/00560.jpg" ], "turn_id": 3, "scan_id": "scene0274_00", "origin_question": "Remember, have you seen any cosmetics(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any cosmetics(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0274_00/00600.jpg", "scene0274_00/00640.jpg", "scene0274_00/00680.jpg", "scene0274_00/00720.jpg", "scene0274_00/00760.jpg" ], "turn_id": 4, "scan_id": "scene0274_00", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0274_00/00800.jpg", "scene0274_00/00840.jpg", "scene0274_00/00880.jpg", "scene0274_00/00920.jpg", "scene0274_00/00960.jpg" ], "turn_id": 5, "scan_id": "scene0274_00", "origin_question": "By the end of the 3 turn, were you getting closer to or farther away from these two objects: X: \"the window\", Y: \"the towel\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 3 turn, were you getting closer to or farther away from these two objects: X: \"the window\", Y: \"the towel\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0274_00/01000.jpg", "scene0274_00/01040.jpg", "scene0274_00/01080.jpg", "scene0274_00/01120.jpg", "scene0274_00/01160.jpg" ], "turn_id": 6, "scan_id": "scene0274_00", "origin_question": "When did you first discover \"the standard rectangular white eraser\"? (index of the turn)", "option": [], "answer": "4", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the standard rectangular white eraser\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0274_00/01200.jpg", "scene0274_00/01240.jpg", "scene0274_00/01280.jpg", "scene0274_00/01320.jpg", "scene0274_00/01360.jpg" ], "turn_id": 7, "scan_id": "scene0274_00", "origin_question": "When did you first discover \"the green cylindrical plastic bin with a black lid\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the green cylindrical plastic bin with a black lid\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0274_00/01400.jpg", "scene0274_00/01440.jpg", "scene0274_00/01480.jpg", "scene0274_00/01520.jpg", "scene0274_00/01560.jpg" ], "turn_id": 8, "scan_id": "scene0274_00", "origin_question": "Please recall \"the large rectangular bright orange cardboard box\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.86", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the large rectangular bright orange cardboard box\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0274_00/01600.jpg", "scene0274_00/01640.jpg", "scene0274_00/01680.jpg", "scene0274_00/01720.jpg", "scene0274_00/01760.jpg" ], "turn_id": 9, "scan_id": "scene0274_00", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the small cylindrical white plastic bottle with a green label\" from your position?", "option": [], "answer": "100.78", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the small cylindrical white plastic bottle with a green label\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "scene0274_00/01800.jpg", "scene0274_00/01840.jpg", "scene0274_00/01880.jpg", "scene0274_00/01920.jpg", "scene0274_00/01960.jpg" ], "turn_id": 10, "scan_id": "scene0274_00", "origin_question": "Among these three objects(\"the window\",\"the towel\",\"the standard rectangular white eraser\" ), which one is the closest to you?", "option": [ "the window", "the towel", "the standard rectangular white eraser" ], "answer": "the standard rectangular white eraser", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the window\",\"the towel\",\"the standard rectangular white eraser\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the window', 'the towel', 'the standard rectangular white eraser'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0451_05/00000.jpg", "scene0451_05/00040.jpg", "scene0451_05/00080.jpg", "scene0451_05/00120.jpg", "scene0451_05/00160.jpg" ], "turn_id": 1, "scan_id": "scene0451_05", "origin_question": "Remember, have you seen any radio(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any radio(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0451_05/00200.jpg", "scene0451_05/00240.jpg", "scene0451_05/00280.jpg", "scene0451_05/00320.jpg", "scene0451_05/00360.jpg" ], "turn_id": 2, "scan_id": "scene0451_05", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the large, comfortable, patterned couch made of soft material\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the large, comfortable, patterned couch made of soft material\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "scene0451_05/00400.jpg", "scene0451_05/00440.jpg", "scene0451_05/00480.jpg", "scene0451_05/00520.jpg", "scene0451_05/00560.jpg" ], "turn_id": 3, "scan_id": "scene0451_05", "origin_question": "Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Right", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0451_05/00600.jpg", "scene0451_05/00640.jpg", "scene0451_05/00680.jpg", "scene0451_05/00720.jpg", "scene0451_05/00760.jpg" ], "turn_id": 4, "scan_id": "scene0451_05", "origin_question": "Which direction is the large, comfortable, patterned couch made of soft material in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the large, comfortable, patterned couch made of soft material in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0451_05/00800.jpg", "scene0451_05/00840.jpg", "scene0451_05/00880.jpg", "scene0451_05/00920.jpg", "scene0451_05/00960.jpg" ], "turn_id": 5, "scan_id": "scene0451_05", "origin_question": "How far is your current position from where you were at the end of the turn 1? (in meters)", "option": [], "answer": "2.36", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 1? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0451_05/01000.jpg", "scene0451_05/01040.jpg", "scene0451_05/01080.jpg", "scene0451_05/01120.jpg", "scene0451_05/01160.jpg" ], "turn_id": 6, "scan_id": "scene0451_05", "origin_question": "What will be the first-time appearance order of the following categories: bag, clothes, toy?", "option": [ "clothes, toy, bag", "toy, clothes, bag", "toy, bag, clothes", "bag, toy, clothes" ], "answer": "clothes, toy, bag", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bag, clothes, toy? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['clothes, toy, bag', 'toy, clothes, bag', 'toy, bag, clothes', 'bag, toy, clothes'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0451_05/01200.jpg", "scene0451_05/01240.jpg", "scene0451_05/01280.jpg", "scene0451_05/01320.jpg", "scene0451_05/01360.jpg" ], "turn_id": 7, "scan_id": "scene0451_05", "origin_question": "Which two objects are on the same side of you now? \"the medium-sized white bag\", \"the large rectangular window with a white frame\", \"the rectangular, white metal radiator in good condition\". ", "option": [ "the medium-sized white bag, the large rectangular window with a white frame", "the medium-sized white bag, the rectangular, white metal radiator in good condition", "the large rectangular window with a white frame, the rectangular, white metal radiator in good condition" ], "answer": "the large rectangular window with a white frame, the rectangular, white metal radiator in good condition", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the medium-sized white bag\", \"the large rectangular window with a white frame\", \"the rectangular, white metal radiator in good condition\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the medium-sized white bag, the large rectangular window with a white frame', 'the medium-sized white bag, the rectangular, white metal radiator in good condition', 'the large rectangular window with a white frame, the rectangular, white metal radiator in good condition'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0451_05/01400.jpg", "scene0451_05/01440.jpg", "scene0451_05/01480.jpg", "scene0451_05/01520.jpg", "scene0451_05/01560.jpg" ], "turn_id": 8, "scan_id": "scene0451_05", "origin_question": "At the end of which turn were both of \"the large, comfortable, patterned couch made of soft material\" and \"the medium-sized white bag\" on your left side?", "option": [], "answer": "5", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the large, comfortable, patterned couch made of soft material\" and \"the medium-sized white bag\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0626_00/00000.jpg", "scene0626_00/00030.jpg", "scene0626_00/00060.jpg", "scene0626_00/00090.jpg" ], "turn_id": 1, "scan_id": "scene0626_00", "origin_question": "Remember, how many chair(s) have you seen so far? ", "option": [], "answer": "3", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many chair(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0626_00/00120.jpg", "scene0626_00/00150.jpg", "scene0626_00/00180.jpg", "scene0626_00/00210.jpg" ], "turn_id": 2, "scan_id": "scene0626_00", "origin_question": "When did you first discover \"the three-seater couch in creamy yellow plush fabric\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the three-seater couch in creamy yellow plush fabric\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0626_00/00240.jpg", "scene0626_00/00270.jpg", "scene0626_00/00300.jpg", "scene0626_00/00330.jpg" ], "turn_id": 3, "scan_id": "scene0626_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0626_00/00360.jpg", "scene0626_00/00390.jpg", "scene0626_00/00420.jpg", "scene0626_00/00450.jpg" ], "turn_id": 4, "scan_id": "scene0626_00", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0626_00/00480.jpg", "scene0626_00/00510.jpg", "scene0626_00/00540.jpg", "scene0626_00/00570.jpg" ], "turn_id": 5, "scan_id": "scene0626_00", "origin_question": "At the end of which turn were both of \"the three-seater couch in creamy yellow plush fabric\" and \"the large rectangular black blackboard\" on your left side?", "option": [], "answer": "1", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the three-seater couch in creamy yellow plush fabric\" and \"the large rectangular black blackboard\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0665_01/00000.jpg", "scene0665_01/00030.jpg", "scene0665_01/00060.jpg", "scene0665_01/00090.jpg" ], "turn_id": 1, "scan_id": "scene0665_01", "origin_question": "Remember, how many table(s) have you seen so far? ", "option": [], "answer": "2", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many table(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0665_01/00120.jpg", "scene0665_01/00150.jpg", "scene0665_01/00180.jpg", "scene0665_01/00210.jpg" ], "turn_id": 2, "scan_id": "scene0665_01", "origin_question": "Remember, have you seen any chair(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any chair(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0665_01/00240.jpg", "scene0665_01/00270.jpg", "scene0665_01/00300.jpg", "scene0665_01/00330.jpg" ], "turn_id": 3, "scan_id": "scene0665_01", "origin_question": "", "option": [], "answer": "", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0665_01/00360.jpg", "scene0665_01/00390.jpg", "scene0665_01/00420.jpg", "scene0665_01/00450.jpg" ], "turn_id": 4, "scan_id": "scene0665_01", "origin_question": "How far is your current position from where you were at the end of the turn 3? (in meters)", "option": [], "answer": "1.64", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 3? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0665_01/00480.jpg", "scene0665_01/00510.jpg", "scene0665_01/00540.jpg", "scene0665_01/00570.jpg" ], "turn_id": 5, "scan_id": "scene0665_01", "origin_question": "What will be the first-time appearance order of the following categories: book, desk, table?", "option": [ "desk, table, book", "table, book, desk", "book, desk, table", "table, desk, book" ], "answer": "table, desk, book", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: book, desk, table? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['desk, table, book', 'table, book, desk', 'book, desk, table', 'table, desk, book'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "scene0665_01/00600.jpg", "scene0665_01/00630.jpg", "scene0665_01/00890.jpg", "scene0665_01/00920.jpg" ], "turn_id": 6, "scan_id": "scene0665_01", "origin_question": "Assuming the direction you are facing at the end of the turn 4 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Right", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 4 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0019_00/00000.jpg", "scene0019_00/00030.jpg", "scene0019_00/00060.jpg", "scene0019_00/00090.jpg" ], "turn_id": 1, "scan_id": "scene0019_00", "origin_question": "Remember, how many hat(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many hat(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0019_00/00120.jpg", "scene0019_00/00150.jpg", "scene0019_00/00180.jpg", "scene0019_00/00210.jpg" ], "turn_id": 2, "scan_id": "scene0019_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0019_00/00240.jpg", "scene0019_00/00270.jpg", "scene0019_00/00300.jpg", "scene0019_00/00330.jpg" ], "turn_id": 3, "scan_id": "scene0019_00", "origin_question": "At the end of which turn were both of \"the rectangular light brown wooden stool with a tapered design\" and \"the small cylindrical white plastic bin\" on your right side?", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the rectangular light brown wooden stool with a tapered design\" and \"the small cylindrical white plastic bin\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0019_00/00360.jpg", "scene0019_00/00390.jpg", "scene0019_00/00420.jpg", "scene0019_00/00450.jpg" ], "turn_id": 4, "scan_id": "scene0019_00", "origin_question": "At the end of which turn were both of \"the rectangular gray two-seater couch made of soft fabric\" and \"the lamp\" on your right side?", "option": [], "answer": "1", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the rectangular gray two-seater couch made of soft fabric\" and \"the lamp\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0100_00/00000.jpg", "scene0100_00/00040.jpg", "scene0100_00/00080.jpg", "scene0100_00/00120.jpg" ], "turn_id": 1, "scan_id": "scene0100_00", "origin_question": "Remember, have you seen any toilet paper(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any toilet paper(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0100_00/00160.jpg", "scene0100_00/00200.jpg", "scene0100_00/00240.jpg", "scene0100_00/00280.jpg" ], "turn_id": 2, "scan_id": "scene0100_00", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the mirror\"; \"the creamy white oval porcelain toilet with a smooth surface\"; \"the cylindrical white plastic bin\" ", "option": [ "the mirror", "the creamy white oval porcelain toilet with a smooth surface", "the cylindrical white plastic bin" ], "answer": "the mirror", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the mirror\"; \"the creamy white oval porcelain toilet with a smooth surface\"; \"the cylindrical white plastic bin\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the mirror', 'the creamy white oval porcelain toilet with a smooth surface', 'the cylindrical white plastic bin'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0100_00/00320.jpg", "scene0100_00/00360.jpg", "scene0100_00/00400.jpg", "scene0100_00/00440.jpg" ], "turn_id": 3, "scan_id": "scene0100_00", "origin_question": "At the end of which turn were both of \"the creamy white oval porcelain toilet with a smooth surface\" and \"the cylindrical white plastic bin\" on your right side?", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the creamy white oval porcelain toilet with a smooth surface\" and \"the cylindrical white plastic bin\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0100_00/00480.jpg", "scene0100_00/00520.jpg", "scene0100_00/00560.jpg", "scene0100_00/00600.jpg" ], "turn_id": 4, "scan_id": "scene0100_00", "origin_question": "Which direction is the cylindrical white plastic bin in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the cylindrical white plastic bin in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0100_00/00640.jpg", "scene0100_00/00680.jpg", "scene0100_00/00720.jpg", "scene0100_00/00760.jpg" ], "turn_id": 5, "scan_id": "scene0100_00", "origin_question": "When did you first discover \"the bathtub\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the bathtub\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0100_00/00800.jpg", "scene0100_00/00840.jpg", "scene0100_00/00880.jpg", "scene0100_00/00920.jpg" ], "turn_id": 6, "scan_id": "scene0100_00", "origin_question": "Which two objects are on the same side of you now? \"the creamy white oval porcelain toilet with a smooth surface\", \"the cylindrical white plastic bin\", \"the sink\". ", "option": [ "the creamy white oval porcelain toilet with a smooth surface, the cylindrical white plastic bin", "the creamy white oval porcelain toilet with a smooth surface, the sink", "the cylindrical white plastic bin, the sink" ], "answer": "the creamy white oval porcelain toilet with a smooth surface, the cylindrical white plastic bin", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the creamy white oval porcelain toilet with a smooth surface\", \"the cylindrical white plastic bin\", \"the sink\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the creamy white oval porcelain toilet with a smooth surface, the cylindrical white plastic bin', 'the creamy white oval porcelain toilet with a smooth surface, the sink', 'the cylindrical white plastic bin, the sink'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0485_00/00000.jpg", "scene0485_00/00030.jpg", "scene0485_00/00060.jpg", "scene0485_00/00090.jpg" ], "turn_id": 1, "scan_id": "scene0485_00", "origin_question": "Remember, have you seen any sheet(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any sheet(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 3 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0485_00/00140.jpg", "scene0485_00/00170.jpg", "scene0485_00/00200.jpg", "scene0485_00/00230.jpg" ], "turn_id": 2, "scan_id": "scene0485_00", "origin_question": "At the end of which turn were both of \"the round chrome basin sink with a central drain\" and \"the toilet paper\" on your right side?", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the round chrome basin sink with a central drain\" and \"the toilet paper\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 3 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0485_00/00260.jpg", "scene0485_00/00290.jpg", "scene0485_00/00320.jpg", "scene0485_00/00350.jpg" ], "turn_id": 3, "scan_id": "scene0485_00", "origin_question": "Please recall \"the rectangular vanity with a white countertop and silver handles\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.09", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the rectangular vanity with a white countertop and silver handles\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 3 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0563_00/00000.jpg", "scene0563_00/00040.jpg", "scene0563_00/00080.jpg", "scene0563_00/00120.jpg", "scene0563_00/00160.jpg" ], "turn_id": 1, "scan_id": "scene0563_00", "origin_question": "Remember, have you seen any blackboard(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any blackboard(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0563_00/00200.jpg", "scene0563_00/00240.jpg", "scene0563_00/00280.jpg", "scene0563_00/00320.jpg", "scene0563_00/00360.jpg" ], "turn_id": 2, "scan_id": "scene0563_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0563_00/00400.jpg", "scene0563_00/00440.jpg", "scene0563_00/00480.jpg", "scene0563_00/00520.jpg", "scene0563_00/00560.jpg" ], "turn_id": 3, "scan_id": "scene0563_00", "origin_question": "At the end of which turn were both of \"the white rectangular wooden table with a smooth surface\" and \"the sturdy blue backpack with black straps and pockets\" on your left side?", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the white rectangular wooden table with a smooth surface\" and \"the sturdy blue backpack with black straps and pockets\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0563_00/00600.jpg", "scene0563_00/00640.jpg", "scene0563_00/00680.jpg", "scene0563_00/00720.jpg", "scene0563_00/00760.jpg" ], "turn_id": 4, "scan_id": "scene0563_00", "origin_question": "When did you first discover \"the rectangular wooden blackboard with a white surface\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the rectangular wooden blackboard with a white surface\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0563_00/00800.jpg", "scene0563_00/00840.jpg", "scene0563_00/00900.jpg", "scene0563_00/00940.jpg", "scene0563_00/00980.jpg" ], "turn_id": 5, "scan_id": "scene0563_00", "origin_question": "At the end of which turn were both of \"the white rectangular wooden table with a smooth surface\" and \"the black metal stand with adjustable arms\" on your right side?", "option": [], "answer": "3", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the white rectangular wooden table with a smooth surface\" and \"the black metal stand with adjustable arms\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0563_00/01020.jpg", "scene0563_00/01060.jpg", "scene0563_00/01100.jpg", "scene0563_00/01140.jpg", "scene0563_00/01180.jpg" ], "turn_id": 6, "scan_id": "scene0563_00", "origin_question": "By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the sturdy blue backpack with black straps and pockets\", Y: \"the black metal stand with adjustable arms\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "farther away from both X and Y", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the sturdy blue backpack with black straps and pockets\", Y: \"the black metal stand with adjustable arms\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "scene0563_00/01220.jpg", "scene0563_00/01260.jpg", "scene0563_00/01300.jpg", "scene0563_00/01340.jpg", "scene0563_00/01380.jpg" ], "turn_id": 7, "scan_id": "scene0563_00", "origin_question": "Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Left", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0563_00/01420.jpg", "scene0563_00/01480.jpg", "scene0563_00/01520.jpg", "scene0563_00/01560.jpg", "scene0563_00/01600.jpg" ], "turn_id": 8, "scan_id": "scene0563_00", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "1.26", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0415_02/00000.jpg", "scene0415_02/00040.jpg", "scene0415_02/00080.jpg", "scene0415_02/00120.jpg" ], "turn_id": 1, "scan_id": "scene0415_02", "origin_question": "Remember, have you seen any bathtub(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any bathtub(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0415_02/00160.jpg", "scene0415_02/00200.jpg", "scene0415_02/00240.jpg", "scene0415_02/00280.jpg" ], "turn_id": 2, "scan_id": "scene0415_02", "origin_question": "When did you first discover \"the medium-sized white rectangular bathtub made of porcelain\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the medium-sized white rectangular bathtub made of porcelain\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0415_02/00320.jpg", "scene0415_02/00360.jpg", "scene0415_02/00400.jpg", "scene0415_02/00440.jpg" ], "turn_id": 3, "scan_id": "scene0415_02", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0415_02/00480.jpg", "scene0415_02/00520.jpg", "scene0415_02/00560.jpg", "scene0415_02/00600.jpg" ], "turn_id": 4, "scan_id": "scene0415_02", "origin_question": "Which direction is the white oval-shaped porcelain toilet bowl in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the white oval-shaped porcelain toilet bowl in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0415_02/00640.jpg", "scene0415_02/00680.jpg", "scene0415_02/00720.jpg", "scene0415_02/00760.jpg" ], "turn_id": 5, "scan_id": "scene0415_02", "origin_question": "Please recall \"the standard-sized, cylindrical roll of toilet paper\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.61", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the standard-sized, cylindrical roll of toilet paper\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0415_02/00800.jpg", "scene0415_02/00840.jpg", "scene0415_02/00880.jpg", "scene0415_02/00920.jpg" ], "turn_id": 6, "scan_id": "scene0415_02", "origin_question": "", "option": [], "answer": "", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0415_02/00960.jpg", "scene0415_02/01000.jpg", "scene0415_02/01040.jpg", "scene0415_02/01080.jpg" ], "turn_id": 7, "scan_id": "scene0415_02", "origin_question": "Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the vertical picture of vivid red and green leaves\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the vertical picture of vivid red and green leaves\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "1mp3d_0004_region10/148ca4f631fe4c06af243a86e008fb67_i1_0.jpg", "1mp3d_0004_region10/148ca4f631fe4c06af243a86e008fb67_i2_0.jpg", "1mp3d_0004_region10/148ca4f631fe4c06af243a86e008fb67_i1_1.jpg", "1mp3d_0004_region10/148ca4f631fe4c06af243a86e008fb67_i2_1.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0004_region10", "origin_question": "Remember, how many player(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many player(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "1mp3d_0004_region10/a17da3e4d65443f89ab161770f40ff0c_i1_5.jpg", "1mp3d_0004_region10/a17da3e4d65443f89ab161770f40ff0c_i2_5.jpg", "1mp3d_0004_region10/e4efb01c4d274ae4bfa779cdd3b1e5d1_i1_4.jpg", "1mp3d_0004_region10/e4efb01c4d274ae4bfa779cdd3b1e5d1_i2_4.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0004_region10", "origin_question": "Among these three objects(\"the large rectangular picture with a black frame\",\"the round, soft beige ceramic flowerpot\",\"the small, oval-shaped white ceramic sink with silver faucets\" ), which one is the closest to you?", "option": [ "the large rectangular picture with a black frame", "the round, soft beige ceramic flowerpot", "the small, oval-shaped white ceramic sink with silver faucets" ], "answer": "the round, soft beige ceramic flowerpot", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects(\"the large rectangular picture with a black frame\",\"the round, soft beige ceramic flowerpot\",\"the small, oval-shaped white ceramic sink with silver faucets\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the large rectangular picture with a black frame', 'the round, soft beige ceramic flowerpot', 'the small, oval-shaped white ceramic sink with silver faucets'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0004_region10/e4efb01c4d274ae4bfa779cdd3b1e5d1_i1_5.jpg", "1mp3d_0004_region10/e4efb01c4d274ae4bfa779cdd3b1e5d1_i2_5.jpg", "1mp3d_0004_region10/a17da3e4d65443f89ab161770f40ff0c_i1_4.jpg", "1mp3d_0004_region10/a17da3e4d65443f89ab161770f40ff0c_i2_4.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0004_region10", "origin_question": "Remember, have you seen any picture(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any picture(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0004_region10/148ca4f631fe4c06af243a86e008fb67_i1_3.jpg", "1mp3d_0004_region10/148ca4f631fe4c06af243a86e008fb67_i2_3.jpg", "1mp3d_0004_region10/148ca4f631fe4c06af243a86e008fb67_i1_2.jpg", "1mp3d_0004_region10/148ca4f631fe4c06af243a86e008fb67_i2_2.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0004_region10", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the white table with a marble countertop\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the white table with a marble countertop\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0004_region10/a17da3e4d65443f89ab161770f40ff0c_i1_0.jpg", "1mp3d_0004_region10/a17da3e4d65443f89ab161770f40ff0c_i2_0.jpg", "1mp3d_0004_region10/a17da3e4d65443f89ab161770f40ff0c_i1_1.jpg", "1mp3d_0004_region10/a17da3e4d65443f89ab161770f40ff0c_i2_1.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0004_region10", "origin_question": "Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the white table with a marble countertop\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the white table with a marble countertop\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "1mp3d_0004_region10/e4efb01c4d274ae4bfa779cdd3b1e5d1_i1_1.jpg", "1mp3d_0004_region10/e4efb01c4d274ae4bfa779cdd3b1e5d1_i2_1.jpg", "1mp3d_0004_region10/e4efb01c4d274ae4bfa779cdd3b1e5d1_i1_0.jpg", "1mp3d_0004_region10/e4efb01c4d274ae4bfa779cdd3b1e5d1_i2_0.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0004_region10", "origin_question": "What will be the first-time appearance order of the following categories: bathtub, doorframe, towel?", "option": [ "towel, bathtub, doorframe", "doorframe, towel, bathtub", "doorframe, bathtub, towel", "towel, doorframe, bathtub" ], "answer": "doorframe, towel, bathtub", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bathtub, doorframe, towel? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['towel, bathtub, doorframe', 'doorframe, towel, bathtub', 'doorframe, bathtub, towel', 'towel, doorframe, bathtub'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0064_00/00000.jpg", "scene0064_00/00040.jpg", "scene0064_00/00080.jpg", "scene0064_00/00120.jpg" ], "turn_id": 1, "scan_id": "scene0064_00", "origin_question": "Remember, have you seen any ball(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any ball(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0064_00/00160.jpg", "scene0064_00/00200.jpg", "scene0064_00/00240.jpg", "scene0064_00/00280.jpg" ], "turn_id": 2, "scan_id": "scene0064_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0064_00/00320.jpg", "scene0064_00/00360.jpg", "scene0064_00/00400.jpg", "scene0064_00/00440.jpg" ], "turn_id": 3, "scan_id": "scene0064_00", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "2.72", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0064_00/00480.jpg", "scene0064_00/00520.jpg", "scene0064_00/00560.jpg", "scene0064_00/00600.jpg" ], "turn_id": 4, "scan_id": "scene0064_00", "origin_question": "In which turn did you see both \"the rectangular brown wooden door\" and \"the medium-sized black and red rectangular backpack\" simultaneously? ", "option": [], "answer": "4", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the rectangular brown wooden door\" and \"the medium-sized black and red rectangular backpack\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0064_00/00640.jpg", "scene0064_00/00680.jpg", "scene0064_00/00720.jpg", "scene0064_00/00760.jpg" ], "turn_id": 5, "scan_id": "scene0064_00", "origin_question": "Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "77.62", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0064_00/00800.jpg", "scene0064_00/00840.jpg", "scene0064_00/00880.jpg", "scene0064_00/00920.jpg" ], "turn_id": 6, "scan_id": "scene0064_00", "origin_question": "Remember, how many shelf(s) have you seen so far? ", "option": [], "answer": "2", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many shelf(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0064_00/00960.jpg", "scene0064_00/01000.jpg", "scene0064_00/01040.jpg", "scene0064_00/01080.jpg" ], "turn_id": 7, "scan_id": "scene0064_00", "origin_question": "When did you first discover \"the medium-sized black and red rectangular backpack\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the medium-sized black and red rectangular backpack\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0004_region7/79961f24d27c4f4e8b4c4e0b62058ea1_i1_3.jpg", "1mp3d_0004_region7/79961f24d27c4f4e8b4c4e0b62058ea1_i2_3.jpg", "1mp3d_0004_region7/d1356be1da164654b63fa549774272e5_i1_4.jpg", "1mp3d_0004_region7/d1356be1da164654b63fa549774272e5_i2_4.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0004_region7", "origin_question": "Remember, have you seen any bed(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any bed(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "1mp3d_0004_region7/d1356be1da164654b63fa549774272e5_i1_3.jpg", "1mp3d_0004_region7/d1356be1da164654b63fa549774272e5_i2_3.jpg", "1mp3d_0004_region7/8e370e8a9f8743b58f3e66b731413d18_i1_0.jpg", "1mp3d_0004_region7/8e370e8a9f8743b58f3e66b731413d18_i2_0.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0004_region7", "origin_question": "When did you first discover \"the thick, light blue knitted blanket\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the thick, light blue knitted blanket\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "1mp3d_0004_region7/816408b1ebd841e99b893b2c2dc240ad_i1_1.jpg", "1mp3d_0004_region7/816408b1ebd841e99b893b2c2dc240ad_i2_1.jpg", "1mp3d_0004_region7/85d2a23222e34cd99e3bc5f719b1cdfc_i1_2.jpg", "1mp3d_0004_region7/85d2a23222e34cd99e3bc5f719b1cdfc_i2_2.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0004_region7", "origin_question": "Please recall \"the rectangular wooden shelf with three levels\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.71", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the rectangular wooden shelf with three levels\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "1mp3d_0004_region7/209cfff028e843749f35039a95d16d97_i1_2.jpg", "1mp3d_0004_region7/209cfff028e843749f35039a95d16d97_i2_2.jpg", "1mp3d_0004_region7/4d059bb6efaf411985e6ee5eba2b0c64_i1_2.jpg", "1mp3d_0004_region7/4d059bb6efaf411985e6ee5eba2b0c64_i2_2.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0004_region7", "origin_question": "Which direction is the stairs in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the stairs in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "1mp3d_0004_region7/4d059bb6efaf411985e6ee5eba2b0c64_i1_3.jpg", "1mp3d_0004_region7/4d059bb6efaf411985e6ee5eba2b0c64_i2_3.jpg", "1mp3d_0004_region7/209cfff028e843749f35039a95d16d97_i1_3.jpg", "1mp3d_0004_region7/209cfff028e843749f35039a95d16d97_i2_3.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0004_region7", "origin_question": "Which direction is the lamp in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the lamp in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "1mp3d_0004_region7/85d2a23222e34cd99e3bc5f719b1cdfc_i1_3.jpg", "1mp3d_0004_region7/85d2a23222e34cd99e3bc5f719b1cdfc_i2_3.jpg", "1mp3d_0004_region7/816408b1ebd841e99b893b2c2dc240ad_i1_3.jpg", "1mp3d_0004_region7/816408b1ebd841e99b893b2c2dc240ad_i2_3.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0004_region7", "origin_question": "Which direction is the classic beige couch with a curved backrest in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the classic beige couch with a curved backrest in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "1mp3d_0004_region7/816408b1ebd841e99b893b2c2dc240ad_i1_2.jpg", "1mp3d_0004_region7/816408b1ebd841e99b893b2c2dc240ad_i2_2.jpg", "1mp3d_0004_region7/8e370e8a9f8743b58f3e66b731413d18_i1_1.jpg", "1mp3d_0004_region7/8e370e8a9f8743b58f3e66b731413d18_i2_1.jpg" ], "turn_id": 7, "scan_id": "1mp3d_0004_region7", "origin_question": "Please recall \"the picture\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.59", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the picture\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "1mp3d_0004_region7/bb1b39f93d884d97a5da9a7c01c52991_i1_3.jpg", "1mp3d_0004_region7/bb1b39f93d884d97a5da9a7c01c52991_i2_3.jpg", "1mp3d_0004_region7/bb1b39f93d884d97a5da9a7c01c52991_i1_2.jpg", "1mp3d_0004_region7/bb1b39f93d884d97a5da9a7c01c52991_i2_2.jpg" ], "turn_id": 8, "scan_id": "1mp3d_0004_region7", "origin_question": "Using your orientation at the end of turn 5 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "76.16", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 5 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0264_02/00000.jpg", "scene0264_02/00040.jpg", "scene0264_02/00080.jpg", "scene0264_02/00120.jpg" ], "turn_id": 1, "scan_id": "scene0264_02", "origin_question": "Remember, have you seen any teapot(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any teapot(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0264_02/00160.jpg", "scene0264_02/00200.jpg", "scene0264_02/00240.jpg", "scene0264_02/00280.jpg" ], "turn_id": 2, "scan_id": "scene0264_02", "origin_question": "Please recall \"the rectangular silver computer tower\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.74", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the rectangular silver computer tower\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0264_02/00320.jpg", "scene0264_02/00360.jpg", "scene0264_02/00400.jpg", "scene0264_02/00440.jpg" ], "turn_id": 3, "scan_id": "scene0264_02", "origin_question": "At the end of which turn were both of \"the rectangular silver computer tower\" and \"the white and black metal printer\" on your left side?", "option": [], "answer": "2", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the rectangular silver computer tower\" and \"the white and black metal printer\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0264_02/00480.jpg", "scene0264_02/00520.jpg", "scene0264_02/00560.jpg", "scene0264_02/00600.jpg" ], "turn_id": 4, "scan_id": "scene0264_02", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the bottle\"; \"the shelf\"; \"the tall, white multifunctional metal copier\" ", "option": [ "the bottle", "the shelf", "the tall, white multifunctional metal copier" ], "answer": "the tall, white multifunctional metal copier", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the bottle\"; \"the shelf\"; \"the tall, white multifunctional metal copier\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the bottle', 'the shelf', 'the tall, white multifunctional metal copier'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0264_02/00640.jpg", "scene0264_02/00680.jpg", "scene0264_02/00720.jpg", "scene0264_02/00760.jpg" ], "turn_id": 5, "scan_id": "scene0264_02", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the bottle\" from your position?", "option": [], "answer": "98.93", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the bottle\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0264_02/00800.jpg", "scene0264_02/00840.jpg", "scene0264_02/00880.jpg", "scene0264_02/00920.jpg" ], "turn_id": 6, "scan_id": "scene0264_02", "origin_question": "In which turn did you see both \"the bottle\" and \"the tall, white multifunctional metal copier\" simultaneously? ", "option": [], "answer": "4", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the bottle\" and \"the tall, white multifunctional metal copier\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0264_02/00960.jpg", "scene0264_02/01000.jpg", "scene0264_02/01040.jpg", "scene0264_02/01080.jpg" ], "turn_id": 7, "scan_id": "scene0264_02", "origin_question": "What will be the first-time appearance order of the following categories: book, copier, door?", "option": [ "copier, book, door", "door, copier, book", "door, book, copier", "book, door, copier" ], "answer": "door, book, copier", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: book, copier, door? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['copier, book, door', 'door, copier, book', 'door, book, copier', 'book, door, copier'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0040_region35/9e486ef6653345158c6489e657f3a6be_i1_5.jpg", "1mp3d_0040_region35/9e486ef6653345158c6489e657f3a6be_i2_5.jpg", "1mp3d_0040_region35/9db4a2dab9304659b414bb92d2e5f248_i1_4.jpg", "1mp3d_0040_region35/9db4a2dab9304659b414bb92d2e5f248_i2_4.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0040_region35", "origin_question": "Remember, have you seen any vacuum cleaner(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any vacuum cleaner(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0040_region35/9c249d7487c3497b970d17e270bfe6ce_i1_5.jpg", "1mp3d_0040_region35/9c249d7487c3497b970d17e270bfe6ce_i2_5.jpg", "1mp3d_0040_region35/9db4a2dab9304659b414bb92d2e5f248_i1_5.jpg", "1mp3d_0040_region35/9db4a2dab9304659b414bb92d2e5f248_i2_5.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0040_region35", "origin_question": "Remember, have you seen any arch(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any arch(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "1mp3d_0040_region35/9e486ef6653345158c6489e657f3a6be_i1_4.jpg", "1mp3d_0040_region35/9e486ef6653345158c6489e657f3a6be_i2_4.jpg", "1mp3d_0040_region35/98b3afcfbd9549bc823fd118719caba0_i1_4.jpg", "1mp3d_0040_region35/98b3afcfbd9549bc823fd118719caba0_i2_4.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0040_region35", "origin_question": "Is the rectangular, beige-colored door with six panels to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the rectangular, beige-colored door with six panels to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "1mp3d_0040_region35/87c114a1c9c848bf866d78a78150550b_i1_4.jpg", "1mp3d_0040_region35/87c114a1c9c848bf866d78a78150550b_i2_4.jpg", "1mp3d_0040_region35/ed32d039ef0d4cb48be9e2d2de8ac133_i1_3.jpg", "1mp3d_0040_region35/ed32d039ef0d4cb48be9e2d2de8ac133_i2_3.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0040_region35", "origin_question": "Please recall \"the rectangular silver stainless steel sink\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.63", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the rectangular silver stainless steel sink\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "1mp3d_0040_region35/89e66ed1e2ab484f9de5cb936c815453_i1_3.jpg", "1mp3d_0040_region35/89e66ed1e2ab484f9de5cb936c815453_i2_3.jpg", "1mp3d_0040_region35/50caa686b5b947d7b6b8e7fcf6bb1152_i1_3.jpg", "1mp3d_0040_region35/50caa686b5b947d7b6b8e7fcf6bb1152_i2_3.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0040_region35", "origin_question": "Remember, how many bag(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many bag(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "1mp3d_0040_region35/50caa686b5b947d7b6b8e7fcf6bb1152_i1_2.jpg", "1mp3d_0040_region35/50caa686b5b947d7b6b8e7fcf6bb1152_i2_2.jpg", "1mp3d_0040_region35/50caa686b5b947d7b6b8e7fcf6bb1152_i1_1.jpg", "1mp3d_0040_region35/50caa686b5b947d7b6b8e7fcf6bb1152_i2_1.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0040_region35", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "4.84", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "1mp3d_0040_region35/89e66ed1e2ab484f9de5cb936c815453_i1_1.jpg", "1mp3d_0040_region35/89e66ed1e2ab484f9de5cb936c815453_i2_1.jpg", "1mp3d_0040_region35/89e66ed1e2ab484f9de5cb936c815453_i1_2.jpg", "1mp3d_0040_region35/89e66ed1e2ab484f9de5cb936c815453_i2_2.jpg" ], "turn_id": 7, "scan_id": "1mp3d_0040_region35", "origin_question": "How far is your current position from where you were at the end of the turn 1? (in meters)", "option": [], "answer": "6.33", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 1? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "1mp3d_0040_region35/ed32d039ef0d4cb48be9e2d2de8ac133_i1_2.jpg", "1mp3d_0040_region35/ed32d039ef0d4cb48be9e2d2de8ac133_i2_2.jpg", "1mp3d_0040_region35/87c114a1c9c848bf866d78a78150550b_i1_3.jpg", "1mp3d_0040_region35/87c114a1c9c848bf866d78a78150550b_i2_3.jpg" ], "turn_id": 8, "scan_id": "1mp3d_0040_region35", "origin_question": "Please recall \"the rectangular silver stainless steel sink\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.32", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the rectangular silver stainless steel sink\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "1mp3d_0040_region35/98b3afcfbd9549bc823fd118719caba0_i1_5.jpg", "1mp3d_0040_region35/98b3afcfbd9549bc823fd118719caba0_i2_5.jpg", "1mp3d_0040_region35/98b3afcfbd9549bc823fd118719caba0_i1_0.jpg", "1mp3d_0040_region35/98b3afcfbd9549bc823fd118719caba0_i2_0.jpg" ], "turn_id": 9, "scan_id": "1mp3d_0040_region35", "origin_question": "Using your orientation at the end of turn 6 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 9 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 6 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0057_region17/9ce062b37b364d078541836381052e9a_i1_1.jpg", "1mp3d_0057_region17/9ce062b37b364d078541836381052e9a_i2_1.jpg", "1mp3d_0057_region17/719c98cc49cd4ea4bdfa1094f30c54d9_i1_2.jpg", "1mp3d_0057_region17/719c98cc49cd4ea4bdfa1094f30c54d9_i2_2.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0057_region17", "origin_question": "Remember, have you seen any toilet paper(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any toilet paper(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "1mp3d_0057_region17/9ce062b37b364d078541836381052e9a_i2_4.jpg", "1mp3d_0057_region17/719c98cc49cd4ea4bdfa1094f30c54d9_i1_3.jpg", "1mp3d_0057_region17/719c98cc49cd4ea4bdfa1094f30c54d9_i2_3.jpg", "1mp3d_0057_region17/719c98cc49cd4ea4bdfa1094f30c54d9_i1_4.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0057_region17", "origin_question": "When did you first discover \"the rectangular white porcelain sink with a silver faucet\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the rectangular white porcelain sink with a silver faucet\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "1mp3d_0057_region17/719c98cc49cd4ea4bdfa1094f30c54d9_i2_4.jpg", "1mp3d_0057_region17/9ce062b37b364d078541836381052e9a_i1_5.jpg", "1mp3d_0057_region17/9ce062b37b364d078541836381052e9a_i2_5.jpg", "1mp3d_0057_region17/9ce062b37b364d078541836381052e9a_i1_0.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0057_region17", "origin_question": "Is the rectangular white porcelain sink with a silver faucet to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the rectangular white porcelain sink with a silver faucet to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "1mp3d_0057_region17/9ce062b37b364d078541836381052e9a_i2_0.jpg", "1mp3d_0057_region17/58eac222ed224afb9241f2f65d189dba_i1_2.jpg", "1mp3d_0057_region17/58eac222ed224afb9241f2f65d189dba_i2_2.jpg", "1mp3d_0057_region17/b03190e8bd2044d6b5361650f69dde54_i1_0.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0057_region17", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the rectangular white porcelain sink with a silver faucet\" from your position?", "option": [], "answer": "145.01", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the rectangular white porcelain sink with a silver faucet\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "1mp3d_0057_region17/b03190e8bd2044d6b5361650f69dde54_i2_0.jpg", "1mp3d_0057_region17/b03190e8bd2044d6b5361650f69dde54_i1_5.jpg", "1mp3d_0057_region17/b03190e8bd2044d6b5361650f69dde54_i2_5.jpg", "1mp3d_0057_region17/b03190e8bd2044d6b5361650f69dde54_i1_4.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0057_region17", "origin_question": "In which turn were you closest to \"the white ceramic standalone toilet in good condition\"? ", "option": [], "answer": "4", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the white ceramic standalone toilet in good condition\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0074_region0/8844f8fdc366462587b70fb7ede46242_i1_4.jpg", "1mp3d_0074_region0/8844f8fdc366462587b70fb7ede46242_i2_4.jpg", "1mp3d_0074_region0/66cb91d800e74042bf34287a76fd36ab_i1_4.jpg", "1mp3d_0074_region0/66cb91d800e74042bf34287a76fd36ab_i2_4.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0074_region0", "origin_question": "Remember, have you seen any wardrobe(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any wardrobe(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "1mp3d_0074_region0/75d7eb848ea84a6dbb717dc231cf2dab_i1_3.jpg", "1mp3d_0074_region0/75d7eb848ea84a6dbb717dc231cf2dab_i2_3.jpg", "1mp3d_0074_region0/66cb91d800e74042bf34287a76fd36ab_i1_1.jpg", "1mp3d_0074_region0/66cb91d800e74042bf34287a76fd36ab_i2_1.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0074_region0", "origin_question": "Which two objects are on the same side of you now? \"the rectangular picture made of shiny stones in brown and gold\", \"the light\", \"the large rectangular white wardrobe with multiple doors\". ", "option": [ "the rectangular picture made of shiny stones in brown and gold, the light", "the rectangular picture made of shiny stones in brown and gold, the large rectangular white wardrobe with multiple doors", "the light, the large rectangular white wardrobe with multiple doors" ], "answer": "the rectangular picture made of shiny stones in brown and gold, the light", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the rectangular picture made of shiny stones in brown and gold\", \"the light\", \"the large rectangular white wardrobe with multiple doors\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular picture made of shiny stones in brown and gold, the light', 'the rectangular picture made of shiny stones in brown and gold, the large rectangular white wardrobe with multiple doors', 'the light, the large rectangular white wardrobe with multiple doors'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "1mp3d_0074_region0/8844f8fdc366462587b70fb7ede46242_i1_5.jpg", "1mp3d_0074_region0/8844f8fdc366462587b70fb7ede46242_i2_5.jpg", "1mp3d_0074_region0/66cb91d800e74042bf34287a76fd36ab_i1_0.jpg", "1mp3d_0074_region0/66cb91d800e74042bf34287a76fd36ab_i2_0.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0074_region0", "origin_question": "In which turn did you see both \"the rectangular medium bed with a white bedsheet\" and \"the curtain\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the rectangular medium bed with a white bedsheet\" and \"the curtain\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "1mp3d_0074_region0/75d7eb848ea84a6dbb717dc231cf2dab_i1_4.jpg", "1mp3d_0074_region0/75d7eb848ea84a6dbb717dc231cf2dab_i2_4.jpg", "1mp3d_0074_region0/b98717151b7b49f59af95a9b7111a658_i1_3.jpg", "1mp3d_0074_region0/b98717151b7b49f59af95a9b7111a658_i2_3.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0074_region0", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "3.55", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "1mp3d_0074_region0/b98717151b7b49f59af95a9b7111a658_i1_2.jpg", "1mp3d_0074_region0/b98717151b7b49f59af95a9b7111a658_i2_2.jpg", "1mp3d_0074_region0/75d7eb848ea84a6dbb717dc231cf2dab_i1_2.jpg", "1mp3d_0074_region0/75d7eb848ea84a6dbb717dc231cf2dab_i2_2.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0074_region0", "origin_question": "At the end of which turn were both of \"the curtain\" and \"the large rectangular white glass window\" on your right side?", "option": [], "answer": "3", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the curtain\" and \"the large rectangular white glass window\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "1mp3d_0074_region0/75d7eb848ea84a6dbb717dc231cf2dab_i1_1.jpg", "1mp3d_0074_region0/75d7eb848ea84a6dbb717dc231cf2dab_i2_1.jpg", "1mp3d_0074_region0/b98717151b7b49f59af95a9b7111a658_i1_1.jpg", "1mp3d_0074_region0/b98717151b7b49f59af95a9b7111a658_i2_1.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0074_region0", "origin_question": "At the end of which turn were both of \"the large rectangular white wardrobe with multiple doors\" and \"the large rectangular white glass window\" on your right side?", "option": [], "answer": "3", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the large rectangular white wardrobe with multiple doors\" and \"the large rectangular white glass window\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0629_01/00000.jpg", "scene0629_01/00040.jpg", "scene0629_01/00080.jpg", "scene0629_01/00120.jpg", "scene0629_01/00160.jpg" ], "turn_id": 1, "scan_id": "scene0629_01", "origin_question": "Remember, have you seen any picture(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any picture(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0629_01/00200.jpg", "scene0629_01/00240.jpg", "scene0629_01/00280.jpg", "scene0629_01/00320.jpg", "scene0629_01/00360.jpg" ], "turn_id": 2, "scan_id": "scene0629_01", "origin_question": "At the end of which turn were both of \"the mattress\" and \"the table\" on your left side?", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the mattress\" and \"the table\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0629_01/00400.jpg", "scene0629_01/00440.jpg", "scene0629_01/00480.jpg", "scene0629_01/00520.jpg", "scene0629_01/00560.jpg" ], "turn_id": 3, "scan_id": "scene0629_01", "origin_question": "Which direction is the mattress in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the mattress in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0629_01/00600.jpg", "scene0629_01/00640.jpg", "scene0629_01/00680.jpg", "scene0629_01/00720.jpg", "scene0629_01/00760.jpg" ], "turn_id": 4, "scan_id": "scene0629_01", "origin_question": "At the end of which turn were both of \"the mattress\" and \"the table\" on your right side?", "option": [], "answer": "4", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the mattress\" and \"the table\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0629_01/00800.jpg", "scene0629_01/00840.jpg", "scene0629_01/00880.jpg", "scene0629_01/00920.jpg", "scene0629_01/00960.jpg" ], "turn_id": 5, "scan_id": "scene0629_01", "origin_question": "In which turn did you see both \"the mattress\" and \"the standard white six-panel wooden door\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the mattress\" and \"the standard white six-panel wooden door\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0629_01/01000.jpg", "scene0629_01/01040.jpg", "scene0629_01/01080.jpg", "scene0629_01/01120.jpg", "scene0629_01/01160.jpg" ], "turn_id": 6, "scan_id": "scene0629_01", "origin_question": "Which direction is the mattress in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the mattress in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0629_01/01200.jpg", "scene0629_01/01240.jpg", "scene0629_01/01280.jpg", "scene0629_01/01320.jpg", "scene0629_01/01360.jpg" ], "turn_id": 7, "scan_id": "scene0629_01", "origin_question": "How far is your current position from where you were at the end of the turn 1? (in meters)", "option": [], "answer": "1.74", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 1? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0629_01/01400.jpg", "scene0629_01/01440.jpg", "scene0629_01/01480.jpg", "scene0629_01/01520.jpg", "scene0629_01/01560.jpg" ], "turn_id": 8, "scan_id": "scene0629_01", "origin_question": "Which two objects are on the same side of you now? \"the small rectangular deep brown wooden desk\", \"the plant\", \"the cylindrical plastic bottle with a blue lid and white body\". ", "option": [ "the small rectangular deep brown wooden desk, the plant", "the small rectangular deep brown wooden desk, the cylindrical plastic bottle with a blue lid and white body", "the plant, the cylindrical plastic bottle with a blue lid and white body" ], "answer": "the small rectangular deep brown wooden desk, the plant", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the small rectangular deep brown wooden desk\", \"the plant\", \"the cylindrical plastic bottle with a blue lid and white body\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the small rectangular deep brown wooden desk, the plant', 'the small rectangular deep brown wooden desk, the cylindrical plastic bottle with a blue lid and white body', 'the plant, the cylindrical plastic bottle with a blue lid and white body'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0629_01/01600.jpg", "scene0629_01/01640.jpg", "scene0629_01/01680.jpg", "scene0629_01/01720.jpg", "scene0629_01/01760.jpg" ], "turn_id": 9, "scan_id": "scene0629_01", "origin_question": "Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "133.74", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0629_01/01800.jpg", "scene0629_01/01840.jpg", "scene0629_01/01880.jpg", "scene0629_01/01920.jpg", "scene0629_01/01960.jpg" ], "turn_id": 10, "scan_id": "scene0629_01", "origin_question": "At the end of which turn were both of \"the cylindrical plastic bottle with a blue lid and white body\" and \"the large rectangular mirror with a black frame\" on your right side?", "option": [], "answer": "4", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the cylindrical plastic bottle with a blue lid and white body\" and \"the large rectangular mirror with a black frame\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "1mp3d_0067_region14/01f2401b5f944914911eece07c0f3ead_i1_2.jpg", "1mp3d_0067_region14/01f2401b5f944914911eece07c0f3ead_i2_2.jpg", "1mp3d_0067_region14/35dde286a5324af49d540f7dd0bc3b74_i1_4.jpg", "1mp3d_0067_region14/35dde286a5324af49d540f7dd0bc3b74_i2_4.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0067_region14", "origin_question": "Remember, how many garbage(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many garbage(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "1mp3d_0067_region14/01f2401b5f944914911eece07c0f3ead_i1_0.jpg", "1mp3d_0067_region14/01f2401b5f944914911eece07c0f3ead_i2_0.jpg", "1mp3d_0067_region14/01f2401b5f944914911eece07c0f3ead_i1_1.jpg", "1mp3d_0067_region14/01f2401b5f944914911eece07c0f3ead_i2_1.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0067_region14", "origin_question": "Is the classic lamp with a creamy white wooden shade to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the classic lamp with a creamy white wooden shade to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "1mp3d_0067_region14/9cc9a6f97b09469e92ada5d6795d125c_i1_4.jpg", "1mp3d_0067_region14/9cc9a6f97b09469e92ada5d6795d125c_i2_4.jpg", "1mp3d_0067_region14/efeef7cc82c84690addb0bf415f075ea_i1_3.jpg", "1mp3d_0067_region14/efeef7cc82c84690addb0bf415f075ea_i2_3.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0067_region14", "origin_question": "Which two objects are on the same side of you now? \"the classic lamp with a creamy white wooden shade\", \"the sturdy rectangular grey metal rack with wheels\", \"the square tray with dark wooden sides and metal handles\". ", "option": [ "the classic lamp with a creamy white wooden shade, the sturdy rectangular grey metal rack with wheels", "the classic lamp with a creamy white wooden shade, the square tray with dark wooden sides and metal handles", "the sturdy rectangular grey metal rack with wheels, the square tray with dark wooden sides and metal handles" ], "answer": "the classic lamp with a creamy white wooden shade, the sturdy rectangular grey metal rack with wheels", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the classic lamp with a creamy white wooden shade\", \"the sturdy rectangular grey metal rack with wheels\", \"the square tray with dark wooden sides and metal handles\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the classic lamp with a creamy white wooden shade, the sturdy rectangular grey metal rack with wheels', 'the classic lamp with a creamy white wooden shade, the square tray with dark wooden sides and metal handles', 'the sturdy rectangular grey metal rack with wheels, the square tray with dark wooden sides and metal handles'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "1mp3d_0067_region14/9cc9a6f97b09469e92ada5d6795d125c_i1_0.jpg", "1mp3d_0067_region14/9cc9a6f97b09469e92ada5d6795d125c_i2_0.jpg", "1mp3d_0067_region14/efeef7cc82c84690addb0bf415f075ea_i1_1.jpg", "1mp3d_0067_region14/efeef7cc82c84690addb0bf415f075ea_i2_1.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0067_region14", "origin_question": "Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance forward or backward from that position?", "option": [ "Forward", "Backward" ], "answer": "Forward", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance forward or backward from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Forward', 'Backward'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "1mp3d_0067_region14/efeef7cc82c84690addb0bf415f075ea_i1_0.jpg", "1mp3d_0067_region14/efeef7cc82c84690addb0bf415f075ea_i2_0.jpg", "1mp3d_0067_region14/efeef7cc82c84690addb0bf415f075ea_i1_5.jpg", "1mp3d_0067_region14/efeef7cc82c84690addb0bf415f075ea_i2_5.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0067_region14", "origin_question": "Is the sturdy rectangular grey metal rack with wheels to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the sturdy rectangular grey metal rack with wheels to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "1mp3d_0067_region14/efeef7cc82c84690addb0bf415f075ea_i1_4.jpg", "1mp3d_0067_region14/efeef7cc82c84690addb0bf415f075ea_i2_4.jpg", "1mp3d_0067_region14/efeef7cc82c84690addb0bf415f075ea_i1_2.jpg", "1mp3d_0067_region14/efeef7cc82c84690addb0bf415f075ea_i2_2.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0067_region14", "origin_question": "Which direction is the square tray with dark wooden sides and metal handles in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the square tray with dark wooden sides and metal handles in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0106_02/00000.jpg", "scene0106_02/00030.jpg", "scene0106_02/00060.jpg", "scene0106_02/00090.jpg" ], "turn_id": 1, "scan_id": "scene0106_02", "origin_question": "Remember, how many monitor(s) have you seen so far? ", "option": [], "answer": "3", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many monitor(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0106_02/00120.jpg", "scene0106_02/00150.jpg", "scene0106_02/00180.jpg", "scene0106_02/00210.jpg" ], "turn_id": 2, "scan_id": "scene0106_02", "origin_question": "In which turn did you see both \"the blue rectangular wooden door with a glass panel\" and \"then oval-shaped white porcelain sink with a curved faucet\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the blue rectangular wooden door with a glass panel\" and \"then oval-shaped white porcelain sink with a curved faucet\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0106_02/00240.jpg", "scene0106_02/00270.jpg", "scene0106_02/00300.jpg", "scene0106_02/00330.jpg" ], "turn_id": 3, "scan_id": "scene0106_02", "origin_question": "What will be the first-time appearance order of the following categories: computer, pillar, window?", "option": [ "pillar, window, computer", "window, pillar, computer", "computer, window, pillar", "computer, pillar, window" ], "answer": "computer, window, pillar", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: computer, pillar, window? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['pillar, window, computer', 'window, pillar, computer', 'computer, window, pillar', 'computer, pillar, window'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0106_02/00360.jpg", "scene0106_02/00390.jpg", "scene0106_02/00420.jpg", "scene0106_02/00450.jpg" ], "turn_id": 4, "scan_id": "scene0106_02", "origin_question": "When did you first discover \"then oval-shaped white porcelain sink with a curved faucet\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"then oval-shaped white porcelain sink with a curved faucet\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0106_02/00480.jpg", "scene0106_02/00510.jpg", "scene0106_02/00540.jpg", "scene0106_02/00570.jpg" ], "turn_id": 5, "scan_id": "scene0106_02", "origin_question": "At the end of which turn were both of \"the large rectangular glass window with a black frame\" and \"the rectangular metal radiator with a simple design\" on your left side?", "option": [], "answer": "1", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the large rectangular glass window with a black frame\" and \"the rectangular metal radiator with a simple design\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0106_02/00600.jpg", "scene0106_02/00630.jpg", "scene0106_02/00660.jpg", "scene0106_02/00690.jpg" ], "turn_id": 6, "scan_id": "scene0106_02", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"then oval-shaped white porcelain sink with a curved faucet\" from your position?", "option": [], "answer": "79.86", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"then oval-shaped white porcelain sink with a curved faucet\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0106_02/00720.jpg", "scene0106_02/00750.jpg", "scene0106_02/00780.jpg", "scene0106_02/00810.jpg" ], "turn_id": 7, "scan_id": "scene0106_02", "origin_question": "Which direction is then oval-shaped white porcelain sink with a curved faucet in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is then oval-shaped white porcelain sink with a curved faucet in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0044_region45/9783920594ec45568fec867a36eea437_i1_1.jpg", "1mp3d_0044_region45/9783920594ec45568fec867a36eea437_i2_1.jpg", "1mp3d_0044_region45/dcbb02cfd24743a8add9e8d221b24084_i1_1.jpg", "1mp3d_0044_region45/dcbb02cfd24743a8add9e8d221b24084_i2_1.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0044_region45", "origin_question": "Remember, have you seen any toothbrush(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any toothbrush(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 2 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "1mp3d_0044_region45/9783920594ec45568fec867a36eea437_i1_4.jpg", "1mp3d_0044_region45/9783920594ec45568fec867a36eea437_i2_4.jpg", "1mp3d_0044_region45/7a7c76c10fc9473fbd7d04a66733b9bc_i1_3.jpg", "1mp3d_0044_region45/7a7c76c10fc9473fbd7d04a66733b9bc_i2_3.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0044_region45", "origin_question": "In which turn did you see both \"the window\" and \"the grand, brown metal roof with light bulbs\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the window\" and \"the grand, brown metal roof with light bulbs\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 2 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0040_region41/a2fc97d027ed4f82971e361e5e476de4_i1_5.jpg", "1mp3d_0040_region41/a2fc97d027ed4f82971e361e5e476de4_i2_5.jpg", "1mp3d_0040_region41/a0e4ec41dae34ccfbfaaf719c22b8388_i1_2.jpg", "1mp3d_0040_region41/a0e4ec41dae34ccfbfaaf719c22b8388_i2_2.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0040_region41", "origin_question": "Remember, have you seen any bottle(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any bottle(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 2 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "1mp3d_0040_region41/a2fc97d027ed4f82971e361e5e476de4_i1_0.jpg", "1mp3d_0040_region41/a2fc97d027ed4f82971e361e5e476de4_i2_0.jpg", "1mp3d_0040_region41/a0e4ec41dae34ccfbfaaf719c22b8388_i1_4.jpg", "1mp3d_0040_region41/a0e4ec41dae34ccfbfaaf719c22b8388_i2_4.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0040_region41", "origin_question": "Please recall \"the counter\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.58", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the counter\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 2 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0044_region31/9127aae77ad348eb8fd02f7cd52a3ce5_i1_5.jpg", "1mp3d_0044_region31/9127aae77ad348eb8fd02f7cd52a3ce5_i2_5.jpg", "1mp3d_0044_region31/8b7469d12514415cada466823f2a7063_i1_5.jpg", "1mp3d_0044_region31/8b7469d12514415cada466823f2a7063_i2_5.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0044_region31", "origin_question": "Remember, have you seen any door(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any door(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0044_region31/9127aae77ad348eb8fd02f7cd52a3ce5_i1_0.jpg", "1mp3d_0044_region31/9127aae77ad348eb8fd02f7cd52a3ce5_i2_0.jpg", "1mp3d_0044_region31/e0047ba33fdb49c499dab9d25da633a8_i1_2.jpg", "1mp3d_0044_region31/e0047ba33fdb49c499dab9d25da633a8_i2_2.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0044_region31", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the medium-sized brown toy with a white shirt\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the medium-sized brown toy with a white shirt\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0044_region31/9177fe5f923242f0b1ba16c7a058e5ac_i1_2.jpg", "1mp3d_0044_region31/9177fe5f923242f0b1ba16c7a058e5ac_i2_2.jpg", "1mp3d_0044_region31/b95772692ede4e9aa2ce4828447e8419_i1_2.jpg", "1mp3d_0044_region31/b95772692ede4e9aa2ce4828447e8419_i2_2.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0044_region31", "origin_question": "", "option": [], "answer": "", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "1mp3d_0044_region31/5bb6a347aea44050b8aa3c0debe93dfc_i1_5.jpg", "1mp3d_0044_region31/5bb6a347aea44050b8aa3c0debe93dfc_i2_5.jpg", "1mp3d_0044_region31/b95772692ede4e9aa2ce4828447e8419_i1_3.jpg", "1mp3d_0044_region31/b95772692ede4e9aa2ce4828447e8419_i2_3.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0044_region31", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the small rectangular picture with a golden frame\" from your position?", "option": [], "answer": "38.1", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the small rectangular picture with a golden frame\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "1mp3d_0044_region31/9177fe5f923242f0b1ba16c7a058e5ac_i1_3.jpg", "1mp3d_0044_region31/9177fe5f923242f0b1ba16c7a058e5ac_i2_3.jpg", "1mp3d_0044_region31/e0047ba33fdb49c499dab9d25da633a8_i1_4.jpg", "1mp3d_0044_region31/e0047ba33fdb49c499dab9d25da633a8_i2_4.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0044_region31", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "1mp3d_0044_region31/e0047ba33fdb49c499dab9d25da633a8_i1_3.jpg", "1mp3d_0044_region31/e0047ba33fdb49c499dab9d25da633a8_i2_3.jpg", "1mp3d_0044_region31/9127aae77ad348eb8fd02f7cd52a3ce5_i1_3.jpg", "1mp3d_0044_region31/9127aae77ad348eb8fd02f7cd52a3ce5_i2_3.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0044_region31", "origin_question": "Which direction is the medium-sized rectangular light grey wooden shelf in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the medium-sized rectangular light grey wooden shelf in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0701_01/00000.jpg", "scene0701_01/00040.jpg", "scene0701_01/00080.jpg", "scene0701_01/00120.jpg" ], "turn_id": 1, "scan_id": "scene0701_01", "origin_question": "Remember, have you seen any table(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any table(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0701_01/00160.jpg", "scene0701_01/00200.jpg", "scene0701_01/00240.jpg", "scene0701_01/00280.jpg" ], "turn_id": 2, "scan_id": "scene0701_01", "origin_question": "Remember, have you seen any backpack(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any backpack(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0701_01/00320.jpg", "scene0701_01/00360.jpg", "scene0701_01/00400.jpg", "scene0701_01/00440.jpg" ], "turn_id": 3, "scan_id": "scene0701_01", "origin_question": "Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "131.95", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0701_01/00480.jpg", "scene0701_01/00520.jpg", "scene0701_01/00560.jpg", "scene0701_01/00600.jpg" ], "turn_id": 4, "scan_id": "scene0701_01", "origin_question": "When did you first discover \"the medium-sized red backpack with black straps\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the medium-sized red backpack with black straps\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0701_01/00640.jpg", "scene0701_01/00680.jpg", "scene0701_01/00720.jpg", "scene0701_01/00760.jpg" ], "turn_id": 5, "scan_id": "scene0701_01", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "1.17", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0701_01/00800.jpg", "scene0701_01/00840.jpg", "scene0701_01/00880.jpg", "scene0701_01/00920.jpg" ], "turn_id": 6, "scan_id": "scene0701_01", "origin_question": "", "option": [], "answer": "", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0040_region5/6459d2b4acd84d8590eeda3114a24885_i1_1.jpg", "1mp3d_0040_region5/6459d2b4acd84d8590eeda3114a24885_i2_1.jpg", "1mp3d_0040_region5/6459d2b4acd84d8590eeda3114a24885_i1_0.jpg", "1mp3d_0040_region5/6459d2b4acd84d8590eeda3114a24885_i2_0.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0040_region5", "origin_question": "Remember, have you seen any person(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any person(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "1mp3d_0040_region5/e8d6fcdc00af4adc93aa78b43804170e_i1_1.jpg", "1mp3d_0040_region5/e8d6fcdc00af4adc93aa78b43804170e_i2_1.jpg", "1mp3d_0040_region5/e8d6fcdc00af4adc93aa78b43804170e_i1_2.jpg", "1mp3d_0040_region5/e8d6fcdc00af4adc93aa78b43804170e_i2_2.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0040_region5", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular light beige window frame\", Y: \"the light\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular light beige window frame\", Y: \"the light\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "1mp3d_0040_region5/b364eb0104d34725810601cc14ebc900_i1_1.jpg", "1mp3d_0040_region5/b364eb0104d34725810601cc14ebc900_i2_1.jpg", "1mp3d_0040_region5/f6458019ed794e9e9b1ab46c800284c2_i1_3.jpg", "1mp3d_0040_region5/f6458019ed794e9e9b1ab46c800284c2_i2_3.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0040_region5", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the tall, ceramic decorative bottle with gold and brown patterns\"; \"the light-colored rectangular step made of wood\"; \"the large, brown rosewood cabinet with curved design\" ", "option": [ "the tall, ceramic decorative bottle with gold and brown patterns", "the light-colored rectangular step made of wood", "the large, brown rosewood cabinet with curved design" ], "answer": "the light-colored rectangular step made of wood", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the tall, ceramic decorative bottle with gold and brown patterns\"; \"the light-colored rectangular step made of wood\"; \"the large, brown rosewood cabinet with curved design\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the tall, ceramic decorative bottle with gold and brown patterns', 'the light-colored rectangular step made of wood', 'the large, brown rosewood cabinet with curved design'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "1mp3d_0040_region5/b364eb0104d34725810601cc14ebc900_i1_3.jpg", "1mp3d_0040_region5/b364eb0104d34725810601cc14ebc900_i2_3.jpg", "1mp3d_0040_region5/f6458019ed794e9e9b1ab46c800284c2_i1_4.jpg", "1mp3d_0040_region5/f6458019ed794e9e9b1ab46c800284c2_i2_4.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0040_region5", "origin_question": "Is the decoration to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the decoration to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "1mp3d_0040_region5/678feb6bbe4745acb80e2a50b6dd9f5a_i1_4.jpg", "1mp3d_0040_region5/678feb6bbe4745acb80e2a50b6dd9f5a_i2_4.jpg", "1mp3d_0040_region5/d2b20ac1680c40799005e91dd3a37de6_i1_3.jpg", "1mp3d_0040_region5/d2b20ac1680c40799005e91dd3a37de6_i2_3.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0040_region5", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "1mp3d_0040_region5/d2b20ac1680c40799005e91dd3a37de6_i1_4.jpg", "1mp3d_0040_region5/d2b20ac1680c40799005e91dd3a37de6_i2_4.jpg", "1mp3d_0040_region5/d2b20ac1680c40799005e91dd3a37de6_i1_5.jpg", "1mp3d_0040_region5/d2b20ac1680c40799005e91dd3a37de6_i2_5.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0040_region5", "origin_question": "Please recall \"the tall, ceramic decorative bottle with gold and brown patterns\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.77", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the tall, ceramic decorative bottle with gold and brown patterns\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "1mp3d_0040_region5/9fd63460efcf4292a96791b9531b6f94_i1_2.jpg", "1mp3d_0040_region5/9fd63460efcf4292a96791b9531b6f94_i2_2.jpg", "1mp3d_0040_region5/fb5d8954758f4bb990c8d78d2a5d63f2_i1_2.jpg", "1mp3d_0040_region5/fb5d8954758f4bb990c8d78d2a5d63f2_i2_2.jpg" ], "turn_id": 7, "scan_id": "1mp3d_0040_region5", "origin_question": "Assuming the direction you are facing at the end of the turn 4 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Right", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 4 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "1mp3d_0040_region5/9fd63460efcf4292a96791b9531b6f94_i1_1.jpg", "1mp3d_0040_region5/9fd63460efcf4292a96791b9531b6f94_i2_1.jpg", "1mp3d_0040_region5/9fd63460efcf4292a96791b9531b6f94_i1_3.jpg", "1mp3d_0040_region5/9fd63460efcf4292a96791b9531b6f94_i2_3.jpg" ], "turn_id": 8, "scan_id": "1mp3d_0040_region5", "origin_question": "Using your orientation at the end of turn 5 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "92.11", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 5 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0353_01/00000.jpg", "scene0353_01/00040.jpg", "scene0353_01/00080.jpg", "scene0353_01/00120.jpg", "scene0353_01/00160.jpg" ], "turn_id": 1, "scan_id": "scene0353_01", "origin_question": "Remember, how many counter(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many counter(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0353_01/00200.jpg", "scene0353_01/00240.jpg", "scene0353_01/00280.jpg", "scene0353_01/00320.jpg", "scene0353_01/00360.jpg" ], "turn_id": 2, "scan_id": "scene0353_01", "origin_question": "Remember, how many bathtub(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many bathtub(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0353_01/00400.jpg", "scene0353_01/00440.jpg", "scene0353_01/00480.jpg", "scene0353_01/00520.jpg", "scene0353_01/00560.jpg" ], "turn_id": 3, "scan_id": "scene0353_01", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the curtain\"; \"the bed\"; \"Deep blue plaid clothes made of cotton and polyester\" ", "option": [ "the curtain", "the bed", "Deep blue plaid clothes made of cotton and polyester" ], "answer": "the curtain", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the curtain\"; \"the bed\"; \"Deep blue plaid clothes made of cotton and polyester\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the curtain', 'the bed', 'Deep blue plaid clothes made of cotton and polyester'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0353_01/00600.jpg", "scene0353_01/00640.jpg", "scene0353_01/00680.jpg", "scene0353_01/00720.jpg", "scene0353_01/00760.jpg" ], "turn_id": 4, "scan_id": "scene0353_01", "origin_question": "Which two objects are on the same side of you now? \"the rectangular white calendar with colorful illustrations\", \"the large blue and white patterned cotton blanket\", \"the small rectangular black plastic bin\". ", "option": [ "the rectangular white calendar with colorful illustrations, the large blue and white patterned cotton blanket", "the rectangular white calendar with colorful illustrations, the small rectangular black plastic bin", "the large blue and white patterned cotton blanket, the small rectangular black plastic bin" ], "answer": "the rectangular white calendar with colorful illustrations, the large blue and white patterned cotton blanket", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the rectangular white calendar with colorful illustrations\", \"the large blue and white patterned cotton blanket\", \"the small rectangular black plastic bin\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular white calendar with colorful illustrations, the large blue and white patterned cotton blanket', 'the rectangular white calendar with colorful illustrations, the small rectangular black plastic bin', 'the large blue and white patterned cotton blanket, the small rectangular black plastic bin'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0353_01/00800.jpg", "scene0353_01/00840.jpg", "scene0353_01/00880.jpg", "scene0353_01/00920.jpg", "scene0353_01/00960.jpg" ], "turn_id": 5, "scan_id": "scene0353_01", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the silver metal laptop with a black keyboard\"; \"the modern flat-panel monitor with a black frame\"; \"the window\" ", "option": [ "the silver metal laptop with a black keyboard", "the modern flat-panel monitor with a black frame", "the window" ], "answer": "the modern flat-panel monitor with a black frame", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the silver metal laptop with a black keyboard\"; \"the modern flat-panel monitor with a black frame\"; \"the window\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the silver metal laptop with a black keyboard', 'the modern flat-panel monitor with a black frame', 'the window'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "scene0353_01/01000.jpg", "scene0353_01/01040.jpg", "scene0353_01/01080.jpg", "scene0353_01/01120.jpg", "scene0353_01/01160.jpg" ], "turn_id": 6, "scan_id": "scene0353_01", "origin_question": "Among these three objects(\"the board\",\"the small rectangular black plastic bin\",\"Deep blue plaid clothes made of cotton and polyester\" ), which one is the closest to you?", "option": [ "the board", "the small rectangular black plastic bin", "Deep blue plaid clothes made of cotton and polyester" ], "answer": "Deep blue plaid clothes made of cotton and polyester", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the board\",\"the small rectangular black plastic bin\",\"Deep blue plaid clothes made of cotton and polyester\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the board', 'the small rectangular black plastic bin', 'Deep blue plaid clothes made of cotton and polyester'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0353_01/01200.jpg", "scene0353_01/01240.jpg", "scene0353_01/01280.jpg", "scene0353_01/01320.jpg", "scene0353_01/01360.jpg" ], "turn_id": 7, "scan_id": "scene0353_01", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the book\" from your position?", "option": [], "answer": "95.4", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the book\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0353_01/01400.jpg", "scene0353_01/01440.jpg", "scene0353_01/01480.jpg", "scene0353_01/01520.jpg", "scene0353_01/01560.jpg" ], "turn_id": 8, "scan_id": "scene0353_01", "origin_question": "Please recall \"the modern flat-panel monitor with a black frame\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.26", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the modern flat-panel monitor with a black frame\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "scene0353_01/01600.jpg", "scene0353_01/01640.jpg", "scene0353_01/01680.jpg", "scene0353_01/01720.jpg", "scene0353_01/01760.jpg" ], "turn_id": 9, "scan_id": "scene0353_01", "origin_question": "Among these three objects(\"the curtain\",\"the refrigerator\",\"the rectangular wooden door with a white surface\" ), which one is the closest to you?", "option": [ "the curtain", "the refrigerator", "the rectangular wooden door with a white surface" ], "answer": "the refrigerator", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the curtain\",\"the refrigerator\",\"the rectangular wooden door with a white surface\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the curtain', 'the refrigerator', 'the rectangular wooden door with a white surface'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0353_01/01800.jpg", "scene0353_01/01840.jpg", "scene0353_01/01880.jpg", "scene0353_01/01920.jpg", "scene0353_01/01960.jpg" ], "turn_id": 10, "scan_id": "scene0353_01", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the small rectangular black plastic bin\" from your position?", "option": [], "answer": "138.6", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the small rectangular black plastic bin\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0695_00/00000.jpg", "scene0695_00/00040.jpg", "scene0695_00/00080.jpg", "scene0695_00/00120.jpg", "scene0695_00/00160.jpg" ], "turn_id": 1, "scan_id": "scene0695_00", "origin_question": "Remember, have you seen any adhesive tape(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any adhesive tape(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0695_00/00200.jpg", "scene0695_00/00240.jpg", "scene0695_00/00380.jpg", "scene0695_00/00420.jpg", "scene0695_00/00460.jpg" ], "turn_id": 2, "scan_id": "scene0695_00", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the rectangular white cotton pillow with gray borders\"; \"the dark flat panel monitor with a matte surface\"; \"the bed\" ", "option": [ "the rectangular white cotton pillow with gray borders", "the dark flat panel monitor with a matte surface", "the bed" ], "answer": "the dark flat panel monitor with a matte surface", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the rectangular white cotton pillow with gray borders\"; \"the dark flat panel monitor with a matte surface\"; \"the bed\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular white cotton pillow with gray borders', 'the dark flat panel monitor with a matte surface', 'the bed'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0695_00/00500.jpg", "scene0695_00/00540.jpg", "scene0695_00/00580.jpg", "scene0695_00/00620.jpg", "scene0695_00/00660.jpg" ], "turn_id": 3, "scan_id": "scene0695_00", "origin_question": "Please recall \"the small rectangular white plastic bin\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.05", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the small rectangular white plastic bin\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "scene0695_00/00700.jpg", "scene0695_00/00740.jpg", "scene0695_00/00780.jpg", "scene0695_00/00820.jpg", "scene0695_00/00860.jpg" ], "turn_id": 4, "scan_id": "scene0695_00", "origin_question": "Among these three objects(\"the cylindrical, white fabric hamper of medium size\",\"the tissue box\",\"the cabinet\" ), which one is the closest to you?", "option": [ "the cylindrical, white fabric hamper of medium size", "the tissue box", "the cabinet" ], "answer": "the cabinet", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the cylindrical, white fabric hamper of medium size\",\"the tissue box\",\"the cabinet\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the cylindrical, white fabric hamper of medium size', 'the tissue box', 'the cabinet'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "scene0695_00/00900.jpg", "scene0695_00/00940.jpg", "scene0695_00/00980.jpg", "scene0695_00/01020.jpg", "scene0695_00/01060.jpg" ], "turn_id": 5, "scan_id": "scene0695_00", "origin_question": "Among these three objects(\"the small rectangular white plastic bin\",\"the cabinet\",\"the sleek white rectangular plastic remote control\" ), which one is the closest to you?", "option": [ "the small rectangular white plastic bin", "the cabinet", "the sleek white rectangular plastic remote control" ], "answer": "the cabinet", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the small rectangular white plastic bin\",\"the cabinet\",\"the sleek white rectangular plastic remote control\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the small rectangular white plastic bin', 'the cabinet', 'the sleek white rectangular plastic remote control'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0695_00/01100.jpg", "scene0695_00/01140.jpg", "scene0695_00/01180.jpg", "scene0695_00/01230.jpg", "scene0695_00/01270.jpg" ], "turn_id": 6, "scan_id": "scene0695_00", "origin_question": "Which direction is the dark flat panel monitor with a matte surface in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the dark flat panel monitor with a matte surface in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0695_00/01310.jpg", "scene0695_00/01350.jpg", "scene0695_00/01390.jpg", "scene0695_00/01430.jpg", "scene0695_00/01470.jpg" ], "turn_id": 7, "scan_id": "scene0695_00", "origin_question": "At the end of which turn were both of \"the dark flat panel monitor with a matte surface\" and \"the cup\" on your left side?", "option": [], "answer": "6", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the dark flat panel monitor with a matte surface\" and \"the cup\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0695_00/01510.jpg", "scene0695_00/01550.jpg", "scene0695_00/01590.jpg", "scene0695_00/01630.jpg", "scene0695_00/01670.jpg" ], "turn_id": 8, "scan_id": "scene0695_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0575_00/00000.jpg", "scene0575_00/00040.jpg", "scene0575_00/00080.jpg", "scene0575_00/00120.jpg", "scene0575_00/00160.jpg" ], "turn_id": 1, "scan_id": "scene0575_00", "origin_question": "Remember, have you seen any chair(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any chair(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0575_00/00200.jpg", "scene0575_00/00240.jpg", "scene0575_00/00280.jpg", "scene0575_00/00320.jpg", "scene0575_00/00360.jpg" ], "turn_id": 2, "scan_id": "scene0575_00", "origin_question": "In which turn were you farthest from \"the light-colored rectangular wooden table\"? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the light-colored rectangular wooden table\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0575_00/00400.jpg", "scene0575_00/00440.jpg", "scene0575_00/00480.jpg", "scene0575_00/00520.jpg", "scene0575_00/00560.jpg" ], "turn_id": 3, "scan_id": "scene0575_00", "origin_question": "When did you first discover \"the light-colored rectangular wooden table\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the light-colored rectangular wooden table\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0575_00/00600.jpg", "scene0575_00/00640.jpg", "scene0575_00/00680.jpg", "scene0575_00/00720.jpg", "scene0575_00/00760.jpg" ], "turn_id": 4, "scan_id": "scene0575_00", "origin_question": "Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "72.11", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0575_00/00800.jpg", "scene0575_00/00840.jpg", "scene0575_00/00880.jpg", "scene0575_00/00920.jpg", "scene0575_00/00960.jpg" ], "turn_id": 5, "scan_id": "scene0575_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0575_00/01000.jpg", "scene0575_00/01040.jpg", "scene0575_00/01080.jpg", "scene0575_00/01120.jpg", "scene0575_00/01160.jpg" ], "turn_id": 6, "scan_id": "scene0575_00", "origin_question": "Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the light-colored rectangular wooden table\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the light-colored rectangular wooden table\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0575_00/01200.jpg", "scene0575_00/01240.jpg", "scene0575_00/01280.jpg", "scene0575_00/01320.jpg", "scene0575_00/01360.jpg" ], "turn_id": 7, "scan_id": "scene0575_00", "origin_question": "In which turn did you see both \"the rectangular door with a frosted glass panel\" and \"the large rectangular glass window with a white frame\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the rectangular door with a frosted glass panel\" and \"the large rectangular glass window with a white frame\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0575_00/01400.jpg", "scene0575_00/01440.jpg", "scene0575_00/01480.jpg", "scene0575_00/01520.jpg", "scene0575_00/01560.jpg" ], "turn_id": 8, "scan_id": "scene0575_00", "origin_question": "In which turn did you see both \"the rectangular door with a frosted glass panel\" and \"the small cylindrical black plastic bin\" simultaneously? ", "option": [], "answer": "8", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the rectangular door with a frosted glass panel\" and \"the small cylindrical black plastic bin\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0575_00/01600.jpg", "scene0575_00/01640.jpg", "scene0575_00/01680.jpg", "scene0575_00/01720.jpg", "scene0575_00/01760.jpg" ], "turn_id": 9, "scan_id": "scene0575_00", "origin_question": "Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the small cylindrical black plastic bin\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the small cylindrical black plastic bin\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0575_00/01800.jpg", "scene0575_00/01840.jpg", "scene0575_00/01880.jpg", "scene0575_00/01920.jpg", "scene0575_00/01960.jpg" ], "turn_id": 10, "scan_id": "scene0575_00", "origin_question": "At the end of which turn were both of \"the light-colored rectangular wooden table\" and \"the small cylindrical black plastic bin\" on your right side?", "option": [], "answer": "2", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the light-colored rectangular wooden table\" and \"the small cylindrical black plastic bin\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0549_01/00000.jpg", "scene0549_01/00040.jpg", "scene0549_01/00080.jpg", "scene0549_01/00120.jpg" ], "turn_id": 1, "scan_id": "scene0549_01", "origin_question": "Remember, have you seen any couch(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any couch(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0549_01/00160.jpg", "scene0549_01/00200.jpg", "scene0549_01/00240.jpg", "scene0549_01/00280.jpg" ], "turn_id": 2, "scan_id": "scene0549_01", "origin_question": "When did you first discover \"the medium-sized, deep blue upholstered chair\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the medium-sized, deep blue upholstered chair\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0549_01/00320.jpg", "scene0549_01/00360.jpg", "scene0549_01/00400.jpg", "scene0549_01/00440.jpg" ], "turn_id": 3, "scan_id": "scene0549_01", "origin_question": "In which turn were you closest to \"the rectangular transparent glass display shelf\"? ", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the rectangular transparent glass display shelf\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0549_01/00480.jpg", "scene0549_01/00520.jpg", "scene0549_01/00560.jpg", "scene0549_01/00600.jpg" ], "turn_id": 4, "scan_id": "scene0549_01", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "2.53", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0549_01/00640.jpg", "scene0549_01/00680.jpg", "scene0549_01/00720.jpg", "scene0549_01/00760.jpg" ], "turn_id": 5, "scan_id": "scene0549_01", "origin_question": "Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "147.53", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0549_01/00800.jpg", "scene0549_01/00840.jpg", "scene0549_01/00880.jpg", "scene0549_01/00920.jpg" ], "turn_id": 6, "scan_id": "scene0549_01", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "135.86", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0549_01/00960.jpg", "scene0549_01/01000.jpg", "scene0549_01/01040.jpg", "scene0549_01/01080.jpg" ], "turn_id": 7, "scan_id": "scene0549_01", "origin_question": "Compared to your position at the end of the 5 turn, are you now closer or farther away from \"the rectangular transparent glass display shelf\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 5 turn, are you now closer or farther away from \"the rectangular transparent glass display shelf\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0338_01/00000.jpg", "scene0338_01/00030.jpg", "scene0338_01/00060.jpg", "scene0338_01/00090.jpg" ], "turn_id": 1, "scan_id": "scene0338_01", "origin_question": "Remember, have you seen any mailbox(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any mailbox(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0338_01/00120.jpg", "scene0338_01/00150.jpg", "scene0338_01/00180.jpg", "scene0338_01/00210.jpg" ], "turn_id": 2, "scan_id": "scene0338_01", "origin_question": "Remember, how many mirror(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many mirror(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0338_01/00240.jpg", "scene0338_01/00270.jpg", "scene0338_01/00300.jpg", "scene0338_01/00330.jpg" ], "turn_id": 3, "scan_id": "scene0338_01", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0338_01/00360.jpg", "scene0338_01/00390.jpg", "scene0338_01/00420.jpg", "scene0338_01/00450.jpg" ], "turn_id": 4, "scan_id": "scene0338_01", "origin_question": "Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "94.87", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0338_01/00480.jpg", "scene0338_01/00510.jpg", "scene0338_01/00540.jpg", "scene0338_01/00570.jpg" ], "turn_id": 5, "scan_id": "scene0338_01", "origin_question": "In which turn did you see both \"the large, modern white copier made of sturdy materials\" and \"the door\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the large, modern white copier made of sturdy materials\" and \"the door\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0338_01/00600.jpg", "scene0338_01/00630.jpg", "scene0338_01/00660.jpg", "scene0338_01/00690.jpg" ], "turn_id": 6, "scan_id": "scene0338_01", "origin_question": "Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0262_00/00000.jpg", "scene0262_00/00040.jpg", "scene0262_00/00080.jpg", "scene0262_00/00120.jpg", "scene0262_00/00190.jpg" ], "turn_id": 1, "scan_id": "scene0262_00", "origin_question": "Remember, have you seen any dress(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any dress(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0262_00/00230.jpg", "scene0262_00/00270.jpg", "scene0262_00/00310.jpg", "scene0262_00/00350.jpg", "scene0262_00/00390.jpg" ], "turn_id": 2, "scan_id": "scene0262_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0262_00/00430.jpg", "scene0262_00/00470.jpg", "scene0262_00/00510.jpg", "scene0262_00/00550.jpg", "scene0262_00/00590.jpg" ], "turn_id": 3, "scan_id": "scene0262_00", "origin_question": "Remember, have you seen any chair(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any chair(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0262_00/00630.jpg", "scene0262_00/00670.jpg", "scene0262_00/00710.jpg", "scene0262_00/00750.jpg", "scene0262_00/00790.jpg" ], "turn_id": 4, "scan_id": "scene0262_00", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the rectangular light grey desk made of sturdy materials\"; \"the red plastic chair with black armrests\"; \"the sleek gray rectangular telephone with a display\" ", "option": [ "the rectangular light grey desk made of sturdy materials", "the red plastic chair with black armrests", "the sleek gray rectangular telephone with a display" ], "answer": "the sleek gray rectangular telephone with a display", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the rectangular light grey desk made of sturdy materials\"; \"the red plastic chair with black armrests\"; \"the sleek gray rectangular telephone with a display\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular light grey desk made of sturdy materials', 'the red plastic chair with black armrests', 'the sleek gray rectangular telephone with a display'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0262_00/00830.jpg", "scene0262_00/00870.jpg", "scene0262_00/00910.jpg", "scene0262_00/00950.jpg", "scene0262_00/00990.jpg" ], "turn_id": 5, "scan_id": "scene0262_00", "origin_question": "At the end of which turn were both of \"the transparent cylindrical plastic bottle with a green cap\" and \"the sleek gray rectangular telephone with a display\" on your left side?", "option": [], "answer": "5", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the transparent cylindrical plastic bottle with a green cap\" and \"the sleek gray rectangular telephone with a display\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0262_00/01030.jpg", "scene0262_00/01070.jpg", "scene0262_00/01110.jpg", "scene0262_00/01150.jpg", "scene0262_00/01190.jpg" ], "turn_id": 6, "scan_id": "scene0262_00", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the sleek gray rectangular telephone with a display\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the sleek gray rectangular telephone with a display\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0262_00/01230.jpg", "scene0262_00/01270.jpg", "scene0262_00/01310.jpg", "scene0262_00/01350.jpg", "scene0262_00/01390.jpg" ], "turn_id": 7, "scan_id": "scene0262_00", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the transparent cylindrical plastic bottle with a green cap\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the transparent cylindrical plastic bottle with a green cap\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "scene0262_00/01430.jpg", "scene0262_00/01470.jpg", "scene0262_00/01510.jpg", "scene0262_00/01550.jpg", "scene0262_00/01590.jpg" ], "turn_id": 8, "scan_id": "scene0262_00", "origin_question": "Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance forward or backward from that position?", "option": [ "Forward", "Backward" ], "answer": "Backward", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance forward or backward from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Forward', 'Backward'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "scene0262_00/01630.jpg", "scene0262_00/01670.jpg", "scene0262_00/01710.jpg", "scene0262_00/01750.jpg", "scene0262_00/01790.jpg" ], "turn_id": 9, "scan_id": "scene0262_00", "origin_question": "Among these three objects(\"the rectangular light grey desk made of sturdy materials\",\"the small round metal clock with a white dial\",\"the square decorative pillow with geometric black and white patterns\" ), which one is the farthest from you?", "option": [ "the rectangular light grey desk made of sturdy materials", "the small round metal clock with a white dial", "the square decorative pillow with geometric black and white patterns" ], "answer": "the small round metal clock with a white dial", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the rectangular light grey desk made of sturdy materials\",\"the small round metal clock with a white dial\",\"the square decorative pillow with geometric black and white patterns\" ), which one is the farthest from you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular light grey desk made of sturdy materials', 'the small round metal clock with a white dial', 'the square decorative pillow with geometric black and white patterns'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0262_00/01830.jpg", "scene0262_00/01870.jpg", "scene0262_00/01910.jpg", "scene0262_00/01950.jpg", "scene0262_00/01990.jpg" ], "turn_id": 10, "scan_id": "scene0262_00", "origin_question": "Is the black rectangular eraser made of plastic or rubber to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the black rectangular eraser made of plastic or rubber to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0379_00/00000.jpg", "scene0379_00/00040.jpg", "scene0379_00/00080.jpg", "scene0379_00/00120.jpg", "scene0379_00/00160.jpg" ], "turn_id": 1, "scan_id": "scene0379_00", "origin_question": "Remember, have you seen any radiator(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any radiator(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0379_00/00200.jpg", "scene0379_00/00240.jpg", "scene0379_00/00280.jpg", "scene0379_00/00320.jpg", "scene0379_00/00360.jpg" ], "turn_id": 2, "scan_id": "scene0379_00", "origin_question": "In which turn were you farthest from \"the large rectangular light-colored table\"? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the large rectangular light-colored table\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0379_00/00400.jpg", "scene0379_00/00440.jpg", "scene0379_00/00480.jpg", "scene0379_00/00520.jpg", "scene0379_00/00560.jpg" ], "turn_id": 3, "scan_id": "scene0379_00", "origin_question": "Which direction is the cylindrical orange-yellow plastic bucket in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the cylindrical orange-yellow plastic bucket in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0379_00/00600.jpg", "scene0379_00/00640.jpg", "scene0379_00/00680.jpg", "scene0379_00/00720.jpg", "scene0379_00/00760.jpg" ], "turn_id": 4, "scan_id": "scene0379_00", "origin_question": "In which turn did you see both \"the large rectangular light-colored table\" and \"the rectangular light-colored wooden or plastic board\" simultaneously? ", "option": [], "answer": "4", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the large rectangular light-colored table\" and \"the rectangular light-colored wooden or plastic board\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0379_00/00800.jpg", "scene0379_00/00840.jpg", "scene0379_00/00880.jpg", "scene0379_00/00920.jpg", "scene0379_00/00960.jpg" ], "turn_id": 5, "scan_id": "scene0379_00", "origin_question": "Which two objects are on the same side of you now? \"the standard-sized beige hardcover book\", \"the cylindrical plastic bin with a smooth surface\", \"the shelf\". ", "option": [ "the standard-sized beige hardcover book, the cylindrical plastic bin with a smooth surface", "the standard-sized beige hardcover book, the shelf", "the cylindrical plastic bin with a smooth surface, the shelf" ], "answer": "the standard-sized beige hardcover book, the cylindrical plastic bin with a smooth surface", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the standard-sized beige hardcover book\", \"the cylindrical plastic bin with a smooth surface\", \"the shelf\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the standard-sized beige hardcover book, the cylindrical plastic bin with a smooth surface', 'the standard-sized beige hardcover book, the shelf', 'the cylindrical plastic bin with a smooth surface, the shelf'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0379_00/01000.jpg", "scene0379_00/01040.jpg", "scene0379_00/01080.jpg", "scene0379_00/01120.jpg", "scene0379_00/01160.jpg" ], "turn_id": 6, "scan_id": "scene0379_00", "origin_question": "In which turn were you closest to \"the shelf\"? ", "option": [], "answer": "6", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the shelf\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0379_00/01200.jpg", "scene0379_00/01240.jpg", "scene0379_00/01280.jpg", "scene0379_00/01320.jpg", "scene0379_00/01360.jpg" ], "turn_id": 7, "scan_id": "scene0379_00", "origin_question": "Which direction is the standard-sized beige hardcover book in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the standard-sized beige hardcover book in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "scene0379_00/01400.jpg", "scene0379_00/01440.jpg", "scene0379_00/01480.jpg", "scene0379_00/01520.jpg", "scene0379_00/01560.jpg" ], "turn_id": 8, "scan_id": "scene0379_00", "origin_question": "Among these three objects(\"the lush green turtle back bamboo plant with wide leaves\",\"the standard-sized beige hardcover book\",\"the cylindrical orange-yellow plastic bucket\" ), which one is the closest to you?", "option": [ "the lush green turtle back bamboo plant with wide leaves", "the standard-sized beige hardcover book", "the cylindrical orange-yellow plastic bucket" ], "answer": "the lush green turtle back bamboo plant with wide leaves", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the lush green turtle back bamboo plant with wide leaves\",\"the standard-sized beige hardcover book\",\"the cylindrical orange-yellow plastic bucket\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the lush green turtle back bamboo plant with wide leaves', 'the standard-sized beige hardcover book', 'the cylindrical orange-yellow plastic bucket'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0379_00/01600.jpg", "scene0379_00/01640.jpg", "scene0379_00/01680.jpg", "scene0379_00/01720.jpg", "scene0379_00/01760.jpg" ], "turn_id": 9, "scan_id": "scene0379_00", "origin_question": "Which direction is the cylindrical orange-yellow plastic bucket in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the cylindrical orange-yellow plastic bucket in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0379_00/01800.jpg", "scene0379_00/01840.jpg", "scene0379_00/01880.jpg", "scene0379_00/01920.jpg", "scene0379_00/01960.jpg" ], "turn_id": 10, "scan_id": "scene0379_00", "origin_question": "Which two objects are on the same side of you now? \"the standard-sized beige hardcover book\", \"the cylindrical orange-yellow plastic bucket\", \"the cylindrical plastic bin with a smooth surface\". ", "option": [ "the standard-sized beige hardcover book, the cylindrical orange-yellow plastic bucket", "the standard-sized beige hardcover book, the cylindrical plastic bin with a smooth surface", "the cylindrical orange-yellow plastic bucket, the cylindrical plastic bin with a smooth surface" ], "answer": "the standard-sized beige hardcover book, the cylindrical orange-yellow plastic bucket", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the standard-sized beige hardcover book\", \"the cylindrical orange-yellow plastic bucket\", \"the cylindrical plastic bin with a smooth surface\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the standard-sized beige hardcover book, the cylindrical orange-yellow plastic bucket', 'the standard-sized beige hardcover book, the cylindrical plastic bin with a smooth surface', 'the cylindrical orange-yellow plastic bucket, the cylindrical plastic bin with a smooth surface'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0251_00/00000.jpg", "scene0251_00/00040.jpg", "scene0251_00/00080.jpg", "scene0251_00/00120.jpg" ], "turn_id": 1, "scan_id": "scene0251_00", "origin_question": "Remember, how many coil(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many coil(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0251_00/00160.jpg", "scene0251_00/00200.jpg", "scene0251_00/00240.jpg", "scene0251_00/00280.jpg" ], "turn_id": 2, "scan_id": "scene0251_00", "origin_question": "Remember, have you seen any cap(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any cap(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0251_00/00320.jpg", "scene0251_00/00360.jpg", "scene0251_00/00400.jpg", "scene0251_00/00440.jpg" ], "turn_id": 3, "scan_id": "scene0251_00", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the rectangular foldable table with a wood grain finish\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the rectangular foldable table with a wood grain finish\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0251_00/00480.jpg", "scene0251_00/00520.jpg", "scene0251_00/00560.jpg", "scene0251_00/00600.jpg" ], "turn_id": 4, "scan_id": "scene0251_00", "origin_question": "When did you first discover \"the rectangular foldable table with a wood grain finish\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the rectangular foldable table with a wood grain finish\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0251_00/00640.jpg", "scene0251_00/00680.jpg", "scene0251_00/00720.jpg", "scene0251_00/00760.jpg" ], "turn_id": 5, "scan_id": "scene0251_00", "origin_question": "Remember, how many laptop(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many laptop(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0251_00/00800.jpg", "scene0251_00/00840.jpg", "scene0251_00/00880.jpg", "scene0251_00/00920.jpg" ], "turn_id": 6, "scan_id": "scene0251_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0251_00/00960.jpg", "scene0251_00/01000.jpg", "scene0251_00/01040.jpg", "scene0251_00/01080.jpg" ], "turn_id": 7, "scan_id": "scene0251_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0251_00/01120.jpg", "scene0251_00/01160.jpg", "scene0251_00/01200.jpg", "scene0251_00/01240.jpg" ], "turn_id": 8, "scan_id": "scene0251_00", "origin_question": "How far is your current position from where you were at the end of the turn 5? (in meters)", "option": [], "answer": "6.04", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 5? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0251_00/01280.jpg", "scene0251_00/01320.jpg", "scene0251_00/01360.jpg", "scene0251_00/01400.jpg" ], "turn_id": 9, "scan_id": "scene0251_00", "origin_question": "How far is your current position from where you were at the end of the turn 6? (in meters)", "option": [], "answer": "1.84", "user_message": "For the 9 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 6? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0697_00/00000.jpg", "scene0697_00/00040.jpg", "scene0697_00/00080.jpg", "scene0697_00/00120.jpg", "scene0697_00/00160.jpg" ], "turn_id": 1, "scan_id": "scene0697_00", "origin_question": "Remember, have you seen any tissue box(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any tissue box(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0697_00/00200.jpg", "scene0697_00/00240.jpg", "scene0697_00/00280.jpg", "scene0697_00/00320.jpg", "scene0697_00/00360.jpg" ], "turn_id": 2, "scan_id": "scene0697_00", "origin_question": "In which turn did you see both \"the small round silver clock with a white face\" and \"the cylindrical beige lamp with a wooden base\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the small round silver clock with a white face\" and \"the cylindrical beige lamp with a wooden base\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0697_00/00400.jpg", "scene0697_00/00440.jpg", "scene0697_00/00490.jpg", "scene0697_00/00530.jpg", "scene0697_00/00570.jpg" ], "turn_id": 3, "scan_id": "scene0697_00", "origin_question": "In which turn did you see both \"the rectangular paper box with a purple zigzag pattern\" and \"the cylindrical beige lamp with a wooden base\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the rectangular paper box with a purple zigzag pattern\" and \"the cylindrical beige lamp with a wooden base\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0697_00/00610.jpg", "scene0697_00/00650.jpg", "scene0697_00/00690.jpg", "scene0697_00/00730.jpg", "scene0697_00/00770.jpg" ], "turn_id": 4, "scan_id": "scene0697_00", "origin_question": "Is the round pink jar to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the round pink jar to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0697_00/00810.jpg", "scene0697_00/00850.jpg", "scene0697_00/00890.jpg", "scene0697_00/00930.jpg", "scene0697_00/00970.jpg" ], "turn_id": 5, "scan_id": "scene0697_00", "origin_question": "At the end of which turn were both of \"the small white bottle with a blue cap\" and \"the round pink jar\" on your left side?", "option": [], "answer": "3", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the small white bottle with a blue cap\" and \"the round pink jar\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0697_00/01010.jpg", "scene0697_00/01050.jpg", "scene0697_00/01090.jpg", "scene0697_00/01130.jpg", "scene0697_00/01170.jpg" ], "turn_id": 6, "scan_id": "scene0697_00", "origin_question": "Please recall \"the medium-sized rectangular picture\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.57", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the medium-sized rectangular picture\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0697_00/01210.jpg", "scene0697_00/01250.jpg", "scene0697_00/01290.jpg", "scene0697_00/01330.jpg", "scene0697_00/01370.jpg" ], "turn_id": 7, "scan_id": "scene0697_00", "origin_question": "Which direction is the small round green plastic bin in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the small round green plastic bin in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0697_00/01410.jpg", "scene0697_00/01450.jpg", "scene0697_00/01490.jpg", "scene0697_00/01530.jpg", "scene0697_00/01570.jpg" ], "turn_id": 8, "scan_id": "scene0697_00", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the white single mattress with a soft sleep surface\" from your position?", "option": [], "answer": "175.45", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the white single mattress with a soft sleep surface\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0697_00/01610.jpg", "scene0697_00/01650.jpg", "scene0697_00/01690.jpg", "scene0697_00/01730.jpg", "scene0697_00/01770.jpg" ], "turn_id": 9, "scan_id": "scene0697_00", "origin_question": "What will be the first-time appearance order of the following categories: clothes, light, tv?", "option": [ "tv, light, clothes", "clothes, tv, light", "light, tv, clothes", "tv, clothes, light" ], "answer": "light, tv, clothes", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: clothes, light, tv? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['tv, light, clothes', 'clothes, tv, light', 'light, tv, clothes', 'tv, clothes, light'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0697_00/01810.jpg", "scene0697_00/01850.jpg", "scene0697_00/01890.jpg", "scene0697_00/01930.jpg", "scene0697_00/01970.jpg" ], "turn_id": 10, "scan_id": "scene0697_00", "origin_question": "At the end of which turn were both of \"the medium-sized rectangular picture\" and \"the mirror\" on your right side?", "option": [], "answer": "2", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the medium-sized rectangular picture\" and \"the mirror\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0422_00/00000.jpg", "scene0422_00/00040.jpg", "scene0422_00/00080.jpg", "scene0422_00/00120.jpg", "scene0422_00/00160.jpg" ], "turn_id": 1, "scan_id": "scene0422_00", "origin_question": "Remember, have you seen any coffee maker(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any coffee maker(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0422_00/00200.jpg", "scene0422_00/00240.jpg", "scene0422_00/00280.jpg", "scene0422_00/00320.jpg", "scene0422_00/00360.jpg" ], "turn_id": 2, "scan_id": "scene0422_00", "origin_question": "Remember, how many ball(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many ball(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0422_00/00400.jpg", "scene0422_00/00440.jpg", "scene0422_00/00480.jpg", "scene0422_00/00520.jpg", "scene0422_00/00560.jpg" ], "turn_id": 3, "scan_id": "scene0422_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "scene0422_00/00600.jpg", "scene0422_00/00640.jpg", "scene0422_00/00680.jpg", "scene0422_00/00720.jpg", "scene0422_00/00760.jpg" ], "turn_id": 4, "scan_id": "scene0422_00", "origin_question": "Among these three objects(\"the medium rectangular wooden table with a smooth surface\",\"the smooth, rectangular white sheet of paper\",\"the light green single leaf door with a window\" ), which one is the closest to you?", "option": [ "the medium rectangular wooden table with a smooth surface", "the smooth, rectangular white sheet of paper", "the light green single leaf door with a window" ], "answer": "the medium rectangular wooden table with a smooth surface", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the medium rectangular wooden table with a smooth surface\",\"the smooth, rectangular white sheet of paper\",\"the light green single leaf door with a window\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the medium rectangular wooden table with a smooth surface', 'the smooth, rectangular white sheet of paper', 'the light green single leaf door with a window'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0422_00/00800.jpg", "scene0422_00/00840.jpg", "scene0422_00/00880.jpg", "scene0422_00/00920.jpg", "scene0422_00/00960.jpg" ], "turn_id": 5, "scan_id": "scene0422_00", "origin_question": "Which direction is the smooth, rectangular white sheet of paper in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the smooth, rectangular white sheet of paper in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0422_00/01000.jpg", "scene0422_00/01040.jpg", "scene0422_00/01080.jpg", "scene0422_00/01120.jpg", "scene0422_00/01160.jpg" ], "turn_id": 6, "scan_id": "scene0422_00", "origin_question": "Please recall \"the smooth, rectangular white sheet of paper\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.67", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the smooth, rectangular white sheet of paper\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0422_00/01200.jpg", "scene0422_00/01240.jpg", "scene0422_00/01280.jpg", "scene0422_00/01320.jpg", "scene0422_00/01360.jpg" ], "turn_id": 7, "scan_id": "scene0422_00", "origin_question": "Which direction is the light green single leaf door with a window in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the light green single leaf door with a window in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0422_00/01400.jpg", "scene0422_00/01440.jpg", "scene0422_00/01480.jpg", "scene0422_00/01520.jpg", "scene0422_00/01560.jpg" ], "turn_id": 8, "scan_id": "scene0422_00", "origin_question": "Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0422_00/01600.jpg", "scene0422_00/01640.jpg", "scene0422_00/01680.jpg", "scene0422_00/01720.jpg", "scene0422_00/01760.jpg" ], "turn_id": 9, "scan_id": "scene0422_00", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0422_00/01800.jpg", "scene0422_00/01840.jpg", "scene0422_00/01880.jpg", "scene0422_00/01920.jpg", "scene0422_00/01960.jpg" ], "turn_id": 10, "scan_id": "scene0422_00", "origin_question": "What will be the first-time appearance order of the following categories: door, paper, table?", "option": [ "door, paper, table", "paper, door, table", "paper, table, door", "table, paper, door" ], "answer": "table, paper, door", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: door, paper, table? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['door, paper, table', 'paper, door, table', 'paper, table, door', 'table, paper, door'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0449_00/00000.jpg", "scene0449_00/00040.jpg", "scene0449_00/00080.jpg", "scene0449_00/00120.jpg" ], "turn_id": 1, "scan_id": "scene0449_00", "origin_question": "Remember, have you seen any chandelier(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any chandelier(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0449_00/00160.jpg", "scene0449_00/00200.jpg", "scene0449_00/00240.jpg", "scene0449_00/00280.jpg" ], "turn_id": 2, "scan_id": "scene0449_00", "origin_question": "When did you first discover \"the curtain\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the curtain\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0449_00/00320.jpg", "scene0449_00/00360.jpg", "scene0449_00/00400.jpg", "scene0449_00/00440.jpg" ], "turn_id": 3, "scan_id": "scene0449_00", "origin_question": "In which turn were you closest to \"the vanity\"? ", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the vanity\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0449_00/00480.jpg", "scene0449_00/00520.jpg", "scene0449_00/00560.jpg", "scene0449_00/00600.jpg" ], "turn_id": 4, "scan_id": "scene0449_00", "origin_question": "What will be the first-time appearance order of the following categories: curtain, door, toilet?", "option": [ "curtain, toilet, door", "toilet, curtain, door", "door, curtain, toilet", "curtain, door, toilet" ], "answer": "toilet, curtain, door", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: curtain, door, toilet? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['curtain, toilet, door', 'toilet, curtain, door', 'door, curtain, toilet', 'curtain, door, toilet'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0449_00/00640.jpg", "scene0449_00/00680.jpg", "scene0449_00/00720.jpg", "scene0449_00/00760.jpg" ], "turn_id": 5, "scan_id": "scene0449_00", "origin_question": "Is the stopcock to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the stopcock to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0449_00/00800.jpg", "scene0449_00/00840.jpg", "scene0449_00/00880.jpg", "scene0449_00/00920.jpg" ], "turn_id": 6, "scan_id": "scene0449_00", "origin_question": "Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the polished stainless steel bar for bathrooms\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the polished stainless steel bar for bathrooms\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0449_00/00960.jpg", "scene0449_00/01000.jpg", "scene0449_00/01040.jpg", "scene0449_00/01080.jpg" ], "turn_id": 7, "scan_id": "scene0449_00", "origin_question": "Is the bathtub to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the bathtub to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0449_00/01120.jpg", "scene0449_00/01160.jpg", "scene0449_00/01200.jpg", "scene0449_00/01240.jpg" ], "turn_id": 8, "scan_id": "scene0449_00", "origin_question": "At the end of which turn were both of \"the vanity\" and \"the curtain\" on your right side?", "option": [], "answer": "8", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the vanity\" and \"the curtain\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0449_00/01280.jpg", "scene0449_00/01320.jpg", "scene0449_00/01360.jpg", "scene0449_00/01400.jpg" ], "turn_id": 9, "scan_id": "scene0449_00", "origin_question": "Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the door\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 9 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the door\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0025_01/00000.jpg", "scene0025_01/00040.jpg", "scene0025_01/00080.jpg", "scene0025_01/00120.jpg", "scene0025_01/00160.jpg" ], "turn_id": 1, "scan_id": "scene0025_01", "origin_question": "Remember, have you seen any table(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any table(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0025_01/00200.jpg", "scene0025_01/00240.jpg", "scene0025_01/00280.jpg", "scene0025_01/00320.jpg", "scene0025_01/00360.jpg" ], "turn_id": 2, "scan_id": "scene0025_01", "origin_question": "In which turn did you see both \"the rectangular, black and white houndstooth pillow\" and \"the cylindrical plastic bottle with a red cap\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the rectangular, black and white houndstooth pillow\" and \"the cylindrical plastic bottle with a red cap\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0025_01/00400.jpg", "scene0025_01/00440.jpg", "scene0025_01/00480.jpg", "scene0025_01/00520.jpg", "scene0025_01/00560.jpg" ], "turn_id": 3, "scan_id": "scene0025_01", "origin_question": "In which turn did you see both \"the standard black rectangular plastic keyboard\" and \"White rectangular blinds\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the standard black rectangular plastic keyboard\" and \"White rectangular blinds\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0025_01/00600.jpg", "scene0025_01/00640.jpg", "scene0025_01/00680.jpg", "scene0025_01/00720.jpg", "scene0025_01/00760.jpg" ], "turn_id": 4, "scan_id": "scene0025_01", "origin_question": "Which direction is the cylindrical plastic bottle with a red cap in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the cylindrical plastic bottle with a red cap in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0025_01/00800.jpg", "scene0025_01/00840.jpg", "scene0025_01/00880.jpg", "scene0025_01/00920.jpg", "scene0025_01/00960.jpg" ], "turn_id": 5, "scan_id": "scene0025_01", "origin_question": "Which direction is the simple white plastic hanger with a triangular shape in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the simple white plastic hanger with a triangular shape in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0025_01/01000.jpg", "scene0025_01/01040.jpg", "scene0025_01/01080.jpg", "scene0025_01/01120.jpg", "scene0025_01/01160.jpg" ], "turn_id": 6, "scan_id": "scene0025_01", "origin_question": "Which direction is the rectangular wooden door with a natural grain surface in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the rectangular wooden door with a natural grain surface in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0025_01/01200.jpg", "scene0025_01/01240.jpg", "scene0025_01/01280.jpg", "scene0025_01/01320.jpg", "scene0025_01/01360.jpg" ], "turn_id": 7, "scan_id": "scene0025_01", "origin_question": "Is the rectangular white telephone with a digital display to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the rectangular white telephone with a digital display to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0025_01/01400.jpg", "scene0025_01/01440.jpg", "scene0025_01/01480.jpg", "scene0025_01/01520.jpg", "scene0025_01/01560.jpg" ], "turn_id": 8, "scan_id": "scene0025_01", "origin_question": "What will be the first-time appearance order of the following categories: bottle, cabinet, couch?", "option": [ "bottle, couch, cabinet", "bottle, cabinet, couch", "couch, bottle, cabinet", "cabinet, bottle, couch" ], "answer": "cabinet, bottle, couch", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bottle, cabinet, couch? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['bottle, couch, cabinet', 'bottle, cabinet, couch', 'couch, bottle, cabinet', 'cabinet, bottle, couch'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0478_00/00000.jpg", "scene0478_00/00040.jpg", "scene0478_00/00080.jpg", "scene0478_00/00120.jpg" ], "turn_id": 1, "scan_id": "scene0478_00", "origin_question": "Remember, have you seen any washbasin(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any washbasin(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0478_00/00160.jpg", "scene0478_00/00200.jpg", "scene0478_00/00240.jpg", "scene0478_00/00280.jpg" ], "turn_id": 2, "scan_id": "scene0478_00", "origin_question": "Is the large, modern white copier made of sturdy materials to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the large, modern white copier made of sturdy materials to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0478_00/00320.jpg", "scene0478_00/00360.jpg", "scene0478_00/00400.jpg", "scene0478_00/00440.jpg" ], "turn_id": 3, "scan_id": "scene0478_00", "origin_question": "In which turn did you see both \"the large black counter made of sturdy materials\" and \"the large, modern white copier made of sturdy materials\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the large black counter made of sturdy materials\" and \"the large, modern white copier made of sturdy materials\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0478_00/00480.jpg", "scene0478_00/00520.jpg", "scene0478_00/00560.jpg", "scene0478_00/00600.jpg" ], "turn_id": 4, "scan_id": "scene0478_00", "origin_question": "Which two objects are on the same side of you now? \"the large black counter made of sturdy materials\", \"the picture\", \"the bin\". ", "option": [ "the large black counter made of sturdy materials, the picture", "the large black counter made of sturdy materials, the bin", "the picture, the bin" ], "answer": "the large black counter made of sturdy materials, the picture", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the large black counter made of sturdy materials\", \"the picture\", \"the bin\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the large black counter made of sturdy materials, the picture', 'the large black counter made of sturdy materials, the bin', 'the picture, the bin'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0478_00/00640.jpg", "scene0478_00/00680.jpg", "scene0478_00/00720.jpg", "scene0478_00/00760.jpg" ], "turn_id": 5, "scan_id": "scene0478_00", "origin_question": "Which direction is the large black counter made of sturdy materials in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the large black counter made of sturdy materials in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0478_00/00800.jpg", "scene0478_00/00840.jpg", "scene0478_00/00880.jpg", "scene0478_00/00920.jpg" ], "turn_id": 6, "scan_id": "scene0478_00", "origin_question": "Using your orientation at the end of turn 4 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "139.71", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 4 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0164_01/00000.jpg", "scene0164_01/00030.jpg", "scene0164_01/00060.jpg", "scene0164_01/00090.jpg" ], "turn_id": 1, "scan_id": "scene0164_01", "origin_question": "Remember, have you seen any switch(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any switch(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0164_01/00120.jpg", "scene0164_01/00150.jpg", "scene0164_01/00180.jpg", "scene0164_01/00210.jpg" ], "turn_id": 2, "scan_id": "scene0164_01", "origin_question": "Is the small rectangular white plastic switch to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the small rectangular white plastic switch to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0164_01/00240.jpg", "scene0164_01/00270.jpg", "scene0164_01/00300.jpg", "scene0164_01/00330.jpg" ], "turn_id": 3, "scan_id": "scene0164_01", "origin_question": "What will be the first-time appearance order of the following categories: bowl, kettle, switch?", "option": [ "bowl, switch, kettle", "switch, kettle, bowl", "switch, bowl, kettle", "kettle, switch, bowl" ], "answer": "switch, bowl, kettle", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bowl, kettle, switch? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['bowl, switch, kettle', 'switch, kettle, bowl', 'switch, bowl, kettle', 'kettle, switch, bowl'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0164_01/00360.jpg", "scene0164_01/00390.jpg", "scene0164_01/00420.jpg", "scene0164_01/00450.jpg" ], "turn_id": 4, "scan_id": "scene0164_01", "origin_question": "When did you first discover \"the small rectangular white plastic dish rack\"? (index of the turn)", "option": [], "answer": "4", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the small rectangular white plastic dish rack\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0164_01/00480.jpg", "scene0164_01/00510.jpg", "scene0164_01/00540.jpg", "scene0164_01/00570.jpg" ], "turn_id": 5, "scan_id": "scene0164_01", "origin_question": "Which two objects are on the same side of you now? \"the microwave\", \"the large, shiny silver stainless steel sink\", \"the standard-sized blue plastic bottle with a white cap\". ", "option": [ "the microwave, the large, shiny silver stainless steel sink", "the microwave, the standard-sized blue plastic bottle with a white cap", "the large, shiny silver stainless steel sink, the standard-sized blue plastic bottle with a white cap" ], "answer": "the microwave, the large, shiny silver stainless steel sink", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the microwave\", \"the large, shiny silver stainless steel sink\", \"the standard-sized blue plastic bottle with a white cap\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the microwave, the large, shiny silver stainless steel sink', 'the microwave, the standard-sized blue plastic bottle with a white cap', 'the large, shiny silver stainless steel sink, the standard-sized blue plastic bottle with a white cap'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0164_01/00600.jpg", "scene0164_01/00630.jpg", "scene0164_01/00660.jpg", "scene0164_01/00690.jpg" ], "turn_id": 6, "scan_id": "scene0164_01", "origin_question": "In which turn were you farthest from \"the refrigerator\"? ", "option": [], "answer": "6", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the refrigerator\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0084_region4/f8a1271047a0471e83c51ec2be8df013_i1_5.jpg", "1mp3d_0084_region4/f8a1271047a0471e83c51ec2be8df013_i2_5.jpg", "1mp3d_0084_region4/a3b941ba54954c9295e26ac438aebbaf_i1_4.jpg", "1mp3d_0084_region4/a3b941ba54954c9295e26ac438aebbaf_i2_4.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0084_region4", "origin_question": "Remember, have you seen any pillow(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any pillow(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0084_region4/a3b941ba54954c9295e26ac438aebbaf_i2_5.jpg", "1mp3d_0084_region4/f8a1271047a0471e83c51ec2be8df013_i1_1.jpg", "1mp3d_0084_region4/f8a1271047a0471e83c51ec2be8df013_i2_1.jpg", "1mp3d_0084_region4/a4d731efebcc4e9f8902a6de63f998fb_i2_2.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0084_region4", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0084_region4/a4d731efebcc4e9f8902a6de63f998fb_i1_1.jpg", "1mp3d_0084_region4/a4d731efebcc4e9f8902a6de63f998fb_i2_1.jpg", "1mp3d_0084_region4/5f3ff024aa0d403691b0f649a7b7a45a_i1_0.jpg", "1mp3d_0084_region4/5f3ff024aa0d403691b0f649a7b7a45a_i2_0.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0084_region4", "origin_question": "", "option": [], "answer": "", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0084_region4/5f3ff024aa0d403691b0f649a7b7a45a_i2_5.jpg", "1mp3d_0084_region4/a4d731efebcc4e9f8902a6de63f998fb_i2_0.jpg", "1mp3d_0084_region4/a4d731efebcc4e9f8902a6de63f998fb_i1_5.jpg", "1mp3d_0084_region4/a4d731efebcc4e9f8902a6de63f998fb_i2_5.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0084_region4", "origin_question": "Remember, have you seen any folder(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any folder(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0552_00/00000.jpg", "scene0552_00/00030.jpg", "scene0552_00/00060.jpg", "scene0552_00/00090.jpg" ], "turn_id": 1, "scan_id": "scene0552_00", "origin_question": "Remember, have you seen any picture(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any picture(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0552_00/00120.jpg", "scene0552_00/00150.jpg", "scene0552_00/00180.jpg", "scene0552_00/00210.jpg" ], "turn_id": 2, "scan_id": "scene0552_00", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the grey plastic copier with a flat top\"; \"the window\"; \"the rectangular green metal paper cutter with a handle\" ", "option": [ "the grey plastic copier with a flat top", "the window", "the rectangular green metal paper cutter with a handle" ], "answer": "the window", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the grey plastic copier with a flat top\"; \"the window\"; \"the rectangular green metal paper cutter with a handle\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the grey plastic copier with a flat top', 'the window', 'the rectangular green metal paper cutter with a handle'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0552_00/00240.jpg", "scene0552_00/00270.jpg", "scene0552_00/00300.jpg", "scene0552_00/00330.jpg" ], "turn_id": 3, "scan_id": "scene0552_00", "origin_question": "Which direction is the standard-sized white plastic bucket with a lid in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the standard-sized white plastic bucket with a lid in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0552_00/00360.jpg", "scene0552_00/00390.jpg", "scene0552_00/00420.jpg", "scene0552_00/00450.jpg" ], "turn_id": 4, "scan_id": "scene0552_00", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the gray rectangular metal paper towel dispenser\"; \"the white rectangular plastic microwave with digital display\"; \"the modern stainless steel double-door refrigerator\" ", "option": [ "the gray rectangular metal paper towel dispenser", "the white rectangular plastic microwave with digital display", "the modern stainless steel double-door refrigerator" ], "answer": "the gray rectangular metal paper towel dispenser", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the gray rectangular metal paper towel dispenser\"; \"the white rectangular plastic microwave with digital display\"; \"the modern stainless steel double-door refrigerator\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the gray rectangular metal paper towel dispenser', 'the white rectangular plastic microwave with digital display', 'the modern stainless steel double-door refrigerator'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "scene0552_00/00480.jpg", "scene0552_00/00510.jpg", "scene0552_00/00540.jpg", "scene0552_00/00570.jpg" ], "turn_id": 5, "scan_id": "scene0552_00", "origin_question": "Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Right", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0552_00/00600.jpg", "scene0552_00/00630.jpg", "scene0552_00/00660.jpg", "scene0552_00/00690.jpg" ], "turn_id": 6, "scan_id": "scene0552_00", "origin_question": "Please recall \"the stylish black and silver metal oven\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.32", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the stylish black and silver metal oven\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0552_00/00720.jpg", "scene0552_00/00750.jpg", "scene0552_00/00780.jpg", "scene0552_00/00810.jpg" ], "turn_id": 7, "scan_id": "scene0552_00", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the small white rectangular plastic soap dispenser\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the small white rectangular plastic soap dispenser\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0693_02/00000.jpg", "scene0693_02/00040.jpg", "scene0693_02/00080.jpg", "scene0693_02/00120.jpg", "scene0693_02/00160.jpg" ], "turn_id": 1, "scan_id": "scene0693_02", "origin_question": "Remember, have you seen any tissue box(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any tissue box(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0693_02/00200.jpg", "scene0693_02/00240.jpg", "scene0693_02/00280.jpg", "scene0693_02/00320.jpg", "scene0693_02/00360.jpg" ], "turn_id": 2, "scan_id": "scene0693_02", "origin_question": "Which direction is the blue ceramic cup of standard size in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the blue ceramic cup of standard size in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0693_02/00400.jpg", "scene0693_02/00440.jpg", "scene0693_02/00480.jpg", "scene0693_02/00520.jpg", "scene0693_02/00560.jpg" ], "turn_id": 3, "scan_id": "scene0693_02", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the small rectangular white plastic storage box\"; \"the blue cylindrical plastic bottle\"; \"the small, metal stopcock to control water flow\" ", "option": [ "the small rectangular white plastic storage box", "the blue cylindrical plastic bottle", "the small, metal stopcock to control water flow" ], "answer": "the small rectangular white plastic storage box", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the small rectangular white plastic storage box\"; \"the blue cylindrical plastic bottle\"; \"the small, metal stopcock to control water flow\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the small rectangular white plastic storage box', 'the blue cylindrical plastic bottle', 'the small, metal stopcock to control water flow'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0693_02/00600.jpg", "scene0693_02/00640.jpg", "scene0693_02/00680.jpg", "scene0693_02/00720.jpg", "scene0693_02/00760.jpg" ], "turn_id": 4, "scan_id": "scene0693_02", "origin_question": "Please recall \"the green and white toothpaste tube made of paper\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.53", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the green and white toothpaste tube made of paper\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0693_02/00800.jpg", "scene0693_02/00840.jpg", "scene0693_02/00880.jpg", "scene0693_02/00920.jpg", "scene0693_02/00960.jpg" ], "turn_id": 5, "scan_id": "scene0693_02", "origin_question": "Please recall \"the blue cylindrical plastic bottle\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.94", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the blue cylindrical plastic bottle\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0693_02/01000.jpg", "scene0693_02/01040.jpg", "scene0693_02/01080.jpg", "scene0693_02/01120.jpg", "scene0693_02/01160.jpg" ], "turn_id": 6, "scan_id": "scene0693_02", "origin_question": "Using your orientation at the end of turn 5 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "25.83", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 5 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "scene0693_02/01200.jpg", "scene0693_02/01240.jpg", "scene0693_02/01280.jpg", "scene0693_02/01320.jpg", "scene0693_02/01360.jpg" ], "turn_id": 7, "scan_id": "scene0693_02", "origin_question": "Among these three objects(\"the rectangular brown wooden tissue box with white tissues\",\"the white rectangular bathtub made of ceramic or porcelain\",\"the towel\" ), which one is the closest to you?", "option": [ "the rectangular brown wooden tissue box with white tissues", "the white rectangular bathtub made of ceramic or porcelain", "the towel" ], "answer": "the towel", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the rectangular brown wooden tissue box with white tissues\",\"the white rectangular bathtub made of ceramic or porcelain\",\"the towel\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular brown wooden tissue box with white tissues', 'the white rectangular bathtub made of ceramic or porcelain', 'the towel'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0693_02/01400.jpg", "scene0693_02/01440.jpg", "scene0693_02/01480.jpg", "scene0693_02/01520.jpg", "scene0693_02/01560.jpg" ], "turn_id": 8, "scan_id": "scene0693_02", "origin_question": "Compared to your position at the end of the 5 turn, are you now closer or farther away from \"the small rectangular white switch\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 5 turn, are you now closer or farther away from \"the small rectangular white switch\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0488_01/00000.jpg", "scene0488_01/00040.jpg", "scene0488_01/00080.jpg", "scene0488_01/00120.jpg" ], "turn_id": 1, "scan_id": "scene0488_01", "origin_question": "Remember, have you seen any sink(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any sink(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0488_01/00160.jpg", "scene0488_01/00200.jpg", "scene0488_01/00240.jpg", "scene0488_01/00280.jpg" ], "turn_id": 2, "scan_id": "scene0488_01", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0488_01/00320.jpg", "scene0488_01/00360.jpg", "scene0488_01/00400.jpg", "scene0488_01/00440.jpg" ], "turn_id": 3, "scan_id": "scene0488_01", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0488_01/00480.jpg", "scene0488_01/00520.jpg", "scene0488_01/00560.jpg", "scene0488_01/00600.jpg" ], "turn_id": 4, "scan_id": "scene0488_01", "origin_question": "", "option": [], "answer": "", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0488_01/00640.jpg", "scene0488_01/00680.jpg", "scene0488_01/00720.jpg", "scene0488_01/00760.jpg" ], "turn_id": 5, "scan_id": "scene0488_01", "origin_question": "Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "91.59", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0488_01/00800.jpg", "scene0488_01/00840.jpg", "scene0488_01/00880.jpg", "scene0488_01/00920.jpg" ], "turn_id": 6, "scan_id": "scene0488_01", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the white rectangular microwave with a digital display\" from your position?", "option": [], "answer": "61.13", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the white rectangular microwave with a digital display\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "1mp3d_0044_region14/6d36fd2333924cf7a8157c5380b7645b_i1_2.jpg", "1mp3d_0044_region14/6d36fd2333924cf7a8157c5380b7645b_i2_2.jpg", "1mp3d_0044_region14/617741b15475497e898493e3d96153fd_i1_4.jpg", "1mp3d_0044_region14/617741b15475497e898493e3d96153fd_i2_4.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0044_region14", "origin_question": "Remember, how many range hood(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many range hood(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0044_region14/f3865d531b6849f497bc9e50a5f74119_i1_1.jpg", "1mp3d_0044_region14/f3865d531b6849f497bc9e50a5f74119_i2_1.jpg", "1mp3d_0044_region14/617741b15475497e898493e3d96153fd_i1_5.jpg", "1mp3d_0044_region14/617741b15475497e898493e3d96153fd_i2_5.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0044_region14", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "1mp3d_0044_region14/617741b15475497e898493e3d96153fd_i1_0.jpg", "1mp3d_0044_region14/617741b15475497e898493e3d96153fd_i2_0.jpg", "1mp3d_0044_region14/6d36fd2333924cf7a8157c5380b7645b_i1_3.jpg", "1mp3d_0044_region14/6d36fd2333924cf7a8157c5380b7645b_i2_3.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0044_region14", "origin_question": "Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance forward or backward from that position?", "option": [ "Forward", "Backward" ], "answer": "Backward", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance forward or backward from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Forward', 'Backward'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "1mp3d_0044_region14/758fe37a60fe444b996ff1be363708a1_i1_1.jpg", "1mp3d_0044_region14/758fe37a60fe444b996ff1be363708a1_i2_1.jpg", "1mp3d_0044_region14/db188148a66d44f8bd85114fd88bf028_i1_3.jpg", "1mp3d_0044_region14/db188148a66d44f8bd85114fd88bf028_i2_3.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0044_region14", "origin_question": "When did you first discover \"the large white stone fireplace with an arched top\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the large white stone fireplace with an arched top\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "1mp3d_0044_region14/db188148a66d44f8bd85114fd88bf028_i1_2.jpg", "1mp3d_0044_region14/db188148a66d44f8bd85114fd88bf028_i2_2.jpg", "1mp3d_0044_region14/db188148a66d44f8bd85114fd88bf028_i1_1.jpg", "1mp3d_0044_region14/db188148a66d44f8bd85114fd88bf028_i2_1.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0044_region14", "origin_question": "At the end of which turn were both of \"the round coffee table with a glass top and wooden base\" and \"the shiny black piano with a polished wooden frame\" on your right side?", "option": [], "answer": "5", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the round coffee table with a glass top and wooden base\" and \"the shiny black piano with a polished wooden frame\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "1mp3d_0044_region14/db188148a66d44f8bd85114fd88bf028_i1_0.jpg", "1mp3d_0044_region14/db188148a66d44f8bd85114fd88bf028_i2_0.jpg", "1mp3d_0044_region14/db188148a66d44f8bd85114fd88bf028_i1_5.jpg", "1mp3d_0044_region14/db188148a66d44f8bd85114fd88bf028_i2_5.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0044_region14", "origin_question": "What will be the first-time appearance order of the following categories: fireplace, piano, table?", "option": [ "fireplace, table, piano", "piano, table, fireplace", "table, fireplace, piano", "fireplace, piano, table" ], "answer": "fireplace, table, piano", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: fireplace, piano, table? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['fireplace, table, piano', 'piano, table, fireplace', 'table, fireplace, piano', 'fireplace, piano, table'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "1mp3d_0044_region14/758fe37a60fe444b996ff1be363708a1_i2_4.jpg", "1mp3d_0044_region14/758fe37a60fe444b996ff1be363708a1_i1_3.jpg", "1mp3d_0044_region14/758fe37a60fe444b996ff1be363708a1_i2_3.jpg", "1mp3d_0044_region14/3a2ee04ca9024979af419e7f9da592ee_i1_4.jpg" ], "turn_id": 7, "scan_id": "1mp3d_0044_region14", "origin_question": "At the end of which turn were both of \"the shiny black piano with a polished wooden frame\" and \"the rectangular dark wooden box without decorative features\" on your right side?", "option": [], "answer": "4", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the shiny black piano with a polished wooden frame\" and \"the rectangular dark wooden box without decorative features\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "1mp3d_0044_region14/3a2ee04ca9024979af419e7f9da592ee_i2_4.jpg", "1mp3d_0044_region14/3a2ee04ca9024979af419e7f9da592ee_i1_5.jpg", "1mp3d_0044_region14/3a2ee04ca9024979af419e7f9da592ee_i2_5.jpg", "1mp3d_0044_region14/9ea4d4f72dab402a8b82585b0e4f0dda_i1_3.jpg" ], "turn_id": 8, "scan_id": "1mp3d_0044_region14", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "4.96", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0204_01/00000.jpg", "scene0204_01/00030.jpg", "scene0204_01/00060.jpg", "scene0204_01/00090.jpg" ], "turn_id": 1, "scan_id": "scene0204_01", "origin_question": "Remember, how many coffee maker(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many coffee maker(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0204_01/00120.jpg", "scene0204_01/00150.jpg", "scene0204_01/00180.jpg", "scene0204_01/00210.jpg" ], "turn_id": 2, "scan_id": "scene0204_01", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0204_01/00240.jpg", "scene0204_01/00270.jpg", "scene0204_01/00300.jpg", "scene0204_01/00330.jpg" ], "turn_id": 3, "scan_id": "scene0204_01", "origin_question": "In which turn were you closest to \"the door\"? ", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the door\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "scene0204_01/00360.jpg", "scene0204_01/00390.jpg", "scene0204_01/00420.jpg", "scene0204_01/00450.jpg" ], "turn_id": 4, "scan_id": "scene0204_01", "origin_question": "Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Right", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0204_01/00480.jpg", "scene0204_01/00510.jpg", "scene0204_01/00610.jpg", "scene0204_01/00640.jpg" ], "turn_id": 5, "scan_id": "scene0204_01", "origin_question": "Please recall \"the door\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "3.66", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the door\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0204_01/00670.jpg", "scene0204_01/00700.jpg", "scene0204_01/00730.jpg", "scene0204_01/00760.jpg" ], "turn_id": 6, "scan_id": "scene0204_01", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the large rectangular wooden table with a polished surface\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the large rectangular wooden table with a polished surface\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0221_01/00000.jpg", "scene0221_01/00030.jpg", "scene0221_01/00060.jpg", "scene0221_01/00090.jpg" ], "turn_id": 1, "scan_id": "scene0221_01", "origin_question": "Remember, have you seen any sack(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any sack(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0221_01/00120.jpg", "scene0221_01/00150.jpg", "scene0221_01/00180.jpg", "scene0221_01/00210.jpg" ], "turn_id": 2, "scan_id": "scene0221_01", "origin_question": "At the end of which turn were both of \"the lamp\" and \"the rectangular white paper with a slanted edge\" on your right side?", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the lamp\" and \"the rectangular white paper with a slanted edge\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0221_01/00240.jpg", "scene0221_01/00270.jpg", "scene0221_01/00300.jpg", "scene0221_01/00330.jpg" ], "turn_id": 3, "scan_id": "scene0221_01", "origin_question": "Is the rectangular white paper with a slanted edge to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the rectangular white paper with a slanted edge to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0221_01/00360.jpg", "scene0221_01/00390.jpg", "scene0221_01/00420.jpg", "scene0221_01/00450.jpg" ], "turn_id": 4, "scan_id": "scene0221_01", "origin_question": "Please recall \"the rectangular white paper with a slanted edge\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "3.18", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the rectangular white paper with a slanted edge\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0221_01/00480.jpg", "scene0221_01/00510.jpg", "scene0221_01/00540.jpg", "scene0221_01/00570.jpg" ], "turn_id": 5, "scan_id": "scene0221_01", "origin_question": "Which two objects are on the same side of you now? \"the rectangular countertop microwave with a black plastic exterior\", \"the rectangular white paper with a slanted edge\", \"the table\". ", "option": [ "the rectangular countertop microwave with a black plastic exterior, the rectangular white paper with a slanted edge", "the rectangular countertop microwave with a black plastic exterior, the table", "the rectangular white paper with a slanted edge, the table" ], "answer": "the rectangular countertop microwave with a black plastic exterior, the rectangular white paper with a slanted edge", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the rectangular countertop microwave with a black plastic exterior\", \"the rectangular white paper with a slanted edge\", \"the table\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular countertop microwave with a black plastic exterior, the rectangular white paper with a slanted edge', 'the rectangular countertop microwave with a black plastic exterior, the table', 'the rectangular white paper with a slanted edge, the table'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0221_01/00600.jpg", "scene0221_01/00630.jpg", "scene0221_01/00660.jpg", "scene0221_01/00690.jpg" ], "turn_id": 6, "scan_id": "scene0221_01", "origin_question": "Is the lamp to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the lamp to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0671_01/00000.jpg", "scene0671_01/00030.jpg", "scene0671_01/00060.jpg", "scene0671_01/00090.jpg" ], "turn_id": 1, "scan_id": "scene0671_01", "origin_question": "Remember, how many pillow(s) have you seen so far? ", "option": [], "answer": "2", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many pillow(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0671_01/00120.jpg", "scene0671_01/00150.jpg", "scene0671_01/00180.jpg", "scene0671_01/00210.jpg" ], "turn_id": 2, "scan_id": "scene0671_01", "origin_question": "In which turn were you farthest from \"the table\"? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the table\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0671_01/00240.jpg", "scene0671_01/00270.jpg", "scene0671_01/00300.jpg", "scene0671_01/00330.jpg" ], "turn_id": 3, "scan_id": "scene0671_01", "origin_question": "In which turn did you see both \"the rectangular black tray made of sturdy material\" and \"the clear round glass cup of standard size\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the rectangular black tray made of sturdy material\" and \"the clear round glass cup of standard size\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0671_01/00360.jpg", "scene0671_01/00390.jpg", "scene0671_01/00420.jpg", "scene0671_01/00450.jpg" ], "turn_id": 4, "scan_id": "scene0671_01", "origin_question": "In which turn were you closest to \"the plant\"? ", "option": [], "answer": "4", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the plant\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0671_01/00480.jpg", "scene0671_01/00510.jpg", "scene0671_01/00540.jpg", "scene0671_01/00570.jpg" ], "turn_id": 5, "scan_id": "scene0671_01", "origin_question": "Is the plant to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the plant to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0260_00/00000.jpg", "scene0260_00/00040.jpg", "scene0260_00/00080.jpg", "scene0260_00/00120.jpg" ], "turn_id": 1, "scan_id": "scene0260_00", "origin_question": "Remember, have you seen any clothes dryer(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any clothes dryer(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0260_00/00160.jpg", "scene0260_00/00200.jpg", "scene0260_00/00240.jpg", "scene0260_00/00280.jpg" ], "turn_id": 2, "scan_id": "scene0260_00", "origin_question": "In which turn did you see both \"the bathtub\" and \"the roll of standard-sized, white cylindrical toilet paper\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the bathtub\" and \"the roll of standard-sized, white cylindrical toilet paper\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0260_00/00320.jpg", "scene0260_00/00360.jpg", "scene0260_00/00400.jpg", "scene0260_00/00440.jpg" ], "turn_id": 3, "scan_id": "scene0260_00", "origin_question": "What will be the first-time appearance order of the following categories: mirror, toilet paper, toothpaste?", "option": [ "toilet paper, mirror, toothpaste", "toothpaste, mirror, toilet paper", "mirror, toilet paper, toothpaste", "mirror, toothpaste, toilet paper" ], "answer": "toilet paper, mirror, toothpaste", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: mirror, toilet paper, toothpaste? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['toilet paper, mirror, toothpaste', 'toothpaste, mirror, toilet paper', 'mirror, toilet paper, toothpaste', 'mirror, toothpaste, toilet paper'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0260_00/00480.jpg", "scene0260_00/00520.jpg", "scene0260_00/00560.jpg", "scene0260_00/00600.jpg" ], "turn_id": 4, "scan_id": "scene0260_00", "origin_question": "Remember, how many lamp(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many lamp(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0260_00/00640.jpg", "scene0260_00/00680.jpg", "scene0260_00/00720.jpg", "scene0260_00/00760.jpg" ], "turn_id": 5, "scan_id": "scene0260_00", "origin_question": "What will be the first-time appearance order of the following categories: doorframe, picture, toilet?", "option": [ "toilet, picture, doorframe", "picture, doorframe, toilet", "picture, toilet, doorframe", "toilet, doorframe, picture" ], "answer": "toilet, picture, doorframe", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: doorframe, picture, toilet? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['toilet, picture, doorframe', 'picture, doorframe, toilet', 'picture, toilet, doorframe', 'toilet, doorframe, picture'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0260_00/00800.jpg", "scene0260_00/00840.jpg", "scene0260_00/00880.jpg", "scene0260_00/00920.jpg" ], "turn_id": 6, "scan_id": "scene0260_00", "origin_question": "Which two objects are on the same side of you now? \"the silver metal bin of standard size\", \"the white rectangular ceramic vanity on a black countertop\", \"the stylish black plastic hair dryer with a modern design\". ", "option": [ "the silver metal bin of standard size, the white rectangular ceramic vanity on a black countertop", "the silver metal bin of standard size, the stylish black plastic hair dryer with a modern design", "the white rectangular ceramic vanity on a black countertop, the stylish black plastic hair dryer with a modern design" ], "answer": "the white rectangular ceramic vanity on a black countertop, the stylish black plastic hair dryer with a modern design", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the silver metal bin of standard size\", \"the white rectangular ceramic vanity on a black countertop\", \"the stylish black plastic hair dryer with a modern design\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the silver metal bin of standard size, the white rectangular ceramic vanity on a black countertop', 'the silver metal bin of standard size, the stylish black plastic hair dryer with a modern design', 'the white rectangular ceramic vanity on a black countertop, the stylish black plastic hair dryer with a modern design'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0260_00/00960.jpg", "scene0260_00/01000.jpg", "scene0260_00/01040.jpg", "scene0260_00/01080.jpg" ], "turn_id": 7, "scan_id": "scene0260_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0260_00/01120.jpg", "scene0260_00/01160.jpg", "scene0260_00/01200.jpg", "scene0260_00/01240.jpg" ], "turn_id": 8, "scan_id": "scene0260_00", "origin_question": "Which two objects are on the same side of you now? \"the light-colored fabric curtain with delicate wavy patterns\", \"the roll of standard-sized, white cylindrical toilet paper\", \"the square canvas picture with a black frame\". ", "option": [ "the light-colored fabric curtain with delicate wavy patterns, the roll of standard-sized, white cylindrical toilet paper", "the light-colored fabric curtain with delicate wavy patterns, the square canvas picture with a black frame", "the roll of standard-sized, white cylindrical toilet paper, the square canvas picture with a black frame" ], "answer": "the roll of standard-sized, white cylindrical toilet paper, the square canvas picture with a black frame", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the light-colored fabric curtain with delicate wavy patterns\", \"the roll of standard-sized, white cylindrical toilet paper\", \"the square canvas picture with a black frame\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the light-colored fabric curtain with delicate wavy patterns, the roll of standard-sized, white cylindrical toilet paper', 'the light-colored fabric curtain with delicate wavy patterns, the square canvas picture with a black frame', 'the roll of standard-sized, white cylindrical toilet paper, the square canvas picture with a black frame'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0260_00/01280.jpg", "scene0260_00/01320.jpg", "scene0260_00/01360.jpg", "scene0260_00/01400.jpg" ], "turn_id": 9, "scan_id": "scene0260_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 9 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0078_region8/e8a34a4b401249a589ed3b571839366e_i1_2.jpg", "1mp3d_0078_region8/e8a34a4b401249a589ed3b571839366e_i2_2.jpg", "1mp3d_0078_region8/22c29c9d052f4439a66194f3f8d8a6e1_i1_2.jpg", "1mp3d_0078_region8/22c29c9d052f4439a66194f3f8d8a6e1_i2_2.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0078_region8", "origin_question": "Remember, have you seen any door(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any door(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "1mp3d_0078_region8/f7dfad43a3a6462cb96145242bd8bdd0_i1_2.jpg", "1mp3d_0078_region8/f7dfad43a3a6462cb96145242bd8bdd0_i2_2.jpg", "1mp3d_0078_region8/f7dfad43a3a6462cb96145242bd8bdd0_i1_1.jpg", "1mp3d_0078_region8/f7dfad43a3a6462cb96145242bd8bdd0_i2_1.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0078_region8", "origin_question": "When did you first discover \"the light brown rectangular wooden door with stripes\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the light brown rectangular wooden door with stripes\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0078_region8/22c29c9d052f4439a66194f3f8d8a6e1_i1_1.jpg", "1mp3d_0078_region8/22c29c9d052f4439a66194f3f8d8a6e1_i2_1.jpg", "1mp3d_0078_region8/e8a34a4b401249a589ed3b571839366e_i1_1.jpg", "1mp3d_0078_region8/e8a34a4b401249a589ed3b571839366e_i2_1.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0078_region8", "origin_question": "", "option": [], "answer": "", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "1mp3d_0078_region8/6cff8d4e873a4dbc8db352872e706bb8_i1_0.jpg", "1mp3d_0078_region8/6cff8d4e873a4dbc8db352872e706bb8_i2_0.jpg", "1mp3d_0078_region8/09f015ea8ca94df8841bcea7030bca99_i1_5.jpg", "1mp3d_0078_region8/09f015ea8ca94df8841bcea7030bca99_i2_5.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0078_region8", "origin_question": "How far is your current position from where you were at the end of the turn 3? (in meters)", "option": [], "answer": "2.18", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 3? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "1mp3d_0078_region8/6cff8d4e873a4dbc8db352872e706bb8_i1_5.jpg", "1mp3d_0078_region8/6cff8d4e873a4dbc8db352872e706bb8_i2_5.jpg", "1mp3d_0078_region8/e057d1f9d9ba42948fc848930f235276_i1_5.jpg", "1mp3d_0078_region8/e057d1f9d9ba42948fc848930f235276_i2_5.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0078_region8", "origin_question": "At the end of which turn were both of \"the rectangular yellow board with a sunflower design\" and \"the white rectangular washing machine made of plastic or steel\" on your left side?", "option": [], "answer": "5", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the rectangular yellow board with a sunflower design\" and \"the white rectangular washing machine made of plastic or steel\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "1mp3d_0078_region8/e057d1f9d9ba42948fc848930f235276_i1_4.jpg", "1mp3d_0078_region8/e057d1f9d9ba42948fc848930f235276_i2_4.jpg", "1mp3d_0078_region8/6cff8d4e873a4dbc8db352872e706bb8_i1_4.jpg", "1mp3d_0078_region8/6cff8d4e873a4dbc8db352872e706bb8_i2_4.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0078_region8", "origin_question": "Please recall \"the box\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.21", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the box\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "1mp3d_0078_region8/e8a34a4b401249a589ed3b571839366e_i1_4.jpg", "1mp3d_0078_region8/e8a34a4b401249a589ed3b571839366e_i2_4.jpg", "1mp3d_0078_region8/22c29c9d052f4439a66194f3f8d8a6e1_i1_4.jpg", "1mp3d_0078_region8/22c29c9d052f4439a66194f3f8d8a6e1_i2_4.jpg" ], "turn_id": 7, "scan_id": "1mp3d_0078_region8", "origin_question": "By the end of the 4 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular yellow board with a sunflower design\", Y: \"the white rectangular washing machine made of plastic or steel\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 4 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular yellow board with a sunflower design\", Y: \"the white rectangular washing machine made of plastic or steel\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0078_region8/f7dfad43a3a6462cb96145242bd8bdd0_i1_3.jpg", "1mp3d_0078_region8/f7dfad43a3a6462cb96145242bd8bdd0_i2_3.jpg", "1mp3d_0078_region8/f7dfad43a3a6462cb96145242bd8bdd0_i1_4.jpg", "1mp3d_0078_region8/f7dfad43a3a6462cb96145242bd8bdd0_i2_4.jpg" ], "turn_id": 8, "scan_id": "1mp3d_0078_region8", "origin_question": "Compared to your position at the end of the 5 turn, are you now closer or farther away from \"the light brown rectangular wooden door with stripes\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 5 turn, are you now closer or farther away from \"the light brown rectangular wooden door with stripes\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0047_region4/ae53fd4386f843b9800fe47f8c1518d1_i1_1.jpg", "1mp3d_0047_region4/ae53fd4386f843b9800fe47f8c1518d1_i2_1.jpg", "1mp3d_0047_region4/c69c3bcba6994b5995d372f7e8f00390_i1_2.jpg", "1mp3d_0047_region4/c69c3bcba6994b5995d372f7e8f00390_i2_2.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0047_region4", "origin_question": "Remember, have you seen any stairs(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any stairs(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "1mp3d_0047_region4/a6555179ea4e45b7af7db8f21d4c3b78_i1_5.jpg", "1mp3d_0047_region4/a6555179ea4e45b7af7db8f21d4c3b78_i2_5.jpg", "1mp3d_0047_region4/9f1591971b10420eafc7424d68339d99_i1_5.jpg", "1mp3d_0047_region4/9f1591971b10420eafc7424d68339d99_i2_5.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0047_region4", "origin_question": "When did you first discover \"the moderate-sized grey concrete stairs\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the moderate-sized grey concrete stairs\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0047_region4/a6555179ea4e45b7af7db8f21d4c3b78_i1_4.jpg", "1mp3d_0047_region4/a6555179ea4e45b7af7db8f21d4c3b78_i2_4.jpg", "1mp3d_0047_region4/9f1591971b10420eafc7424d68339d99_i1_0.jpg", "1mp3d_0047_region4/9f1591971b10420eafc7424d68339d99_i2_0.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0047_region4", "origin_question": "", "option": [], "answer": "", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0047_region4/a6555179ea4e45b7af7db8f21d4c3b78_i1_1.jpg", "1mp3d_0047_region4/a6555179ea4e45b7af7db8f21d4c3b78_i2_1.jpg", "1mp3d_0047_region4/c69c3bcba6994b5995d372f7e8f00390_i1_5.jpg", "1mp3d_0047_region4/c69c3bcba6994b5995d372f7e8f00390_i2_5.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0047_region4", "origin_question": "", "option": [], "answer": "", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "1mp3d_0047_region4/c69c3bcba6994b5995d372f7e8f00390_i1_1.jpg", "1mp3d_0047_region4/c69c3bcba6994b5995d372f7e8f00390_i2_1.jpg", "1mp3d_0047_region4/ae53fd4386f843b9800fe47f8c1518d1_i1_0.jpg", "1mp3d_0047_region4/ae53fd4386f843b9800fe47f8c1518d1_i2_0.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0047_region4", "origin_question": "How far is your current position from where you were at the end of the turn 3? (in meters)", "option": [], "answer": "3.72", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 3? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0143_01/00000.jpg", "scene0143_01/00040.jpg", "scene0143_01/00080.jpg", "scene0143_01/00120.jpg", "scene0143_01/00160.jpg" ], "turn_id": 1, "scan_id": "scene0143_01", "origin_question": "Remember, have you seen any candle(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any candle(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0143_01/00200.jpg", "scene0143_01/00240.jpg", "scene0143_01/00390.jpg", "scene0143_01/00430.jpg", "scene0143_01/00470.jpg" ], "turn_id": 2, "scan_id": "scene0143_01", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the medium black soft fabric jacket\"; \"the long, textured beige curtain with pleats\"; \"the bed\" ", "option": [ "the medium black soft fabric jacket", "the long, textured beige curtain with pleats", "the bed" ], "answer": "the medium black soft fabric jacket", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the medium black soft fabric jacket\"; \"the long, textured beige curtain with pleats\"; \"the bed\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the medium black soft fabric jacket', 'the long, textured beige curtain with pleats', 'the bed'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0143_01/00510.jpg", "scene0143_01/00550.jpg", "scene0143_01/00590.jpg", "scene0143_01/00630.jpg", "scene0143_01/00670.jpg" ], "turn_id": 3, "scan_id": "scene0143_01", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the cylindrical, dark-colored bin\"; \"the long, textured beige curtain with pleats\"; \"the bed\" ", "option": [ "the cylindrical, dark-colored bin", "the long, textured beige curtain with pleats", "the bed" ], "answer": "the cylindrical, dark-colored bin", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the cylindrical, dark-colored bin\"; \"the long, textured beige curtain with pleats\"; \"the bed\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the cylindrical, dark-colored bin', 'the long, textured beige curtain with pleats', 'the bed'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0143_01/00710.jpg", "scene0143_01/00750.jpg", "scene0143_01/00790.jpg", "scene0143_01/00830.jpg", "scene0143_01/00870.jpg" ], "turn_id": 4, "scan_id": "scene0143_01", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "1.21", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0143_01/00910.jpg", "scene0143_01/00950.jpg", "scene0143_01/00990.jpg", "scene0143_01/01030.jpg", "scene0143_01/01070.jpg" ], "turn_id": 5, "scan_id": "scene0143_01", "origin_question": "Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0143_01/01110.jpg", "scene0143_01/01150.jpg", "scene0143_01/01190.jpg", "scene0143_01/01230.jpg", "scene0143_01/01270.jpg" ], "turn_id": 6, "scan_id": "scene0143_01", "origin_question": "Which two objects are on the same side of you now? \"the vibrant red, modern single-seat couch\", \"the case\", \"the dark television set made of smooth materials\". ", "option": [ "the vibrant red, modern single-seat couch, the case", "the vibrant red, modern single-seat couch, the dark television set made of smooth materials", "the case, the dark television set made of smooth materials" ], "answer": "the vibrant red, modern single-seat couch, the case", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the vibrant red, modern single-seat couch\", \"the case\", \"the dark television set made of smooth materials\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the vibrant red, modern single-seat couch, the case', 'the vibrant red, modern single-seat couch, the dark television set made of smooth materials', 'the case, the dark television set made of smooth materials'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0143_01/01310.jpg", "scene0143_01/01350.jpg", "scene0143_01/01390.jpg", "scene0143_01/01470.jpg", "scene0143_01/01510.jpg" ], "turn_id": 7, "scan_id": "scene0143_01", "origin_question": "Is the small rectangular white plastic switch to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the small rectangular white plastic switch to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0143_01/01550.jpg", "scene0143_01/01630.jpg", "scene0143_01/01670.jpg", "scene0143_01/01720.jpg", "scene0143_01/01760.jpg" ], "turn_id": 8, "scan_id": "scene0143_01", "origin_question": "Please recall \"the backpack\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.91", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the backpack\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0143_01/01800.jpg", "scene0143_01/01840.jpg", "scene0143_01/01880.jpg", "scene0143_01/01920.jpg", "scene0143_01/01960.jpg" ], "turn_id": 9, "scan_id": "scene0143_01", "origin_question": "Using your orientation at the end of turn 5 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "39.72", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 5 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0143_01/02000.jpg", "scene0143_01/02040.jpg", "scene0143_01/02080.jpg", "scene0143_01/02120.jpg", "scene0143_01/02160.jpg" ], "turn_id": 10, "scan_id": "scene0143_01", "origin_question": "Compared to your position at the end of the 6 turn, are you now closer or farther away from \"the radiator\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 6 turn, are you now closer or farther away from \"the radiator\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0072_00/00000.jpg", "scene0072_00/00040.jpg", "scene0072_00/00080.jpg", "scene0072_00/00120.jpg" ], "turn_id": 1, "scan_id": "scene0072_00", "origin_question": "Remember, have you seen any drawer(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any drawer(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0072_00/00160.jpg", "scene0072_00/00200.jpg", "scene0072_00/00240.jpg", "scene0072_00/00280.jpg" ], "turn_id": 2, "scan_id": "scene0072_00", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the window\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the window\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0072_00/00320.jpg", "scene0072_00/00360.jpg", "scene0072_00/00400.jpg", "scene0072_00/00440.jpg" ], "turn_id": 3, "scan_id": "scene0072_00", "origin_question": "Which direction is the curtain in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the curtain in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0072_00/00480.jpg", "scene0072_00/00520.jpg", "scene0072_00/00560.jpg", "scene0072_00/00600.jpg" ], "turn_id": 4, "scan_id": "scene0072_00", "origin_question": "Is the rectangular bed covered with a green cotton blanket to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the rectangular bed covered with a green cotton blanket to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0072_00/00640.jpg", "scene0072_00/00680.jpg", "scene0072_00/00720.jpg", "scene0072_00/00760.jpg" ], "turn_id": 5, "scan_id": "scene0072_00", "origin_question": "Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the small white rectangular plastic switch\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the small white rectangular plastic switch\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0072_00/00800.jpg", "scene0072_00/00840.jpg", "scene0072_00/00880.jpg", "scene0072_00/00920.jpg" ], "turn_id": 6, "scan_id": "scene0072_00", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "1.29", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0072_00/00960.jpg", "scene0072_00/01000.jpg", "scene0072_00/01040.jpg", "scene0072_00/01080.jpg" ], "turn_id": 7, "scan_id": "scene0072_00", "origin_question": "Is the small white rectangular shelf made of lightweight material to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the small white rectangular shelf made of lightweight material to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0057_region21/f0fea79c1a36425b9979f781ecd4d092_i1_2.jpg", "1mp3d_0057_region21/f0fea79c1a36425b9979f781ecd4d092_i2_2.jpg", "1mp3d_0057_region21/f0fea79c1a36425b9979f781ecd4d092_i1_1.jpg", "1mp3d_0057_region21/f0fea79c1a36425b9979f781ecd4d092_i2_1.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0057_region21", "origin_question": "Remember, have you seen any faucet(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any faucet(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 3 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0057_region21/1ad77d6ebc304b7894e4608ac7738b8a_i2_0.jpg", "1mp3d_0057_region21/1ad77d6ebc304b7894e4608ac7738b8a_i2_1.jpg", "1mp3d_0057_region21/1ad77d6ebc304b7894e4608ac7738b8a_i2_5.jpg", "1mp3d_0057_region21/6d0b2a9abbed4e6b812b898f027fc9f9_i1_1.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0057_region21", "origin_question": "Remember, have you seen any towel(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any towel(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 3 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "1mp3d_0057_region21/6d0b2a9abbed4e6b812b898f027fc9f9_i2_1.jpg", "1mp3d_0057_region21/6d0b2a9abbed4e6b812b898f027fc9f9_i1_0.jpg", "1mp3d_0057_region21/6d0b2a9abbed4e6b812b898f027fc9f9_i2_0.jpg", "1mp3d_0057_region21/f5a31e39d8fa40e89bbbc110bf25de2c_i1_5.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0057_region21", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "1.78", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 3 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0598_02/00000.jpg", "scene0598_02/00040.jpg", "scene0598_02/00080.jpg", "scene0598_02/00120.jpg", "scene0598_02/00160.jpg" ], "turn_id": 1, "scan_id": "scene0598_02", "origin_question": "Remember, have you seen any shelf(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any shelf(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0598_02/00200.jpg", "scene0598_02/00240.jpg", "scene0598_02/00280.jpg", "scene0598_02/00320.jpg", "scene0598_02/00360.jpg" ], "turn_id": 2, "scan_id": "scene0598_02", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0598_02/00400.jpg", "scene0598_02/00440.jpg", "scene0598_02/00480.jpg", "scene0598_02/00520.jpg", "scene0598_02/00560.jpg" ], "turn_id": 3, "scan_id": "scene0598_02", "origin_question": "What will be the first-time appearance order of the following categories: door, shelf, stool?", "option": [ "shelf, stool, door", "stool, shelf, door", "stool, door, shelf", "door, shelf, stool" ], "answer": "shelf, stool, door", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: door, shelf, stool? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['shelf, stool, door', 'stool, shelf, door', 'stool, door, shelf', 'door, shelf, stool'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "scene0598_02/00600.jpg", "scene0598_02/00640.jpg", "scene0598_02/00680.jpg", "scene0598_02/00720.jpg", "scene0598_02/00760.jpg" ], "turn_id": 4, "scan_id": "scene0598_02", "origin_question": "Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Left", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0598_02/00800.jpg", "scene0598_02/00840.jpg", "scene0598_02/00880.jpg", "scene0598_02/00920.jpg", "scene0598_02/00960.jpg" ], "turn_id": 5, "scan_id": "scene0598_02", "origin_question": "Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0598_02/01000.jpg", "scene0598_02/01040.jpg", "scene0598_02/01080.jpg", "scene0598_02/01120.jpg", "scene0598_02/01160.jpg" ], "turn_id": 6, "scan_id": "scene0598_02", "origin_question": "Remember, have you seen any book(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any book(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0598_02/01200.jpg", "scene0598_02/01240.jpg", "scene0598_02/01280.jpg", "scene0598_02/01320.jpg", "scene0598_02/01360.jpg" ], "turn_id": 7, "scan_id": "scene0598_02", "origin_question": "Please recall \"the door\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.15", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the door\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0598_02/01400.jpg", "scene0598_02/01440.jpg", "scene0598_02/01480.jpg", "scene0598_02/01520.jpg", "scene0598_02/01560.jpg" ], "turn_id": 8, "scan_id": "scene0598_02", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0040_region10/91e99aaa58e548a4b9db01e0f6863567_i1_1.jpg", "1mp3d_0040_region10/91e99aaa58e548a4b9db01e0f6863567_i2_1.jpg", "1mp3d_0040_region10/0f8ff514226040b6bb8c1a2c9cb168f2_i1_5.jpg", "1mp3d_0040_region10/0f8ff514226040b6bb8c1a2c9cb168f2_i2_5.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0040_region10", "origin_question": "Remember, have you seen any drum(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any drum(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "1mp3d_0040_region10/649674968430426d81caa4ab2e2b60d5_i1_4.jpg", "1mp3d_0040_region10/649674968430426d81caa4ab2e2b60d5_i2_4.jpg", "1mp3d_0040_region10/cdd11a8615a140ebbe4d0177d5aec1ab_i1_3.jpg", "1mp3d_0040_region10/cdd11a8615a140ebbe4d0177d5aec1ab_i2_3.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0040_region10", "origin_question": "In which turn did you see both \"the large rectangular window with a black frame\" and \"the rectangular wooden bed covered with a blue blanket\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the large rectangular window with a black frame\" and \"the rectangular wooden bed covered with a blue blanket\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "1mp3d_0040_region10/cdd11a8615a140ebbe4d0177d5aec1ab_i1_4.jpg", "1mp3d_0040_region10/cdd11a8615a140ebbe4d0177d5aec1ab_i2_4.jpg", "1mp3d_0040_region10/649674968430426d81caa4ab2e2b60d5_i1_2.jpg", "1mp3d_0040_region10/649674968430426d81caa4ab2e2b60d5_i2_2.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0040_region10", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the modern round black stool with a silver metal frame\"; \"the rectangular wooden bed covered with a blue blanket\"; \"the red antique-style car wall art piece\" ", "option": [ "the modern round black stool with a silver metal frame", "the rectangular wooden bed covered with a blue blanket", "the red antique-style car wall art piece" ], "answer": "the red antique-style car wall art piece", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the modern round black stool with a silver metal frame\"; \"the rectangular wooden bed covered with a blue blanket\"; \"the red antique-style car wall art piece\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the modern round black stool with a silver metal frame', 'the rectangular wooden bed covered with a blue blanket', 'the red antique-style car wall art piece'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "1mp3d_0040_region10/0f8ff514226040b6bb8c1a2c9cb168f2_i1_3.jpg", "1mp3d_0040_region10/0f8ff514226040b6bb8c1a2c9cb168f2_i2_3.jpg", "1mp3d_0040_region10/0f8ff514226040b6bb8c1a2c9cb168f2_i1_4.jpg", "1mp3d_0040_region10/0f8ff514226040b6bb8c1a2c9cb168f2_i2_4.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0040_region10", "origin_question": "Which two objects are on the same side of you now? \"the modern round black stool with a silver metal frame\", \"the rectangular wooden bed covered with a blue blanket\", \"the square picture in brown and beige tones\". ", "option": [ "the modern round black stool with a silver metal frame, the rectangular wooden bed covered with a blue blanket", "the modern round black stool with a silver metal frame, the square picture in brown and beige tones", "the rectangular wooden bed covered with a blue blanket, the square picture in brown and beige tones" ], "answer": "the rectangular wooden bed covered with a blue blanket, the square picture in brown and beige tones", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the modern round black stool with a silver metal frame\", \"the rectangular wooden bed covered with a blue blanket\", \"the square picture in brown and beige tones\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the modern round black stool with a silver metal frame, the rectangular wooden bed covered with a blue blanket', 'the modern round black stool with a silver metal frame, the square picture in brown and beige tones', 'the rectangular wooden bed covered with a blue blanket, the square picture in brown and beige tones'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "1mp3d_0040_region10/91e99aaa58e548a4b9db01e0f6863567_i1_5.jpg", "1mp3d_0040_region10/91e99aaa58e548a4b9db01e0f6863567_i2_5.jpg", "1mp3d_0040_region10/91e99aaa58e548a4b9db01e0f6863567_i1_4.jpg", "1mp3d_0040_region10/91e99aaa58e548a4b9db01e0f6863567_i2_4.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0040_region10", "origin_question": "Which direction is the modern round black stool with a silver metal frame in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the modern round black stool with a silver metal frame in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0196_00/00000.jpg", "scene0196_00/00040.jpg", "scene0196_00/00080.jpg", "scene0196_00/00120.jpg", "scene0196_00/00160.jpg" ], "turn_id": 1, "scan_id": "scene0196_00", "origin_question": "Remember, have you seen any bin(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any bin(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0196_00/00200.jpg", "scene0196_00/00240.jpg", "scene0196_00/00280.jpg", "scene0196_00/00320.jpg", "scene0196_00/00360.jpg" ], "turn_id": 2, "scan_id": "scene0196_00", "origin_question": "When did you first discover \"the large black cylindrical bin with a shiny surface\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the large black cylindrical bin with a shiny surface\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0196_00/00400.jpg", "scene0196_00/00440.jpg", "scene0196_00/00480.jpg", "scene0196_00/00520.jpg", "scene0196_00/00560.jpg" ], "turn_id": 3, "scan_id": "scene0196_00", "origin_question": "In which turn were you closest to \"the red and black nylon backpack with a zipper closure\"? ", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the red and black nylon backpack with a zipper closure\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0196_00/00600.jpg", "scene0196_00/00640.jpg", "scene0196_00/00680.jpg", "scene0196_00/00720.jpg", "scene0196_00/00760.jpg" ], "turn_id": 4, "scan_id": "scene0196_00", "origin_question": "In which turn did you see both \"the standard-sized, vertical brown wooden door\" and \"the large black cylindrical bin with a shiny surface\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the standard-sized, vertical brown wooden door\" and \"the large black cylindrical bin with a shiny surface\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0196_00/00800.jpg", "scene0196_00/00840.jpg", "scene0196_00/00880.jpg", "scene0196_00/00920.jpg", "scene0196_00/00960.jpg" ], "turn_id": 5, "scan_id": "scene0196_00", "origin_question": "Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0196_00/01000.jpg", "scene0196_00/01040.jpg", "scene0196_00/01080.jpg", "scene0196_00/01120.jpg", "scene0196_00/01160.jpg" ], "turn_id": 6, "scan_id": "scene0196_00", "origin_question": "Using your orientation at the end of turn 5 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "35.55", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 5 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0196_00/01200.jpg", "scene0196_00/01240.jpg", "scene0196_00/01340.jpg", "scene0196_00/01380.jpg", "scene0196_00/01420.jpg" ], "turn_id": 7, "scan_id": "scene0196_00", "origin_question": "Using your orientation at the end of turn 6 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "155.3", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 6 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0196_00/01460.jpg", "scene0196_00/01500.jpg", "scene0196_00/01540.jpg", "scene0196_00/01580.jpg", "scene0196_00/01620.jpg" ], "turn_id": 8, "scan_id": "scene0196_00", "origin_question": "Please recall \"the red and black nylon backpack with a zipper closure\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.11", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the red and black nylon backpack with a zipper closure\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0004_region19/10f096d877cc45f79f47f82bcc3e7ffe_i1_5.jpg", "1mp3d_0004_region19/10f096d877cc45f79f47f82bcc3e7ffe_i2_5.jpg", "1mp3d_0004_region19/10f096d877cc45f79f47f82bcc3e7ffe_i1_4.jpg", "1mp3d_0004_region19/10f096d877cc45f79f47f82bcc3e7ffe_i2_4.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0004_region19", "origin_question": "Remember, have you seen any picture(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any picture(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0004_region19/990f7ecd608442cfac24fa3e0db8e612_i1_4.jpg", "1mp3d_0004_region19/990f7ecd608442cfac24fa3e0db8e612_i2_4.jpg", "1mp3d_0004_region19/ad4a92079e654de79e28c6cb45d6c2a4_i1_0.jpg", "1mp3d_0004_region19/ad4a92079e654de79e28c6cb45d6c2a4_i2_0.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0004_region19", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "1mp3d_0004_region19/5fa2885d9a1747c4b557fd346ce46707_i1_2.jpg", "1mp3d_0004_region19/5fa2885d9a1747c4b557fd346ce46707_i2_2.jpg", "1mp3d_0004_region19/bd5ad4194bb44a6794529278f17e1185_i1_1.jpg", "1mp3d_0004_region19/bd5ad4194bb44a6794529278f17e1185_i2_1.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0004_region19", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "1.25", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0004_region19/5fa2885d9a1747c4b557fd346ce46707_i1_3.jpg", "1mp3d_0004_region19/5fa2885d9a1747c4b557fd346ce46707_i2_3.jpg", "1mp3d_0004_region19/ad4a92079e654de79e28c6cb45d6c2a4_i1_1.jpg", "1mp3d_0004_region19/ad4a92079e654de79e28c6cb45d6c2a4_i2_1.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0004_region19", "origin_question": "", "option": [], "answer": "", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0004_region19/ad4a92079e654de79e28c6cb45d6c2a4_i1_2.jpg", "1mp3d_0004_region19/990f7ecd608442cfac24fa3e0db8e612_i1_2.jpg", "1mp3d_0004_region19/990f7ecd608442cfac24fa3e0db8e612_i2_2.jpg", "1mp3d_0004_region19/990f7ecd608442cfac24fa3e0db8e612_i1_1.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0004_region19", "origin_question": "Remember, have you seen any picture(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any picture(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_45663105/Validation_45663105_45663105_54744.357.png", "Validation_45663105/Validation_45663105_45663105_54746.356.png", "Validation_45663105/Validation_45663105_45663105_54748.355.png", "Validation_45663105/Validation_45663105_45663105_54750.354.png" ], "turn_id": 1, "scan_id": "Validation_45663105", "origin_question": "Remember, have you seen any plant(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any plant(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_45663105/Validation_45663105_45663105_54752.353.png", "Validation_45663105/Validation_45663105_45663105_54754.353.png", "Validation_45663105/Validation_45663105_45663105_54756.352.png", "Validation_45663105/Validation_45663105_45663105_54758.351.png" ], "turn_id": 2, "scan_id": "Validation_45663105", "origin_question": "Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance forward or backward from that position?", "option": [ "Forward", "Backward" ], "answer": "Forward", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance forward or backward from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Forward', 'Backward'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_45663105/Validation_45663105_45663105_54760.350.png", "Validation_45663105/Validation_45663105_45663105_54762.349.png", "Validation_45663105/Validation_45663105_45663105_54764.349.png", "Validation_45663105/Validation_45663105_45663105_54766.348.png" ], "turn_id": 3, "scan_id": "Validation_45663105", "origin_question": "", "option": [], "answer": "", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_45663105/Validation_45663105_45663105_54768.347.png", "Validation_45663105/Validation_45663105_45663105_54770.346.png", "Validation_45663105/Validation_45663105_45663105_54772.345.png", "Validation_45663105/Validation_45663105_45663105_54774.344.png" ], "turn_id": 4, "scan_id": "Validation_45663105", "origin_question": "How far is your current position from where you were at the end of the turn 1? (in meters)", "option": [], "answer": "1.97", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 1? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_45663105/Validation_45663105_45663105_54776.344.png", "Validation_45663105/Validation_45663105_45663105_54778.359.png", "Validation_45663105/Validation_45663105_45663105_54780.359.png", "Validation_45663105/Validation_45663105_45663105_54782.358.png" ], "turn_id": 5, "scan_id": "Validation_45663105", "origin_question": "Please recall \"the compact vertical white steel radiator\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.08", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the compact vertical white steel radiator\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_45663105/Validation_45663105_45663105_54784.357.png", "Validation_45663105/Validation_45663105_45663105_54786.356.png", "Validation_45663105/Validation_45663105_45663105_54788.355.png", "Validation_45663105/Validation_45663105_45663105_54790.354.png" ], "turn_id": 6, "scan_id": "Validation_45663105", "origin_question": "Please recall \"the large rectangular couch in warm brown fabric\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "3.42", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the large rectangular couch in warm brown fabric\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47895541/Validation_47895541_47895541_455510.552.png", "Validation_47895541/Validation_47895541_47895541_455512.034.png", "Validation_47895541/Validation_47895541_47895541_455513.534.png", "Validation_47895541/Validation_47895541_47895541_455515.033.png" ], "turn_id": 1, "scan_id": "Validation_47895541", "origin_question": "", "option": [], "answer": "", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47895541/Validation_47895541_47895541_455516.549.png", "Validation_47895541/Validation_47895541_47895541_455518.049.png", "Validation_47895541/Validation_47895541_47895541_455519.548.png", "Validation_47895541/Validation_47895541_47895541_455521.047.png" ], "turn_id": 2, "scan_id": "Validation_47895541", "origin_question": "Remember, have you seen any jacket(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any jacket(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_47895541/Validation_47895541_47895541_455522.547.png", "Validation_47895541/Validation_47895541_47895541_455524.046.png", "Validation_47895541/Validation_47895541_47895541_455525.546.png", "Validation_47895541/Validation_47895541_47895541_455527.045.png" ], "turn_id": 3, "scan_id": "Validation_47895541", "origin_question": "How far is your current position from where you were at the end of the turn 1? (in meters)", "option": [], "answer": "1.87", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 1? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_47895541/Validation_47895541_47895541_455528.544.png", "Validation_47895541/Validation_47895541_47895541_455530.044.png", "Validation_47895541/Validation_47895541_47895541_455531.543.png", "Validation_47895541/Validation_47895541_47895541_455533.042.png" ], "turn_id": 4, "scan_id": "Validation_47895541", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47895541/Validation_47895541_47895541_455534.542.png", "Validation_47895541/Validation_47895541_47895541_455536.041.png", "Validation_47895541/Validation_47895541_47895541_455537.541.png", "Validation_47895541/Validation_47895541_47895541_455539.040.png" ], "turn_id": 5, "scan_id": "Validation_47895541", "origin_question": "Which direction is the window in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the window in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_47895541/Validation_47895541_47895541_455540.540.png", "Validation_47895541/Validation_47895541_47895541_455542.039.png", "Validation_47895541/Validation_47895541_47895541_455543.538.png", "Validation_47895541/Validation_47895541_47895541_455545.038.png" ], "turn_id": 6, "scan_id": "Validation_47895541", "origin_question": "Is the window to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the window to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_47895541/Validation_47895541_47895541_455546.537.png", "Validation_47895541/Validation_47895541_47895541_455548.036.png", "Validation_47895541/Validation_47895541_47895541_455549.536.png", "Validation_47895541/Validation_47895541_47895541_455551.035.png" ], "turn_id": 7, "scan_id": "Validation_47895541", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the sleek white steel radiator\" from your position?", "option": [], "answer": "140.7", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the sleek white steel radiator\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42444976/Validation_42444976_42444976_40141.410.png", "Validation_42444976/Validation_42444976_42444976_40143.393.png", "Validation_42444976/Validation_42444976_42444976_40145.392.png", "Validation_42444976/Validation_42444976_42444976_40147.391.png", "Validation_42444976/Validation_42444976_42444976_40149.390.png" ], "turn_id": 1, "scan_id": "Validation_42444976", "origin_question": "Remember, have you seen any range hood(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any range hood(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_42444976/Validation_42444976_42444976_40151.389.png", "Validation_42444976/Validation_42444976_42444976_40153.389.png", "Validation_42444976/Validation_42444976_42444976_40155.388.png", "Validation_42444976/Validation_42444976_42444976_40157.387.png", "Validation_42444976/Validation_42444976_42444976_40159.386.png" ], "turn_id": 2, "scan_id": "Validation_42444976", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the sleek white front-loading washing machine\" from your position?", "option": [], "answer": "85.07", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the sleek white front-loading washing machine\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_42444976/Validation_42444976_42444976_40161.385.png", "Validation_42444976/Validation_42444976_42444976_40163.384.png", "Validation_42444976/Validation_42444976_42444976_40165.384.png", "Validation_42444976/Validation_42444976_42444976_40167.383.png", "Validation_42444976/Validation_42444976_42444976_40169.382.png" ], "turn_id": 3, "scan_id": "Validation_42444976", "origin_question": "In which turn did you see both \"the contemporary rectangular metallic oven with the black finish\" and \"the sleek white wooden door with the metallic handle\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the contemporary rectangular metallic oven with the black finish\" and \"the sleek white wooden door with the metallic handle\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_42444976/Validation_42444976_42444976_40171.398.png", "Validation_42444976/Validation_42444976_42444976_40173.397.png", "Validation_42444976/Validation_42444976_42444976_40175.396.png", "Validation_42444976/Validation_42444976_42444976_40177.395.png", "Validation_42444976/Validation_42444976_42444976_40179.395.png" ], "turn_id": 4, "scan_id": "Validation_42444976", "origin_question": "Which direction is the sleek white wooden door with the metallic handle in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the sleek white wooden door with the metallic handle in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_42444976/Validation_42444976_42444976_40181.394.png", "Validation_42444976/Validation_42444976_42444976_40183.393.png", "Validation_42444976/Validation_42444976_42444976_40185.392.png", "Validation_42444976/Validation_42444976_42444976_40187.391.png", "Validation_42444976/Validation_42444976_42444976_40189.390.png" ], "turn_id": 5, "scan_id": "Validation_42444976", "origin_question": "Which direction is the contemporary stainless steel rectangular range hood in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the contemporary stainless steel rectangular range hood in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "Validation_42444976/Validation_42444976_42444976_40191.389.png", "Validation_42444976/Validation_42444976_42444976_40193.389.png", "Validation_42444976/Validation_42444976_42444976_40195.388.png", "Validation_42444976/Validation_42444976_42444976_40197.387.png", "Validation_42444976/Validation_42444976_42444976_40199.386.png" ], "turn_id": 6, "scan_id": "Validation_42444976", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the cylindrical chandelier with the soft gold color\"; \"the sleek white front-loading washing machine\"; \"the rectangular light gray carpet with dark gray patterns\" ", "option": [ "the cylindrical chandelier with the soft gold color", "the sleek white front-loading washing machine", "the rectangular light gray carpet with dark gray patterns" ], "answer": "the cylindrical chandelier with the soft gold color", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the cylindrical chandelier with the soft gold color\"; \"the sleek white front-loading washing machine\"; \"the rectangular light gray carpet with dark gray patterns\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the cylindrical chandelier with the soft gold color', 'the sleek white front-loading washing machine', 'the rectangular light gray carpet with dark gray patterns'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "Validation_42444976/Validation_42444976_42444976_40201.385.png", "Validation_42444976/Validation_42444976_42444976_40203.385.png", "Validation_42444976/Validation_42444976_42444976_40205.384.png", "Validation_42444976/Validation_42444976_42444976_40207.383.png", "Validation_42444976/Validation_42444976_42444976_40209.382.png" ], "turn_id": 7, "scan_id": "Validation_42444976", "origin_question": "Among these three objects(\"the sleek white front-loading washing machine\",\"the silver stainless steel kettle with the sleek cylindrical shape\",\"the contemporary stainless steel rectangular range hood\" ), which one is the closest to you?", "option": [ "the sleek white front-loading washing machine", "the silver stainless steel kettle with the sleek cylindrical shape", "the contemporary stainless steel rectangular range hood" ], "answer": "the sleek white front-loading washing machine", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the sleek white front-loading washing machine\",\"the silver stainless steel kettle with the sleek cylindrical shape\",\"the contemporary stainless steel rectangular range hood\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the sleek white front-loading washing machine', 'the silver stainless steel kettle with the sleek cylindrical shape', 'the contemporary stainless steel rectangular range hood'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_42444976/Validation_42444976_42444976_40211.381.png", "Validation_42444976/Validation_42444976_42444976_40213.397.png", "Validation_42444976/Validation_42444976_42444976_40215.396.png", "Validation_42444976/Validation_42444976_42444976_40217.395.png", "Validation_42444976/Validation_42444976_42444976_40219.395.png" ], "turn_id": 8, "scan_id": "Validation_42444976", "origin_question": "What will be the first-time appearance order of the following categories: carpet, refrigerator, washing machine?", "option": [ "refrigerator, washing machine, carpet", "refrigerator, carpet, washing machine", "carpet, washing machine, refrigerator", "washing machine, refrigerator, carpet" ], "answer": "washing machine, refrigerator, carpet", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: carpet, refrigerator, washing machine? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['refrigerator, washing machine, carpet', 'refrigerator, carpet, washing machine', 'carpet, washing machine, refrigerator', 'washing machine, refrigerator, carpet'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "Validation_42444976/Validation_42444976_42444976_40221.394.png", "Validation_42444976/Validation_42444976_42444976_40223.393.png", "Validation_42444976/Validation_42444976_42444976_40225.392.png", "Validation_42444976/Validation_42444976_42444976_40227.391.png", "Validation_42444976/Validation_42444976_42444976_40229.390.png" ], "turn_id": 9, "scan_id": "Validation_42444976", "origin_question": "Among these three objects(\"the refrigerator\",\"the silver stainless steel kettle with the sleek cylindrical shape\",\"the cylindrical chandelier with the soft gold color\" ), which one is the closest to you?", "option": [ "the refrigerator", "the silver stainless steel kettle with the sleek cylindrical shape", "the cylindrical chandelier with the soft gold color" ], "answer": "the cylindrical chandelier with the soft gold color", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the refrigerator\",\"the silver stainless steel kettle with the sleek cylindrical shape\",\"the cylindrical chandelier with the soft gold color\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the refrigerator', 'the silver stainless steel kettle with the sleek cylindrical shape', 'the cylindrical chandelier with the soft gold color'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47331337/Validation_47331337_47331337_183471.438.png", "Validation_47331337/Validation_47331337_47331337_183472.921.png", "Validation_47331337/Validation_47331337_47331337_183474.420.png", "Validation_47331337/Validation_47331337_47331337_183475.919.png" ], "turn_id": 1, "scan_id": "Validation_47331337", "origin_question": "Remember, have you seen any letter(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any letter(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47331337/Validation_47331337_47331337_183477.419.png", "Validation_47331337/Validation_47331337_47331337_183478.918.png", "Validation_47331337/Validation_47331337_47331337_183480.418.png", "Validation_47331337/Validation_47331337_47331337_183481.917.png" ], "turn_id": 2, "scan_id": "Validation_47331337", "origin_question": "In which turn were you farthest from \"the sleek, rounded white porcelain toilet\"? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the sleek, rounded white porcelain toilet\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_47331337/Validation_47331337_47331337_183483.433.png", "Validation_47331337/Validation_47331337_47331337_183484.932.png", "Validation_47331337/Validation_47331337_47331337_183486.432.png", "Validation_47331337/Validation_47331337_47331337_183487.931.png" ], "turn_id": 3, "scan_id": "Validation_47331337", "origin_question": "Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "88.89", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_47331337/Validation_47331337_47331337_183489.431.png", "Validation_47331337/Validation_47331337_47331337_183490.930.png", "Validation_47331337/Validation_47331337_47331337_183492.429.png", "Validation_47331337/Validation_47331337_47331337_183493.929.png" ], "turn_id": 4, "scan_id": "Validation_47331337", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_47331337/Validation_47331337_47331337_183495.428.png", "Validation_47331337/Validation_47331337_47331337_183496.927.png", "Validation_47331337/Validation_47331337_47331337_183498.427.png", "Validation_47331337/Validation_47331337_47331337_183499.926.png" ], "turn_id": 5, "scan_id": "Validation_47331337", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47331337/Validation_47331337_47331337_183501.426.png", "Validation_47331337/Validation_47331337_47331337_183502.925.png", "Validation_47331337/Validation_47331337_47331337_183504.424.png", "Validation_47331337/Validation_47331337_47331337_183505.924.png" ], "turn_id": 6, "scan_id": "Validation_47331337", "origin_question": "In which turn were you closest to \"the glossy white oval sink made of porcelain\"? ", "option": [], "answer": "6", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the glossy white oval sink made of porcelain\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_47331337/Validation_47331337_47331337_183507.423.png", "Validation_47331337/Validation_47331337_47331337_183508.923.png", "Validation_47331337/Validation_47331337_47331337_183510.422.png", "Validation_47331337/Validation_47331337_47331337_183511.921.png" ], "turn_id": 7, "scan_id": "Validation_47331337", "origin_question": "Using your orientation at the end of turn 5 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "60.78", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 5 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_47331337/Validation_47331337_47331337_183513.421.png", "Validation_47331337/Validation_47331337_47331337_183514.920.png", "Validation_47331337/Validation_47331337_47331337_183516.420.png", "Validation_47331337/Validation_47331337_47331337_183517.919.png" ], "turn_id": 8, "scan_id": "Validation_47331337", "origin_question": "Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "147.31", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42899737/Validation_42899737_42899737_801525.249.png", "Validation_42899737/Validation_42899737_42899737_801527.232.png", "Validation_42899737/Validation_42899737_42899737_801529.231.png", "Validation_42899737/Validation_42899737_42899737_801531.230.png", "Validation_42899737/Validation_42899737_42899737_801533.229.png" ], "turn_id": 1, "scan_id": "Validation_42899737", "origin_question": "Remember, have you seen any cosmetics(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any cosmetics(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_42899737/Validation_42899737_42899737_801535.228.png", "Validation_42899737/Validation_42899737_42899737_801537.227.png", "Validation_42899737/Validation_42899737_42899737_801539.227.png", "Validation_42899737/Validation_42899737_42899737_801541.242.png", "Validation_42899737/Validation_42899737_42899737_801543.242.png" ], "turn_id": 2, "scan_id": "Validation_42899737", "origin_question": "In which turn did you see both \"the rectangular double bed with the wooden frame\" and \"the classic white, neatly folded cotton towel\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the rectangular double bed with the wooden frame\" and \"the classic white, neatly folded cotton towel\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_42899737/Validation_42899737_42899737_801545.241.png", "Validation_42899737/Validation_42899737_42899737_801547.240.png", "Validation_42899737/Validation_42899737_42899737_801549.239.png", "Validation_42899737/Validation_42899737_42899737_801551.238.png", "Validation_42899737/Validation_42899737_42899737_801553.237.png" ], "turn_id": 3, "scan_id": "Validation_42899737", "origin_question": "In which turn were you closest to \"the rectangular double bed with the wooden frame\"? ", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the rectangular double bed with the wooden frame\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_42899737/Validation_42899737_42899737_801555.237.png", "Validation_42899737/Validation_42899737_42899737_801557.236.png", "Validation_42899737/Validation_42899737_42899737_801559.235.png", "Validation_42899737/Validation_42899737_42899737_801561.234.png", "Validation_42899737/Validation_42899737_42899737_801563.233.png" ], "turn_id": 4, "scan_id": "Validation_42899737", "origin_question": "When did you first discover \"the modern lamp with the cylindrical neutral-toned shade\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the modern lamp with the cylindrical neutral-toned shade\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_42899737/Validation_42899737_42899737_801565.232.png", "Validation_42899737/Validation_42899737_42899737_801567.232.png", "Validation_42899737/Validation_42899737_42899737_801569.231.png", "Validation_42899737/Validation_42899737_42899737_801571.230.png", "Validation_42899737/Validation_42899737_42899737_801573.229.png" ], "turn_id": 5, "scan_id": "Validation_42899737", "origin_question": "", "option": [], "answer": "", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_42899737/Validation_42899737_42899737_801575.228.png", "Validation_42899737/Validation_42899737_42899737_801577.228.png", "Validation_42899737/Validation_42899737_42899737_801579.227.png", "Validation_42899737/Validation_42899737_42899737_801581.242.png", "Validation_42899737/Validation_42899737_42899737_801583.242.png" ], "turn_id": 6, "scan_id": "Validation_42899737", "origin_question": "Assuming the direction you are facing at the end of the turn 3 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Right", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 3 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_42899737/Validation_42899737_42899737_801585.241.png", "Validation_42899737/Validation_42899737_42899737_801587.240.png", "Validation_42899737/Validation_42899737_42899737_801589.239.png", "Validation_42899737/Validation_42899737_42899737_801591.238.png", "Validation_42899737/Validation_42899737_42899737_801593.238.png" ], "turn_id": 7, "scan_id": "Validation_42899737", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "38.57", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_42899737/Validation_42899737_42899737_801595.237.png", "Validation_42899737/Validation_42899737_42899737_801597.236.png", "Validation_42899737/Validation_42899737_42899737_801599.235.png", "Validation_42899737/Validation_42899737_42899737_801601.234.png", "Validation_42899737/Validation_42899737_42899737_801603.233.png" ], "turn_id": 8, "scan_id": "Validation_42899737", "origin_question": "How far is your current position from where you were at the end of the turn 6? (in meters)", "option": [], "answer": "3.75", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 6? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_41069050/Validation_41069050_41069050_5045.334.png", "Validation_41069050/Validation_41069050_41069050_5046.816.png", "Validation_41069050/Validation_41069050_41069050_5048.316.png", "Validation_41069050/Validation_41069050_41069050_5049.815.png" ], "turn_id": 1, "scan_id": "Validation_41069050", "origin_question": "Remember, have you seen any door(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any door(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_41069050/Validation_41069050_41069050_5051.315.png", "Validation_41069050/Validation_41069050_41069050_5052.814.png", "Validation_41069050/Validation_41069050_41069050_5054.313.png", "Validation_41069050/Validation_41069050_41069050_5055.813.png" ], "turn_id": 2, "scan_id": "Validation_41069050", "origin_question": "In which turn did you see both \"the sleek, modern toilet with the glossy white finish\" and \"the soft, absorbent white cotton towel\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the sleek, modern toilet with the glossy white finish\" and \"the soft, absorbent white cotton towel\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_41069050/Validation_41069050_41069050_5057.312.png", "Validation_41069050/Validation_41069050_41069050_5058.811.png", "Validation_41069050/Validation_41069050_41069050_5060.311.png", "Validation_41069050/Validation_41069050_41069050_5061.810.png" ], "turn_id": 3, "scan_id": "Validation_41069050", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_41069050/Validation_41069050_41069050_5063.310.png", "Validation_41069050/Validation_41069050_41069050_5064.809.png", "Validation_41069050/Validation_41069050_41069050_5066.308.png", "Validation_41069050/Validation_41069050_41069050_5067.808.png" ], "turn_id": 4, "scan_id": "Validation_41069050", "origin_question": "Is the soft, absorbent white cotton towel to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the soft, absorbent white cotton towel to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_41069050/Validation_41069050_41069050_5069.307.png", "Validation_41069050/Validation_41069050_41069050_5070.806.png", "Validation_41069050/Validation_41069050_41069050_5072.306.png", "Validation_41069050/Validation_41069050_41069050_5073.805.png" ], "turn_id": 5, "scan_id": "Validation_41069050", "origin_question": "Which two objects are on the same side of you now? \"the medium angular glossy white ceramic sink\", \"the cylindrical white toilet paper roll\", \"the compact, polished stainless steel bin\". ", "option": [ "the medium angular glossy white ceramic sink, the cylindrical white toilet paper roll", "the medium angular glossy white ceramic sink, the compact, polished stainless steel bin", "the cylindrical white toilet paper roll, the compact, polished stainless steel bin" ], "answer": "the medium angular glossy white ceramic sink, the cylindrical white toilet paper roll", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the medium angular glossy white ceramic sink\", \"the cylindrical white toilet paper roll\", \"the compact, polished stainless steel bin\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the medium angular glossy white ceramic sink, the cylindrical white toilet paper roll', 'the medium angular glossy white ceramic sink, the compact, polished stainless steel bin', 'the cylindrical white toilet paper roll, the compact, polished stainless steel bin'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47331265/Validation_47331265_47331265_90013.317.png", "Validation_47331265/Validation_47331265_47331265_90015.299.png", "Validation_47331265/Validation_47331265_47331265_90017.299.png", "Validation_47331265/Validation_47331265_47331265_90019.298.png" ], "turn_id": 1, "scan_id": "Validation_47331265", "origin_question": "Remember, have you seen any flowerpot(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any flowerpot(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47331265/Validation_47331265_47331265_90021.297.png", "Validation_47331265/Validation_47331265_47331265_90023.296.png", "Validation_47331265/Validation_47331265_47331265_90025.295.png", "Validation_47331265/Validation_47331265_47331265_90027.295.png" ], "turn_id": 2, "scan_id": "Validation_47331265", "origin_question": "In which turn were you farthest from \"the tall, dark wooden shelf with glass doors\"? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the tall, dark wooden shelf with glass doors\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "Validation_47331265/Validation_47331265_47331265_90029.294.png", "Validation_47331265/Validation_47331265_47331265_90031.293.png", "Validation_47331265/Validation_47331265_47331265_90033.292.png", "Validation_47331265/Validation_47331265_47331265_90035.291.png" ], "turn_id": 3, "scan_id": "Validation_47331265", "origin_question": "Among these three objects(\"the modern rectangular black matte monitor\",\"the sleek white metal radiator with ribbed panels\",\"the sleek cylindrical black flowerpot\" ), which one is the closest to you?", "option": [ "the modern rectangular black matte monitor", "the sleek white metal radiator with ribbed panels", "the sleek cylindrical black flowerpot" ], "answer": "the sleek white metal radiator with ribbed panels", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects(\"the modern rectangular black matte monitor\",\"the sleek white metal radiator with ribbed panels\",\"the sleek cylindrical black flowerpot\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the modern rectangular black matte monitor', 'the sleek white metal radiator with ribbed panels', 'the sleek cylindrical black flowerpot'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47331265/Validation_47331265_47331265_90037.291.png", "Validation_47331265/Validation_47331265_47331265_90039.290.png", "Validation_47331265/Validation_47331265_47331265_90041.289.png", "Validation_47331265/Validation_47331265_47331265_90043.305.png" ], "turn_id": 4, "scan_id": "Validation_47331265", "origin_question": "Which direction is the sleek cylindrical black flowerpot in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the sleek cylindrical black flowerpot in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_47331265/Validation_47331265_47331265_90045.304.png", "Validation_47331265/Validation_47331265_47331265_90047.303.png", "Validation_47331265/Validation_47331265_47331265_90049.302.png", "Validation_47331265/Validation_47331265_47331265_90051.302.png" ], "turn_id": 5, "scan_id": "Validation_47331265", "origin_question": "By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the modern rectangular black matte monitor\", Y: \"the tall, dark wooden shelf with glass doors\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "farther away from both X and Y", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the modern rectangular black matte monitor\", Y: \"the tall, dark wooden shelf with glass doors\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47331265/Validation_47331265_47331265_90053.301.png", "Validation_47331265/Validation_47331265_47331265_90055.300.png", "Validation_47331265/Validation_47331265_47331265_90057.299.png", "Validation_47331265/Validation_47331265_47331265_90059.298.png" ], "turn_id": 6, "scan_id": "Validation_47331265", "origin_question": "Which direction is the sleek rectangular bag made of water-resistant fabric in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the sleek rectangular bag made of water-resistant fabric in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_47331265/Validation_47331265_47331265_90061.298.png", "Validation_47331265/Validation_47331265_47331265_90063.297.png", "Validation_47331265/Validation_47331265_47331265_90065.296.png", "Validation_47331265/Validation_47331265_47331265_90067.295.png" ], "turn_id": 7, "scan_id": "Validation_47331265", "origin_question": "Is the sleek white metal radiator with ribbed panels to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the sleek white metal radiator with ribbed panels to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47331063/Validation_47331063_47331063_129401.163.png", "Validation_47331063/Validation_47331063_47331063_129403.145.png", "Validation_47331063/Validation_47331063_47331063_129405.144.png", "Validation_47331063/Validation_47331063_47331063_129407.144.png", "Validation_47331063/Validation_47331063_47331063_129409.143.png" ], "turn_id": 1, "scan_id": "Validation_47331063", "origin_question": "Remember, have you seen any bed(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any bed(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47331063/Validation_47331063_47331063_129411.142.png", "Validation_47331063/Validation_47331063_47331063_129413.158.png", "Validation_47331063/Validation_47331063_47331063_129415.157.png", "Validation_47331063/Validation_47331063_47331063_129417.156.png", "Validation_47331063/Validation_47331063_47331063_129419.155.png" ], "turn_id": 2, "scan_id": "Validation_47331063", "origin_question": "In which turn were you closest to \"the bed\"? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the bed\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_47331063/Validation_47331063_47331063_129421.155.png", "Validation_47331063/Validation_47331063_47331063_129423.154.png", "Validation_47331063/Validation_47331063_47331063_129425.153.png", "Validation_47331063/Validation_47331063_47331063_129427.152.png", "Validation_47331063/Validation_47331063_47331063_129429.151.png" ], "turn_id": 3, "scan_id": "Validation_47331063", "origin_question": "What will be the first-time appearance order of the following categories: bed, monitor, mop?", "option": [ "bed, mop, monitor", "monitor, bed, mop", "monitor, mop, bed", "bed, monitor, mop" ], "answer": "bed, mop, monitor", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bed, monitor, mop? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['bed, mop, monitor', 'monitor, bed, mop', 'monitor, mop, bed', 'bed, monitor, mop'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_47331063/Validation_47331063_47331063_129431.151.png", "Validation_47331063/Validation_47331063_47331063_129433.150.png", "Validation_47331063/Validation_47331063_47331063_129435.149.png", "Validation_47331063/Validation_47331063_47331063_129437.148.png", "Validation_47331063/Validation_47331063_47331063_129439.147.png" ], "turn_id": 4, "scan_id": "Validation_47331063", "origin_question": "Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "82.53", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47331063/Validation_47331063_47331063_129441.147.png", "Validation_47331063/Validation_47331063_47331063_129443.146.png", "Validation_47331063/Validation_47331063_47331063_129445.145.png", "Validation_47331063/Validation_47331063_47331063_129447.144.png", "Validation_47331063/Validation_47331063_47331063_129449.143.png" ], "turn_id": 5, "scan_id": "Validation_47331063", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the large, rectangular window with the light cream frame\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the large, rectangular window with the light cream frame\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_47331063/Validation_47331063_47331063_129451.143.png", "Validation_47331063/Validation_47331063_47331063_129453.142.png", "Validation_47331063/Validation_47331063_47331063_129455.158.png", "Validation_47331063/Validation_47331063_47331063_129457.157.png", "Validation_47331063/Validation_47331063_47331063_129459.156.png" ], "turn_id": 6, "scan_id": "Validation_47331063", "origin_question": "Assuming the direction you are facing at the end of the turn 3 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Left", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 3 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_47331063/Validation_47331063_47331063_129461.155.png", "Validation_47331063/Validation_47331063_47331063_129463.154.png", "Validation_47331063/Validation_47331063_47331063_129465.154.png", "Validation_47331063/Validation_47331063_47331063_129467.153.png", "Validation_47331063/Validation_47331063_47331063_129469.152.png" ], "turn_id": 7, "scan_id": "Validation_47331063", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the bed\" from your position?", "option": [], "answer": "132.56", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the bed\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47331063/Validation_47331063_47331063_129471.151.png", "Validation_47331063/Validation_47331063_47331063_129473.150.png", "Validation_47331063/Validation_47331063_47331063_129475.149.png", "Validation_47331063/Validation_47331063_47331063_129477.149.png", "Validation_47331063/Validation_47331063_47331063_129479.148.png" ], "turn_id": 8, "scan_id": "Validation_47331063", "origin_question": "", "option": [], "answer": "", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_47331063/Validation_47331063_47331063_129481.147.png", "Validation_47331063/Validation_47331063_47331063_129483.146.png", "Validation_47331063/Validation_47331063_47331063_129485.145.png", "Validation_47331063/Validation_47331063_47331063_129487.145.png", "Validation_47331063/Validation_47331063_47331063_129489.144.png" ], "turn_id": 9, "scan_id": "Validation_47331063", "origin_question": "At the end of which turn were both of \"the large, rectangular window with the light cream frame\" and \"the stainless steel oven with the metallic finish\" on your left side?", "option": [], "answer": "3", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the large, rectangular window with the light cream frame\" and \"the stainless steel oven with the metallic finish\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47331063/Validation_47331063_47331063_129491.143.png", "Validation_47331063/Validation_47331063_47331063_129493.142.png", "Validation_47331063/Validation_47331063_47331063_129495.158.png", "Validation_47331063/Validation_47331063_47331063_129497.157.png", "Validation_47331063/Validation_47331063_47331063_129499.156.png" ], "turn_id": 10, "scan_id": "Validation_47331063", "origin_question": "In which turn were you closest to \"the medium-sized metal and plastic microwave\"? ", "option": [], "answer": "6", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the medium-sized metal and plastic microwave\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42446116/Validation_42446116_42446116_349958.471.png", "Validation_42446116/Validation_42446116_42446116_349960.470.png", "Validation_42446116/Validation_42446116_42446116_349962.469.png", "Validation_42446116/Validation_42446116_42446116_349964.468.png", "Validation_42446116/Validation_42446116_42446116_349966.467.png" ], "turn_id": 1, "scan_id": "Validation_42446116", "origin_question": "Remember, have you seen any panel(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any panel(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_42446116/Validation_42446116_42446116_349968.466.png", "Validation_42446116/Validation_42446116_42446116_349970.466.png", "Validation_42446116/Validation_42446116_42446116_349972.465.png", "Validation_42446116/Validation_42446116_42446116_349974.464.png", "Validation_42446116/Validation_42446116_42446116_349976.463.png" ], "turn_id": 2, "scan_id": "Validation_42446116", "origin_question": "In which turn were you closest to \"an elegant, cylindrical white fabric chandelier\"? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"an elegant, cylindrical white fabric chandelier\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_42446116/Validation_42446116_42446116_349978.462.png", "Validation_42446116/Validation_42446116_42446116_349980.462.png", "Validation_42446116/Validation_42446116_42446116_349982.461.png", "Validation_42446116/Validation_42446116_42446116_349984.460.png", "Validation_42446116/Validation_42446116_42446116_349986.476.png" ], "turn_id": 3, "scan_id": "Validation_42446116", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"an elegant, cylindrical white fabric chandelier\" from your position?", "option": [], "answer": "105.61", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"an elegant, cylindrical white fabric chandelier\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_42446116/Validation_42446116_42446116_349988.475.png", "Validation_42446116/Validation_42446116_42446116_349990.474.png", "Validation_42446116/Validation_42446116_42446116_349992.473.png", "Validation_42446116/Validation_42446116_42446116_349994.472.png", "Validation_42446116/Validation_42446116_42446116_349996.472.png" ], "turn_id": 4, "scan_id": "Validation_42446116", "origin_question": "When did you first discover \"the sleek white rectangular cabinet with the closed door\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the sleek white rectangular cabinet with the closed door\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_42446116/Validation_42446116_42446116_349998.471.png", "Validation_42446116/Validation_42446116_42446116_350000.470.png", "Validation_42446116/Validation_42446116_42446116_350002.469.png", "Validation_42446116/Validation_42446116_42446116_350004.468.png", "Validation_42446116/Validation_42446116_42446116_350006.468.png" ], "turn_id": 5, "scan_id": "Validation_42446116", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_42446116/Validation_42446116_42446116_350008.467.png", "Validation_42446116/Validation_42446116_42446116_350010.466.png", "Validation_42446116/Validation_42446116_42446116_350012.465.png", "Validation_42446116/Validation_42446116_42446116_350014.464.png", "Validation_42446116/Validation_42446116_42446116_350016.464.png" ], "turn_id": 6, "scan_id": "Validation_42446116", "origin_question": "Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance forward or backward from that position?", "option": [ "Forward", "Backward" ], "answer": "Forward", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance forward or backward from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Forward', 'Backward'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_42446116/Validation_42446116_42446116_350018.463.png", "Validation_42446116/Validation_42446116_42446116_350020.462.png", "Validation_42446116/Validation_42446116_42446116_350022.461.png", "Validation_42446116/Validation_42446116_42446116_350024.460.png", "Validation_42446116/Validation_42446116_42446116_350026.459.png" ], "turn_id": 7, "scan_id": "Validation_42446116", "origin_question": "What will be the first-time appearance order of the following categories: bed, frame, window?", "option": [ "frame, window, bed", "frame, bed, window", "bed, frame, window", "bed, window, frame" ], "answer": "bed, frame, window", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bed, frame, window? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['frame, window, bed', 'frame, bed, window', 'bed, frame, window', 'bed, window, frame'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_42446116/Validation_42446116_42446116_350028.475.png", "Validation_42446116/Validation_42446116_42446116_350030.474.png", "Validation_42446116/Validation_42446116_42446116_350032.474.png", "Validation_42446116/Validation_42446116_42446116_350034.473.png", "Validation_42446116/Validation_42446116_42446116_350036.472.png" ], "turn_id": 8, "scan_id": "Validation_42446116", "origin_question": "Compared to your position at the end of the 6 turn, are you now closer or farther away from \"the bed\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 6 turn, are you now closer or farther away from \"the bed\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_42446116/Validation_42446116_42446116_350038.471.png", "Validation_42446116/Validation_42446116_42446116_350040.470.png", "Validation_42446116/Validation_42446116_42446116_350042.470.png", "Validation_42446116/Validation_42446116_42446116_350044.469.png", "Validation_42446116/Validation_42446116_42446116_350046.468.png" ], "turn_id": 9, "scan_id": "Validation_42446116", "origin_question": "", "option": [], "answer": "", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_42446116/Validation_42446116_42446116_350048.467.png", "Validation_42446116/Validation_42446116_42446116_350050.466.png", "Validation_42446116/Validation_42446116_42446116_350052.465.png", "Validation_42446116/Validation_42446116_42446116_350054.465.png", "Validation_42446116/Validation_42446116_42446116_350056.464.png" ], "turn_id": 10, "scan_id": "Validation_42446116", "origin_question": "Using your orientation at the end of turn 7 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 7 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42899694/Validation_42899694_42899694_792451.797.png", "Validation_42899694/Validation_42899694_42899694_792453.780.png", "Validation_42899694/Validation_42899694_42899694_792455.779.png", "Validation_42899694/Validation_42899694_42899694_792457.778.png" ], "turn_id": 1, "scan_id": "Validation_42899694", "origin_question": "Remember, have you seen any pillow(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any pillow(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_42899694/Validation_42899694_42899694_792459.777.png", "Validation_42899694/Validation_42899694_42899694_792461.776.png", "Validation_42899694/Validation_42899694_42899694_792463.775.png", "Validation_42899694/Validation_42899694_42899694_792465.775.png" ], "turn_id": 2, "scan_id": "Validation_42899694", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_42899694/Validation_42899694_42899694_792467.774.png", "Validation_42899694/Validation_42899694_42899694_792469.773.png", "Validation_42899694/Validation_42899694_42899694_792471.772.png", "Validation_42899694/Validation_42899694_42899694_792473.771.png" ], "turn_id": 3, "scan_id": "Validation_42899694", "origin_question": "When did you first discover \"the small, cubic, minimalist white table\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the small, cubic, minimalist white table\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_42899694/Validation_42899694_42899694_792475.770.png", "Validation_42899694/Validation_42899694_42899694_792477.786.png", "Validation_42899694/Validation_42899694_42899694_792479.785.png", "Validation_42899694/Validation_42899694_42899694_792481.785.png" ], "turn_id": 4, "scan_id": "Validation_42899694", "origin_question": "Which direction is the small, cubic, minimalist white table in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the small, cubic, minimalist white table in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_42899694/Validation_42899694_42899694_792483.784.png", "Validation_42899694/Validation_42899694_42899694_792485.783.png", "Validation_42899694/Validation_42899694_42899694_792487.782.png", "Validation_42899694/Validation_42899694_42899694_792489.781.png" ], "turn_id": 5, "scan_id": "Validation_42899694", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the bed\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the bed\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_42899694/Validation_42899694_42899694_792491.780.png", "Validation_42899694/Validation_42899694_42899694_792493.780.png", "Validation_42899694/Validation_42899694_42899694_792495.779.png", "Validation_42899694/Validation_42899694_42899694_792497.778.png" ], "turn_id": 6, "scan_id": "Validation_42899694", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the bed\", Y: \"the tall, full-length mirror with the matte black frame\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to Y and farther away from X", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the bed\", Y: \"the tall, full-length mirror with the matte black frame\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_42899694/Validation_42899694_42899694_792499.777.png", "Validation_42899694/Validation_42899694_42899694_792501.776.png", "Validation_42899694/Validation_42899694_42899694_792503.775.png", "Validation_42899694/Validation_42899694_42899694_792505.775.png" ], "turn_id": 7, "scan_id": "Validation_42899694", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the sleek, off-white wooden door\", Y: \"the bathtub\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the sleek, off-white wooden door\", Y: \"the bathtub\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47204607/Validation_47204607_47204607_2520.819.png", "Validation_47204607/Validation_47204607_47204607_2522.802.png", "Validation_47204607/Validation_47204607_47204607_2524.801.png", "Validation_47204607/Validation_47204607_47204607_2526.800.png", "Validation_47204607/Validation_47204607_47204607_2528.799.png" ], "turn_id": 1, "scan_id": "Validation_47204607", "origin_question": "Remember, have you seen any monitor(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any monitor(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_47204607/Validation_47204607_47204607_2530.798.png", "Validation_47204607/Validation_47204607_47204607_2532.798.png", "Validation_47204607/Validation_47204607_47204607_2534.797.png", "Validation_47204607/Validation_47204607_47204607_2536.796.png", "Validation_47204607/Validation_47204607_47204607_2538.795.png" ], "turn_id": 2, "scan_id": "Validation_47204607", "origin_question": "At the end of which turn were both of \"the sleek, cylindrical silver lamp with glass elements\" and \"the shiny metallic spherical ball\" on your left side?", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the sleek, cylindrical silver lamp with glass elements\" and \"the shiny metallic spherical ball\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_47204607/Validation_47204607_47204607_2540.795.png", "Validation_47204607/Validation_47204607_47204607_2542.794.png", "Validation_47204607/Validation_47204607_47204607_2544.793.png", "Validation_47204607/Validation_47204607_47204607_2546.792.png", "Validation_47204607/Validation_47204607_47204607_2548.791.png" ], "turn_id": 3, "scan_id": "Validation_47204607", "origin_question": "What will be the first-time appearance order of the following categories: ball, monitor, table?", "option": [ "ball, monitor, table", "table, monitor, ball", "monitor, ball, table", "ball, table, monitor" ], "answer": "table, monitor, ball", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: ball, monitor, table? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['ball, monitor, table', 'table, monitor, ball', 'monitor, ball, table', 'ball, table, monitor'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47204607/Validation_47204607_47204607_2550.791.png", "Validation_47204607/Validation_47204607_47204607_2552.790.png", "Validation_47204607/Validation_47204607_47204607_2554.806.png", "Validation_47204607/Validation_47204607_47204607_2556.805.png", "Validation_47204607/Validation_47204607_47204607_2558.804.png" ], "turn_id": 4, "scan_id": "Validation_47204607", "origin_question": "Which direction is the shiny metallic spherical ball in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the shiny metallic spherical ball in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_47204607/Validation_47204607_47204607_2560.803.png", "Validation_47204607/Validation_47204607_47204607_2562.803.png", "Validation_47204607/Validation_47204607_47204607_2564.802.png", "Validation_47204607/Validation_47204607_47204607_2566.801.png", "Validation_47204607/Validation_47204607_47204607_2568.800.png" ], "turn_id": 5, "scan_id": "Validation_47204607", "origin_question": "What will be the first-time appearance order of the following categories: door, monitor, window?", "option": [ "door, monitor, window", "door, window, monitor", "monitor, door, window", "window, monitor, door" ], "answer": "door, monitor, window", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: door, monitor, window? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['door, monitor, window', 'door, window, monitor', 'monitor, door, window', 'window, monitor, door'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_47204607/Validation_47204607_47204607_2570.799.png", "Validation_47204607/Validation_47204607_47204607_2572.799.png", "Validation_47204607/Validation_47204607_47204607_2574.798.png", "Validation_47204607/Validation_47204607_47204607_2576.797.png", "Validation_47204607/Validation_47204607_47204607_2578.796.png" ], "turn_id": 6, "scan_id": "Validation_47204607", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the light brown wooden door with the metallic handle\" from your position?", "option": [], "answer": "176.76", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the light brown wooden door with the metallic handle\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_47204607/Validation_47204607_47204607_2580.795.png", "Validation_47204607/Validation_47204607_47204607_2582.795.png", "Validation_47204607/Validation_47204607_47204607_2584.794.png", "Validation_47204607/Validation_47204607_47204607_2586.793.png", "Validation_47204607/Validation_47204607_47204607_2588.792.png" ], "turn_id": 7, "scan_id": "Validation_47204607", "origin_question": "What will be the first-time appearance order of the following categories: couch, door, window?", "option": [ "window, door, couch", "couch, door, window", "door, couch, window", "door, window, couch" ], "answer": "door, couch, window", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: couch, door, window? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['window, door, couch', 'couch, door, window', 'door, couch, window', 'door, window, couch'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_47204607/Validation_47204607_47204607_2590.791.png", "Validation_47204607/Validation_47204607_47204607_2592.791.png", "Validation_47204607/Validation_47204607_47204607_2594.790.png", "Validation_47204607/Validation_47204607_47204607_2596.789.png", "Validation_47204607/Validation_47204607_47204607_2598.805.png" ], "turn_id": 8, "scan_id": "Validation_47204607", "origin_question": "Assuming the direction you are facing at the end of the turn 6 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Left", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 6 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_47204607/Validation_47204607_47204607_2600.804.png", "Validation_47204607/Validation_47204607_47204607_2602.803.png", "Validation_47204607/Validation_47204607_47204607_2604.803.png", "Validation_47204607/Validation_47204607_47204607_2606.802.png", "Validation_47204607/Validation_47204607_47204607_2608.801.png" ], "turn_id": 9, "scan_id": "Validation_47204607", "origin_question": "Which two objects are on the same side of you now? \"the sleek black monitor with the slim bezel\", \"the plush gray rectangular pillow made of microfiber\", \"the shiny metallic spherical ball\". ", "option": [ "the sleek black monitor with the slim bezel, the plush gray rectangular pillow made of microfiber", "the sleek black monitor with the slim bezel, the shiny metallic spherical ball", "the plush gray rectangular pillow made of microfiber, the shiny metallic spherical ball" ], "answer": "the sleek black monitor with the slim bezel, the plush gray rectangular pillow made of microfiber", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the sleek black monitor with the slim bezel\", \"the plush gray rectangular pillow made of microfiber\", \"the shiny metallic spherical ball\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the sleek black monitor with the slim bezel, the plush gray rectangular pillow made of microfiber', 'the sleek black monitor with the slim bezel, the shiny metallic spherical ball', 'the plush gray rectangular pillow made of microfiber, the shiny metallic spherical ball'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47204607/Validation_47204607_47204607_2610.800.png", "Validation_47204607/Validation_47204607_47204607_2612.800.png", "Validation_47204607/Validation_47204607_47204607_2614.799.png", "Validation_47204607/Validation_47204607_47204607_2616.798.png", "Validation_47204607/Validation_47204607_47204607_2618.797.png" ], "turn_id": 10, "scan_id": "Validation_47204607", "origin_question": "Compared to your position at the end of the 5 turn, are you now closer or farther away from \"the sleek black monitor with the slim bezel\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 5 turn, are you now closer or farther away from \"the sleek black monitor with the slim bezel\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "Validation_41159508/Validation_41159508_41159508_34807.899.png", "Validation_41159508/Validation_41159508_41159508_34809.882.png", "Validation_41159508/Validation_41159508_41159508_34811.881.png", "Validation_41159508/Validation_41159508_41159508_34813.880.png", "Validation_41159508/Validation_41159508_41159508_34815.879.png" ], "turn_id": 1, "scan_id": "Validation_41159508", "origin_question": "Remember, how many carpet(s) have you seen so far? ", "option": [], "answer": "2", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many carpet(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_41159508/Validation_41159508_41159508_34817.878.png", "Validation_41159508/Validation_41159508_41159508_34819.877.png", "Validation_41159508/Validation_41159508_41159508_34821.876.png", "Validation_41159508/Validation_41159508_41159508_34823.876.png", "Validation_41159508/Validation_41159508_41159508_34825.875.png" ], "turn_id": 2, "scan_id": "Validation_41159508", "origin_question": "When did you first discover \"the spacious, oval white acrylic bathtub\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the spacious, oval white acrylic bathtub\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_41159508/Validation_41159508_41159508_34827.874.png", "Validation_41159508/Validation_41159508_41159508_34829.873.png", "Validation_41159508/Validation_41159508_41159508_34831.889.png", "Validation_41159508/Validation_41159508_41159508_34833.888.png", "Validation_41159508/Validation_41159508_41159508_34835.887.png" ], "turn_id": 3, "scan_id": "Validation_41159508", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_41159508/Validation_41159508_41159508_34837.887.png", "Validation_41159508/Validation_41159508_41159508_34839.886.png", "Validation_41159508/Validation_41159508_41159508_34841.885.png", "Validation_41159508/Validation_41159508_41159508_34843.884.png", "Validation_41159508/Validation_41159508_41159508_34845.883.png" ], "turn_id": 4, "scan_id": "Validation_41159508", "origin_question": "Which direction is the contemporary white oval toilet made of porcelain in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the contemporary white oval toilet made of porcelain in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "Validation_41159508/Validation_41159508_41159508_34847.882.png", "Validation_41159508/Validation_41159508_41159508_34849.882.png", "Validation_41159508/Validation_41159508_41159508_34851.881.png", "Validation_41159508/Validation_41159508_41159508_34853.880.png", "Validation_41159508/Validation_41159508_41159508_34855.879.png" ], "turn_id": 5, "scan_id": "Validation_41159508", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the vertical wooden cabinet with the mirrored front panel\"; \"the sturdy wooden bed with neutral-colored bedding\"; \"the compact rectangular dark wood grain table\" ", "option": [ "the vertical wooden cabinet with the mirrored front panel", "the sturdy wooden bed with neutral-colored bedding", "the compact rectangular dark wood grain table" ], "answer": "the sturdy wooden bed with neutral-colored bedding", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the vertical wooden cabinet with the mirrored front panel\"; \"the sturdy wooden bed with neutral-colored bedding\"; \"the compact rectangular dark wood grain table\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the vertical wooden cabinet with the mirrored front panel', 'the sturdy wooden bed with neutral-colored bedding', 'the compact rectangular dark wood grain table'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_41159508/Validation_41159508_41159508_34857.878.png", "Validation_41159508/Validation_41159508_41159508_34859.877.png", "Validation_41159508/Validation_41159508_41159508_34861.877.png", "Validation_41159508/Validation_41159508_41159508_34863.876.png", "Validation_41159508/Validation_41159508_41159508_34865.875.png" ], "turn_id": 6, "scan_id": "Validation_41159508", "origin_question": "Using your orientation at the end of turn 5 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 5 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_41159508/Validation_41159508_41159508_34867.874.png", "Validation_41159508/Validation_41159508_41159508_34869.873.png", "Validation_41159508/Validation_41159508_41159508_34871.889.png", "Validation_41159508/Validation_41159508_41159508_34873.888.png", "Validation_41159508/Validation_41159508_41159508_34875.887.png" ], "turn_id": 7, "scan_id": "Validation_41159508", "origin_question": "Which two objects are on the same side of you now? \"the contemporary white oval toilet made of porcelain\", \"the window\", \"the compact rectangular dark wood grain table\". ", "option": [ "the contemporary white oval toilet made of porcelain, the window", "the contemporary white oval toilet made of porcelain, the compact rectangular dark wood grain table", "the window, the compact rectangular dark wood grain table" ], "answer": "the contemporary white oval toilet made of porcelain, the window", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the contemporary white oval toilet made of porcelain\", \"the window\", \"the compact rectangular dark wood grain table\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the contemporary white oval toilet made of porcelain, the window', 'the contemporary white oval toilet made of porcelain, the compact rectangular dark wood grain table', 'the window, the compact rectangular dark wood grain table'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "Validation_41159508/Validation_41159508_41159508_34877.887.png", "Validation_41159508/Validation_41159508_41159508_34879.886.png", "Validation_41159508/Validation_41159508_41159508_34881.885.png", "Validation_41159508/Validation_41159508_41159508_34883.884.png", "Validation_41159508/Validation_41159508_41159508_34885.883.png" ], "turn_id": 8, "scan_id": "Validation_41159508", "origin_question": "Among these three objects(\"the medium wood-toned rectangular door with the smooth finish\",\"the rectangular mirror with the medium-brown wooden frame\",\"the sturdy wooden bed with neutral-colored bedding\" ), which one is the farthest from you?", "option": [ "the medium wood-toned rectangular door with the smooth finish", "the rectangular mirror with the medium-brown wooden frame", "the sturdy wooden bed with neutral-colored bedding" ], "answer": "the sturdy wooden bed with neutral-colored bedding", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the medium wood-toned rectangular door with the smooth finish\",\"the rectangular mirror with the medium-brown wooden frame\",\"the sturdy wooden bed with neutral-colored bedding\" ), which one is the farthest from you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the medium wood-toned rectangular door with the smooth finish', 'the rectangular mirror with the medium-brown wooden frame', 'the sturdy wooden bed with neutral-colored bedding'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_41159508/Validation_41159508_41159508_34887.882.png", "Validation_41159508/Validation_41159508_41159508_34889.882.png", "Validation_41159508/Validation_41159508_41159508_34891.881.png", "Validation_41159508/Validation_41159508_41159508_34893.880.png", "Validation_41159508/Validation_41159508_41159508_34895.879.png" ], "turn_id": 9, "scan_id": "Validation_41159508", "origin_question": "How far is your current position from where you were at the end of the turn 6? (in meters)", "option": [], "answer": "1.64", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 6? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_41159508/Validation_41159508_41159508_34897.878.png", "Validation_41159508/Validation_41159508_41159508_34899.878.png", "Validation_41159508/Validation_41159508_41159508_34901.877.png", "Validation_41159508/Validation_41159508_41159508_34903.876.png", "Validation_41159508/Validation_41159508_41159508_34905.875.png" ], "turn_id": 10, "scan_id": "Validation_41159508", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the rectangular light brown woven basket\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the rectangular light brown woven basket\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47429975/Validation_47429975_47429975_16187.094.png", "Validation_47429975/Validation_47429975_47429975_16188.560.png", "Validation_47429975/Validation_47429975_47429975_16190.059.png", "Validation_47429975/Validation_47429975_47429975_16191.575.png" ], "turn_id": 1, "scan_id": "Validation_47429975", "origin_question": "Remember, have you seen any statue(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any statue(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_47429975/Validation_47429975_47429975_16193.075.png", "Validation_47429975/Validation_47429975_47429975_16194.574.png", "Validation_47429975/Validation_47429975_47429975_16196.073.png", "Validation_47429975/Validation_47429975_47429975_16197.573.png" ], "turn_id": 2, "scan_id": "Validation_47429975", "origin_question": "When did you first discover \"the window\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the window\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_47429975/Validation_47429975_47429975_16199.072.png", "Validation_47429975/Validation_47429975_47429975_16200.572.png", "Validation_47429975/Validation_47429975_47429975_16202.071.png", "Validation_47429975/Validation_47429975_47429975_16203.570.png" ], "turn_id": 3, "scan_id": "Validation_47429975", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47429975/Validation_47429975_47429975_16205.070.png", "Validation_47429975/Validation_47429975_47429975_16206.569.png", "Validation_47429975/Validation_47429975_47429975_16208.068.png", "Validation_47429975/Validation_47429975_47429975_16209.568.png" ], "turn_id": 4, "scan_id": "Validation_47429975", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the sleek cylindrical white ceramic flowerpot\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the sleek cylindrical white ceramic flowerpot\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47429975/Validation_47429975_47429975_16211.067.png", "Validation_47429975/Validation_47429975_47429975_16212.567.png", "Validation_47429975/Validation_47429975_47429975_16214.066.png", "Validation_47429975/Validation_47429975_47429975_16215.565.png" ], "turn_id": 5, "scan_id": "Validation_47429975", "origin_question": "", "option": [], "answer": "", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_47429975/Validation_47429975_47429975_16217.065.png", "Validation_47429975/Validation_47429975_47429975_16218.564.png", "Validation_47429975/Validation_47429975_47429975_16220.064.png", "Validation_47429975/Validation_47429975_47429975_16221.563.png" ], "turn_id": 6, "scan_id": "Validation_47429975", "origin_question": "What will be the first-time appearance order of the following categories: couch, oven, window?", "option": [ "couch, oven, window", "window, couch, oven", "window, oven, couch", "oven, window, couch" ], "answer": "window, oven, couch", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: couch, oven, window? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['couch, oven, window', 'window, couch, oven', 'window, oven, couch', 'oven, window, couch'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_47429975/Validation_47429975_47429975_16223.062.png", "Validation_47429975/Validation_47429975_47429975_16224.562.png", "Validation_47429975/Validation_47429975_47429975_16226.061.png", "Validation_47429975/Validation_47429975_47429975_16227.560.png" ], "turn_id": 7, "scan_id": "Validation_47429975", "origin_question": "How far is your current position from where you were at the end of the turn 6? (in meters)", "option": [], "answer": "1.66", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 6? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47429912/Validation_47429912_47429912_136023.603.png", "Validation_47429912/Validation_47429912_47429912_136025.086.png", "Validation_47429912/Validation_47429912_47429912_136026.585.png", "Validation_47429912/Validation_47429912_47429912_136028.084.png" ], "turn_id": 1, "scan_id": "Validation_47429912", "origin_question": "Remember, have you seen any broom(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any broom(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47429912/Validation_47429912_47429912_136029.584.png", "Validation_47429912/Validation_47429912_47429912_136031.083.png", "Validation_47429912/Validation_47429912_47429912_136032.582.png", "Validation_47429912/Validation_47429912_47429912_136034.082.png" ], "turn_id": 2, "scan_id": "Validation_47429912", "origin_question": "Remember, have you seen any poster(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any poster(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_47429912/Validation_47429912_47429912_136035.581.png", "Validation_47429912/Validation_47429912_47429912_136037.081.png", "Validation_47429912/Validation_47429912_47429912_136038.580.png", "Validation_47429912/Validation_47429912_47429912_136040.079.png" ], "turn_id": 3, "scan_id": "Validation_47429912", "origin_question": "When did you first discover \"the modern white ceramic toilet with the sleek design\"? (index of the turn)", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the modern white ceramic toilet with the sleek design\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47429912/Validation_47429912_47429912_136041.579.png", "Validation_47429912/Validation_47429912_47429912_136043.095.png", "Validation_47429912/Validation_47429912_47429912_136044.594.png", "Validation_47429912/Validation_47429912_47429912_136046.094.png" ], "turn_id": 4, "scan_id": "Validation_47429912", "origin_question": "Remember, have you seen any notebook(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any notebook(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_47429912/Validation_47429912_47429912_136047.593.png", "Validation_47429912/Validation_47429912_47429912_136049.092.png", "Validation_47429912/Validation_47429912_47429912_136050.592.png", "Validation_47429912/Validation_47429912_47429912_136052.091.png" ], "turn_id": 5, "scan_id": "Validation_47429912", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "1.16", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_47429912/Validation_47429912_47429912_136053.590.png", "Validation_47429912/Validation_47429912_47429912_136055.090.png", "Validation_47429912/Validation_47429912_47429912_136056.589.png", "Validation_47429912/Validation_47429912_47429912_136058.089.png" ], "turn_id": 6, "scan_id": "Validation_47429912", "origin_question": "At the end of which turn were both of \"the modern white ceramic toilet with the sleek design\" and \"the bathtub\" on your right side?", "option": [], "answer": "5", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the modern white ceramic toilet with the sleek design\" and \"the bathtub\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_47429912/Validation_47429912_47429912_136059.588.png", "Validation_47429912/Validation_47429912_47429912_136061.087.png", "Validation_47429912/Validation_47429912_47429912_136062.587.png", "Validation_47429912/Validation_47429912_47429912_136064.086.png" ], "turn_id": 7, "scan_id": "Validation_47429912", "origin_question": "Using your orientation at the end of turn 5 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 5 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_48018741/Validation_48018741_48018741_3388.843.png", "Validation_48018741/Validation_48018741_48018741_3390.326.png", "Validation_48018741/Validation_48018741_48018741_3391.825.png", "Validation_48018741/Validation_48018741_48018741_3393.325.png" ], "turn_id": 1, "scan_id": "Validation_48018741", "origin_question": "Remember, have you seen any spoon(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any spoon(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_48018741/Validation_48018741_48018741_3394.824.png", "Validation_48018741/Validation_48018741_48018741_3396.324.png", "Validation_48018741/Validation_48018741_48018741_3397.823.png", "Validation_48018741/Validation_48018741_48018741_3399.323.png" ], "turn_id": 2, "scan_id": "Validation_48018741", "origin_question": "In which turn were you closest to \"the large rectangular box in neutral tones\"? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the large rectangular box in neutral tones\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_48018741/Validation_48018741_48018741_3400.822.png", "Validation_48018741/Validation_48018741_48018741_3402.321.png", "Validation_48018741/Validation_48018741_48018741_3403.821.png", "Validation_48018741/Validation_48018741_48018741_3405.320.png" ], "turn_id": 3, "scan_id": "Validation_48018741", "origin_question": "Is the wall-mounted light with the frosted glass shade to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the wall-mounted light with the frosted glass shade to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_48018741/Validation_48018741_48018741_3406.819.png", "Validation_48018741/Validation_48018741_48018741_3408.319.png", "Validation_48018741/Validation_48018741_48018741_3409.818.png", "Validation_48018741/Validation_48018741_48018741_3411.318.png" ], "turn_id": 4, "scan_id": "Validation_48018741", "origin_question": "Which direction is the sleek white rectangular acrylic bathtub in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the sleek white rectangular acrylic bathtub in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_48018741/Validation_48018741_48018741_3412.817.png", "Validation_48018741/Validation_48018741_48018741_3414.316.png", "Validation_48018741/Validation_48018741_48018741_3415.816.png", "Validation_48018741/Validation_48018741_48018741_3417.315.png" ], "turn_id": 5, "scan_id": "Validation_48018741", "origin_question": "What will be the first-time appearance order of the following categories: bin, door, mirror?", "option": [ "mirror, bin, door", "door, bin, mirror", "mirror, door, bin", "door, mirror, bin" ], "answer": "door, bin, mirror", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bin, door, mirror? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['mirror, bin, door', 'door, bin, mirror', 'mirror, door, bin', 'door, mirror, bin'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_42444953/Validation_42444953_42444953_224960.185.png", "Validation_42444953/Validation_42444953_42444953_224962.167.png", "Validation_42444953/Validation_42444953_42444953_224964.166.png", "Validation_42444953/Validation_42444953_42444953_224966.166.png" ], "turn_id": 1, "scan_id": "Validation_42444953", "origin_question": "", "option": [], "answer": "", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42444953/Validation_42444953_42444953_224968.165.png", "Validation_42444953/Validation_42444953_42444953_224970.164.png", "Validation_42444953/Validation_42444953_42444953_224972.163.png", "Validation_42444953/Validation_42444953_42444953_224974.162.png" ], "turn_id": 2, "scan_id": "Validation_42444953", "origin_question": "Remember, have you seen any coat hanger(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any coat hanger(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_42444953/Validation_42444953_42444953_224976.162.png", "Validation_42444953/Validation_42444953_42444953_224978.161.png", "Validation_42444953/Validation_42444953_42444953_224980.160.png", "Validation_42444953/Validation_42444953_42444953_224982.159.png" ], "turn_id": 3, "scan_id": "Validation_42444953", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the white oval ceramic sink with smooth edges\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the white oval ceramic sink with smooth edges\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_42444953/Validation_42444953_42444953_224984.158.png", "Validation_42444953/Validation_42444953_42444953_224986.158.png", "Validation_42444953/Validation_42444953_42444953_224988.157.png", "Validation_42444953/Validation_42444953_42444953_224990.156.png" ], "turn_id": 4, "scan_id": "Validation_42444953", "origin_question": "Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_42444953/Validation_42444953_42444953_224992.155.png", "Validation_42444953/Validation_42444953_42444953_224994.154.png", "Validation_42444953/Validation_42444953_42444953_224996.170.png", "Validation_42444953/Validation_42444953_42444953_224998.169.png" ], "turn_id": 5, "scan_id": "Validation_42444953", "origin_question": "What will be the first-time appearance order of the following categories: bathtub, radiator, sink?", "option": [ "radiator, sink, bathtub", "sink, radiator, bathtub", "bathtub, sink, radiator", "bathtub, radiator, sink" ], "answer": "bathtub, sink, radiator", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bathtub, radiator, sink? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['radiator, sink, bathtub', 'sink, radiator, bathtub', 'bathtub, sink, radiator', 'bathtub, radiator, sink'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_42444953/Validation_42444953_42444953_225000.168.png", "Validation_42444953/Validation_42444953_42444953_225002.168.png", "Validation_42444953/Validation_42444953_42444953_225004.167.png", "Validation_42444953/Validation_42444953_42444953_225006.166.png" ], "turn_id": 6, "scan_id": "Validation_42444953", "origin_question": "In which turn were you closest to \"the classic rectangular white enamel bathtub\"? ", "option": [], "answer": "3", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the classic rectangular white enamel bathtub\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_42444953/Validation_42444953_42444953_225008.165.png", "Validation_42444953/Validation_42444953_42444953_225010.164.png", "Validation_42444953/Validation_42444953_42444953_225012.164.png", "Validation_42444953/Validation_42444953_42444953_225014.163.png" ], "turn_id": 7, "scan_id": "Validation_42444953", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the standard white porcelain toilet with the classic shape\" from your position?", "option": [], "answer": "94.84", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the standard white porcelain toilet with the classic shape\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_42444953/Validation_42444953_42444953_225016.162.png", "Validation_42444953/Validation_42444953_42444953_225018.161.png", "Validation_42444953/Validation_42444953_42444953_225020.160.png", "Validation_42444953/Validation_42444953_42444953_225022.160.png" ], "turn_id": 8, "scan_id": "Validation_42444953", "origin_question": "Is the radiator to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the radiator to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_42444953/Validation_42444953_42444953_225024.159.png", "Validation_42444953/Validation_42444953_42444953_225026.158.png", "Validation_42444953/Validation_42444953_42444953_225028.157.png", "Validation_42444953/Validation_42444953_42444953_225030.156.png" ], "turn_id": 9, "scan_id": "Validation_42444953", "origin_question": "Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "39.83", "user_message": "For the 9 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47895771/Validation_47895771_47895771_386486.687.png", "Validation_47895771/Validation_47895771_47895771_386488.669.png", "Validation_47895771/Validation_47895771_47895771_386490.669.png", "Validation_47895771/Validation_47895771_47895771_386492.668.png" ], "turn_id": 1, "scan_id": "Validation_47895771", "origin_question": "Remember, have you seen any shirt(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any shirt(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_47895771/Validation_47895771_47895771_386494.667.png", "Validation_47895771/Validation_47895771_47895771_386496.666.png", "Validation_47895771/Validation_47895771_47895771_386498.665.png", "Validation_47895771/Validation_47895771_47895771_386500.665.png" ], "turn_id": 2, "scan_id": "Validation_47895771", "origin_question": "When did you first discover \"the medium-sized light green towel\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the medium-sized light green towel\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_47895771/Validation_47895771_47895771_386502.664.png", "Validation_47895771/Validation_47895771_47895771_386504.663.png", "Validation_47895771/Validation_47895771_47895771_386506.662.png", "Validation_47895771/Validation_47895771_47895771_386508.678.png" ], "turn_id": 3, "scan_id": "Validation_47895771", "origin_question": "At the end of which turn were both of \"the medium-sized light green towel\" and \"the rectangular stainless steel sink with the spacious basin\" on your left side?", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the medium-sized light green towel\" and \"the rectangular stainless steel sink with the spacious basin\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47895771/Validation_47895771_47895771_386510.677.png", "Validation_47895771/Validation_47895771_47895771_386512.676.png", "Validation_47895771/Validation_47895771_47895771_386514.676.png", "Validation_47895771/Validation_47895771_47895771_386516.675.png" ], "turn_id": 4, "scan_id": "Validation_47895771", "origin_question": "Which direction is the rectangular stainless steel sink with the spacious basin in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the rectangular stainless steel sink with the spacious basin in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47895771/Validation_47895771_47895771_386518.674.png", "Validation_47895771/Validation_47895771_47895771_386520.673.png", "Validation_47895771/Validation_47895771_47895771_386522.672.png", "Validation_47895771/Validation_47895771_47895771_386524.671.png" ], "turn_id": 5, "scan_id": "Validation_47895771", "origin_question": "In which turn were you closest to \"the classic white ceramic cylindrical cup\"? ", "option": [], "answer": "3", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the classic white ceramic cylindrical cup\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_47895771/Validation_47895771_47895771_386526.671.png", "Validation_47895771/Validation_47895771_47895771_386528.670.png", "Validation_47895771/Validation_47895771_47895771_386530.669.png", "Validation_47895771/Validation_47895771_47895771_386532.668.png" ], "turn_id": 6, "scan_id": "Validation_47895771", "origin_question": "Is the rectangular stainless steel sink with the spacious basin to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the rectangular stainless steel sink with the spacious basin to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_47895771/Validation_47895771_47895771_386534.667.png", "Validation_47895771/Validation_47895771_47895771_386536.666.png", "Validation_47895771/Validation_47895771_47895771_386538.666.png", "Validation_47895771/Validation_47895771_47895771_386540.665.png" ], "turn_id": 7, "scan_id": "Validation_47895771", "origin_question": "By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the cozy rectangular bed with earthy toned bedding\", Y: \"the sleek, foldable shelf with the white finish\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the cozy rectangular bed with earthy toned bedding\", Y: \"the sleek, foldable shelf with the white finish\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42446519/Validation_42446519_42446519_197484.887.png", "Validation_42446519/Validation_42446519_42446519_197486.870.png", "Validation_42446519/Validation_42446519_42446519_197488.869.png", "Validation_42446519/Validation_42446519_42446519_197490.868.png", "Validation_42446519/Validation_42446519_42446519_197492.867.png" ], "turn_id": 1, "scan_id": "Validation_42446519", "origin_question": "Remember, have you seen any bench(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any bench(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_42446519/Validation_42446519_42446519_197494.883.png", "Validation_42446519/Validation_42446519_42446519_197496.883.png", "Validation_42446519/Validation_42446519_42446519_197498.882.png", "Validation_42446519/Validation_42446519_42446519_197500.881.png", "Validation_42446519/Validation_42446519_42446519_197502.880.png" ], "turn_id": 2, "scan_id": "Validation_42446519", "origin_question": "In which turn did you see both \"the bed\" and \"the window\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the bed\" and \"the window\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_42446519/Validation_42446519_42446519_197504.879.png", "Validation_42446519/Validation_42446519_42446519_197506.878.png", "Validation_42446519/Validation_42446519_42446519_197508.878.png", "Validation_42446519/Validation_42446519_42446519_197510.877.png", "Validation_42446519/Validation_42446519_42446519_197512.876.png" ], "turn_id": 3, "scan_id": "Validation_42446519", "origin_question": "In which turn did you see both \"the vibrant red freestanding shelf\" and \"the black and gray plastic bin with an open-top design\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the vibrant red freestanding shelf\" and \"the black and gray plastic bin with an open-top design\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_42446519/Validation_42446519_42446519_197514.875.png", "Validation_42446519/Validation_42446519_42446519_197516.874.png", "Validation_42446519/Validation_42446519_42446519_197518.873.png", "Validation_42446519/Validation_42446519_42446519_197520.873.png", "Validation_42446519/Validation_42446519_42446519_197522.872.png" ], "turn_id": 4, "scan_id": "Validation_42446519", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the sleek black ergonomic chair with the high backrest\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the sleek black ergonomic chair with the high backrest\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_42446519/Validation_42446519_42446519_197524.871.png", "Validation_42446519/Validation_42446519_42446519_197526.870.png", "Validation_42446519/Validation_42446519_42446519_197528.869.png", "Validation_42446519/Validation_42446519_42446519_197530.869.png", "Validation_42446519/Validation_42446519_42446519_197532.868.png" ], "turn_id": 5, "scan_id": "Validation_42446519", "origin_question": "", "option": [], "answer": "", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_42446519/Validation_42446519_42446519_197534.883.png", "Validation_42446519/Validation_42446519_42446519_197536.883.png", "Validation_42446519/Validation_42446519_42446519_197538.882.png", "Validation_42446519/Validation_42446519_42446519_197540.881.png", "Validation_42446519/Validation_42446519_42446519_197542.880.png" ], "turn_id": 6, "scan_id": "Validation_42446519", "origin_question": "When did you first discover \"the vibrant red freestanding shelf\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the vibrant red freestanding shelf\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_42446519/Validation_42446519_42446519_197544.879.png", "Validation_42446519/Validation_42446519_42446519_197546.879.png", "Validation_42446519/Validation_42446519_42446519_197548.878.png", "Validation_42446519/Validation_42446519_42446519_197550.877.png", "Validation_42446519/Validation_42446519_42446519_197552.876.png" ], "turn_id": 7, "scan_id": "Validation_42446519", "origin_question": "Please recall \"the window\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.99", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the window\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_42446519/Validation_42446519_42446519_197554.875.png", "Validation_42446519/Validation_42446519_42446519_197556.874.png", "Validation_42446519/Validation_42446519_42446519_197558.874.png", "Validation_42446519/Validation_42446519_42446519_197560.873.png", "Validation_42446519/Validation_42446519_42446519_197562.872.png" ], "turn_id": 8, "scan_id": "Validation_42446519", "origin_question": "In which turn were you closest to \"the smooth rectangular door with the neutral light color\"? ", "option": [], "answer": "1", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the smooth rectangular door with the neutral light color\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_42446519/Validation_42446519_42446519_197564.871.png", "Validation_42446519/Validation_42446519_42446519_197566.870.png", "Validation_42446519/Validation_42446519_42446519_197568.870.png", "Validation_42446519/Validation_42446519_42446519_197570.869.png", "Validation_42446519/Validation_42446519_42446519_197572.868.png" ], "turn_id": 9, "scan_id": "Validation_42446519", "origin_question": "In which turn were you closest to \"the vibrant red freestanding shelf\"? ", "option": [], "answer": "6", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the vibrant red freestanding shelf\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_42446519/Validation_42446519_42446519_197574.867.png", "Validation_42446519/Validation_42446519_42446519_197576.883.png", "Validation_42446519/Validation_42446519_42446519_197578.882.png", "Validation_42446519/Validation_42446519_42446519_197580.881.png", "Validation_42446519/Validation_42446519_42446519_197582.880.png" ], "turn_id": 10, "scan_id": "Validation_42446519", "origin_question": "Is the vibrant red freestanding shelf to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the vibrant red freestanding shelf to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_41254386/Validation_41254386_41254386_95038.790.png", "Validation_41254386/Validation_41254386_41254386_95040.773.png", "Validation_41254386/Validation_41254386_41254386_95042.772.png", "Validation_41254386/Validation_41254386_41254386_95044.771.png", "Validation_41254386/Validation_41254386_41254386_95046.770.png" ], "turn_id": 1, "scan_id": "Validation_41254386", "origin_question": "Remember, have you seen any door(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any door(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_41254386/Validation_41254386_41254386_95048.769.png", "Validation_41254386/Validation_41254386_41254386_95050.769.png", "Validation_41254386/Validation_41254386_41254386_95052.768.png", "Validation_41254386/Validation_41254386_41254386_95054.767.png", "Validation_41254386/Validation_41254386_41254386_95056.766.png" ], "turn_id": 2, "scan_id": "Validation_41254386", "origin_question": "Is the glossy white rectangular bathtub with the door to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the glossy white rectangular bathtub with the door to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_41254386/Validation_41254386_41254386_95058.765.png", "Validation_41254386/Validation_41254386_41254386_95060.765.png", "Validation_41254386/Validation_41254386_41254386_95062.764.png", "Validation_41254386/Validation_41254386_41254386_95064.763.png", "Validation_41254386/Validation_41254386_41254386_95066.762.png" ], "turn_id": 3, "scan_id": "Validation_41254386", "origin_question": "By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the glossy white rectangular bathtub with the door\", Y: \"the white towel with gray and black patterns\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to X and farther away from Y", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the glossy white rectangular bathtub with the door\", Y: \"the white towel with gray and black patterns\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_41254386/Validation_41254386_41254386_95068.761.png", "Validation_41254386/Validation_41254386_41254386_95070.761.png", "Validation_41254386/Validation_41254386_41254386_95072.777.png", "Validation_41254386/Validation_41254386_41254386_95074.776.png", "Validation_41254386/Validation_41254386_41254386_95076.775.png" ], "turn_id": 4, "scan_id": "Validation_41254386", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the classic rectangular white door with paneling\", Y: \"the window\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to Y and farther away from X", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the classic rectangular white door with paneling\", Y: \"the window\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_41254386/Validation_41254386_41254386_95078.774.png", "Validation_41254386/Validation_41254386_41254386_95080.773.png", "Validation_41254386/Validation_41254386_41254386_95082.773.png", "Validation_41254386/Validation_41254386_41254386_95084.772.png", "Validation_41254386/Validation_41254386_41254386_95086.771.png" ], "turn_id": 5, "scan_id": "Validation_41254386", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_41254386/Validation_41254386_41254386_95088.770.png", "Validation_41254386/Validation_41254386_41254386_95090.769.png", "Validation_41254386/Validation_41254386_41254386_95092.769.png", "Validation_41254386/Validation_41254386_41254386_95094.768.png", "Validation_41254386/Validation_41254386_41254386_95096.767.png" ], "turn_id": 6, "scan_id": "Validation_41254386", "origin_question": "Please recall \"the classic rectangular white door with paneling\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.84", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the classic rectangular white door with paneling\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_41254386/Validation_41254386_41254386_95098.766.png", "Validation_41254386/Validation_41254386_41254386_95100.765.png", "Validation_41254386/Validation_41254386_41254386_95102.765.png", "Validation_41254386/Validation_41254386_41254386_95104.764.png", "Validation_41254386/Validation_41254386_41254386_95106.763.png" ], "turn_id": 7, "scan_id": "Validation_41254386", "origin_question": "Which direction is the glossy white rectangular bathtub with the door in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the glossy white rectangular bathtub with the door in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_41254386/Validation_41254386_41254386_95108.762.png", "Validation_41254386/Validation_41254386_41254386_95110.762.png", "Validation_41254386/Validation_41254386_41254386_95112.761.png", "Validation_41254386/Validation_41254386_41254386_95114.760.png", "Validation_41254386/Validation_41254386_41254386_95116.776.png" ], "turn_id": 8, "scan_id": "Validation_41254386", "origin_question": "Which direction is the classic white oval ceramic toilet in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the classic white oval ceramic toilet in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_41254386/Validation_41254386_41254386_95118.775.png", "Validation_41254386/Validation_41254386_41254386_95120.774.png", "Validation_41254386/Validation_41254386_41254386_95122.773.png", "Validation_41254386/Validation_41254386_41254386_95124.773.png", "Validation_41254386/Validation_41254386_41254386_95126.772.png" ], "turn_id": 9, "scan_id": "Validation_41254386", "origin_question": "In which turn were you farthest from \"the white towel with gray and black patterns\"? ", "option": [], "answer": "1", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the white towel with gray and black patterns\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_44358515/Validation_44358515_44358515_52554.326.png", "Validation_44358515/Validation_44358515_44358515_52556.309.png", "Validation_44358515/Validation_44358515_44358515_52558.308.png", "Validation_44358515/Validation_44358515_44358515_52560.307.png" ], "turn_id": 1, "scan_id": "Validation_44358515", "origin_question": "Remember, have you seen any chair(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any chair(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_44358515/Validation_44358515_44358515_52562.306.png", "Validation_44358515/Validation_44358515_44358515_52564.305.png", "Validation_44358515/Validation_44358515_44358515_52566.304.png", "Validation_44358515/Validation_44358515_44358515_52568.304.png" ], "turn_id": 2, "scan_id": "Validation_44358515", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the sleek rectangular glass table with metallic legs\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the sleek rectangular glass table with metallic legs\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_44358515/Validation_44358515_44358515_52570.319.png", "Validation_44358515/Validation_44358515_44358515_52572.319.png", "Validation_44358515/Validation_44358515_44358515_52574.318.png", "Validation_44358515/Validation_44358515_44358515_52576.317.png" ], "turn_id": 3, "scan_id": "Validation_44358515", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the classic rectangular white fireplace made of wood and stone\", Y: \"the plant with long, slender vibrant green leaves\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the classic rectangular white fireplace made of wood and stone\", Y: \"the plant with long, slender vibrant green leaves\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_44358515/Validation_44358515_44358515_52578.316.png", "Validation_44358515/Validation_44358515_44358515_52580.315.png", "Validation_44358515/Validation_44358515_44358515_52582.315.png", "Validation_44358515/Validation_44358515_44358515_52584.314.png" ], "turn_id": 4, "scan_id": "Validation_44358515", "origin_question": "What will be the first-time appearance order of the following categories: chair, light, plant?", "option": [ "light, chair, plant", "plant, chair, light", "chair, plant, light", "chair, light, plant" ], "answer": "chair, plant, light", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: chair, light, plant? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['light, chair, plant', 'plant, chair, light', 'chair, plant, light', 'chair, light, plant'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_44358515/Validation_44358515_44358515_52586.313.png", "Validation_44358515/Validation_44358515_44358515_52588.312.png", "Validation_44358515/Validation_44358515_44358515_52590.311.png", "Validation_44358515/Validation_44358515_44358515_52592.310.png" ], "turn_id": 5, "scan_id": "Validation_44358515", "origin_question": "When did you first discover \"the classic rectangular white fireplace made of wood and stone\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the classic rectangular white fireplace made of wood and stone\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_44358515/Validation_44358515_44358515_52594.310.png", "Validation_44358515/Validation_44358515_44358515_52596.309.png", "Validation_44358515/Validation_44358515_44358515_52598.308.png", "Validation_44358515/Validation_44358515_44358515_52600.307.png" ], "turn_id": 6, "scan_id": "Validation_44358515", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_44358515/Validation_44358515_44358515_52602.306.png", "Validation_44358515/Validation_44358515_44358515_52604.305.png", "Validation_44358515/Validation_44358515_44358515_52606.304.png", "Validation_44358515/Validation_44358515_44358515_52608.304.png" ], "turn_id": 7, "scan_id": "Validation_44358515", "origin_question": "Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance forward or backward from that position?", "option": [ "Forward", "Backward" ], "answer": "Forward", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance forward or backward from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Forward', 'Backward'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_44358515/Validation_44358515_44358515_52610.319.png", "Validation_44358515/Validation_44358515_44358515_52612.319.png", "Validation_44358515/Validation_44358515_44358515_52614.318.png", "Validation_44358515/Validation_44358515_44358515_52616.317.png" ], "turn_id": 8, "scan_id": "Validation_44358515", "origin_question": "Which direction is the plant with long, slender vibrant green leaves in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the plant with long, slender vibrant green leaves in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_44358515/Validation_44358515_44358515_52618.316.png", "Validation_44358515/Validation_44358515_44358515_52620.315.png", "Validation_44358515/Validation_44358515_44358515_52622.315.png", "Validation_44358515/Validation_44358515_44358515_52624.314.png" ], "turn_id": 9, "scan_id": "Validation_44358515", "origin_question": "", "option": [], "answer": "", "user_message": "For the 9 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_48458417/Validation_48458417_48458417_115251.765.png", "Validation_48458417/Validation_48458417_48458417_115253.748.png", "Validation_48458417/Validation_48458417_48458417_115255.747.png", "Validation_48458417/Validation_48458417_48458417_115257.746.png", "Validation_48458417/Validation_48458417_48458417_115259.745.png" ], "turn_id": 1, "scan_id": "Validation_48458417", "origin_question": "Remember, have you seen any light(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any light(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "Validation_48458417/Validation_48458417_48458417_115261.744.png", "Validation_48458417/Validation_48458417_48458417_115263.744.png", "Validation_48458417/Validation_48458417_48458417_115265.759.png", "Validation_48458417/Validation_48458417_48458417_115267.759.png", "Validation_48458417/Validation_48458417_48458417_115269.758.png" ], "turn_id": 2, "scan_id": "Validation_48458417", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the sleek black rectangular monitor with the glossy screen\"; \"the modern light with the wood base and fabric shade\"; \"the sleek rectangular case made of lightweight material\" ", "option": [ "the sleek black rectangular monitor with the glossy screen", "the modern light with the wood base and fabric shade", "the sleek rectangular case made of lightweight material" ], "answer": "the sleek black rectangular monitor with the glossy screen", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the sleek black rectangular monitor with the glossy screen\"; \"the modern light with the wood base and fabric shade\"; \"the sleek rectangular case made of lightweight material\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the sleek black rectangular monitor with the glossy screen', 'the modern light with the wood base and fabric shade', 'the sleek rectangular case made of lightweight material'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_48458417/Validation_48458417_48458417_115271.757.png", "Validation_48458417/Validation_48458417_48458417_115273.756.png", "Validation_48458417/Validation_48458417_48458417_115275.755.png", "Validation_48458417/Validation_48458417_48458417_115277.754.png", "Validation_48458417/Validation_48458417_48458417_115279.754.png" ], "turn_id": 3, "scan_id": "Validation_48458417", "origin_question": "Please recall \"the warm wood finished rectangular door\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.64", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the warm wood finished rectangular door\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_48458417/Validation_48458417_48458417_115281.753.png", "Validation_48458417/Validation_48458417_48458417_115283.752.png", "Validation_48458417/Validation_48458417_48458417_115285.751.png", "Validation_48458417/Validation_48458417_48458417_115287.750.png", "Validation_48458417/Validation_48458417_48458417_115289.750.png" ], "turn_id": 4, "scan_id": "Validation_48458417", "origin_question": "What will be the first-time appearance order of the following categories: chair, monitor, radiator?", "option": [ "radiator, monitor, chair", "chair, monitor, radiator", "radiator, chair, monitor", "chair, radiator, monitor" ], "answer": "chair, monitor, radiator", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: chair, monitor, radiator? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['radiator, monitor, chair', 'chair, monitor, radiator', 'radiator, chair, monitor', 'chair, radiator, monitor'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_48458417/Validation_48458417_48458417_115291.749.png", "Validation_48458417/Validation_48458417_48458417_115293.748.png", "Validation_48458417/Validation_48458417_48458417_115295.747.png", "Validation_48458417/Validation_48458417_48458417_115297.746.png", "Validation_48458417/Validation_48458417_48458417_115299.745.png" ], "turn_id": 5, "scan_id": "Validation_48458417", "origin_question": "How far is your current position from where you were at the end of the turn 4? (in meters)", "option": [], "answer": "1.82", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 4? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "Validation_48458417/Validation_48458417_48458417_115301.745.png", "Validation_48458417/Validation_48458417_48458417_115303.744.png", "Validation_48458417/Validation_48458417_48458417_115305.760.png", "Validation_48458417/Validation_48458417_48458417_115307.759.png", "Validation_48458417/Validation_48458417_48458417_115309.758.png" ], "turn_id": 6, "scan_id": "Validation_48458417", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the vibrant yellow stool with the white tubular frame\"; \"the rectangular wooden table with the yellow tablecloth\"; \"the sleek, white rectangular metal radiator\" ", "option": [ "the vibrant yellow stool with the white tubular frame", "the rectangular wooden table with the yellow tablecloth", "the sleek, white rectangular metal radiator" ], "answer": "the vibrant yellow stool with the white tubular frame", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the vibrant yellow stool with the white tubular frame\"; \"the rectangular wooden table with the yellow tablecloth\"; \"the sleek, white rectangular metal radiator\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the vibrant yellow stool with the white tubular frame', 'the rectangular wooden table with the yellow tablecloth', 'the sleek, white rectangular metal radiator'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_48458417/Validation_48458417_48458417_115311.757.png", "Validation_48458417/Validation_48458417_48458417_115313.756.png", "Validation_48458417/Validation_48458417_48458417_115315.755.png", "Validation_48458417/Validation_48458417_48458417_115317.754.png", "Validation_48458417/Validation_48458417_48458417_115319.754.png" ], "turn_id": 7, "scan_id": "Validation_48458417", "origin_question": "Compared to your position at the end of the 5 turn, are you now closer or farther away from \"the white metallic stove with four burners\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 5 turn, are you now closer or farther away from \"the white metallic stove with four burners\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_48458417/Validation_48458417_48458417_115321.753.png", "Validation_48458417/Validation_48458417_48458417_115323.752.png", "Validation_48458417/Validation_48458417_48458417_115325.751.png", "Validation_48458417/Validation_48458417_48458417_115327.750.png", "Validation_48458417/Validation_48458417_48458417_115329.750.png" ], "turn_id": 8, "scan_id": "Validation_48458417", "origin_question": "Which direction is the large rectangular window with the neutral-colored frame in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the large rectangular window with the neutral-colored frame in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_48458417/Validation_48458417_48458417_115331.749.png", "Validation_48458417/Validation_48458417_48458417_115333.748.png", "Validation_48458417/Validation_48458417_48458417_115335.747.png", "Validation_48458417/Validation_48458417_48458417_115337.746.png", "Validation_48458417/Validation_48458417_48458417_115339.745.png" ], "turn_id": 9, "scan_id": "Validation_48458417", "origin_question": "Which direction is the sleek rectangular case made of lightweight material in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the sleek rectangular case made of lightweight material in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47895364/Validation_47895364_47895364_100169.962.png", "Validation_47895364/Validation_47895364_47895364_100171.945.png", "Validation_47895364/Validation_47895364_47895364_100173.944.png", "Validation_47895364/Validation_47895364_47895364_100175.943.png", "Validation_47895364/Validation_47895364_47895364_100177.942.png" ], "turn_id": 1, "scan_id": "Validation_47895364", "origin_question": "Remember, have you seen any sack(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any sack(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47895364/Validation_47895364_47895364_100179.942.png", "Validation_47895364/Validation_47895364_47895364_100181.941.png", "Validation_47895364/Validation_47895364_47895364_100183.940.png", "Validation_47895364/Validation_47895364_47895364_100185.939.png", "Validation_47895364/Validation_47895364_47895364_100187.938.png" ], "turn_id": 2, "scan_id": "Validation_47895364", "origin_question": "In which turn were you closest to \"the sleek modern front-loading washing machine\"? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the sleek modern front-loading washing machine\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47895364/Validation_47895364_47895364_100189.938.png", "Validation_47895364/Validation_47895364_47895364_100191.937.png", "Validation_47895364/Validation_47895364_47895364_100193.936.png", "Validation_47895364/Validation_47895364_47895364_100195.935.png", "Validation_47895364/Validation_47895364_47895364_100197.934.png" ], "turn_id": 3, "scan_id": "Validation_47895364", "origin_question": "In which turn were you closest to \"the sleek modern front-loading washing machine\"? ", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the sleek modern front-loading washing machine\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47895364/Validation_47895364_47895364_100199.933.png", "Validation_47895364/Validation_47895364_47895364_100201.933.png", "Validation_47895364/Validation_47895364_47895364_100203.932.png", "Validation_47895364/Validation_47895364_47895364_100205.931.png", "Validation_47895364/Validation_47895364_47895364_100207.930.png" ], "turn_id": 4, "scan_id": "Validation_47895364", "origin_question": "Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the sleek modern front-loading washing machine\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the sleek modern front-loading washing machine\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47895364/Validation_47895364_47895364_100209.929.png", "Validation_47895364/Validation_47895364_47895364_100211.945.png", "Validation_47895364/Validation_47895364_47895364_100213.944.png", "Validation_47895364/Validation_47895364_47895364_100215.944.png", "Validation_47895364/Validation_47895364_47895364_100217.943.png" ], "turn_id": 5, "scan_id": "Validation_47895364", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the sleek modern front-loading washing machine\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the sleek modern front-loading washing machine\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47895364/Validation_47895364_47895364_100219.942.png", "Validation_47895364/Validation_47895364_47895364_100221.941.png", "Validation_47895364/Validation_47895364_47895364_100223.940.png", "Validation_47895364/Validation_47895364_47895364_100225.940.png", "Validation_47895364/Validation_47895364_47895364_100227.939.png" ], "turn_id": 6, "scan_id": "Validation_47895364", "origin_question": "Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the medium-sized cylindrical white plastic bottle\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the medium-sized cylindrical white plastic bottle\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_47895364/Validation_47895364_47895364_100229.938.png", "Validation_47895364/Validation_47895364_47895364_100231.937.png", "Validation_47895364/Validation_47895364_47895364_100233.936.png", "Validation_47895364/Validation_47895364_47895364_100235.936.png", "Validation_47895364/Validation_47895364_47895364_100237.935.png" ], "turn_id": 7, "scan_id": "Validation_47895364", "origin_question": "Please recall \"the sleek modern front-loading washing machine\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.8", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the sleek modern front-loading washing machine\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47895364/Validation_47895364_47895364_100239.934.png", "Validation_47895364/Validation_47895364_47895364_100241.933.png", "Validation_47895364/Validation_47895364_47895364_100243.932.png", "Validation_47895364/Validation_47895364_47895364_100245.931.png", "Validation_47895364/Validation_47895364_47895364_100247.931.png" ], "turn_id": 8, "scan_id": "Validation_47895364", "origin_question": "Which direction is the sleek modern front-loading washing machine in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the sleek modern front-loading washing machine in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_47895364/Validation_47895364_47895364_100249.930.png", "Validation_47895364/Validation_47895364_47895364_100251.946.png", "Validation_47895364/Validation_47895364_47895364_100253.945.png", "Validation_47895364/Validation_47895364_47895364_100255.944.png", "Validation_47895364/Validation_47895364_47895364_100257.943.png" ], "turn_id": 9, "scan_id": "Validation_47895364", "origin_question": "When did you first discover \"the sleek modern front-loading washing machine\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the sleek modern front-loading washing machine\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_47895364/Validation_47895364_47895364_100259.942.png", "Validation_47895364/Validation_47895364_47895364_100261.942.png", "Validation_47895364/Validation_47895364_47895364_100263.941.png", "Validation_47895364/Validation_47895364_47895364_100265.940.png", "Validation_47895364/Validation_47895364_47895364_100267.939.png" ], "turn_id": 10, "scan_id": "Validation_47895364", "origin_question": "Using your orientation at the end of turn 6 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "144.95", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 6 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_45662944/Validation_45662944_45662944_22942.767.png", "Validation_45662944/Validation_45662944_45662944_22944.766.png", "Validation_45662944/Validation_45662944_45662944_22946.765.png", "Validation_45662944/Validation_45662944_45662944_22948.764.png" ], "turn_id": 1, "scan_id": "Validation_45662944", "origin_question": "Remember, have you seen any lamp(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any lamp(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_45662944/Validation_45662944_45662944_22950.763.png", "Validation_45662944/Validation_45662944_45662944_22952.762.png", "Validation_45662944/Validation_45662944_45662944_22954.762.png", "Validation_45662944/Validation_45662944_45662944_22956.761.png" ], "turn_id": 2, "scan_id": "Validation_45662944", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the sleek gray upholstered bed with white duvet cover\", Y: \"the sleek white chair with dark accents\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the sleek gray upholstered bed with white duvet cover\", Y: \"the sleek white chair with dark accents\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "Validation_45662944/Validation_45662944_45662944_22958.760.png", "Validation_45662944/Validation_45662944_45662944_22960.759.png", "Validation_45662944/Validation_45662944_45662944_22962.758.png", "Validation_45662944/Validation_45662944_45662944_22964.757.png" ], "turn_id": 3, "scan_id": "Validation_45662944", "origin_question": "Among these three objects(\"the sleek gray upholstered bed with white duvet cover\",\"the sleek white rectangular table made of composite material\",\"the sleek white chair with dark accents\" ), which one is the farthest from you?", "option": [ "the sleek gray upholstered bed with white duvet cover", "the sleek white rectangular table made of composite material", "the sleek white chair with dark accents" ], "answer": "the sleek white chair with dark accents", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects(\"the sleek gray upholstered bed with white duvet cover\",\"the sleek white rectangular table made of composite material\",\"the sleek white chair with dark accents\" ), which one is the farthest from you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the sleek gray upholstered bed with white duvet cover', 'the sleek white rectangular table made of composite material', 'the sleek white chair with dark accents'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_45662944/Validation_45662944_45662944_22966.757.png", "Validation_45662944/Validation_45662944_45662944_22968.756.png", "Validation_45662944/Validation_45662944_45662944_22970.755.png", "Validation_45662944/Validation_45662944_45662944_22972.771.png" ], "turn_id": 4, "scan_id": "Validation_45662944", "origin_question": "When did you first discover \"the sleek white chair with dark accents\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the sleek white chair with dark accents\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_45662944/Validation_45662944_45662944_22974.770.png", "Validation_45662944/Validation_45662944_45662944_22976.769.png", "Validation_45662944/Validation_45662944_45662944_22978.768.png", "Validation_45662944/Validation_45662944_45662944_22980.767.png" ], "turn_id": 5, "scan_id": "Validation_45662944", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the sleek, rectangular mirror with the thin metallic frame\" from your position?", "option": [], "answer": "138.68", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the sleek, rectangular mirror with the thin metallic frame\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_45662944/Validation_45662944_45662944_22982.767.png", "Validation_45662944/Validation_45662944_45662944_22984.766.png", "Validation_45662944/Validation_45662944_45662944_22986.765.png", "Validation_45662944/Validation_45662944_45662944_22988.764.png" ], "turn_id": 6, "scan_id": "Validation_45662944", "origin_question": "Please recall \"the sleek, rectangular mirror with the thin metallic frame\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.12", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the sleek, rectangular mirror with the thin metallic frame\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_45662944/Validation_45662944_45662944_22990.763.png", "Validation_45662944/Validation_45662944_45662944_22992.762.png", "Validation_45662944/Validation_45662944_45662944_22994.762.png", "Validation_45662944/Validation_45662944_45662944_22996.761.png" ], "turn_id": 7, "scan_id": "Validation_45662944", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the sleek white chair with dark accents\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the sleek white chair with dark accents\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_45662944/Validation_45662944_45662944_22998.760.png", "Validation_45662944/Validation_45662944_45662944_23000.759.png", "Validation_45662944/Validation_45662944_45662944_23002.758.png", "Validation_45662944/Validation_45662944_45662944_23004.757.png" ], "turn_id": 8, "scan_id": "Validation_45662944", "origin_question": "How far is your current position from where you were at the end of the turn 1? (in meters)", "option": [], "answer": "2.43", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 1? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_48018562/Validation_48018562_48018562_3449.156.png", "Validation_48018562/Validation_48018562_48018562_3450.639.png", "Validation_48018562/Validation_48018562_48018562_3452.139.png", "Validation_48018562/Validation_48018562_48018562_3453.655.png" ], "turn_id": 1, "scan_id": "Validation_48018562", "origin_question": "Remember, have you seen any cabinet(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any cabinet(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_48018562/Validation_48018562_48018562_3455.154.png", "Validation_48018562/Validation_48018562_48018562_3456.653.png", "Validation_48018562/Validation_48018562_48018562_3458.153.png", "Validation_48018562/Validation_48018562_48018562_3459.652.png" ], "turn_id": 2, "scan_id": "Validation_48018562", "origin_question": "Please recall \"the sleek stainless steel kettle\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.89", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the sleek stainless steel kettle\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_48018562/Validation_48018562_48018562_3461.152.png", "Validation_48018562/Validation_48018562_48018562_3462.651.png", "Validation_48018562/Validation_48018562_48018562_3464.150.png", "Validation_48018562/Validation_48018562_48018562_3465.650.png" ], "turn_id": 3, "scan_id": "Validation_48018562", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the compact wooden step stool with the light brown finish\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the compact wooden step stool with the light brown finish\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_48018562/Validation_48018562_48018562_3467.149.png", "Validation_48018562/Validation_48018562_48018562_3468.648.png", "Validation_48018562/Validation_48018562_48018562_3470.148.png", "Validation_48018562/Validation_48018562_48018562_3471.647.png" ], "turn_id": 4, "scan_id": "Validation_48018562", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the muted gray, patterned towel\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the muted gray, patterned towel\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_45261581/Validation_45261581_45261581_77516.524.png", "Validation_45261581/Validation_45261581_45261581_77518.506.png", "Validation_45261581/Validation_45261581_45261581_77520.505.png", "Validation_45261581/Validation_45261581_45261581_77522.505.png", "Validation_45261581/Validation_45261581_45261581_77524.504.png" ], "turn_id": 1, "scan_id": "Validation_45261581", "origin_question": "Remember, have you seen any monitor(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any monitor(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_45261581/Validation_45261581_45261581_77526.503.png", "Validation_45261581/Validation_45261581_45261581_77528.519.png", "Validation_45261581/Validation_45261581_45261581_77530.518.png", "Validation_45261581/Validation_45261581_45261581_77532.517.png", "Validation_45261581/Validation_45261581_45261581_77534.516.png" ], "turn_id": 2, "scan_id": "Validation_45261581", "origin_question": "When did you first discover \"the compact single bed with the white wooden frame\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the compact single bed with the white wooden frame\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_45261581/Validation_45261581_45261581_77536.515.png", "Validation_45261581/Validation_45261581_45261581_77538.515.png", "Validation_45261581/Validation_45261581_45261581_77540.514.png", "Validation_45261581/Validation_45261581_45261581_77542.513.png", "Validation_45261581/Validation_45261581_45261581_77544.512.png" ], "turn_id": 3, "scan_id": "Validation_45261581", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the large, sleek black monitor with thin bezels\" from your position?", "option": [], "answer": "118.36", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the large, sleek black monitor with thin bezels\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_45261581/Validation_45261581_45261581_77546.511.png", "Validation_45261581/Validation_45261581_45261581_77548.510.png", "Validation_45261581/Validation_45261581_45261581_77550.510.png", "Validation_45261581/Validation_45261581_45261581_77552.509.png", "Validation_45261581/Validation_45261581_45261581_77554.508.png" ], "turn_id": 4, "scan_id": "Validation_45261581", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the compact single bed with the white wooden frame\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the compact single bed with the white wooden frame\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_45261581/Validation_45261581_45261581_77556.507.png", "Validation_45261581/Validation_45261581_45261581_77558.506.png", "Validation_45261581/Validation_45261581_45261581_77560.505.png", "Validation_45261581/Validation_45261581_45261581_77562.505.png", "Validation_45261581/Validation_45261581_45261581_77564.504.png" ], "turn_id": 5, "scan_id": "Validation_45261581", "origin_question": "At the end of which turn were both of \"the compact single bed with the white wooden frame\" and \"the large, sleek black monitor with thin bezels\" on your left side?", "option": [], "answer": "5", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the compact single bed with the white wooden frame\" and \"the large, sleek black monitor with thin bezels\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_45261581/Validation_45261581_45261581_77566.503.png", "Validation_45261581/Validation_45261581_45261581_77568.519.png", "Validation_45261581/Validation_45261581_45261581_77570.518.png", "Validation_45261581/Validation_45261581_45261581_77572.517.png", "Validation_45261581/Validation_45261581_45261581_77574.516.png" ], "turn_id": 6, "scan_id": "Validation_45261581", "origin_question": "Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the compact single bed with the white wooden frame\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the compact single bed with the white wooden frame\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_45261581/Validation_45261581_45261581_77576.515.png", "Validation_45261581/Validation_45261581_45261581_77578.515.png", "Validation_45261581/Validation_45261581_45261581_77580.514.png", "Validation_45261581/Validation_45261581_45261581_77582.513.png", "Validation_45261581/Validation_45261581_45261581_77584.512.png" ], "turn_id": 7, "scan_id": "Validation_45261581", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the compact single bed with the white wooden frame\", Y: \"the large, sleek black monitor with thin bezels\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the compact single bed with the white wooden frame\", Y: \"the large, sleek black monitor with thin bezels\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_45261581/Validation_45261581_45261581_77586.511.png", "Validation_45261581/Validation_45261581_45261581_77588.511.png", "Validation_45261581/Validation_45261581_45261581_77590.510.png", "Validation_45261581/Validation_45261581_45261581_77592.509.png", "Validation_45261581/Validation_45261581_45261581_77594.508.png" ], "turn_id": 8, "scan_id": "Validation_45261581", "origin_question": "Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the compact single bed with the white wooden frame\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the compact single bed with the white wooden frame\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_45261581/Validation_45261581_45261581_77596.507.png", "Validation_45261581/Validation_45261581_45261581_77598.506.png", "Validation_45261581/Validation_45261581_45261581_77600.505.png", "Validation_45261581/Validation_45261581_45261581_77602.505.png", "Validation_45261581/Validation_45261581_45261581_77604.504.png" ], "turn_id": 9, "scan_id": "Validation_45261581", "origin_question": "", "option": [], "answer": "", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_48458665/Validation_48458665_48458665_8155.176.png", "Validation_48458665/Validation_48458665_48458665_8156.658.png", "Validation_48458665/Validation_48458665_48458665_8158.158.png", "Validation_48458665/Validation_48458665_48458665_8159.657.png" ], "turn_id": 1, "scan_id": "Validation_48458665", "origin_question": "Remember, have you seen any jacket(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any jacket(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_48458665/Validation_48458665_48458665_8161.157.png", "Validation_48458665/Validation_48458665_48458665_8162.656.png", "Validation_48458665/Validation_48458665_48458665_8164.155.png", "Validation_48458665/Validation_48458665_48458665_8165.655.png" ], "turn_id": 2, "scan_id": "Validation_48458665", "origin_question": "When did you first discover \"the dark ergonomic chair with the contoured backrest\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the dark ergonomic chair with the contoured backrest\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_48458665/Validation_48458665_48458665_8167.154.png", "Validation_48458665/Validation_48458665_48458665_8168.654.png", "Validation_48458665/Validation_48458665_48458665_8170.153.png", "Validation_48458665/Validation_48458665_48458665_8171.652.png" ], "turn_id": 3, "scan_id": "Validation_48458665", "origin_question": "Which two objects are on the same side of you now? \"the bed\", \"the glossy poster with soft earth tones and pastels\", \"the sleek black rectangular monitor with the slim frame\". ", "option": [ "the bed, the glossy poster with soft earth tones and pastels", "the bed, the sleek black rectangular monitor with the slim frame", "the glossy poster with soft earth tones and pastels, the sleek black rectangular monitor with the slim frame" ], "answer": "the bed, the glossy poster with soft earth tones and pastels", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the bed\", \"the glossy poster with soft earth tones and pastels\", \"the sleek black rectangular monitor with the slim frame\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the bed, the glossy poster with soft earth tones and pastels', 'the bed, the sleek black rectangular monitor with the slim frame', 'the glossy poster with soft earth tones and pastels, the sleek black rectangular monitor with the slim frame'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "Validation_48458665/Validation_48458665_48458665_8173.152.png", "Validation_48458665/Validation_48458665_48458665_8174.651.png", "Validation_48458665/Validation_48458665_48458665_8176.150.png", "Validation_48458665/Validation_48458665_48458665_8177.650.png" ], "turn_id": 4, "scan_id": "Validation_48458665", "origin_question": "Among these three objects(\"the bed\",\"the dark ergonomic chair with the contoured backrest\",\"the glossy poster with soft earth tones and pastels\" ), which one is the farthest from you?", "option": [ "the bed", "the dark ergonomic chair with the contoured backrest", "the glossy poster with soft earth tones and pastels" ], "answer": "the glossy poster with soft earth tones and pastels", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects(\"the bed\",\"the dark ergonomic chair with the contoured backrest\",\"the glossy poster with soft earth tones and pastels\" ), which one is the farthest from you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the bed', 'the dark ergonomic chair with the contoured backrest', 'the glossy poster with soft earth tones and pastels'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_48018347/Validation_48018347_48018347_388724.150.png", "Validation_48018347/Validation_48018347_48018347_388725.633.png", "Validation_48018347/Validation_48018347_48018347_388727.149.png", "Validation_48018347/Validation_48018347_48018347_388728.648.png" ], "turn_id": 1, "scan_id": "Validation_48018347", "origin_question": "", "option": [], "answer": "", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_48018347/Validation_48018347_48018347_388730.148.png", "Validation_48018347/Validation_48018347_48018347_388731.647.png", "Validation_48018347/Validation_48018347_48018347_388733.146.png", "Validation_48018347/Validation_48018347_48018347_388734.646.png" ], "turn_id": 2, "scan_id": "Validation_48018347", "origin_question": "When did you first discover \"the compact bed with the colorful striped pattern\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the compact bed with the colorful striped pattern\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_48018347/Validation_48018347_48018347_388736.145.png", "Validation_48018347/Validation_48018347_48018347_388737.644.png", "Validation_48018347/Validation_48018347_48018347_388739.144.png", "Validation_48018347/Validation_48018347_48018347_388740.643.png" ], "turn_id": 3, "scan_id": "Validation_48018347", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "Validation_48018347/Validation_48018347_48018347_388742.143.png", "Validation_48018347/Validation_48018347_48018347_388743.642.png", "Validation_48018347/Validation_48018347_48018347_388745.141.png", "Validation_48018347/Validation_48018347_48018347_388746.641.png" ], "turn_id": 4, "scan_id": "Validation_48018347", "origin_question": "Among these three objects(\"the compact bed with the colorful striped pattern\",\"the sleek, narrow beige wood case\",\"the classic cylindrical brown flowerpot made of clay\" ), which one is the closest to you?", "option": [ "the compact bed with the colorful striped pattern", "the sleek, narrow beige wood case", "the classic cylindrical brown flowerpot made of clay" ], "answer": "the sleek, narrow beige wood case", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects(\"the compact bed with the colorful striped pattern\",\"the sleek, narrow beige wood case\",\"the classic cylindrical brown flowerpot made of clay\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the compact bed with the colorful striped pattern', 'the sleek, narrow beige wood case', 'the classic cylindrical brown flowerpot made of clay'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_48018347/Validation_48018347_48018347_388748.140.png", "Validation_48018347/Validation_48018347_48018347_388749.639.png", "Validation_48018347/Validation_48018347_48018347_388751.139.png", "Validation_48018347/Validation_48018347_48018347_388752.638.png" ], "turn_id": 5, "scan_id": "Validation_48018347", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the compact bed with the colorful striped pattern\", Y: \"the compact calendar with the sleek white surface\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the compact bed with the colorful striped pattern\", Y: \"the compact calendar with the sleek white surface\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_48018347/Validation_48018347_48018347_388754.138.png", "Validation_48018347/Validation_48018347_48018347_388755.637.png", "Validation_48018347/Validation_48018347_48018347_388757.136.png", "Validation_48018347/Validation_48018347_48018347_388758.636.png" ], "turn_id": 6, "scan_id": "Validation_48018347", "origin_question": "By the end of the 3 turn, were you getting closer to or farther away from these two objects: X: \"the sleek, narrow beige wood case\", Y: \"the compact minimalist chair with the neutral color\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 3 turn, were you getting closer to or farther away from these two objects: X: \"the sleek, narrow beige wood case\", Y: \"the compact minimalist chair with the neutral color\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_48018347/Validation_48018347_48018347_388760.135.png", "Validation_48018347/Validation_48018347_48018347_388761.634.png", "Validation_48018347/Validation_48018347_48018347_388763.134.png", "Validation_48018347/Validation_48018347_48018347_388764.633.png" ], "turn_id": 7, "scan_id": "Validation_48018347", "origin_question": "In which turn were you closest to \"the light\"? ", "option": [], "answer": "6", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the light\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "Validation_41125696/Validation_41125696_41125696_3575.959.png", "Validation_41125696/Validation_41125696_41125696_3577.942.png", "Validation_41125696/Validation_41125696_41125696_3579.941.png", "Validation_41125696/Validation_41125696_41125696_3581.940.png", "Validation_41125696/Validation_41125696_41125696_3583.939.png" ], "turn_id": 1, "scan_id": "Validation_41125696", "origin_question": "Remember, how many bin(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many bin(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "Validation_41125696/Validation_41125696_41125696_3585.939.png", "Validation_41125696/Validation_41125696_41125696_3587.938.png", "Validation_41125696/Validation_41125696_41125696_3589.937.png", "Validation_41125696/Validation_41125696_41125696_3591.936.png", "Validation_41125696/Validation_41125696_41125696_3593.936.png" ], "turn_id": 2, "scan_id": "Validation_41125696", "origin_question": "Remember, how many table(s) have you seen so far? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many table(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_41125696/Validation_41125696_41125696_3595.951.png", "Validation_41125696/Validation_41125696_41125696_3597.950.png", "Validation_41125696/Validation_41125696_41125696_3599.950.png", "Validation_41125696/Validation_41125696_41125696_3601.949.png", "Validation_41125696/Validation_41125696_41125696_3603.948.png" ], "turn_id": 3, "scan_id": "Validation_41125696", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the large, sleek black monitor with the glossy screen\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the large, sleek black monitor with the glossy screen\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_41125696/Validation_41125696_41125696_3605.947.png", "Validation_41125696/Validation_41125696_41125696_3607.947.png", "Validation_41125696/Validation_41125696_41125696_3609.946.png", "Validation_41125696/Validation_41125696_41125696_3611.945.png", "Validation_41125696/Validation_41125696_41125696_3613.944.png" ], "turn_id": 4, "scan_id": "Validation_41125696", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the large, sleek black monitor with the glossy screen\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the large, sleek black monitor with the glossy screen\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_41125696/Validation_41125696_41125696_3615.943.png", "Validation_41125696/Validation_41125696_41125696_3617.942.png", "Validation_41125696/Validation_41125696_41125696_3619.942.png", "Validation_41125696/Validation_41125696_41125696_3621.941.png", "Validation_41125696/Validation_41125696_41125696_3623.940.png" ], "turn_id": 5, "scan_id": "Validation_41125696", "origin_question": "At the end of which turn were both of \"the large, sleek black monitor with the glossy screen\" and \"the oven\" on your right side?", "option": [], "answer": "1", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the large, sleek black monitor with the glossy screen\" and \"the oven\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_41125696/Validation_41125696_41125696_3625.939.png", "Validation_41125696/Validation_41125696_41125696_3627.938.png", "Validation_41125696/Validation_41125696_41125696_3629.938.png", "Validation_41125696/Validation_41125696_41125696_3631.937.png", "Validation_41125696/Validation_41125696_41125696_3633.936.png" ], "turn_id": 6, "scan_id": "Validation_41125696", "origin_question": "In which turn did you see both \"the large, sleek black monitor with the glossy screen\" and \"the modern stainless steel range hood with the sleek gray finish\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the large, sleek black monitor with the glossy screen\" and \"the modern stainless steel range hood with the sleek gray finish\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_41125696/Validation_41125696_41125696_3635.952.png", "Validation_41125696/Validation_41125696_41125696_3637.951.png", "Validation_41125696/Validation_41125696_41125696_3639.950.png", "Validation_41125696/Validation_41125696_41125696_3641.949.png", "Validation_41125696/Validation_41125696_41125696_3643.949.png" ], "turn_id": 7, "scan_id": "Validation_41125696", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the large, sleek black monitor with the glossy screen\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the large, sleek black monitor with the glossy screen\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_41125696/Validation_41125696_41125696_3645.948.png", "Validation_41125696/Validation_41125696_41125696_3647.947.png", "Validation_41125696/Validation_41125696_41125696_3649.946.png", "Validation_41125696/Validation_41125696_41125696_3651.945.png", "Validation_41125696/Validation_41125696_41125696_3653.945.png" ], "turn_id": 8, "scan_id": "Validation_41125696", "origin_question": "Using your orientation at the end of turn 7 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 7 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_41125696/Validation_41125696_41125696_3655.944.png", "Validation_41125696/Validation_41125696_41125696_3657.943.png", "Validation_41125696/Validation_41125696_41125696_3659.942.png", "Validation_41125696/Validation_41125696_41125696_3661.941.png", "Validation_41125696/Validation_41125696_41125696_3663.941.png" ], "turn_id": 9, "scan_id": "Validation_41125696", "origin_question": "", "option": [], "answer": "", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_41125696/Validation_41125696_41125696_3665.940.png", "Validation_41125696/Validation_41125696_41125696_3667.939.png", "Validation_41125696/Validation_41125696_41125696_3669.938.png", "Validation_41125696/Validation_41125696_41125696_3671.937.png", "Validation_41125696/Validation_41125696_41125696_3673.936.png" ], "turn_id": 10, "scan_id": "Validation_41125696", "origin_question": "Is the oven to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the oven to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47895552/Validation_47895552_47895552_457686.071.png", "Validation_47895552/Validation_47895552_47895552_457687.554.png", "Validation_47895552/Validation_47895552_47895552_457689.053.png", "Validation_47895552/Validation_47895552_47895552_457690.552.png" ], "turn_id": 1, "scan_id": "Validation_47895552", "origin_question": "Remember, have you seen any teapot(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any teapot(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_47895552/Validation_47895552_47895552_457692.052.png", "Validation_47895552/Validation_47895552_47895552_457693.551.png", "Validation_47895552/Validation_47895552_47895552_457695.051.png", "Validation_47895552/Validation_47895552_47895552_457696.550.png" ], "turn_id": 2, "scan_id": "Validation_47895552", "origin_question": "In which turn did you see both \"the glossy white oval-shaped toilet made of vitreous china\" and \"the radiator\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the glossy white oval-shaped toilet made of vitreous china\" and \"the radiator\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47895552/Validation_47895552_47895552_457698.049.png", "Validation_47895552/Validation_47895552_47895552_457699.549.png", "Validation_47895552/Validation_47895552_47895552_457701.048.png", "Validation_47895552/Validation_47895552_47895552_457702.548.png" ], "turn_id": 3, "scan_id": "Validation_47895552", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the glossy white oval-shaped toilet made of vitreous china\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the glossy white oval-shaped toilet made of vitreous china\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_47895552/Validation_47895552_47895552_457704.047.png", "Validation_47895552/Validation_47895552_47895552_457705.546.png", "Validation_47895552/Validation_47895552_47895552_457707.046.png", "Validation_47895552/Validation_47895552_47895552_457708.545.png" ], "turn_id": 4, "scan_id": "Validation_47895552", "origin_question": "Is the sleek rectangular white washing machine to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the sleek rectangular white washing machine to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_47895552/Validation_47895552_47895552_457710.045.png", "Validation_47895552/Validation_47895552_47895552_457711.544.png", "Validation_47895552/Validation_47895552_47895552_457713.043.png", "Validation_47895552/Validation_47895552_47895552_457714.543.png" ], "turn_id": 5, "scan_id": "Validation_47895552", "origin_question": "At the end of which turn were both of \"the sleek rectangular white washing machine\" and \"the radiator\" on your left side?", "option": [], "answer": "2", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the sleek rectangular white washing machine\" and \"the radiator\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47895552/Validation_47895552_47895552_457716.042.png", "Validation_47895552/Validation_47895552_47895552_457717.558.png", "Validation_47895552/Validation_47895552_47895552_457719.058.png", "Validation_47895552/Validation_47895552_47895552_457720.557.png" ], "turn_id": 6, "scan_id": "Validation_47895552", "origin_question": "Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the glossy white oval-shaped toilet made of vitreous china\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the glossy white oval-shaped toilet made of vitreous china\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_47895552/Validation_47895552_47895552_457722.056.png", "Validation_47895552/Validation_47895552_47895552_457723.556.png", "Validation_47895552/Validation_47895552_47895552_457725.055.png", "Validation_47895552/Validation_47895552_47895552_457726.555.png" ], "turn_id": 7, "scan_id": "Validation_47895552", "origin_question": "Please recall \"the radiator\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.75", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the radiator\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_47895552/Validation_47895552_47895552_457728.054.png", "Validation_47895552/Validation_47895552_47895552_457729.553.png", "Validation_47895552/Validation_47895552_47895552_457731.053.png", "Validation_47895552/Validation_47895552_47895552_457732.552.png" ], "turn_id": 8, "scan_id": "Validation_47895552", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the red cylindrical plastic bucket with the handle\" from your position?", "option": [], "answer": "139.0", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the red cylindrical plastic bucket with the handle\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_44358583/Validation_44358583_44358583_55706.438.png", "Validation_44358583/Validation_44358583_44358583_55708.420.png", "Validation_44358583/Validation_44358583_44358583_55710.420.png", "Validation_44358583/Validation_44358583_44358583_55712.435.png" ], "turn_id": 1, "scan_id": "Validation_44358583", "origin_question": "Remember, have you seen any cabinet(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any cabinet(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_44358583/Validation_44358583_44358583_55714.435.png", "Validation_44358583/Validation_44358583_44358583_55716.434.png", "Validation_44358583/Validation_44358583_44358583_55718.433.png", "Validation_44358583/Validation_44358583_44358583_55720.432.png" ], "turn_id": 2, "scan_id": "Validation_44358583", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_44358583/Validation_44358583_44358583_55722.431.png", "Validation_44358583/Validation_44358583_44358583_55724.431.png", "Validation_44358583/Validation_44358583_44358583_55726.430.png", "Validation_44358583/Validation_44358583_44358583_55728.429.png" ], "turn_id": 3, "scan_id": "Validation_44358583", "origin_question": "Remember, have you seen any machine(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any machine(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_44358583/Validation_44358583_44358583_55730.428.png", "Validation_44358583/Validation_44358583_44358583_55732.427.png", "Validation_44358583/Validation_44358583_44358583_55734.427.png", "Validation_44358583/Validation_44358583_44358583_55736.426.png" ], "turn_id": 4, "scan_id": "Validation_44358583", "origin_question": "Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "137.83", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_44358583/Validation_44358583_44358583_55738.425.png", "Validation_44358583/Validation_44358583_44358583_55740.424.png", "Validation_44358583/Validation_44358583_44358583_55742.423.png", "Validation_44358583/Validation_44358583_44358583_55744.423.png" ], "turn_id": 5, "scan_id": "Validation_44358583", "origin_question": "Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_44358583/Validation_44358583_44358583_55746.422.png", "Validation_44358583/Validation_44358583_44358583_55748.421.png", "Validation_44358583/Validation_44358583_44358583_55750.420.png", "Validation_44358583/Validation_44358583_44358583_55752.420.png" ], "turn_id": 6, "scan_id": "Validation_44358583", "origin_question": "Using your orientation at the end of turn 4 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "24.74", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 4 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_44358583/Validation_44358583_44358583_55754.435.png", "Validation_44358583/Validation_44358583_44358583_55756.435.png", "Validation_44358583/Validation_44358583_44358583_55758.434.png", "Validation_44358583/Validation_44358583_44358583_55760.433.png" ], "turn_id": 7, "scan_id": "Validation_44358583", "origin_question": "Compared to your position at the end of the 6 turn, are you now closer or farther away from \"the sleek white wooden bed with the minimalist design\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 6 turn, are you now closer or farther away from \"the sleek white wooden bed with the minimalist design\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47334360/Validation_47334360_47334360_61948.335.png", "Validation_47334360/Validation_47334360_47334360_61950.317.png", "Validation_47334360/Validation_47334360_47334360_61952.316.png", "Validation_47334360/Validation_47334360_47334360_61954.316.png", "Validation_47334360/Validation_47334360_47334360_61956.315.png" ], "turn_id": 1, "scan_id": "Validation_47334360", "origin_question": "Remember, have you seen any curtain(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any curtain(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47334360/Validation_47334360_47334360_61958.314.png", "Validation_47334360/Validation_47334360_47334360_61960.330.png", "Validation_47334360/Validation_47334360_47334360_61962.329.png", "Validation_47334360/Validation_47334360_47334360_61964.328.png", "Validation_47334360/Validation_47334360_47334360_61966.327.png" ], "turn_id": 2, "scan_id": "Validation_47334360", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_47334360/Validation_47334360_47334360_61968.327.png", "Validation_47334360/Validation_47334360_47334360_61970.326.png", "Validation_47334360/Validation_47334360_47334360_61972.325.png", "Validation_47334360/Validation_47334360_47334360_61974.324.png", "Validation_47334360/Validation_47334360_47334360_61976.323.png" ], "turn_id": 3, "scan_id": "Validation_47334360", "origin_question": "Which two objects are on the same side of you now? \"the soft beige couch with the medium-height backrest\", \"the rectangular soft grey wooden door with the metallic handle\", \"the sleek black fabric bag with the vertical design\". ", "option": [ "the soft beige couch with the medium-height backrest, the rectangular soft grey wooden door with the metallic handle", "the soft beige couch with the medium-height backrest, the sleek black fabric bag with the vertical design", "the rectangular soft grey wooden door with the metallic handle, the sleek black fabric bag with the vertical design" ], "answer": "the rectangular soft grey wooden door with the metallic handle, the sleek black fabric bag with the vertical design", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the soft beige couch with the medium-height backrest\", \"the rectangular soft grey wooden door with the metallic handle\", \"the sleek black fabric bag with the vertical design\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the soft beige couch with the medium-height backrest, the rectangular soft grey wooden door with the metallic handle', 'the soft beige couch with the medium-height backrest, the sleek black fabric bag with the vertical design', 'the rectangular soft grey wooden door with the metallic handle, the sleek black fabric bag with the vertical design'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_47334360/Validation_47334360_47334360_61978.323.png", "Validation_47334360/Validation_47334360_47334360_61980.322.png", "Validation_47334360/Validation_47334360_47334360_61982.321.png", "Validation_47334360/Validation_47334360_47334360_61984.320.png", "Validation_47334360/Validation_47334360_47334360_61986.319.png" ], "turn_id": 4, "scan_id": "Validation_47334360", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "103.5", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47334360/Validation_47334360_47334360_61988.319.png", "Validation_47334360/Validation_47334360_47334360_61990.318.png", "Validation_47334360/Validation_47334360_47334360_61992.317.png", "Validation_47334360/Validation_47334360_47334360_61994.316.png", "Validation_47334360/Validation_47334360_47334360_61996.315.png" ], "turn_id": 5, "scan_id": "Validation_47334360", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the floor-length sheer white curtain made of gauzy material\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the floor-length sheer white curtain made of gauzy material\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47334360/Validation_47334360_47334360_61998.314.png", "Validation_47334360/Validation_47334360_47334360_62000.330.png", "Validation_47334360/Validation_47334360_47334360_62002.329.png", "Validation_47334360/Validation_47334360_47334360_62004.329.png", "Validation_47334360/Validation_47334360_47334360_62006.328.png" ], "turn_id": 6, "scan_id": "Validation_47334360", "origin_question": "In which turn were you farthest from \"the rectangular soft grey wooden door with the metallic handle\"? ", "option": [], "answer": "6", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the rectangular soft grey wooden door with the metallic handle\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47334360/Validation_47334360_47334360_62008.327.png", "Validation_47334360/Validation_47334360_47334360_62010.326.png", "Validation_47334360/Validation_47334360_47334360_62012.325.png", "Validation_47334360/Validation_47334360_47334360_62014.325.png", "Validation_47334360/Validation_47334360_47334360_62016.324.png" ], "turn_id": 7, "scan_id": "Validation_47334360", "origin_question": "Which direction is the rectangular soft grey wooden door with the metallic handle in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the rectangular soft grey wooden door with the metallic handle in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_47334360/Validation_47334360_47334360_62018.323.png", "Validation_47334360/Validation_47334360_47334360_62020.322.png", "Validation_47334360/Validation_47334360_47334360_62022.321.png", "Validation_47334360/Validation_47334360_47334360_62024.321.png", "Validation_47334360/Validation_47334360_47334360_62026.320.png" ], "turn_id": 8, "scan_id": "Validation_47334360", "origin_question": "At the end of which turn were both of \"the rectangular soft grey wooden door with the metallic handle\" and \"the floor-length sheer white curtain made of gauzy material\" on your right side?", "option": [], "answer": "8", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the rectangular soft grey wooden door with the metallic handle\" and \"the floor-length sheer white curtain made of gauzy material\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47334360/Validation_47334360_47334360_62028.319.png", "Validation_47334360/Validation_47334360_47334360_62030.318.png", "Validation_47334360/Validation_47334360_47334360_62032.317.png", "Validation_47334360/Validation_47334360_47334360_62034.317.png", "Validation_47334360/Validation_47334360_47334360_62036.316.png" ], "turn_id": 9, "scan_id": "Validation_47334360", "origin_question": "Which direction is the sleek black fabric bag with the vertical design in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the sleek black fabric bag with the vertical design in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_47334360/Validation_47334360_47334360_62038.315.png", "Validation_47334360/Validation_47334360_47334360_62040.314.png", "Validation_47334360/Validation_47334360_47334360_62042.330.png", "Validation_47334360/Validation_47334360_47334360_62044.329.png", "Validation_47334360/Validation_47334360_47334360_62046.328.png" ], "turn_id": 10, "scan_id": "Validation_47334360", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the sleek black fabric bag with the vertical design\" from your position?", "option": [], "answer": "146.56", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the sleek black fabric bag with the vertical design\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42897549/Validation_42897549_42897549_276488.376.png", "Validation_42897549/Validation_42897549_42897549_276489.859.png", "Validation_42897549/Validation_42897549_42897549_276491.358.png", "Validation_42897549/Validation_42897549_42897549_276492.874.png" ], "turn_id": 1, "scan_id": "Validation_42897549", "origin_question": "Remember, have you seen any window(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any window(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_42897549/Validation_42897549_42897549_276494.373.png", "Validation_42897549/Validation_42897549_42897549_276495.873.png", "Validation_42897549/Validation_42897549_42897549_276497.372.png", "Validation_42897549/Validation_42897549_42897549_276498.872.png" ], "turn_id": 2, "scan_id": "Validation_42897549", "origin_question": "When did you first discover \"the deep blue sporty shoe with white accents\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the deep blue sporty shoe with white accents\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_42897549/Validation_42897549_42897549_276500.371.png", "Validation_42897549/Validation_42897549_42897549_276501.870.png", "Validation_42897549/Validation_42897549_42897549_276503.370.png", "Validation_42897549/Validation_42897549_42897549_276504.869.png" ], "turn_id": 3, "scan_id": "Validation_42897549", "origin_question": "What will be the first-time appearance order of the following categories: door, shoe, toilet?", "option": [ "toilet, shoe, door", "shoe, door, toilet", "toilet, door, shoe", "door, shoe, toilet" ], "answer": "shoe, door, toilet", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: door, shoe, toilet? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['toilet, shoe, door', 'shoe, door, toilet', 'toilet, door, shoe', 'door, shoe, toilet'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_42897549/Validation_42897549_42897549_276506.368.png", "Validation_42897549/Validation_42897549_42897549_276507.868.png", "Validation_42897549/Validation_42897549_42897549_276509.367.png", "Validation_42897549/Validation_42897549_42897549_276510.867.png" ], "turn_id": 4, "scan_id": "Validation_42897549", "origin_question": "Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_42897549/Validation_42897549_42897549_276512.366.png", "Validation_42897549/Validation_42897549_42897549_276513.865.png", "Validation_42897549/Validation_42897549_42897549_276515.365.png", "Validation_42897549/Validation_42897549_42897549_276516.864.png" ], "turn_id": 5, "scan_id": "Validation_42897549", "origin_question": "Which two objects are on the same side of you now? \"the bathtub\", \"the window\", \"the white rounded porcelain toilet\". ", "option": [ "the bathtub, the window", "the bathtub, the white rounded porcelain toilet", "the window, the white rounded porcelain toilet" ], "answer": "the bathtub, the window", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the bathtub\", \"the window\", \"the white rounded porcelain toilet\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the bathtub, the window', 'the bathtub, the white rounded porcelain toilet', 'the window, the white rounded porcelain toilet'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_48458428/Validation_48458428_48458428_116469.967.png", "Validation_48458428/Validation_48458428_48458428_116471.449.png", "Validation_48458428/Validation_48458428_48458428_116472.949.png", "Validation_48458428/Validation_48458428_48458428_116474.448.png" ], "turn_id": 1, "scan_id": "Validation_48458428", "origin_question": "Remember, have you seen any sink(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any sink(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_48458428/Validation_48458428_48458428_116475.947.png", "Validation_48458428/Validation_48458428_48458428_116477.447.png", "Validation_48458428/Validation_48458428_48458428_116478.946.png", "Validation_48458428/Validation_48458428_48458428_116480.446.png" ], "turn_id": 2, "scan_id": "Validation_48458428", "origin_question": "When did you first discover \"the bathtub\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the bathtub\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_48458428/Validation_48458428_48458428_116481.945.png", "Validation_48458428/Validation_48458428_48458428_116483.444.png", "Validation_48458428/Validation_48458428_48458428_116484.944.png", "Validation_48458428/Validation_48458428_48458428_116486.443.png" ], "turn_id": 3, "scan_id": "Validation_48458428", "origin_question": "Which direction is the rectangular bathroom window with the white frame in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the rectangular bathroom window with the white frame in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_48458428/Validation_48458428_48458428_116487.943.png", "Validation_48458428/Validation_48458428_48458428_116489.442.png", "Validation_48458428/Validation_48458428_48458428_116490.941.png", "Validation_48458428/Validation_48458428_48458428_116492.441.png" ], "turn_id": 4, "scan_id": "Validation_48458428", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the modern rectangular white ceramic basin\", Y: \"the rectangular bathroom window with the white frame\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the modern rectangular white ceramic basin\", Y: \"the rectangular bathroom window with the white frame\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_48458428/Validation_48458428_48458428_116493.940.png", "Validation_48458428/Validation_48458428_48458428_116495.439.png", "Validation_48458428/Validation_48458428_48458428_116496.939.png", "Validation_48458428/Validation_48458428_48458428_116498.438.png" ], "turn_id": 5, "scan_id": "Validation_48458428", "origin_question": "", "option": [], "answer": "", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47332904/Validation_47332904_47332904_11078.408.png", "Validation_47332904/Validation_47332904_47332904_11080.391.png", "Validation_47332904/Validation_47332904_47332904_11082.390.png", "Validation_47332904/Validation_47332904_47332904_11084.389.png", "Validation_47332904/Validation_47332904_47332904_11086.389.png" ], "turn_id": 1, "scan_id": "Validation_47332904", "origin_question": "Remember, have you seen any tissue box(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any tissue box(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "Validation_47332904/Validation_47332904_47332904_11088.388.png", "Validation_47332904/Validation_47332904_47332904_11090.387.png", "Validation_47332904/Validation_47332904_47332904_11092.386.png", "Validation_47332904/Validation_47332904_47332904_11094.385.png", "Validation_47332904/Validation_47332904_47332904_11096.385.png" ], "turn_id": 2, "scan_id": "Validation_47332904", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the window\"; \"the sleek stainless steel refrigerator in the compact design\"; \"the compact stainless steel stove with the silver finish\" ", "option": [ "the window", "the sleek stainless steel refrigerator in the compact design", "the compact stainless steel stove with the silver finish" ], "answer": "the compact stainless steel stove with the silver finish", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the window\"; \"the sleek stainless steel refrigerator in the compact design\"; \"the compact stainless steel stove with the silver finish\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the window', 'the sleek stainless steel refrigerator in the compact design', 'the compact stainless steel stove with the silver finish'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_47332904/Validation_47332904_47332904_11098.400.png", "Validation_47332904/Validation_47332904_47332904_11100.400.png", "Validation_47332904/Validation_47332904_47332904_11102.399.png", "Validation_47332904/Validation_47332904_47332904_11104.398.png", "Validation_47332904/Validation_47332904_47332904_11106.397.png" ], "turn_id": 3, "scan_id": "Validation_47332904", "origin_question": "At the end of which turn were both of \"the sleek stainless steel refrigerator in the compact design\" and \"the window\" on your left side?", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the sleek stainless steel refrigerator in the compact design\" and \"the window\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47332904/Validation_47332904_47332904_11108.396.png", "Validation_47332904/Validation_47332904_47332904_11110.396.png", "Validation_47332904/Validation_47332904_47332904_11112.395.png", "Validation_47332904/Validation_47332904_47332904_11114.394.png", "Validation_47332904/Validation_47332904_47332904_11116.393.png" ], "turn_id": 4, "scan_id": "Validation_47332904", "origin_question": "Which direction is the rectangular light-colored engineered wood board in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the rectangular light-colored engineered wood board in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_47332904/Validation_47332904_47332904_11118.392.png", "Validation_47332904/Validation_47332904_47332904_11120.391.png", "Validation_47332904/Validation_47332904_47332904_11122.391.png", "Validation_47332904/Validation_47332904_47332904_11124.390.png", "Validation_47332904/Validation_47332904_47332904_11126.389.png" ], "turn_id": 5, "scan_id": "Validation_47332904", "origin_question": "In which turn did you see both \"the compact black microwave with the glossy front\" and \"the rectangular light-colored engineered wood board\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the compact black microwave with the glossy front\" and \"the rectangular light-colored engineered wood board\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_47332904/Validation_47332904_47332904_11128.388.png", "Validation_47332904/Validation_47332904_47332904_11130.388.png", "Validation_47332904/Validation_47332904_47332904_11132.387.png", "Validation_47332904/Validation_47332904_47332904_11134.386.png", "Validation_47332904/Validation_47332904_47332904_11136.385.png" ], "turn_id": 6, "scan_id": "Validation_47332904", "origin_question": "Please recall \"the rectangular light-colored wooden table\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.86", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the rectangular light-colored wooden table\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47332904/Validation_47332904_47332904_11138.401.png", "Validation_47332904/Validation_47332904_47332904_11140.400.png", "Validation_47332904/Validation_47332904_47332904_11142.399.png", "Validation_47332904/Validation_47332904_47332904_11144.398.png", "Validation_47332904/Validation_47332904_47332904_11146.398.png" ], "turn_id": 7, "scan_id": "Validation_47332904", "origin_question": "Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the sleek stainless steel refrigerator in the compact design\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the sleek stainless steel refrigerator in the compact design\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_47332904/Validation_47332904_47332904_11148.397.png", "Validation_47332904/Validation_47332904_47332904_11150.396.png", "Validation_47332904/Validation_47332904_47332904_11152.395.png", "Validation_47332904/Validation_47332904_47332904_11154.394.png", "Validation_47332904/Validation_47332904_47332904_11156.394.png" ], "turn_id": 8, "scan_id": "Validation_47332904", "origin_question": "What will be the first-time appearance order of the following categories: oven, refrigerator, stove?", "option": [ "refrigerator, stove, oven", "stove, oven, refrigerator", "oven, refrigerator, stove", "oven, stove, refrigerator" ], "answer": "refrigerator, stove, oven", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: oven, refrigerator, stove? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['refrigerator, stove, oven', 'stove, oven, refrigerator', 'oven, refrigerator, stove', 'oven, stove, refrigerator'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_47332904/Validation_47332904_47332904_11158.393.png", "Validation_47332904/Validation_47332904_47332904_11160.392.png", "Validation_47332904/Validation_47332904_47332904_11162.391.png", "Validation_47332904/Validation_47332904_47332904_11164.390.png", "Validation_47332904/Validation_47332904_47332904_11166.389.png" ], "turn_id": 9, "scan_id": "Validation_47332904", "origin_question": "Is sturdy orange rectangular case made of cardboard to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Is sturdy orange rectangular case made of cardboard to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47331324/Validation_47331324_47331324_179185.452.png", "Validation_47331324/Validation_47331324_47331324_179186.935.png", "Validation_47331324/Validation_47331324_47331324_179188.434.png", "Validation_47331324/Validation_47331324_47331324_179189.934.png" ], "turn_id": 1, "scan_id": "Validation_47331324", "origin_question": "Remember, have you seen any range hood(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any range hood(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47331324/Validation_47331324_47331324_179191.433.png", "Validation_47331324/Validation_47331324_47331324_179192.933.png", "Validation_47331324/Validation_47331324_47331324_179194.432.png", "Validation_47331324/Validation_47331324_47331324_179195.931.png" ], "turn_id": 2, "scan_id": "Validation_47331324", "origin_question": "In which turn were you closest to \"the modern rectangular stainless steel oven\"? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the modern rectangular stainless steel oven\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_47331324/Validation_47331324_47331324_179197.431.png", "Validation_47331324/Validation_47331324_47331324_179198.930.png", "Validation_47331324/Validation_47331324_47331324_179200.430.png", "Validation_47331324/Validation_47331324_47331324_179201.929.png" ], "turn_id": 3, "scan_id": "Validation_47331324", "origin_question": "What will be the first-time appearance order of the following categories: flowerpot, range hood, sink?", "option": [ "sink, range hood, flowerpot", "range hood, sink, flowerpot", "flowerpot, range hood, sink", "range hood, flowerpot, sink" ], "answer": "range hood, sink, flowerpot", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: flowerpot, range hood, sink? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['sink, range hood, flowerpot', 'range hood, sink, flowerpot', 'flowerpot, range hood, sink', 'range hood, flowerpot, sink'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_47331324/Validation_47331324_47331324_179203.428.png", "Validation_47331324/Validation_47331324_47331324_179204.928.png", "Validation_47331324/Validation_47331324_47331324_179206.427.png", "Validation_47331324/Validation_47331324_47331324_179207.927.png" ], "turn_id": 4, "scan_id": "Validation_47331324", "origin_question": "What will be the first-time appearance order of the following categories: fruit, range hood, sink?", "option": [ "range hood, sink, fruit", "sink, range hood, fruit", "range hood, fruit, sink", "sink, fruit, range hood" ], "answer": "range hood, sink, fruit", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: fruit, range hood, sink? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['range hood, sink, fruit', 'sink, range hood, fruit', 'range hood, fruit, sink', 'sink, fruit, range hood'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_47331324/Validation_47331324_47331324_179209.426.png", "Validation_47331324/Validation_47331324_47331324_179210.925.png", "Validation_47331324/Validation_47331324_47331324_179212.425.png", "Validation_47331324/Validation_47331324_47331324_179213.924.png" ], "turn_id": 5, "scan_id": "Validation_47331324", "origin_question": "At the end of which turn were both of \"the sleek, medium-sized stainless steel pan\" and \"the sleek, white plastic kettle with the glossy finish\" on your right side?", "option": [], "answer": "1", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the sleek, medium-sized stainless steel pan\" and \"the sleek, white plastic kettle with the glossy finish\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_47331324/Validation_47331324_47331324_179215.440.png", "Validation_47331324/Validation_47331324_47331324_179216.940.png", "Validation_47331324/Validation_47331324_47331324_179218.439.png", "Validation_47331324/Validation_47331324_47331324_179219.938.png" ], "turn_id": 6, "scan_id": "Validation_47331324", "origin_question": "Is the smooth, round yellow flowerpot made of ceramic to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the smooth, round yellow flowerpot made of ceramic to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42444949/Validation_42444949_42444949_223706.690.png", "Validation_42444949/Validation_42444949_42444949_223708.673.png", "Validation_42444949/Validation_42444949_42444949_223710.672.png", "Validation_42444949/Validation_42444949_42444949_223712.671.png", "Validation_42444949/Validation_42444949_42444949_223714.670.png" ], "turn_id": 1, "scan_id": "Validation_42444949", "origin_question": "Remember, have you seen any bread(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any bread(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_42444949/Validation_42444949_42444949_223716.669.png", "Validation_42444949/Validation_42444949_42444949_223718.669.png", "Validation_42444949/Validation_42444949_42444949_223720.668.png", "Validation_42444949/Validation_42444949_42444949_223722.667.png", "Validation_42444949/Validation_42444949_42444949_223724.666.png" ], "turn_id": 2, "scan_id": "Validation_42444949", "origin_question": "When did you first discover \"the compact round table made of rich brown wood\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the compact round table made of rich brown wood\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_42444949/Validation_42444949_42444949_223726.665.png", "Validation_42444949/Validation_42444949_42444949_223728.665.png", "Validation_42444949/Validation_42444949_42444949_223730.664.png", "Validation_42444949/Validation_42444949_42444949_223732.679.png", "Validation_42444949/Validation_42444949_42444949_223734.679.png" ], "turn_id": 3, "scan_id": "Validation_42444949", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the classic wooden chair with the warm wood tone\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the classic wooden chair with the warm wood tone\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_42444949/Validation_42444949_42444949_223736.678.png", "Validation_42444949/Validation_42444949_42444949_223738.677.png", "Validation_42444949/Validation_42444949_42444949_223740.676.png", "Validation_42444949/Validation_42444949_42444949_223742.675.png", "Validation_42444949/Validation_42444949_42444949_223744.675.png" ], "turn_id": 4, "scan_id": "Validation_42444949", "origin_question": "", "option": [], "answer": "", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_42444949/Validation_42444949_42444949_223746.674.png", "Validation_42444949/Validation_42444949_42444949_223748.673.png", "Validation_42444949/Validation_42444949_42444949_223750.672.png", "Validation_42444949/Validation_42444949_42444949_223752.671.png", "Validation_42444949/Validation_42444949_42444949_223754.671.png" ], "turn_id": 5, "scan_id": "Validation_42444949", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the sleek rectangular radiator with vertical ridges\" from your position?", "option": [], "answer": "162.12", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the sleek rectangular radiator with vertical ridges\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_42444949/Validation_42444949_42444949_223756.670.png", "Validation_42444949/Validation_42444949_42444949_223758.669.png", "Validation_42444949/Validation_42444949_42444949_223760.668.png", "Validation_42444949/Validation_42444949_42444949_223762.667.png", "Validation_42444949/Validation_42444949_42444949_223764.666.png" ], "turn_id": 6, "scan_id": "Validation_42444949", "origin_question": "Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_42444949/Validation_42444949_42444949_223766.666.png", "Validation_42444949/Validation_42444949_42444949_223768.665.png", "Validation_42444949/Validation_42444949_42444949_223770.664.png", "Validation_42444949/Validation_42444949_42444949_223772.663.png", "Validation_42444949/Validation_42444949_42444949_223774.679.png" ], "turn_id": 7, "scan_id": "Validation_42444949", "origin_question": "Which two objects are on the same side of you now? \"the sleek rectangular radiator with vertical ridges\", \"an elegant rectangular fireplace with cream-colored wood detailing\", \"the couch\". ", "option": [ "the sleek rectangular radiator with vertical ridges, an elegant rectangular fireplace with cream-colored wood detailing", "the sleek rectangular radiator with vertical ridges, the couch", "an elegant rectangular fireplace with cream-colored wood detailing, the couch" ], "answer": "an elegant rectangular fireplace with cream-colored wood detailing, the couch", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the sleek rectangular radiator with vertical ridges\", \"an elegant rectangular fireplace with cream-colored wood detailing\", \"the couch\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the sleek rectangular radiator with vertical ridges, an elegant rectangular fireplace with cream-colored wood detailing', 'the sleek rectangular radiator with vertical ridges, the couch', 'an elegant rectangular fireplace with cream-colored wood detailing, the couch'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_42444949/Validation_42444949_42444949_223776.678.png", "Validation_42444949/Validation_42444949_42444949_223778.677.png", "Validation_42444949/Validation_42444949_42444949_223780.677.png", "Validation_42444949/Validation_42444949_42444949_223782.676.png", "Validation_42444949/Validation_42444949_42444949_223784.675.png" ], "turn_id": 8, "scan_id": "Validation_42444949", "origin_question": "Which direction is the couch in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the couch in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_45663099/Validation_45663099_45663099_54872.870.png", "Validation_45663099/Validation_45663099_45663099_54874.853.png", "Validation_45663099/Validation_45663099_45663099_54876.852.png", "Validation_45663099/Validation_45663099_45663099_54878.851.png", "Validation_45663099/Validation_45663099_45663099_54880.850.png" ], "turn_id": 1, "scan_id": "Validation_45663099", "origin_question": "Remember, have you seen any chair(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any chair(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_45663099/Validation_45663099_45663099_54882.849.png", "Validation_45663099/Validation_45663099_45663099_54884.849.png", "Validation_45663099/Validation_45663099_45663099_54886.848.png", "Validation_45663099/Validation_45663099_45663099_54888.847.png", "Validation_45663099/Validation_45663099_45663099_54890.846.png" ], "turn_id": 2, "scan_id": "Validation_45663099", "origin_question": "In which turn did you see both \"the natural-colored rectangular cabinet\" and \"the wooden-framed rectangular mirror with the warm charm\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the natural-colored rectangular cabinet\" and \"the wooden-framed rectangular mirror with the warm charm\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_45663099/Validation_45663099_45663099_54892.845.png", "Validation_45663099/Validation_45663099_45663099_54894.844.png", "Validation_45663099/Validation_45663099_45663099_54896.844.png", "Validation_45663099/Validation_45663099_45663099_54898.843.png", "Validation_45663099/Validation_45663099_45663099_54900.859.png" ], "turn_id": 3, "scan_id": "Validation_45663099", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_45663099/Validation_45663099_45663099_54902.858.png", "Validation_45663099/Validation_45663099_45663099_54904.857.png", "Validation_45663099/Validation_45663099_45663099_54906.856.png", "Validation_45663099/Validation_45663099_45663099_54908.855.png", "Validation_45663099/Validation_45663099_45663099_54910.855.png" ], "turn_id": 4, "scan_id": "Validation_45663099", "origin_question": "What will be the first-time appearance order of the following categories: door, picture, stool?", "option": [ "door, picture, stool", "picture, stool, door", "stool, door, picture", "picture, door, stool" ], "answer": "door, picture, stool", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: door, picture, stool? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['door, picture, stool', 'picture, stool, door', 'stool, door, picture', 'picture, door, stool'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_45663099/Validation_45663099_45663099_54912.854.png", "Validation_45663099/Validation_45663099_45663099_54914.853.png", "Validation_45663099/Validation_45663099_45663099_54916.852.png", "Validation_45663099/Validation_45663099_45663099_54918.851.png", "Validation_45663099/Validation_45663099_45663099_54920.850.png" ], "turn_id": 5, "scan_id": "Validation_45663099", "origin_question": "In which turn were you closest to \"the lamp\"? ", "option": [], "answer": "5", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the lamp\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_45663099/Validation_45663099_45663099_54922.850.png", "Validation_45663099/Validation_45663099_45663099_54924.849.png", "Validation_45663099/Validation_45663099_45663099_54926.848.png", "Validation_45663099/Validation_45663099_45663099_54928.847.png", "Validation_45663099/Validation_45663099_45663099_54930.846.png" ], "turn_id": 6, "scan_id": "Validation_45663099", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the wooden-framed rectangular mirror with the warm charm\" from your position?", "option": [], "answer": "60.68", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the wooden-framed rectangular mirror with the warm charm\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_45663099/Validation_45663099_45663099_54932.845.png", "Validation_45663099/Validation_45663099_45663099_54934.845.png", "Validation_45663099/Validation_45663099_45663099_54936.844.png", "Validation_45663099/Validation_45663099_45663099_54938.843.png", "Validation_45663099/Validation_45663099_45663099_54940.842.png" ], "turn_id": 7, "scan_id": "Validation_45663099", "origin_question": "Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_45663099/Validation_45663099_45663099_54942.841.png", "Validation_45663099/Validation_45663099_45663099_54944.857.png", "Validation_45663099/Validation_45663099_45663099_54946.856.png", "Validation_45663099/Validation_45663099_45663099_54948.855.png", "Validation_45663099/Validation_45663099_45663099_54950.855.png" ], "turn_id": 8, "scan_id": "Validation_45663099", "origin_question": "Compared to your position at the end of the 6 turn, are you now closer or farther away from \"the lamp\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 6 turn, are you now closer or farther away from \"the lamp\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_48018571/Validation_48018571_48018571_3900.055.png", "Validation_48018571/Validation_48018571_48018571_3901.538.png", "Validation_48018571/Validation_48018571_48018571_3903.037.png", "Validation_48018571/Validation_48018571_48018571_3904.536.png" ], "turn_id": 1, "scan_id": "Validation_48018571", "origin_question": "Remember, have you seen any panel(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any panel(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "Validation_48018571/Validation_48018571_48018571_3906.036.png", "Validation_48018571/Validation_48018571_48018571_3907.535.png", "Validation_48018571/Validation_48018571_48018571_3909.035.png", "Validation_48018571/Validation_48018571_48018571_3910.534.png" ], "turn_id": 2, "scan_id": "Validation_48018571", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the light-colored rectangular wooden table\"; \"the radiator\"; \"the cozy rectangular couch with tan upholstery\" ", "option": [ "the light-colored rectangular wooden table", "the radiator", "the cozy rectangular couch with tan upholstery" ], "answer": "the radiator", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the light-colored rectangular wooden table\"; \"the radiator\"; \"the cozy rectangular couch with tan upholstery\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the light-colored rectangular wooden table', 'the radiator', 'the cozy rectangular couch with tan upholstery'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_48018571/Validation_48018571_48018571_3912.033.png", "Validation_48018571/Validation_48018571_48018571_3913.533.png", "Validation_48018571/Validation_48018571_48018571_3915.032.png", "Validation_48018571/Validation_48018571_48018571_3916.532.png" ], "turn_id": 3, "scan_id": "Validation_48018571", "origin_question": "Which two objects are on the same side of you now? \"the light-colored rectangular wooden table\", \"the radiator\", \"the window\". ", "option": [ "the light-colored rectangular wooden table, the radiator", "the light-colored rectangular wooden table, the window", "the radiator, the window" ], "answer": "the radiator, the window", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the light-colored rectangular wooden table\", \"the radiator\", \"the window\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the light-colored rectangular wooden table, the radiator', 'the light-colored rectangular wooden table, the window', 'the radiator, the window'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_48018571/Validation_48018571_48018571_3918.031.png", "Validation_48018571/Validation_48018571_48018571_3919.530.png", "Validation_48018571/Validation_48018571_48018571_3921.030.png", "Validation_48018571/Validation_48018571_48018571_3922.546.png" ], "turn_id": 4, "scan_id": "Validation_48018571", "origin_question": "Remember, have you seen any couch(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any couch(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_48018571/Validation_48018571_48018571_3924.045.png", "Validation_48018571/Validation_48018571_48018571_3925.545.png", "Validation_48018571/Validation_48018571_48018571_3927.044.png", "Validation_48018571/Validation_48018571_48018571_3928.543.png" ], "turn_id": 5, "scan_id": "Validation_48018571", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the radiator\" from your position?", "option": [], "answer": "42.59", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the radiator\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "Validation_45261140/Validation_45261140_45261140_3947.290.png", "Validation_45261140/Validation_45261140_45261140_3949.273.png", "Validation_45261140/Validation_45261140_45261140_3951.272.png", "Validation_45261140/Validation_45261140_45261140_3953.271.png" ], "turn_id": 1, "scan_id": "Validation_45261140", "origin_question": "Remember, how many shower(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many shower(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_45261140/Validation_45261140_45261140_3955.270.png", "Validation_45261140/Validation_45261140_45261140_3957.269.png", "Validation_45261140/Validation_45261140_45261140_3959.268.png", "Validation_45261140/Validation_45261140_45261140_3961.267.png" ], "turn_id": 2, "scan_id": "Validation_45261140", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the rectangular burgundy pillow with white text\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the rectangular burgundy pillow with white text\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_45261140/Validation_45261140_45261140_3963.267.png", "Validation_45261140/Validation_45261140_45261140_3965.266.png", "Validation_45261140/Validation_45261140_45261140_3967.265.png", "Validation_45261140/Validation_45261140_45261140_3969.264.png" ], "turn_id": 3, "scan_id": "Validation_45261140", "origin_question": "In which turn were you farthest from \"the sleek white rectangular washing machine with the circular door\"? ", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the sleek white rectangular washing machine with the circular door\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "Validation_45261140/Validation_45261140_45261140_3971.263.png", "Validation_45261140/Validation_45261140_45261140_3973.262.png", "Validation_45261140/Validation_45261140_45261140_3975.262.png", "Validation_45261140/Validation_45261140_45261140_3977.261.png" ], "turn_id": 4, "scan_id": "Validation_45261140", "origin_question": "Remember, how many couch(s) have you seen so far? ", "option": [], "answer": "2", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many couch(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_45261140/Validation_45261140_45261140_3979.260.png", "Validation_45261140/Validation_45261140_45261140_3981.259.png", "Validation_45261140/Validation_45261140_45261140_3983.275.png", "Validation_45261140/Validation_45261140_45261140_3985.274.png" ], "turn_id": 5, "scan_id": "Validation_45261140", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "1.59", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_45261140/Validation_45261140_45261140_3987.273.png", "Validation_45261140/Validation_45261140_45261140_3989.273.png", "Validation_45261140/Validation_45261140_45261140_3991.272.png", "Validation_45261140/Validation_45261140_45261140_3993.271.png" ], "turn_id": 6, "scan_id": "Validation_45261140", "origin_question": "Assuming the direction you are facing at the end of the turn 5 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Left", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 5 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47333440/Validation_47333440_47333440_53794.247.png", "Validation_47333440/Validation_47333440_47333440_53796.229.png", "Validation_47333440/Validation_47333440_47333440_53798.228.png", "Validation_47333440/Validation_47333440_47333440_53800.227.png" ], "turn_id": 1, "scan_id": "Validation_47333440", "origin_question": "Remember, have you seen any pillow(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any pillow(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_47333440/Validation_47333440_47333440_53802.227.png", "Validation_47333440/Validation_47333440_47333440_53804.226.png", "Validation_47333440/Validation_47333440_47333440_53806.225.png", "Validation_47333440/Validation_47333440_47333440_53808.224.png" ], "turn_id": 2, "scan_id": "Validation_47333440", "origin_question": "In which turn did you see both \"the small, see-through cylindrical plastic bottle\" and \"the medium-sized rectangular bag in muted purple\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the small, see-through cylindrical plastic bottle\" and \"the medium-sized rectangular bag in muted purple\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47333440/Validation_47333440_47333440_53810.223.png", "Validation_47333440/Validation_47333440_47333440_53812.223.png", "Validation_47333440/Validation_47333440_47333440_53814.222.png", "Validation_47333440/Validation_47333440_47333440_53816.221.png" ], "turn_id": 3, "scan_id": "Validation_47333440", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the sleek silver metal leaning shelf with open bars\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the sleek silver metal leaning shelf with open bars\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47333440/Validation_47333440_47333440_53818.220.png", "Validation_47333440/Validation_47333440_47333440_53820.219.png", "Validation_47333440/Validation_47333440_47333440_53822.218.png", "Validation_47333440/Validation_47333440_47333440_53824.218.png" ], "turn_id": 4, "scan_id": "Validation_47333440", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the medium-sized rectangular bag in muted purple\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the medium-sized rectangular bag in muted purple\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47333440/Validation_47333440_47333440_53826.217.png", "Validation_47333440/Validation_47333440_47333440_53828.233.png", "Validation_47333440/Validation_47333440_47333440_53830.232.png", "Validation_47333440/Validation_47333440_47333440_53832.231.png" ], "turn_id": 5, "scan_id": "Validation_47333440", "origin_question": "Remember, have you seen any bag(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any bag(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47333440/Validation_47333440_47333440_53834.230.png", "Validation_47333440/Validation_47333440_47333440_53836.229.png", "Validation_47333440/Validation_47333440_47333440_53838.228.png", "Validation_47333440/Validation_47333440_47333440_53840.228.png" ], "turn_id": 6, "scan_id": "Validation_47333440", "origin_question": "Which direction is the vase-like, earth-toned light with the fabric shade in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the vase-like, earth-toned light with the fabric shade in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42446535/Validation_42446535_42446535_200479.513.png", "Validation_42446535/Validation_42446535_42446535_200481.495.png", "Validation_42446535/Validation_42446535_42446535_200483.494.png", "Validation_42446535/Validation_42446535_42446535_200485.494.png" ], "turn_id": 1, "scan_id": "Validation_42446535", "origin_question": "Remember, have you seen any toilet(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any toilet(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_42446535/Validation_42446535_42446535_200487.493.png", "Validation_42446535/Validation_42446535_42446535_200489.492.png", "Validation_42446535/Validation_42446535_42446535_200491.491.png", "Validation_42446535/Validation_42446535_42446535_200493.490.png" ], "turn_id": 2, "scan_id": "Validation_42446535", "origin_question": "In which turn did you see both \"the sleek, rounded white ceramic toilet\" and \"the glossy white oval ceramic sink\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the sleek, rounded white ceramic toilet\" and \"the glossy white oval ceramic sink\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_42446535/Validation_42446535_42446535_200495.489.png", "Validation_42446535/Validation_42446535_42446535_200497.505.png", "Validation_42446535/Validation_42446535_42446535_200499.505.png", "Validation_42446535/Validation_42446535_42446535_200501.504.png" ], "turn_id": 3, "scan_id": "Validation_42446535", "origin_question": "In which turn did you see both \"the bathtub\" and \"the sleek, frameless rectangular mirror with polished edges\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the bathtub\" and \"the sleek, frameless rectangular mirror with polished edges\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_42446535/Validation_42446535_42446535_200503.503.png", "Validation_42446535/Validation_42446535_42446535_200505.502.png", "Validation_42446535/Validation_42446535_42446535_200507.501.png", "Validation_42446535/Validation_42446535_42446535_200509.500.png" ], "turn_id": 4, "scan_id": "Validation_42446535", "origin_question": "Which direction is the glossy white oval ceramic sink in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the glossy white oval ceramic sink in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_42446535/Validation_42446535_42446535_200511.500.png", "Validation_42446535/Validation_42446535_42446535_200513.499.png", "Validation_42446535/Validation_42446535_42446535_200515.498.png", "Validation_42446535/Validation_42446535_42446535_200517.497.png" ], "turn_id": 5, "scan_id": "Validation_42446535", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the sleek white wall-mounted shelf made of composite material\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the sleek white wall-mounted shelf made of composite material\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_42446535/Validation_42446535_42446535_200519.496.png", "Validation_42446535/Validation_42446535_42446535_200521.495.png", "Validation_42446535/Validation_42446535_42446535_200523.495.png", "Validation_42446535/Validation_42446535_42446535_200525.494.png" ], "turn_id": 6, "scan_id": "Validation_42446535", "origin_question": "Is the glossy white oval ceramic sink to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the glossy white oval ceramic sink to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47333925/Validation_47333925_47333925_83541.626.png", "Validation_47333925/Validation_47333925_47333925_83543.609.png", "Validation_47333925/Validation_47333925_47333925_83545.608.png", "Validation_47333925/Validation_47333925_47333925_83547.607.png" ], "turn_id": 1, "scan_id": "Validation_47333925", "origin_question": "Remember, have you seen any telephone(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any telephone(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_47333925/Validation_47333925_47333925_83549.606.png", "Validation_47333925/Validation_47333925_47333925_83551.606.png", "Validation_47333925/Validation_47333925_47333925_83553.605.png", "Validation_47333925/Validation_47333925_47333925_83555.604.png" ], "turn_id": 2, "scan_id": "Validation_47333925", "origin_question": "In which turn did you see both \"the compact, white washing machine\" and \"the vibrant blue plastic basin with the carrying handle\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the compact, white washing machine\" and \"the vibrant blue plastic basin with the carrying handle\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "Validation_47333925/Validation_47333925_47333925_83557.603.png", "Validation_47333925/Validation_47333925_47333925_83559.602.png", "Validation_47333925/Validation_47333925_47333925_83561.602.png", "Validation_47333925/Validation_47333925_47333925_83563.601.png" ], "turn_id": 3, "scan_id": "Validation_47333925", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the bathtub\"; \"the medium-sized green plastic bottle with the cylindrical shape\"; \"the compact wall-mounted white ceramic sink\" ", "option": [ "the bathtub", "the medium-sized green plastic bottle with the cylindrical shape", "the compact wall-mounted white ceramic sink" ], "answer": "the medium-sized green plastic bottle with the cylindrical shape", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the bathtub\"; \"the medium-sized green plastic bottle with the cylindrical shape\"; \"the compact wall-mounted white ceramic sink\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the bathtub', 'the medium-sized green plastic bottle with the cylindrical shape', 'the compact wall-mounted white ceramic sink'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_47333925/Validation_47333925_47333925_83565.600.png", "Validation_47333925/Validation_47333925_47333925_83567.616.png", "Validation_47333925/Validation_47333925_47333925_83569.615.png", "Validation_47333925/Validation_47333925_47333925_83571.614.png" ], "turn_id": 4, "scan_id": "Validation_47333925", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the compact, white washing machine\" from your position?", "option": [], "answer": "125.57", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the compact, white washing machine\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_47333925/Validation_47333925_47333925_83573.613.png", "Validation_47333925/Validation_47333925_47333925_83575.613.png", "Validation_47333925/Validation_47333925_47333925_83577.612.png", "Validation_47333925/Validation_47333925_47333925_83579.611.png" ], "turn_id": 5, "scan_id": "Validation_47333925", "origin_question": "Please recall \"the medium-sized green plastic bottle with the cylindrical shape\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.29", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the medium-sized green plastic bottle with the cylindrical shape\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_47333925/Validation_47333925_47333925_83581.610.png", "Validation_47333925/Validation_47333925_47333925_83583.609.png", "Validation_47333925/Validation_47333925_47333925_83585.609.png", "Validation_47333925/Validation_47333925_47333925_83587.608.png" ], "turn_id": 6, "scan_id": "Validation_47333925", "origin_question": "Please recall \"the vibrant blue plastic basin with the carrying handle\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.54", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the vibrant blue plastic basin with the carrying handle\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_48018970/Validation_48018970_48018970_556723.870.png", "Validation_48018970/Validation_48018970_48018970_556725.353.png", "Validation_48018970/Validation_48018970_48018970_556726.852.png", "Validation_48018970/Validation_48018970_48018970_556728.352.png" ], "turn_id": 1, "scan_id": "Validation_48018970", "origin_question": "Remember, have you seen any stopcock(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any stopcock(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_48018970/Validation_48018970_48018970_556729.851.png", "Validation_48018970/Validation_48018970_48018970_556731.351.png", "Validation_48018970/Validation_48018970_48018970_556732.850.png", "Validation_48018970/Validation_48018970_48018970_556734.349.png" ], "turn_id": 2, "scan_id": "Validation_48018970", "origin_question": "Remember, have you seen any mirror(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any mirror(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_48018970/Validation_48018970_48018970_556735.849.png", "Validation_48018970/Validation_48018970_48018970_556737.348.png", "Validation_48018970/Validation_48018970_48018970_556738.848.png", "Validation_48018970/Validation_48018970_48018970_556740.347.png" ], "turn_id": 3, "scan_id": "Validation_48018970", "origin_question": "When did you first discover \"the white interior door with the sleek lever handle\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the white interior door with the sleek lever handle\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_48018970/Validation_48018970_48018970_556741.863.png", "Validation_48018970/Validation_48018970_48018970_556743.362.png", "Validation_48018970/Validation_48018970_48018970_556744.862.png", "Validation_48018970/Validation_48018970_48018970_556746.361.png" ], "turn_id": 4, "scan_id": "Validation_48018970", "origin_question": "What will be the first-time appearance order of the following categories: bed, chair, door?", "option": [ "door, chair, bed", "chair, door, bed", "door, bed, chair", "chair, bed, door" ], "answer": "door, chair, bed", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bed, chair, door? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['door, chair, bed', 'chair, door, bed', 'door, bed, chair', 'chair, bed, door'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_45663154/Validation_45663154_45663154_63399.880.png", "Validation_45663154/Validation_45663154_45663154_63401.862.png", "Validation_45663154/Validation_45663154_45663154_63403.861.png", "Validation_45663154/Validation_45663154_45663154_63405.861.png" ], "turn_id": 1, "scan_id": "Validation_45663154", "origin_question": "Remember, have you seen any fruit(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any fruit(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_45663154/Validation_45663154_45663154_63407.860.png", "Validation_45663154/Validation_45663154_45663154_63409.859.png", "Validation_45663154/Validation_45663154_45663154_63411.858.png", "Validation_45663154/Validation_45663154_45663154_63413.857.png" ], "turn_id": 2, "scan_id": "Validation_45663154", "origin_question": "Please recall \"the sleek black clock with the white face and bold numbers\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "4.4", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the sleek black clock with the white face and bold numbers\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_45663154/Validation_45663154_45663154_63415.857.png", "Validation_45663154/Validation_45663154_45663154_63417.856.png", "Validation_45663154/Validation_45663154_45663154_63419.855.png", "Validation_45663154/Validation_45663154_45663154_63421.854.png" ], "turn_id": 3, "scan_id": "Validation_45663154", "origin_question": "In which turn were you farthest from \"the dark matte black wood-burning fireplace with stone surround\"? ", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the dark matte black wood-burning fireplace with stone surround\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_45663154/Validation_45663154_45663154_63423.853.png", "Validation_45663154/Validation_45663154_45663154_63425.853.png", "Validation_45663154/Validation_45663154_45663154_63427.852.png", "Validation_45663154/Validation_45663154_45663154_63429.851.png" ], "turn_id": 4, "scan_id": "Validation_45663154", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "1.89", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_45663154/Validation_45663154_45663154_63431.850.png", "Validation_45663154/Validation_45663154_45663154_63433.850.png", "Validation_45663154/Validation_45663154_45663154_63435.849.png", "Validation_45663154/Validation_45663154_45663154_63437.848.png" ], "turn_id": 5, "scan_id": "Validation_45663154", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the dark matte black wood-burning fireplace with stone surround\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the dark matte black wood-burning fireplace with stone surround\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_45663154/Validation_45663154_45663154_63439.864.png", "Validation_45663154/Validation_45663154_45663154_63441.863.png", "Validation_45663154/Validation_45663154_45663154_63443.862.png", "Validation_45663154/Validation_45663154_45663154_63445.861.png" ], "turn_id": 6, "scan_id": "Validation_45663154", "origin_question": "Is the small square light wood table with minimalist design to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the small square light wood table with minimalist design to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_45663154/Validation_45663154_45663154_63447.861.png", "Validation_45663154/Validation_45663154_45663154_63449.860.png", "Validation_45663154/Validation_45663154_45663154_63451.859.png", "Validation_45663154/Validation_45663154_45663154_63453.858.png" ], "turn_id": 7, "scan_id": "Validation_45663154", "origin_question": "In which turn were you farthest from \"the dark matte black wood-burning fireplace with stone surround\"? ", "option": [], "answer": "1", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the dark matte black wood-burning fireplace with stone surround\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_45663154/Validation_45663154_45663154_63455.857.png", "Validation_45663154/Validation_45663154_45663154_63457.857.png", "Validation_45663154/Validation_45663154_45663154_63459.856.png", "Validation_45663154/Validation_45663154_45663154_63461.855.png" ], "turn_id": 8, "scan_id": "Validation_45663154", "origin_question": "Which direction is the sleek black clock with the white face and bold numbers in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the sleek black clock with the white face and bold numbers in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42446083/Validation_42446083_42446083_348825.502.png", "Validation_42446083/Validation_42446083_42446083_348827.484.png", "Validation_42446083/Validation_42446083_42446083_348829.483.png", "Validation_42446083/Validation_42446083_42446083_348831.482.png", "Validation_42446083/Validation_42446083_42446083_348833.482.png" ], "turn_id": 1, "scan_id": "Validation_42446083", "origin_question": "Remember, have you seen any window(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any window(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_42446083/Validation_42446083_42446083_348835.481.png", "Validation_42446083/Validation_42446083_42446083_348837.480.png", "Validation_42446083/Validation_42446083_42446083_348839.479.png", "Validation_42446083/Validation_42446083_42446083_348841.479.png", "Validation_42446083/Validation_42446083_42446083_348843.478.png" ], "turn_id": 2, "scan_id": "Validation_42446083", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the compact pink plastic table with rounded edges\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the compact pink plastic table with rounded edges\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_42446083/Validation_42446083_42446083_348845.477.png", "Validation_42446083/Validation_42446083_42446083_348847.476.png", "Validation_42446083/Validation_42446083_42446083_348849.475.png", "Validation_42446083/Validation_42446083_42446083_348851.491.png", "Validation_42446083/Validation_42446083_42446083_348853.490.png" ], "turn_id": 3, "scan_id": "Validation_42446083", "origin_question": "Please recall \"the rectangular plastic bin in the neutral shade\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.32", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the rectangular plastic bin in the neutral shade\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_42446083/Validation_42446083_42446083_348855.489.png", "Validation_42446083/Validation_42446083_42446083_348857.489.png", "Validation_42446083/Validation_42446083_42446083_348859.488.png", "Validation_42446083/Validation_42446083_42446083_348861.487.png", "Validation_42446083/Validation_42446083_42446083_348863.486.png" ], "turn_id": 4, "scan_id": "Validation_42446083", "origin_question": "How far is your current position from where you were at the end of the turn 1? (in meters)", "option": [], "answer": "3.02", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 1? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_42446083/Validation_42446083_42446083_348865.485.png", "Validation_42446083/Validation_42446083_42446083_348867.485.png", "Validation_42446083/Validation_42446083_42446083_348869.484.png", "Validation_42446083/Validation_42446083_42446083_348871.483.png", "Validation_42446083/Validation_42446083_42446083_348873.482.png" ], "turn_id": 5, "scan_id": "Validation_42446083", "origin_question": "Which direction is the rectangular, colorful picture of planets in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the rectangular, colorful picture of planets in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_42446083/Validation_42446083_42446083_348875.481.png", "Validation_42446083/Validation_42446083_42446083_348877.480.png", "Validation_42446083/Validation_42446083_42446083_348879.480.png", "Validation_42446083/Validation_42446083_42446083_348881.479.png", "Validation_42446083/Validation_42446083_42446083_348883.478.png" ], "turn_id": 6, "scan_id": "Validation_42446083", "origin_question": "How far is your current position from where you were at the end of the turn 3? (in meters)", "option": [], "answer": "1.46", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 3? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_42446083/Validation_42446083_42446083_348885.477.png", "Validation_42446083/Validation_42446083_42446083_348887.476.png", "Validation_42446083/Validation_42446083_42446083_348889.476.png", "Validation_42446083/Validation_42446083_42446083_348891.475.png", "Validation_42446083/Validation_42446083_42446083_348893.491.png" ], "turn_id": 7, "scan_id": "Validation_42446083", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the rectangular plastic bin in the neutral shade\" from your position?", "option": [], "answer": "64.31", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the rectangular plastic bin in the neutral shade\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_42446083/Validation_42446083_42446083_348895.490.png", "Validation_42446083/Validation_42446083_42446083_348897.489.png", "Validation_42446083/Validation_42446083_42446083_348899.488.png", "Validation_42446083/Validation_42446083_42446083_348901.487.png", "Validation_42446083/Validation_42446083_42446083_348903.486.png" ], "turn_id": 8, "scan_id": "Validation_42446083", "origin_question": "What will be the first-time appearance order of the following categories: bed, picture, window?", "option": [ "picture, bed, window", "window, picture, bed", "picture, window, bed", "window, bed, picture" ], "answer": "picture, window, bed", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bed, picture, window? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['picture, bed, window', 'window, picture, bed', 'picture, window, bed', 'window, bed, picture'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_42446083/Validation_42446083_42446083_348905.486.png", "Validation_42446083/Validation_42446083_42446083_348907.485.png", "Validation_42446083/Validation_42446083_42446083_348909.484.png", "Validation_42446083/Validation_42446083_42446083_348911.483.png", "Validation_42446083/Validation_42446083_42446083_348913.482.png" ], "turn_id": 9, "scan_id": "Validation_42446083", "origin_question": "Is the rectangular, colorful picture of planets to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the rectangular, colorful picture of planets to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_42446083/Validation_42446083_42446083_348915.481.png", "Validation_42446083/Validation_42446083_42446083_348917.481.png", "Validation_42446083/Validation_42446083_42446083_348919.480.png", "Validation_42446083/Validation_42446083_42446083_348921.479.png", "Validation_42446083/Validation_42446083_42446083_348923.478.png" ], "turn_id": 10, "scan_id": "Validation_42446083", "origin_question": "What will be the first-time appearance order of the following categories: bed, bin, picture?", "option": [ "bin, bed, picture", "bin, picture, bed", "picture, bin, bed", "picture, bed, bin" ], "answer": "picture, bin, bed", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bed, bin, picture? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['bin, bed, picture', 'bin, picture, bed', 'picture, bin, bed', 'picture, bed, bin'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42898519/Validation_42898519_42898519_750471.780.png", "Validation_42898519/Validation_42898519_42898519_750473.762.png", "Validation_42898519/Validation_42898519_42898519_750475.778.png", "Validation_42898519/Validation_42898519_42898519_750477.777.png", "Validation_42898519/Validation_42898519_42898519_750479.777.png" ], "turn_id": 1, "scan_id": "Validation_42898519", "origin_question": "Remember, have you seen any bin(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any bin(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_42898519/Validation_42898519_42898519_750481.776.png", "Validation_42898519/Validation_42898519_42898519_750483.775.png", "Validation_42898519/Validation_42898519_42898519_750485.774.png", "Validation_42898519/Validation_42898519_42898519_750487.773.png", "Validation_42898519/Validation_42898519_42898519_750489.773.png" ], "turn_id": 2, "scan_id": "Validation_42898519", "origin_question": "In which turn were you closest to \"the medium-sized, rectangular, soft cotton towel\"? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the medium-sized, rectangular, soft cotton towel\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_42898519/Validation_42898519_42898519_750491.772.png", "Validation_42898519/Validation_42898519_42898519_750493.771.png", "Validation_42898519/Validation_42898519_42898519_750495.770.png", "Validation_42898519/Validation_42898519_42898519_750497.769.png", "Validation_42898519/Validation_42898519_42898519_750499.769.png" ], "turn_id": 3, "scan_id": "Validation_42898519", "origin_question": "Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Left", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_42898519/Validation_42898519_42898519_750501.768.png", "Validation_42898519/Validation_42898519_42898519_750503.767.png", "Validation_42898519/Validation_42898519_42898519_750505.766.png", "Validation_42898519/Validation_42898519_42898519_750507.765.png", "Validation_42898519/Validation_42898519_42898519_750509.764.png" ], "turn_id": 4, "scan_id": "Validation_42898519", "origin_question": "How far is your current position from where you were at the end of the turn 1? (in meters)", "option": [], "answer": "3.29", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 1? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_42898519/Validation_42898519_42898519_750511.764.png", "Validation_42898519/Validation_42898519_42898519_750513.763.png", "Validation_42898519/Validation_42898519_42898519_750515.779.png", "Validation_42898519/Validation_42898519_42898519_750517.778.png", "Validation_42898519/Validation_42898519_42898519_750519.777.png" ], "turn_id": 5, "scan_id": "Validation_42898519", "origin_question": "Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "149.59", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_42898519/Validation_42898519_42898519_750521.776.png", "Validation_42898519/Validation_42898519_42898519_750523.775.png", "Validation_42898519/Validation_42898519_42898519_750525.775.png", "Validation_42898519/Validation_42898519_42898519_750527.774.png", "Validation_42898519/Validation_42898519_42898519_750529.773.png" ], "turn_id": 6, "scan_id": "Validation_42898519", "origin_question": "In which turn did you see both \"the sleek black plastic cuboid bin\" and \"the sleek white dishwasher with the glossy finish\" simultaneously? ", "option": [], "answer": "6", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the sleek black plastic cuboid bin\" and \"the sleek white dishwasher with the glossy finish\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_42898519/Validation_42898519_42898519_750531.772.png", "Validation_42898519/Validation_42898519_42898519_750533.771.png", "Validation_42898519/Validation_42898519_42898519_750535.770.png", "Validation_42898519/Validation_42898519_42898519_750537.770.png", "Validation_42898519/Validation_42898519_42898519_750539.769.png" ], "turn_id": 7, "scan_id": "Validation_42898519", "origin_question": "Which direction is the sleek white dishwasher with the glossy finish in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the sleek white dishwasher with the glossy finish in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_42898519/Validation_42898519_42898519_750541.768.png", "Validation_42898519/Validation_42898519_42898519_750543.767.png", "Validation_42898519/Validation_42898519_42898519_750545.766.png", "Validation_42898519/Validation_42898519_42898519_750547.765.png", "Validation_42898519/Validation_42898519_42898519_750549.765.png" ], "turn_id": 8, "scan_id": "Validation_42898519", "origin_question": "How far is your current position from where you were at the end of the turn 4? (in meters)", "option": [], "answer": "2.2", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 4? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_42898519/Validation_42898519_42898519_750551.764.png", "Validation_42898519/Validation_42898519_42898519_750553.763.png", "Validation_42898519/Validation_42898519_42898519_750555.762.png", "Validation_42898519/Validation_42898519_42898519_750557.778.png", "Validation_42898519/Validation_42898519_42898519_750559.777.png" ], "turn_id": 9, "scan_id": "Validation_42898519", "origin_question": "Which direction is the sleek black plastic cuboid bin in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the sleek black plastic cuboid bin in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "Validation_42898519/Validation_42898519_42898519_750561.776.png", "Validation_42898519/Validation_42898519_42898519_750563.776.png", "Validation_42898519/Validation_42898519_42898519_750565.775.png", "Validation_42898519/Validation_42898519_42898519_750567.774.png", "Validation_42898519/Validation_42898519_42898519_750569.773.png" ], "turn_id": 10, "scan_id": "Validation_42898519", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the modern rectangular stainless steel oven\"; \"the modern rectangular stainless steel stove\"; \"the medium-sized, rectangular, soft cotton towel\" ", "option": [ "the modern rectangular stainless steel oven", "the modern rectangular stainless steel stove", "the medium-sized, rectangular, soft cotton towel" ], "answer": "the modern rectangular stainless steel oven", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the modern rectangular stainless steel oven\"; \"the modern rectangular stainless steel stove\"; \"the medium-sized, rectangular, soft cotton towel\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the modern rectangular stainless steel oven', 'the modern rectangular stainless steel stove', 'the medium-sized, rectangular, soft cotton towel'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47332000/Validation_47332000_47332000_172001.140.png", "Validation_47332000/Validation_47332000_47332000_172003.139.png", "Validation_47332000/Validation_47332000_47332000_172005.138.png", "Validation_47332000/Validation_47332000_47332000_172007.137.png" ], "turn_id": 1, "scan_id": "Validation_47332000", "origin_question": "", "option": [], "answer": "", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47332000/Validation_47332000_47332000_172009.136.png", "Validation_47332000/Validation_47332000_47332000_172011.136.png", "Validation_47332000/Validation_47332000_47332000_172013.135.png", "Validation_47332000/Validation_47332000_47332000_172015.134.png" ], "turn_id": 2, "scan_id": "Validation_47332000", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_47332000/Validation_47332000_47332000_172017.133.png", "Validation_47332000/Validation_47332000_47332000_172019.132.png", "Validation_47332000/Validation_47332000_47332000_172021.131.png", "Validation_47332000/Validation_47332000_47332000_172023.131.png" ], "turn_id": 3, "scan_id": "Validation_47332000", "origin_question": "When did you first discover \"the large white panel-style radiator\"? (index of the turn)", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the large white panel-style radiator\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47332000/Validation_47332000_47332000_172025.130.png", "Validation_47332000/Validation_47332000_47332000_172027.129.png", "Validation_47332000/Validation_47332000_47332000_172029.128.png", "Validation_47332000/Validation_47332000_47332000_172031.128.png" ], "turn_id": 4, "scan_id": "Validation_47332000", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the large white panel-style radiator\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the large white panel-style radiator\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47332000/Validation_47332000_47332000_172033.143.png", "Validation_47332000/Validation_47332000_47332000_172035.142.png", "Validation_47332000/Validation_47332000_47332000_172037.142.png", "Validation_47332000/Validation_47332000_47332000_172039.141.png" ], "turn_id": 5, "scan_id": "Validation_47332000", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the large white panel-style radiator\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the large white panel-style radiator\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_47332000/Validation_47332000_47332000_172041.140.png", "Validation_47332000/Validation_47332000_47332000_172043.139.png", "Validation_47332000/Validation_47332000_47332000_172045.139.png", "Validation_47332000/Validation_47332000_47332000_172047.138.png" ], "turn_id": 6, "scan_id": "Validation_47332000", "origin_question": "In which turn did you see both \"the large white panel-style radiator\" and \"the sleek gray cabinet with rectangular shape and handles\" simultaneously? ", "option": [], "answer": "5", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the large white panel-style radiator\" and \"the sleek gray cabinet with rectangular shape and handles\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47332000/Validation_47332000_47332000_172049.137.png", "Validation_47332000/Validation_47332000_47332000_172051.136.png", "Validation_47332000/Validation_47332000_47332000_172053.135.png", "Validation_47332000/Validation_47332000_47332000_172055.134.png" ], "turn_id": 7, "scan_id": "Validation_47332000", "origin_question": "", "option": [], "answer": "", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_48458481/Validation_48458481_48458481_111504.308.png", "Validation_48458481/Validation_48458481_48458481_111505.808.png", "Validation_48458481/Validation_48458481_48458481_111507.307.png", "Validation_48458481/Validation_48458481_48458481_111508.807.png" ], "turn_id": 1, "scan_id": "Validation_48458481", "origin_question": "Remember, have you seen any light(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any light(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_48458481/Validation_48458481_48458481_111510.306.png", "Validation_48458481/Validation_48458481_48458481_111511.805.png", "Validation_48458481/Validation_48458481_48458481_111513.305.png", "Validation_48458481/Validation_48458481_48458481_111514.804.png" ], "turn_id": 2, "scan_id": "Validation_48458481", "origin_question": "Which direction is the spherical light fixture with the frosted glass surface in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the spherical light fixture with the frosted glass surface in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_48458481/Validation_48458481_48458481_111516.304.png", "Validation_48458481/Validation_48458481_48458481_111517.803.png", "Validation_48458481/Validation_48458481_48458481_111519.302.png", "Validation_48458481/Validation_48458481_48458481_111520.802.png" ], "turn_id": 3, "scan_id": "Validation_48458481", "origin_question": "In which turn did you see both \"the front-loading dishwasher with the sleek white exterior\" and \"the spherical light fixture with the frosted glass surface\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the front-loading dishwasher with the sleek white exterior\" and \"the spherical light fixture with the frosted glass surface\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_48458481/Validation_48458481_48458481_111522.301.png", "Validation_48458481/Validation_48458481_48458481_111523.801.png", "Validation_48458481/Validation_48458481_48458481_111525.300.png", "Validation_48458481/Validation_48458481_48458481_111526.799.png" ], "turn_id": 4, "scan_id": "Validation_48458481", "origin_question": "What will be the first-time appearance order of the following categories: box, light, washing machine?", "option": [ "light, box, washing machine", "light, washing machine, box", "box, light, washing machine", "washing machine, light, box" ], "answer": "washing machine, light, box", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: box, light, washing machine? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['light, box, washing machine', 'light, washing machine, box', 'box, light, washing machine', 'washing machine, light, box'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_41254425/Validation_41254425_41254425_95813.104.png", "Validation_41254425/Validation_41254425_41254425_95815.087.png", "Validation_41254425/Validation_41254425_41254425_95817.086.png", "Validation_41254425/Validation_41254425_41254425_95819.085.png" ], "turn_id": 1, "scan_id": "Validation_41254425", "origin_question": "Remember, have you seen any bucket(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any bucket(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_41254425/Validation_41254425_41254425_95821.084.png", "Validation_41254425/Validation_41254425_41254425_95823.084.png", "Validation_41254425/Validation_41254425_41254425_95825.083.png", "Validation_41254425/Validation_41254425_41254425_95827.082.png" ], "turn_id": 2, "scan_id": "Validation_41254425", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the sleek white laminated wood cabinet\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the sleek white laminated wood cabinet\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_41254425/Validation_41254425_41254425_95829.081.png", "Validation_41254425/Validation_41254425_41254425_95831.080.png", "Validation_41254425/Validation_41254425_41254425_95833.080.png", "Validation_41254425/Validation_41254425_41254425_95835.096.png" ], "turn_id": 3, "scan_id": "Validation_41254425", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "117.12", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_41254425/Validation_41254425_41254425_95837.095.png", "Validation_41254425/Validation_41254425_41254425_95839.094.png", "Validation_41254425/Validation_41254425_41254425_95841.093.png", "Validation_41254425/Validation_41254425_41254425_95843.092.png" ], "turn_id": 4, "scan_id": "Validation_41254425", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_41254425/Validation_41254425_41254425_95845.092.png", "Validation_41254425/Validation_41254425_41254425_95847.091.png", "Validation_41254425/Validation_41254425_41254425_95849.090.png", "Validation_41254425/Validation_41254425_41254425_95851.089.png" ], "turn_id": 5, "scan_id": "Validation_41254425", "origin_question": "Is the sleek, medium gray rectangular table to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the sleek, medium gray rectangular table to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_41254425/Validation_41254425_41254425_95853.088.png", "Validation_41254425/Validation_41254425_41254425_95855.088.png", "Validation_41254425/Validation_41254425_41254425_95857.087.png", "Validation_41254425/Validation_41254425_41254425_95859.086.png" ], "turn_id": 6, "scan_id": "Validation_41254425", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the bed\", Y: \"the rectangular white door with raised panels\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to X and farther away from Y", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the bed\", Y: \"the rectangular white door with raised panels\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_41254425/Validation_41254425_41254425_95861.085.png", "Validation_41254425/Validation_41254425_41254425_95863.084.png", "Validation_41254425/Validation_41254425_41254425_95865.084.png", "Validation_41254425/Validation_41254425_41254425_95867.083.png" ], "turn_id": 7, "scan_id": "Validation_41254425", "origin_question": "Compared to your position at the end of the 6 turn, are you now closer or farther away from \"the bed\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 6 turn, are you now closer or farther away from \"the bed\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_41069043/Validation_41069043_41069043_2799.640.png", "Validation_41069043/Validation_41069043_41069043_2801.639.png", "Validation_41069043/Validation_41069043_41069043_2803.638.png", "Validation_41069043/Validation_41069043_41069043_2805.637.png", "Validation_41069043/Validation_41069043_41069043_2807.636.png" ], "turn_id": 1, "scan_id": "Validation_41069043", "origin_question": "Remember, have you seen any table(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any table(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_41069043/Validation_41069043_41069043_2809.636.png", "Validation_41069043/Validation_41069043_41069043_2811.635.png", "Validation_41069043/Validation_41069043_41069043_2813.634.png", "Validation_41069043/Validation_41069043_41069043_2815.633.png", "Validation_41069043/Validation_41069043_41069043_2817.632.png" ], "turn_id": 2, "scan_id": "Validation_41069043", "origin_question": "In which turn were you closest to \"the sleek double-sized bed with light-colored sheets\"? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the sleek double-sized bed with light-colored sheets\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_41069043/Validation_41069043_41069043_2819.631.png", "Validation_41069043/Validation_41069043_41069043_2821.631.png", "Validation_41069043/Validation_41069043_41069043_2823.630.png", "Validation_41069043/Validation_41069043_41069043_2825.629.png", "Validation_41069043/Validation_41069043_41069043_2827.628.png" ], "turn_id": 3, "scan_id": "Validation_41069043", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "25.91", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_41069043/Validation_41069043_41069043_2829.627.png", "Validation_41069043/Validation_41069043_41069043_2831.626.png", "Validation_41069043/Validation_41069043_41069043_2833.626.png", "Validation_41069043/Validation_41069043_41069043_2835.625.png", "Validation_41069043/Validation_41069043_41069043_2837.624.png" ], "turn_id": 4, "scan_id": "Validation_41069043", "origin_question": "What will be the first-time appearance order of the following categories: curtain, light, pillow?", "option": [ "light, pillow, curtain", "curtain, light, pillow", "pillow, light, curtain", "curtain, pillow, light" ], "answer": "light, pillow, curtain", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: curtain, light, pillow? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['light, pillow, curtain', 'curtain, light, pillow', 'pillow, light, curtain', 'curtain, pillow, light'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_41069043/Validation_41069043_41069043_2839.640.png", "Validation_41069043/Validation_41069043_41069043_2841.639.png", "Validation_41069043/Validation_41069043_41069043_2843.638.png", "Validation_41069043/Validation_41069043_41069043_2845.637.png", "Validation_41069043/Validation_41069043_41069043_2847.637.png" ], "turn_id": 5, "scan_id": "Validation_41069043", "origin_question": "Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the rectangular neutral-colored picture frame displaying an image\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the rectangular neutral-colored picture frame displaying an image\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_41069043/Validation_41069043_41069043_2849.636.png", "Validation_41069043/Validation_41069043_41069043_2851.635.png", "Validation_41069043/Validation_41069043_41069043_2853.634.png", "Validation_41069043/Validation_41069043_41069043_2855.633.png", "Validation_41069043/Validation_41069043_41069043_2857.632.png" ], "turn_id": 6, "scan_id": "Validation_41069043", "origin_question": "", "option": [], "answer": "", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_41069043/Validation_41069043_41069043_2859.632.png", "Validation_41069043/Validation_41069043_41069043_2861.631.png", "Validation_41069043/Validation_41069043_41069043_2863.630.png", "Validation_41069043/Validation_41069043_41069043_2865.629.png", "Validation_41069043/Validation_41069043_41069043_2867.628.png" ], "turn_id": 7, "scan_id": "Validation_41069043", "origin_question": "How far is your current position from where you were at the end of the turn 4? (in meters)", "option": [], "answer": "1.78", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 4? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_41069043/Validation_41069043_41069043_2869.628.png", "Validation_41069043/Validation_41069043_41069043_2871.627.png", "Validation_41069043/Validation_41069043_41069043_2873.626.png", "Validation_41069043/Validation_41069043_41069043_2875.625.png", "Validation_41069043/Validation_41069043_41069043_2877.624.png" ], "turn_id": 8, "scan_id": "Validation_41069043", "origin_question": "Assuming the direction you are facing at the end of the turn 3 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Left", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 3 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "Validation_41069043/Validation_41069043_41069043_2879.623.png", "Validation_41069043/Validation_41069043_41069043_2881.639.png", "Validation_41069043/Validation_41069043_41069043_2883.638.png", "Validation_41069043/Validation_41069043_41069043_2885.638.png", "Validation_41069043/Validation_41069043_41069043_2887.637.png" ], "turn_id": 9, "scan_id": "Validation_41069043", "origin_question": "Remember, how many table(s) have you seen so far? ", "option": [], "answer": "2", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many table(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_45261637/Validation_45261637_45261637_55956.500.png", "Validation_45261637/Validation_45261637_45261637_55958.499.png", "Validation_45261637/Validation_45261637_45261637_55960.498.png", "Validation_45261637/Validation_45261637_45261637_55962.497.png" ], "turn_id": 1, "scan_id": "Validation_45261637", "origin_question": "Remember, have you seen any pillow(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any pillow(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_45261637/Validation_45261637_45261637_55964.497.png", "Validation_45261637/Validation_45261637_45261637_55966.496.png", "Validation_45261637/Validation_45261637_45261637_55968.495.png", "Validation_45261637/Validation_45261637_45261637_55970.494.png" ], "turn_id": 2, "scan_id": "Validation_45261637", "origin_question": "In which turn were you farthest from \"the light blue and white rectangular plush pillow\"? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the light blue and white rectangular plush pillow\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_45261637/Validation_45261637_45261637_55972.493.png", "Validation_45261637/Validation_45261637_45261637_55974.493.png", "Validation_45261637/Validation_45261637_45261637_55976.492.png", "Validation_45261637/Validation_45261637_45261637_55978.491.png" ], "turn_id": 3, "scan_id": "Validation_45261637", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "Validation_45261637/Validation_45261637_45261637_55980.490.png", "Validation_45261637/Validation_45261637_45261637_55982.490.png", "Validation_45261637/Validation_45261637_45261637_55984.489.png", "Validation_45261637/Validation_45261637_45261637_55986.488.png" ], "turn_id": 4, "scan_id": "Validation_45261637", "origin_question": "Among these three objects(\"the sturdy bed with the light-colored bedspread and pillows\",\"the light blue and white rectangular plush pillow\",\"the window\" ), which one is the farthest from you?", "option": [ "the sturdy bed with the light-colored bedspread and pillows", "the light blue and white rectangular plush pillow", "the window" ], "answer": "the window", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects(\"the sturdy bed with the light-colored bedspread and pillows\",\"the light blue and white rectangular plush pillow\",\"the window\" ), which one is the farthest from you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the sturdy bed with the light-colored bedspread and pillows', 'the light blue and white rectangular plush pillow', 'the window'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_45261637/Validation_45261637_45261637_55988.504.png", "Validation_45261637/Validation_45261637_45261637_55990.503.png", "Validation_45261637/Validation_45261637_45261637_55992.502.png", "Validation_45261637/Validation_45261637_45261637_55994.501.png" ], "turn_id": 5, "scan_id": "Validation_45261637", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the sleek black rectangular monitor with the vivid display\" from your position?", "option": [], "answer": "132.1", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the sleek black rectangular monitor with the vivid display\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_45261637/Validation_45261637_45261637_55996.501.png", "Validation_45261637/Validation_45261637_45261637_55998.500.png", "Validation_45261637/Validation_45261637_45261637_56000.499.png", "Validation_45261637/Validation_45261637_45261637_56002.498.png" ], "turn_id": 6, "scan_id": "Validation_45261637", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the sleek black rectangular monitor with the vivid display\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the sleek black rectangular monitor with the vivid display\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_45261637/Validation_45261637_45261637_56004.497.png", "Validation_45261637/Validation_45261637_45261637_56006.497.png", "Validation_45261637/Validation_45261637_45261637_56008.496.png", "Validation_45261637/Validation_45261637_45261637_56010.495.png" ], "turn_id": 7, "scan_id": "Validation_45261637", "origin_question": "By the end of the 6 turn, were you getting closer to or farther away from these two objects: X: \"the sturdy bed with the light-colored bedspread and pillows\", Y: \"the light blue and white rectangular plush pillow\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 6 turn, were you getting closer to or farther away from these two objects: X: \"the sturdy bed with the light-colored bedspread and pillows\", Y: \"the light blue and white rectangular plush pillow\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47895348/Validation_47895348_47895348_96459.535.png", "Validation_47895348/Validation_47895348_47895348_96461.518.png", "Validation_47895348/Validation_47895348_47895348_96463.517.png", "Validation_47895348/Validation_47895348_47895348_96465.516.png", "Validation_47895348/Validation_47895348_47895348_96467.515.png" ], "turn_id": 1, "scan_id": "Validation_47895348", "origin_question": "Remember, have you seen any globe(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any globe(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "Validation_47895348/Validation_47895348_47895348_96469.514.png", "Validation_47895348/Validation_47895348_47895348_96471.514.png", "Validation_47895348/Validation_47895348_47895348_96473.513.png", "Validation_47895348/Validation_47895348_47895348_96475.529.png", "Validation_47895348/Validation_47895348_47895348_96477.528.png" ], "turn_id": 2, "scan_id": "Validation_47895348", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the rectangular, muted brown plush carpet with wavy stripes\"; \"the light wood-toned rectangular minimalist table\"; \"the sleek, modern bed with the wooden frame and headboard\" ", "option": [ "the rectangular, muted brown plush carpet with wavy stripes", "the light wood-toned rectangular minimalist table", "the sleek, modern bed with the wooden frame and headboard" ], "answer": "the sleek, modern bed with the wooden frame and headboard", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the rectangular, muted brown plush carpet with wavy stripes\"; \"the light wood-toned rectangular minimalist table\"; \"the sleek, modern bed with the wooden frame and headboard\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular, muted brown plush carpet with wavy stripes', 'the light wood-toned rectangular minimalist table', 'the sleek, modern bed with the wooden frame and headboard'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47895348/Validation_47895348_47895348_96479.527.png", "Validation_47895348/Validation_47895348_47895348_96481.526.png", "Validation_47895348/Validation_47895348_47895348_96483.525.png", "Validation_47895348/Validation_47895348_47895348_96485.525.png", "Validation_47895348/Validation_47895348_47895348_96487.524.png" ], "turn_id": 3, "scan_id": "Validation_47895348", "origin_question": "In which turn were you farthest from \"the rectangular, medium-brown oak door\"? ", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the rectangular, medium-brown oak door\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47895348/Validation_47895348_47895348_96489.523.png", "Validation_47895348/Validation_47895348_47895348_96491.522.png", "Validation_47895348/Validation_47895348_47895348_96493.521.png", "Validation_47895348/Validation_47895348_47895348_96495.521.png", "Validation_47895348/Validation_47895348_47895348_96497.520.png" ], "turn_id": 4, "scan_id": "Validation_47895348", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the stylish white ergonomic chair with the high backrest\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the stylish white ergonomic chair with the high backrest\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_47895348/Validation_47895348_47895348_96499.519.png", "Validation_47895348/Validation_47895348_47895348_96501.518.png", "Validation_47895348/Validation_47895348_47895348_96503.517.png", "Validation_47895348/Validation_47895348_47895348_96505.516.png", "Validation_47895348/Validation_47895348_47895348_96507.516.png" ], "turn_id": 5, "scan_id": "Validation_47895348", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the rectangular, medium-brown oak door\" from your position?", "option": [], "answer": "60.6", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the rectangular, medium-brown oak door\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_47895348/Validation_47895348_47895348_96509.515.png", "Validation_47895348/Validation_47895348_47895348_96511.514.png", "Validation_47895348/Validation_47895348_47895348_96513.513.png", "Validation_47895348/Validation_47895348_47895348_96515.512.png", "Validation_47895348/Validation_47895348_47895348_96517.528.png" ], "turn_id": 6, "scan_id": "Validation_47895348", "origin_question": "How far is your current position from where you were at the end of the turn 1? (in meters)", "option": [], "answer": "1.2", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 1? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_47895348/Validation_47895348_47895348_96519.527.png", "Validation_47895348/Validation_47895348_47895348_96521.527.png", "Validation_47895348/Validation_47895348_47895348_96523.526.png", "Validation_47895348/Validation_47895348_47895348_96525.525.png", "Validation_47895348/Validation_47895348_47895348_96527.524.png" ], "turn_id": 7, "scan_id": "Validation_47895348", "origin_question": "In which turn did you see both \"the rectangular, muted brown plush carpet with wavy stripes\" and \"the sleek blue rectangular frame with the glossy finish\" simultaneously? ", "option": [], "answer": "6", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the rectangular, muted brown plush carpet with wavy stripes\" and \"the sleek blue rectangular frame with the glossy finish\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_47895348/Validation_47895348_47895348_96529.523.png", "Validation_47895348/Validation_47895348_47895348_96531.523.png", "Validation_47895348/Validation_47895348_47895348_96533.522.png", "Validation_47895348/Validation_47895348_47895348_96535.521.png", "Validation_47895348/Validation_47895348_47895348_96537.520.png" ], "turn_id": 8, "scan_id": "Validation_47895348", "origin_question": "Please recall \"the unique blue rectangular pillow with white protrusions\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.74", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the unique blue rectangular pillow with white protrusions\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47895348/Validation_47895348_47895348_96539.519.png", "Validation_47895348/Validation_47895348_47895348_96541.519.png", "Validation_47895348/Validation_47895348_47895348_96543.518.png", "Validation_47895348/Validation_47895348_47895348_96545.517.png", "Validation_47895348/Validation_47895348_47895348_96547.516.png" ], "turn_id": 9, "scan_id": "Validation_47895348", "origin_question": "Compared to your position at the end of the 6 turn, are you now closer or farther away from \"the white rectangular book with the well-kept spine\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 6 turn, are you now closer or farther away from \"the white rectangular book with the well-kept spine\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_41254400/Validation_41254400_41254400_95667.495.png", "Validation_41254400/Validation_41254400_41254400_95669.495.png", "Validation_41254400/Validation_41254400_41254400_95671.494.png", "Validation_41254400/Validation_41254400_41254400_95673.493.png" ], "turn_id": 1, "scan_id": "Validation_41254400", "origin_question": "Remember, have you seen any hat(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any hat(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_41254400/Validation_41254400_41254400_95675.492.png", "Validation_41254400/Validation_41254400_41254400_95677.491.png", "Validation_41254400/Validation_41254400_41254400_95679.491.png", "Validation_41254400/Validation_41254400_41254400_95681.490.png" ], "turn_id": 2, "scan_id": "Validation_41254400", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the bed\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the bed\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_41254400/Validation_41254400_41254400_95683.489.png", "Validation_41254400/Validation_41254400_41254400_95685.488.png", "Validation_41254400/Validation_41254400_41254400_95687.487.png", "Validation_41254400/Validation_41254400_41254400_95689.487.png" ], "turn_id": 3, "scan_id": "Validation_41254400", "origin_question": "Please recall \"the compact, modern cabinet with the matte white finish\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.26", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the compact, modern cabinet with the matte white finish\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_41254400/Validation_41254400_41254400_95691.486.png", "Validation_41254400/Validation_41254400_41254400_95693.485.png", "Validation_41254400/Validation_41254400_41254400_95695.484.png", "Validation_41254400/Validation_41254400_41254400_95697.484.png" ], "turn_id": 4, "scan_id": "Validation_41254400", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the compact, modern cabinet with the matte white finish\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the compact, modern cabinet with the matte white finish\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_41254400/Validation_41254400_41254400_95699.483.png", "Validation_41254400/Validation_41254400_41254400_95701.482.png", "Validation_41254400/Validation_41254400_41254400_95703.481.png", "Validation_41254400/Validation_41254400_41254400_95705.480.png" ], "turn_id": 5, "scan_id": "Validation_41254400", "origin_question": "By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular, light-colored pillow\", Y: \"the compact, modern cabinet with the matte white finish\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to X and farther away from Y", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular, light-colored pillow\", Y: \"the compact, modern cabinet with the matte white finish\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_41254400/Validation_41254400_41254400_95707.480.png", "Validation_41254400/Validation_41254400_41254400_95709.495.png", "Validation_41254400/Validation_41254400_41254400_95711.495.png", "Validation_41254400/Validation_41254400_41254400_95713.494.png" ], "turn_id": 6, "scan_id": "Validation_41254400", "origin_question": "Is the modern silver-framed picture for reflections to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the modern silver-framed picture for reflections to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42899712/Validation_42899712_42899712_795948.430.png", "Validation_42899712/Validation_42899712_42899712_795950.412.png", "Validation_42899712/Validation_42899712_42899712_795952.411.png", "Validation_42899712/Validation_42899712_42899712_795954.411.png", "Validation_42899712/Validation_42899712_42899712_795956.410.png" ], "turn_id": 1, "scan_id": "Validation_42899712", "origin_question": "Remember, have you seen any chair(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any chair(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_42899712/Validation_42899712_42899712_795958.409.png", "Validation_42899712/Validation_42899712_42899712_795960.408.png", "Validation_42899712/Validation_42899712_42899712_795962.407.png", "Validation_42899712/Validation_42899712_42899712_795964.406.png", "Validation_42899712/Validation_42899712_42899712_795966.406.png" ], "turn_id": 2, "scan_id": "Validation_42899712", "origin_question": "In which turn were you closest to \"the table\"? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the table\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_42899712/Validation_42899712_42899712_795968.405.png", "Validation_42899712/Validation_42899712_42899712_795970.404.png", "Validation_42899712/Validation_42899712_42899712_795972.403.png", "Validation_42899712/Validation_42899712_42899712_795974.402.png", "Validation_42899712/Validation_42899712_42899712_795976.402.png" ], "turn_id": 3, "scan_id": "Validation_42899712", "origin_question": "Please recall \"the table\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.87", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the table\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_42899712/Validation_42899712_42899712_795978.417.png", "Validation_42899712/Validation_42899712_42899712_795980.417.png", "Validation_42899712/Validation_42899712_42899712_795982.416.png", "Validation_42899712/Validation_42899712_42899712_795984.415.png", "Validation_42899712/Validation_42899712_42899712_795986.414.png" ], "turn_id": 4, "scan_id": "Validation_42899712", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the modern stainless steel kettle with the round body\", Y: \"the compact silver machine with user-friendly design\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the modern stainless steel kettle with the round body\", Y: \"the compact silver machine with user-friendly design\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "Validation_42899712/Validation_42899712_42899712_795988.413.png", "Validation_42899712/Validation_42899712_42899712_795990.412.png", "Validation_42899712/Validation_42899712_42899712_795992.412.png", "Validation_42899712/Validation_42899712_42899712_795994.411.png", "Validation_42899712/Validation_42899712_42899712_795996.410.png" ], "turn_id": 5, "scan_id": "Validation_42899712", "origin_question": "Among these three objects(\"the table\",\"the sleek silver rectangular microwave made of metal and plastic\",\"the rectangular stainless steel sink with dual faucets\" ), which one is the closest to you?", "option": [ "the table", "the sleek silver rectangular microwave made of metal and plastic", "the rectangular stainless steel sink with dual faucets" ], "answer": "the table", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the table\",\"the sleek silver rectangular microwave made of metal and plastic\",\"the rectangular stainless steel sink with dual faucets\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the table', 'the sleek silver rectangular microwave made of metal and plastic', 'the rectangular stainless steel sink with dual faucets'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "Validation_42899712/Validation_42899712_42899712_795998.409.png", "Validation_42899712/Validation_42899712_42899712_796000.408.png", "Validation_42899712/Validation_42899712_42899712_796002.407.png", "Validation_42899712/Validation_42899712_42899712_796004.407.png", "Validation_42899712/Validation_42899712_42899712_796006.406.png" ], "turn_id": 6, "scan_id": "Validation_42899712", "origin_question": "Among these three objects(\"the table\",\"the sleek silver rectangular microwave made of metal and plastic\",\"the compact silver machine with user-friendly design\" ), which one is the closest to you?", "option": [ "the table", "the sleek silver rectangular microwave made of metal and plastic", "the compact silver machine with user-friendly design" ], "answer": "the table", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the table\",\"the sleek silver rectangular microwave made of metal and plastic\",\"the compact silver machine with user-friendly design\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the table', 'the sleek silver rectangular microwave made of metal and plastic', 'the compact silver machine with user-friendly design'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_42899712/Validation_42899712_42899712_796008.405.png", "Validation_42899712/Validation_42899712_42899712_796010.404.png", "Validation_42899712/Validation_42899712_42899712_796012.403.png", "Validation_42899712/Validation_42899712_42899712_796014.402.png", "Validation_42899712/Validation_42899712_42899712_796016.402.png" ], "turn_id": 7, "scan_id": "Validation_42899712", "origin_question": "Please recall \"the modern stainless steel oven with the black glass front\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "5.43", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the modern stainless steel oven with the black glass front\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_42899712/Validation_42899712_42899712_796018.401.png", "Validation_42899712/Validation_42899712_42899712_796020.417.png", "Validation_42899712/Validation_42899712_42899712_796022.416.png", "Validation_42899712/Validation_42899712_42899712_796024.415.png", "Validation_42899712/Validation_42899712_42899712_796026.414.png" ], "turn_id": 8, "scan_id": "Validation_42899712", "origin_question": "By the end of the 4 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular stainless steel sink with dual faucets\", Y: \"the modern stainless steel kettle with the round body\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "farther away from both X and Y", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 4 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular stainless steel sink with dual faucets\", Y: \"the modern stainless steel kettle with the round body\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_42899712/Validation_42899712_42899712_796028.413.png", "Validation_42899712/Validation_42899712_42899712_796030.412.png", "Validation_42899712/Validation_42899712_42899712_796032.412.png", "Validation_42899712/Validation_42899712_42899712_796034.411.png", "Validation_42899712/Validation_42899712_42899712_796036.410.png" ], "turn_id": 9, "scan_id": "Validation_42899712", "origin_question": "How far is your current position from where you were at the end of the turn 5? (in meters)", "option": [], "answer": "3.76", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 5? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_42899712/Validation_42899712_42899712_796038.409.png", "Validation_42899712/Validation_42899712_42899712_796040.408.png", "Validation_42899712/Validation_42899712_42899712_796042.407.png", "Validation_42899712/Validation_42899712_42899712_796044.407.png", "Validation_42899712/Validation_42899712_42899712_796046.406.png" ], "turn_id": 10, "scan_id": "Validation_42899712", "origin_question": "Please recall \"the table\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "4.62", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the table\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47331974/Validation_47331974_47331974_168802.297.png", "Validation_47331974/Validation_47331974_47331974_168804.279.png", "Validation_47331974/Validation_47331974_47331974_168806.278.png", "Validation_47331974/Validation_47331974_47331974_168808.277.png", "Validation_47331974/Validation_47331974_47331974_168810.277.png" ], "turn_id": 1, "scan_id": "Validation_47331974", "origin_question": "", "option": [], "answer": "", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_47331974/Validation_47331974_47331974_168812.276.png", "Validation_47331974/Validation_47331974_47331974_168814.275.png", "Validation_47331974/Validation_47331974_47331974_168816.274.png", "Validation_47331974/Validation_47331974_47331974_168818.274.png", "Validation_47331974/Validation_47331974_47331974_168820.273.png" ], "turn_id": 2, "scan_id": "Validation_47331974", "origin_question": "When did you first discover \"the modern rectangular white ceramic sink\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the modern rectangular white ceramic sink\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_47331974/Validation_47331974_47331974_168822.272.png", "Validation_47331974/Validation_47331974_47331974_168824.271.png", "Validation_47331974/Validation_47331974_47331974_168826.270.png", "Validation_47331974/Validation_47331974_47331974_168828.269.png", "Validation_47331974/Validation_47331974_47331974_168830.269.png" ], "turn_id": 3, "scan_id": "Validation_47331974", "origin_question": "Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Left", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_47331974/Validation_47331974_47331974_168832.268.png", "Validation_47331974/Validation_47331974_47331974_168834.267.png", "Validation_47331974/Validation_47331974_47331974_168836.266.png", "Validation_47331974/Validation_47331974_47331974_168838.265.png", "Validation_47331974/Validation_47331974_47331974_168840.265.png" ], "turn_id": 4, "scan_id": "Validation_47331974", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the large rectangular black monitor with the glossy screen\", Y: \"the rectangular off-white pillow\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the large rectangular black monitor with the glossy screen\", Y: \"the rectangular off-white pillow\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_47331974/Validation_47331974_47331974_168842.280.png", "Validation_47331974/Validation_47331974_47331974_168844.280.png", "Validation_47331974/Validation_47331974_47331974_168846.279.png", "Validation_47331974/Validation_47331974_47331974_168848.278.png", "Validation_47331974/Validation_47331974_47331974_168850.277.png" ], "turn_id": 5, "scan_id": "Validation_47331974", "origin_question": "At the end of which turn were both of \"the modern rectangular white ceramic sink\" and \"the medium-sized, classic iron fireplace with glass door\" on your left side?", "option": [], "answer": "1", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the modern rectangular white ceramic sink\" and \"the medium-sized, classic iron fireplace with glass door\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_47331974/Validation_47331974_47331974_168852.276.png", "Validation_47331974/Validation_47331974_47331974_168854.276.png", "Validation_47331974/Validation_47331974_47331974_168856.275.png", "Validation_47331974/Validation_47331974_47331974_168858.274.png", "Validation_47331974/Validation_47331974_47331974_168860.273.png" ], "turn_id": 6, "scan_id": "Validation_47331974", "origin_question": "Which two objects are on the same side of you now? \"the rectangular off-white pillow\", \"the large rectangular black monitor with the glossy screen\", \"the medium-sized, classic iron fireplace with glass door\". ", "option": [ "the rectangular off-white pillow, the large rectangular black monitor with the glossy screen", "the rectangular off-white pillow, the medium-sized, classic iron fireplace with glass door", "the large rectangular black monitor with the glossy screen, the medium-sized, classic iron fireplace with glass door" ], "answer": "the large rectangular black monitor with the glossy screen, the medium-sized, classic iron fireplace with glass door", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the rectangular off-white pillow\", \"the large rectangular black monitor with the glossy screen\", \"the medium-sized, classic iron fireplace with glass door\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular off-white pillow, the large rectangular black monitor with the glossy screen', 'the rectangular off-white pillow, the medium-sized, classic iron fireplace with glass door', 'the large rectangular black monitor with the glossy screen, the medium-sized, classic iron fireplace with glass door'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_47331974/Validation_47331974_47331974_168862.272.png", "Validation_47331974/Validation_47331974_47331974_168864.272.png", "Validation_47331974/Validation_47331974_47331974_168866.271.png", "Validation_47331974/Validation_47331974_47331974_168868.270.png", "Validation_47331974/Validation_47331974_47331974_168870.269.png" ], "turn_id": 7, "scan_id": "Validation_47331974", "origin_question": "Please recall \"the rectangular off-white pillow\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.62", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the rectangular off-white pillow\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_47331974/Validation_47331974_47331974_168872.268.png", "Validation_47331974/Validation_47331974_47331974_168874.267.png", "Validation_47331974/Validation_47331974_47331974_168876.267.png", "Validation_47331974/Validation_47331974_47331974_168878.266.png", "Validation_47331974/Validation_47331974_47331974_168880.265.png" ], "turn_id": 8, "scan_id": "Validation_47331974", "origin_question": "By the end of the 3 turn, were you getting closer to or farther away from these two objects: X: \"the modern rectangular white ceramic sink\", Y: \"the plush, curved modern couch in warm tone\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "farther away from both X and Y", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 3 turn, were you getting closer to or farther away from these two objects: X: \"the modern rectangular white ceramic sink\", Y: \"the plush, curved modern couch in warm tone\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_47331974/Validation_47331974_47331974_168882.281.png", "Validation_47331974/Validation_47331974_47331974_168884.280.png", "Validation_47331974/Validation_47331974_47331974_168886.279.png", "Validation_47331974/Validation_47331974_47331974_168888.278.png", "Validation_47331974/Validation_47331974_47331974_168890.278.png" ], "turn_id": 9, "scan_id": "Validation_47331974", "origin_question": "Is the large rectangular black monitor with the glossy screen to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the large rectangular black monitor with the glossy screen to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_47331974/Validation_47331974_47331974_168892.277.png", "Validation_47331974/Validation_47331974_47331974_168894.276.png", "Validation_47331974/Validation_47331974_47331974_168896.275.png", "Validation_47331974/Validation_47331974_47331974_168898.274.png", "Validation_47331974/Validation_47331974_47331974_168900.274.png" ], "turn_id": 10, "scan_id": "Validation_47331974", "origin_question": "How far is your current position from where you were at the end of the turn 8? (in meters)", "option": [], "answer": "1.9", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 8? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_45261588/Validation_45261588_45261588_78064.630.png", "Validation_45261588/Validation_45261588_45261588_78066.096.png", "Validation_45261588/Validation_45261588_45261588_78067.595.png", "Validation_45261588/Validation_45261588_45261588_78069.094.png" ], "turn_id": 1, "scan_id": "Validation_45261588", "origin_question": "Remember, have you seen any pillow(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any pillow(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "Validation_45261588/Validation_45261588_45261588_78070.594.png", "Validation_45261588/Validation_45261588_45261588_78072.093.png", "Validation_45261588/Validation_45261588_45261588_78073.609.png", "Validation_45261588/Validation_45261588_45261588_78075.109.png" ], "turn_id": 2, "scan_id": "Validation_45261588", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the small oval polished white ceramic sink\"; \"the compact, clean white ceramic toilet\"; \"an elegant, light-colored door\" ", "option": [ "the small oval polished white ceramic sink", "the compact, clean white ceramic toilet", "an elegant, light-colored door" ], "answer": "the compact, clean white ceramic toilet", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the small oval polished white ceramic sink\"; \"the compact, clean white ceramic toilet\"; \"an elegant, light-colored door\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the small oval polished white ceramic sink', 'the compact, clean white ceramic toilet', 'an elegant, light-colored door'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_45261588/Validation_45261588_45261588_78076.608.png", "Validation_45261588/Validation_45261588_45261588_78078.107.png", "Validation_45261588/Validation_45261588_45261588_78079.607.png", "Validation_45261588/Validation_45261588_45261588_78081.106.png" ], "turn_id": 3, "scan_id": "Validation_45261588", "origin_question": "When did you first discover \"an elegant, light-colored door\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"an elegant, light-colored door\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_45261588/Validation_45261588_45261588_78082.606.png", "Validation_45261588/Validation_45261588_45261588_78084.105.png", "Validation_45261588/Validation_45261588_45261588_78085.604.png", "Validation_45261588/Validation_45261588_45261588_78087.104.png" ], "turn_id": 4, "scan_id": "Validation_45261588", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the small oval polished white ceramic sink\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the small oval polished white ceramic sink\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_45261588/Validation_45261588_45261588_78088.603.png", "Validation_45261588/Validation_45261588_45261588_78090.102.png", "Validation_45261588/Validation_45261588_45261588_78091.602.png", "Validation_45261588/Validation_45261588_45261588_78093.101.png" ], "turn_id": 5, "scan_id": "Validation_45261588", "origin_question": "In which turn did you see both \"an elegant, light-colored door\" and \"the compact, clean white ceramic toilet\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"an elegant, light-colored door\" and \"the compact, clean white ceramic toilet\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "Validation_45662983/Validation_45662983_45662983_27796.905.png", "Validation_45662983/Validation_45662983_45662983_27798.888.png", "Validation_45662983/Validation_45662983_45662983_27800.887.png", "Validation_45662983/Validation_45662983_45662983_27802.886.png" ], "turn_id": 1, "scan_id": "Validation_45662983", "origin_question": "Remember, how many wine(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many wine(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_45662983/Validation_45662983_45662983_27804.885.png", "Validation_45662983/Validation_45662983_45662983_27806.885.png", "Validation_45662983/Validation_45662983_45662983_27808.884.png", "Validation_45662983/Validation_45662983_45662983_27810.883.png" ], "turn_id": 2, "scan_id": "Validation_45662983", "origin_question": "Which two objects are on the same side of you now? \"the chic beige couch with plush fabric and rectangular shape\", \"the large, sleek rectangular monitor with glossy black casing\", \"the robust green plant with broad, sleek leaves\". ", "option": [ "the chic beige couch with plush fabric and rectangular shape, the large, sleek rectangular monitor with glossy black casing", "the chic beige couch with plush fabric and rectangular shape, the robust green plant with broad, sleek leaves", "the large, sleek rectangular monitor with glossy black casing, the robust green plant with broad, sleek leaves" ], "answer": "the large, sleek rectangular monitor with glossy black casing, the robust green plant with broad, sleek leaves", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the chic beige couch with plush fabric and rectangular shape\", \"the large, sleek rectangular monitor with glossy black casing\", \"the robust green plant with broad, sleek leaves\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the chic beige couch with plush fabric and rectangular shape, the large, sleek rectangular monitor with glossy black casing', 'the chic beige couch with plush fabric and rectangular shape, the robust green plant with broad, sleek leaves', 'the large, sleek rectangular monitor with glossy black casing, the robust green plant with broad, sleek leaves'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_45662983/Validation_45662983_45662983_27812.882.png", "Validation_45662983/Validation_45662983_45662983_27814.898.png", "Validation_45662983/Validation_45662983_45662983_27816.897.png", "Validation_45662983/Validation_45662983_45662983_27818.896.png" ], "turn_id": 3, "scan_id": "Validation_45662983", "origin_question": "At the end of which turn were both of \"the large, white, tall multi-pane window\" and \"the robust green plant with broad, sleek leaves\" on your left side?", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the large, white, tall multi-pane window\" and \"the robust green plant with broad, sleek leaves\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_45662983/Validation_45662983_45662983_27820.896.png", "Validation_45662983/Validation_45662983_45662983_27822.895.png", "Validation_45662983/Validation_45662983_45662983_27824.894.png", "Validation_45662983/Validation_45662983_45662983_27826.893.png" ], "turn_id": 4, "scan_id": "Validation_45662983", "origin_question": "How far is your current position from where you were at the end of the turn 3? (in meters)", "option": [], "answer": "2.25", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 3? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_45662983/Validation_45662983_45662983_27828.892.png", "Validation_45662983/Validation_45662983_45662983_27830.892.png", "Validation_45662983/Validation_45662983_45662983_27832.891.png", "Validation_45662983/Validation_45662983_45662983_27834.890.png" ], "turn_id": 5, "scan_id": "Validation_45662983", "origin_question": "At the end of which turn were both of \"the large, white, tall multi-pane window\" and \"the sleek rectangular mirror with the metallic silver frame\" on your left side?", "option": [], "answer": "1", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the large, white, tall multi-pane window\" and \"the sleek rectangular mirror with the metallic silver frame\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_45662983/Validation_45662983_45662983_27836.889.png", "Validation_45662983/Validation_45662983_45662983_27838.889.png", "Validation_45662983/Validation_45662983_45662983_27840.888.png", "Validation_45662983/Validation_45662983_45662983_27842.887.png" ], "turn_id": 6, "scan_id": "Validation_45662983", "origin_question": "Which direction is the large, white, tall multi-pane window in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the large, white, tall multi-pane window in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_45662983/Validation_45662983_45662983_27844.886.png", "Validation_45662983/Validation_45662983_45662983_27846.885.png", "Validation_45662983/Validation_45662983_45662983_27848.885.png", "Validation_45662983/Validation_45662983_45662983_27850.884.png" ], "turn_id": 7, "scan_id": "Validation_45662983", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the sleek rectangular mirror with the metallic silver frame\" from your position?", "option": [], "answer": "81.6", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the sleek rectangular mirror with the metallic silver frame\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_45662983/Validation_45662983_45662983_27852.883.png", "Validation_45662983/Validation_45662983_45662983_27854.882.png", "Validation_45662983/Validation_45662983_45662983_27856.898.png", "Validation_45662983/Validation_45662983_45662983_27858.897.png" ], "turn_id": 8, "scan_id": "Validation_45662983", "origin_question": "How far is your current position from where you were at the end of the turn 4? (in meters)", "option": [], "answer": "2.9", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 4? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47895341/Validation_47895341_47895341_96386.348.png", "Validation_47895341/Validation_47895341_47895341_96388.314.png", "Validation_47895341/Validation_47895341_47895341_96390.313.png", "Validation_47895341/Validation_47895341_47895341_96392.312.png" ], "turn_id": 1, "scan_id": "Validation_47895341", "origin_question": "Remember, have you seen any toy(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any toy(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_47895341/Validation_47895341_47895341_96394.328.png", "Validation_47895341/Validation_47895341_47895341_96396.327.png", "Validation_47895341/Validation_47895341_47895341_96398.326.png", "Validation_47895341/Validation_47895341_47895341_96400.326.png" ], "turn_id": 2, "scan_id": "Validation_47895341", "origin_question": "When did you first discover \"the dark, minimalist shelf\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the dark, minimalist shelf\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "Validation_47895341/Validation_47895341_47895341_96402.325.png", "Validation_47895341/Validation_47895341_47895341_96404.324.png", "Validation_47895341/Validation_47895341_47895341_96406.323.png", "Validation_47895341/Validation_47895341_47895341_96408.322.png" ], "turn_id": 3, "scan_id": "Validation_47895341", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the modern white ergonomic chair with the high backrest\"; \"the sturdy, brown wooden door with the metallic handle\"; \"the medium-sized rectangular light wood table\" ", "option": [ "the modern white ergonomic chair with the high backrest", "the sturdy, brown wooden door with the metallic handle", "the medium-sized rectangular light wood table" ], "answer": "the medium-sized rectangular light wood table", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the modern white ergonomic chair with the high backrest\"; \"the sturdy, brown wooden door with the metallic handle\"; \"the medium-sized rectangular light wood table\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the modern white ergonomic chair with the high backrest', 'the sturdy, brown wooden door with the metallic handle', 'the medium-sized rectangular light wood table'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_47895341/Validation_47895341_47895341_96410.322.png", "Validation_47895341/Validation_47895341_47895341_96412.321.png", "Validation_47895341/Validation_47895341_47895341_96414.320.png", "Validation_47895341/Validation_47895341_47895341_96416.319.png" ], "turn_id": 4, "scan_id": "Validation_47895341", "origin_question": "Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Right", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_47895341/Validation_47895341_47895341_96418.318.png", "Validation_47895341/Validation_47895341_47895341_96420.318.png", "Validation_47895341/Validation_47895341_47895341_96422.317.png", "Validation_47895341/Validation_47895341_47895341_96424.316.png" ], "turn_id": 5, "scan_id": "Validation_47895341", "origin_question": "Which two objects are on the same side of you now? \"the modern white ergonomic chair with the high backrest\", \"the classic wooden bed with the light-colored sheet\", \"the rectangular blue pillow with white spikes\". ", "option": [ "the modern white ergonomic chair with the high backrest, the classic wooden bed with the light-colored sheet", "the modern white ergonomic chair with the high backrest, the rectangular blue pillow with white spikes", "the classic wooden bed with the light-colored sheet, the rectangular blue pillow with white spikes" ], "answer": "the classic wooden bed with the light-colored sheet, the rectangular blue pillow with white spikes", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the modern white ergonomic chair with the high backrest\", \"the classic wooden bed with the light-colored sheet\", \"the rectangular blue pillow with white spikes\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the modern white ergonomic chair with the high backrest, the classic wooden bed with the light-colored sheet', 'the modern white ergonomic chair with the high backrest, the rectangular blue pillow with white spikes', 'the classic wooden bed with the light-colored sheet, the rectangular blue pillow with white spikes'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "Validation_47895341/Validation_47895341_47895341_96426.315.png", "Validation_47895341/Validation_47895341_47895341_96428.314.png", "Validation_47895341/Validation_47895341_47895341_96430.314.png", "Validation_47895341/Validation_47895341_47895341_96432.313.png" ], "turn_id": 6, "scan_id": "Validation_47895341", "origin_question": "Among these three objects(\"the modern white ergonomic chair with the high backrest\",\"the dark, minimalist shelf\",\"the classic wooden bed with the light-colored sheet\" ), which one is the closest to you?", "option": [ "the modern white ergonomic chair with the high backrest", "the dark, minimalist shelf", "the classic wooden bed with the light-colored sheet" ], "answer": "the classic wooden bed with the light-colored sheet", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects(\"the modern white ergonomic chair with the high backrest\",\"the dark, minimalist shelf\",\"the classic wooden bed with the light-colored sheet\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the modern white ergonomic chair with the high backrest', 'the dark, minimalist shelf', 'the classic wooden bed with the light-colored sheet'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47895341/Validation_47895341_47895341_96434.329.png", "Validation_47895341/Validation_47895341_47895341_96436.328.png", "Validation_47895341/Validation_47895341_47895341_96438.327.png", "Validation_47895341/Validation_47895341_47895341_96440.326.png" ], "turn_id": 7, "scan_id": "Validation_47895341", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the rectangular blue pillow with white spikes\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the rectangular blue pillow with white spikes\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_47895341/Validation_47895341_47895341_96442.325.png", "Validation_47895341/Validation_47895341_47895341_96444.325.png", "Validation_47895341/Validation_47895341_47895341_96446.324.png", "Validation_47895341/Validation_47895341_47895341_96448.323.png" ], "turn_id": 8, "scan_id": "Validation_47895341", "origin_question": "By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the sturdy, brown wooden door with the metallic handle\", Y: \"the classic wooden bed with the light-colored sheet\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to Y and farther away from X", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the sturdy, brown wooden door with the metallic handle\", Y: \"the classic wooden bed with the light-colored sheet\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_47895341/Validation_47895341_47895341_96450.322.png", "Validation_47895341/Validation_47895341_47895341_96452.321.png", "Validation_47895341/Validation_47895341_47895341_96454.321.png", "Validation_47895341/Validation_47895341_47895341_96456.320.png" ], "turn_id": 9, "scan_id": "Validation_47895341", "origin_question": "Please recall \"the rectangular blue pillow with white spikes\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "4.35", "user_message": "For the 9 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the rectangular blue pillow with white spikes\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "Validation_47115463/Validation_47115463_47115463_13219.484.png", "Validation_47115463/Validation_47115463_47115463_13221.466.png", "Validation_47115463/Validation_47115463_47115463_13223.465.png", "Validation_47115463/Validation_47115463_47115463_13225.465.png", "Validation_47115463/Validation_47115463_47115463_13227.480.png" ], "turn_id": 1, "scan_id": "Validation_47115463", "origin_question": "Remember, how many chair(s) have you seen so far? ", "option": [], "answer": "2", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many chair(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_47115463/Validation_47115463_47115463_13229.480.png", "Validation_47115463/Validation_47115463_47115463_13231.479.png", "Validation_47115463/Validation_47115463_47115463_13233.478.png", "Validation_47115463/Validation_47115463_47115463_13235.477.png", "Validation_47115463/Validation_47115463_47115463_13237.476.png" ], "turn_id": 2, "scan_id": "Validation_47115463", "origin_question": "When did you first discover \"the polished, rectangular stainless steel oven\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the polished, rectangular stainless steel oven\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_47115463/Validation_47115463_47115463_13239.475.png", "Validation_47115463/Validation_47115463_47115463_13241.475.png", "Validation_47115463/Validation_47115463_47115463_13243.474.png", "Validation_47115463/Validation_47115463_47115463_13245.473.png", "Validation_47115463/Validation_47115463_47115463_13247.472.png" ], "turn_id": 3, "scan_id": "Validation_47115463", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47115463/Validation_47115463_47115463_13249.471.png", "Validation_47115463/Validation_47115463_47115463_13251.470.png", "Validation_47115463/Validation_47115463_47115463_13253.470.png", "Validation_47115463/Validation_47115463_47115463_13255.469.png", "Validation_47115463/Validation_47115463_47115463_13257.468.png" ], "turn_id": 4, "scan_id": "Validation_47115463", "origin_question": "", "option": [], "answer": "", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_47115463/Validation_47115463_47115463_13259.467.png", "Validation_47115463/Validation_47115463_47115463_13261.466.png", "Validation_47115463/Validation_47115463_47115463_13263.466.png", "Validation_47115463/Validation_47115463_47115463_13265.465.png", "Validation_47115463/Validation_47115463_47115463_13267.480.png" ], "turn_id": 5, "scan_id": "Validation_47115463", "origin_question": "Please recall \"the rectangular, polished stainless steel stove\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.98", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the rectangular, polished stainless steel stove\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_47115463/Validation_47115463_47115463_13269.480.png", "Validation_47115463/Validation_47115463_47115463_13271.479.png", "Validation_47115463/Validation_47115463_47115463_13273.478.png", "Validation_47115463/Validation_47115463_47115463_13275.477.png", "Validation_47115463/Validation_47115463_47115463_13277.476.png" ], "turn_id": 6, "scan_id": "Validation_47115463", "origin_question": "When did you first discover \"the natural wooden door with visible wood grain\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the natural wooden door with visible wood grain\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_47115463/Validation_47115463_47115463_13279.476.png", "Validation_47115463/Validation_47115463_47115463_13281.475.png", "Validation_47115463/Validation_47115463_47115463_13283.474.png", "Validation_47115463/Validation_47115463_47115463_13285.473.png", "Validation_47115463/Validation_47115463_47115463_13287.472.png" ], "turn_id": 7, "scan_id": "Validation_47115463", "origin_question": "When did you first discover \"the polished, rectangular stainless steel oven\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the polished, rectangular stainless steel oven\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_47115463/Validation_47115463_47115463_13289.471.png", "Validation_47115463/Validation_47115463_47115463_13291.470.png", "Validation_47115463/Validation_47115463_47115463_13293.470.png", "Validation_47115463/Validation_47115463_47115463_13295.469.png", "Validation_47115463/Validation_47115463_47115463_13297.468.png" ], "turn_id": 8, "scan_id": "Validation_47115463", "origin_question": "By the end of the 7 turn, were you getting closer to or farther away from these two objects: X: \"the round glass table with the metal frame\", Y: \"the polished, rectangular stainless steel oven\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "farther away from both X and Y", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 7 turn, were you getting closer to or farther away from these two objects: X: \"the round glass table with the metal frame\", Y: \"the polished, rectangular stainless steel oven\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47115463/Validation_47115463_47115463_13299.467.png", "Validation_47115463/Validation_47115463_47115463_13301.466.png", "Validation_47115463/Validation_47115463_47115463_13303.466.png", "Validation_47115463/Validation_47115463_47115463_13305.465.png", "Validation_47115463/Validation_47115463_47115463_13307.464.png" ], "turn_id": 9, "scan_id": "Validation_47115463", "origin_question": "Which direction is the polished, rectangular stainless steel oven in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the polished, rectangular stainless steel oven in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42899611/Validation_42899611_42899611_228349.736.png", "Validation_42899611/Validation_42899611_42899611_228351.719.png", "Validation_42899611/Validation_42899611_42899611_228353.718.png", "Validation_42899611/Validation_42899611_42899611_228355.717.png" ], "turn_id": 1, "scan_id": "Validation_42899611", "origin_question": "Remember, have you seen any pillow(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any pillow(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_42899611/Validation_42899611_42899611_228357.717.png", "Validation_42899611/Validation_42899611_42899611_228359.716.png", "Validation_42899611/Validation_42899611_42899611_228361.715.png", "Validation_42899611/Validation_42899611_42899611_228363.714.png" ], "turn_id": 2, "scan_id": "Validation_42899611", "origin_question": "When did you first discover \"the bed\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the bed\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_42899611/Validation_42899611_42899611_228365.713.png", "Validation_42899611/Validation_42899611_42899611_228367.712.png", "Validation_42899611/Validation_42899611_42899611_228369.712.png", "Validation_42899611/Validation_42899611_42899611_228371.728.png" ], "turn_id": 3, "scan_id": "Validation_42899611", "origin_question": "What will be the first-time appearance order of the following categories: chair, light, vase?", "option": [ "vase, chair, light", "light, chair, vase", "light, vase, chair", "vase, light, chair" ], "answer": "light, chair, vase", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: chair, light, vase? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['vase, chair, light', 'light, chair, vase', 'light, vase, chair', 'vase, light, chair'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_42899611/Validation_42899611_42899611_228373.727.png", "Validation_42899611/Validation_42899611_42899611_228375.726.png", "Validation_42899611/Validation_42899611_42899611_228377.725.png", "Validation_42899611/Validation_42899611_42899611_228379.724.png" ], "turn_id": 4, "scan_id": "Validation_42899611", "origin_question": "What will be the first-time appearance order of the following categories: bed, cabinet, vase?", "option": [ "bed, vase, cabinet", "vase, cabinet, bed", "cabinet, bed, vase", "bed, cabinet, vase" ], "answer": "bed, cabinet, vase", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bed, cabinet, vase? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['bed, vase, cabinet', 'vase, cabinet, bed', 'cabinet, bed, vase', 'bed, cabinet, vase'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_42899611/Validation_42899611_42899611_228381.724.png", "Validation_42899611/Validation_42899611_42899611_228383.723.png", "Validation_42899611/Validation_42899611_42899611_228385.722.png", "Validation_42899611/Validation_42899611_42899611_228387.721.png" ], "turn_id": 5, "scan_id": "Validation_42899611", "origin_question": "", "option": [], "answer": "", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_42899611/Validation_42899611_42899611_228389.720.png", "Validation_42899611/Validation_42899611_42899611_228391.720.png", "Validation_42899611/Validation_42899611_42899611_228393.719.png", "Validation_42899611/Validation_42899611_42899611_228395.718.png" ], "turn_id": 6, "scan_id": "Validation_42899611", "origin_question": "In which turn did you see both \"the bed\" and \"the deep brown wrought iron light\" simultaneously? ", "option": [], "answer": "6", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the bed\" and \"the deep brown wrought iron light\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_42899611/Validation_42899611_42899611_228397.717.png", "Validation_42899611/Validation_42899611_42899611_228399.717.png", "Validation_42899611/Validation_42899611_42899611_228401.716.png", "Validation_42899611/Validation_42899611_42899611_228403.715.png" ], "turn_id": 7, "scan_id": "Validation_42899611", "origin_question": "", "option": [], "answer": "", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_42899611/Validation_42899611_42899611_228405.714.png", "Validation_42899611/Validation_42899611_42899611_228407.713.png", "Validation_42899611/Validation_42899611_42899611_228409.713.png", "Validation_42899611/Validation_42899611_42899611_228411.712.png" ], "turn_id": 8, "scan_id": "Validation_42899611", "origin_question": "Is the modern dark chair with plush leather and wood details to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the modern dark chair with plush leather and wood details to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42446541/Validation_42446541_42446541_201550.462.png", "Validation_42446541/Validation_42446541_42446541_201552.444.png", "Validation_42446541/Validation_42446541_42446541_201554.443.png", "Validation_42446541/Validation_42446541_42446541_201556.443.png", "Validation_42446541/Validation_42446541_42446541_201558.442.png" ], "turn_id": 1, "scan_id": "Validation_42446541", "origin_question": "Remember, have you seen any couch(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any couch(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_42446541/Validation_42446541_42446541_201560.441.png", "Validation_42446541/Validation_42446541_42446541_201562.440.png", "Validation_42446541/Validation_42446541_42446541_201564.439.png", "Validation_42446541/Validation_42446541_42446541_201566.439.png", "Validation_42446541/Validation_42446541_42446541_201568.438.png" ], "turn_id": 2, "scan_id": "Validation_42446541", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the sleek rectangular table with light wood tones\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the sleek rectangular table with light wood tones\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_42446541/Validation_42446541_42446541_201570.437.png", "Validation_42446541/Validation_42446541_42446541_201572.436.png", "Validation_42446541/Validation_42446541_42446541_201574.435.png", "Validation_42446541/Validation_42446541_42446541_201576.435.png", "Validation_42446541/Validation_42446541_42446541_201578.434.png" ], "turn_id": 3, "scan_id": "Validation_42446541", "origin_question": "In which turn were you farthest from \"the large flat-screen monitor with the black bezel\"? ", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the large flat-screen monitor with the black bezel\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_42446541/Validation_42446541_42446541_201580.449.png", "Validation_42446541/Validation_42446541_42446541_201582.449.png", "Validation_42446541/Validation_42446541_42446541_201584.448.png", "Validation_42446541/Validation_42446541_42446541_201586.447.png", "Validation_42446541/Validation_42446541_42446541_201588.446.png" ], "turn_id": 4, "scan_id": "Validation_42446541", "origin_question": "Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_42446541/Validation_42446541_42446541_201590.445.png", "Validation_42446541/Validation_42446541_42446541_201592.445.png", "Validation_42446541/Validation_42446541_42446541_201594.444.png", "Validation_42446541/Validation_42446541_42446541_201596.443.png", "Validation_42446541/Validation_42446541_42446541_201598.442.png" ], "turn_id": 5, "scan_id": "Validation_42446541", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_42446541/Validation_42446541_42446541_201600.441.png", "Validation_42446541/Validation_42446541_42446541_201602.440.png", "Validation_42446541/Validation_42446541_42446541_201604.440.png", "Validation_42446541/Validation_42446541_42446541_201606.439.png", "Validation_42446541/Validation_42446541_42446541_201608.438.png" ], "turn_id": 6, "scan_id": "Validation_42446541", "origin_question": "Using your orientation at the end of turn 5 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "90.66", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 5 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_42446541/Validation_42446541_42446541_201610.437.png", "Validation_42446541/Validation_42446541_42446541_201612.436.png", "Validation_42446541/Validation_42446541_42446541_201614.435.png", "Validation_42446541/Validation_42446541_42446541_201616.435.png", "Validation_42446541/Validation_42446541_42446541_201618.434.png" ], "turn_id": 7, "scan_id": "Validation_42446541", "origin_question": "When did you first discover \"the sleek white stainless steel refrigerator\"? (index of the turn)", "option": [], "answer": "7", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the sleek white stainless steel refrigerator\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_42446541/Validation_42446541_42446541_201620.450.png", "Validation_42446541/Validation_42446541_42446541_201622.449.png", "Validation_42446541/Validation_42446541_42446541_201624.448.png", "Validation_42446541/Validation_42446541_42446541_201626.447.png", "Validation_42446541/Validation_42446541_42446541_201628.446.png" ], "turn_id": 8, "scan_id": "Validation_42446541", "origin_question": "Compared to your position at the end of the 7 turn, are you now closer or farther away from \"the sleek white washing machine with stainless steel\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 7 turn, are you now closer or farther away from \"the sleek white washing machine with stainless steel\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_42446541/Validation_42446541_42446541_201630.446.png", "Validation_42446541/Validation_42446541_42446541_201632.445.png", "Validation_42446541/Validation_42446541_42446541_201634.444.png", "Validation_42446541/Validation_42446541_42446541_201636.443.png", "Validation_42446541/Validation_42446541_42446541_201638.442.png" ], "turn_id": 9, "scan_id": "Validation_42446541", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the sleek white washing machine with stainless steel\" from your position?", "option": [], "answer": "147.84", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the sleek white washing machine with stainless steel\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_42446541/Validation_42446541_42446541_201640.441.png", "Validation_42446541/Validation_42446541_42446541_201642.441.png", "Validation_42446541/Validation_42446541_42446541_201644.440.png", "Validation_42446541/Validation_42446541_42446541_201646.439.png", "Validation_42446541/Validation_42446541_42446541_201648.438.png" ], "turn_id": 10, "scan_id": "Validation_42446541", "origin_question": "", "option": [], "answer": "", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_41159566/Validation_41159566_41159566_35197.443.png", "Validation_41159566/Validation_41159566_41159566_35199.426.png", "Validation_41159566/Validation_41159566_41159566_35201.425.png", "Validation_41159566/Validation_41159566_41159566_35203.424.png", "Validation_41159566/Validation_41159566_41159566_35205.423.png" ], "turn_id": 1, "scan_id": "Validation_41159566", "origin_question": "Remember, have you seen any curtain(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any curtain(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_41159566/Validation_41159566_41159566_35207.423.png", "Validation_41159566/Validation_41159566_41159566_35209.422.png", "Validation_41159566/Validation_41159566_41159566_35211.421.png", "Validation_41159566/Validation_41159566_41159566_35213.420.png", "Validation_41159566/Validation_41159566_41159566_35215.420.png" ], "turn_id": 2, "scan_id": "Validation_41159566", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_41159566/Validation_41159566_41159566_35217.419.png", "Validation_41159566/Validation_41159566_41159566_35219.418.png", "Validation_41159566/Validation_41159566_41159566_35221.417.png", "Validation_41159566/Validation_41159566_41159566_35223.416.png", "Validation_41159566/Validation_41159566_41159566_35225.416.png" ], "turn_id": 3, "scan_id": "Validation_41159566", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_41159566/Validation_41159566_41159566_35227.415.png", "Validation_41159566/Validation_41159566_41159566_35229.414.png", "Validation_41159566/Validation_41159566_41159566_35231.413.png", "Validation_41159566/Validation_41159566_41159566_35233.412.png", "Validation_41159566/Validation_41159566_41159566_35235.412.png" ], "turn_id": 4, "scan_id": "Validation_41159566", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_41159566/Validation_41159566_41159566_35237.411.png", "Validation_41159566/Validation_41159566_41159566_35239.410.png", "Validation_41159566/Validation_41159566_41159566_35241.426.png", "Validation_41159566/Validation_41159566_41159566_35243.425.png", "Validation_41159566/Validation_41159566_41159566_35245.424.png" ], "turn_id": 5, "scan_id": "Validation_41159566", "origin_question": "By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the fireplace\", Y: \"the tall, sleek white wooden shelf with compartments\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the fireplace\", Y: \"the tall, sleek white wooden shelf with compartments\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "Validation_41159566/Validation_41159566_41159566_35247.423.png", "Validation_41159566/Validation_41159566_41159566_35249.423.png", "Validation_41159566/Validation_41159566_41159566_35251.422.png", "Validation_41159566/Validation_41159566_41159566_35253.421.png", "Validation_41159566/Validation_41159566_41159566_35255.420.png" ], "turn_id": 6, "scan_id": "Validation_41159566", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the natural wood stool with the medium-brown square top\"; \"the medium plant with lush, glossy green leaves\"; \"the large rectangular book with the colorful glossy cover\" ", "option": [ "the natural wood stool with the medium-brown square top", "the medium plant with lush, glossy green leaves", "the large rectangular book with the colorful glossy cover" ], "answer": "the large rectangular book with the colorful glossy cover", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the natural wood stool with the medium-brown square top\"; \"the medium plant with lush, glossy green leaves\"; \"the large rectangular book with the colorful glossy cover\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the natural wood stool with the medium-brown square top', 'the medium plant with lush, glossy green leaves', 'the large rectangular book with the colorful glossy cover'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_41159566/Validation_41159566_41159566_35257.420.png", "Validation_41159566/Validation_41159566_41159566_35259.419.png", "Validation_41159566/Validation_41159566_41159566_35261.418.png", "Validation_41159566/Validation_41159566_41159566_35263.417.png", "Validation_41159566/Validation_41159566_41159566_35265.416.png" ], "turn_id": 7, "scan_id": "Validation_41159566", "origin_question": "At the end of which turn were both of \"the stairs\" and \"the fireplace\" on your left side?", "option": [], "answer": "1", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the stairs\" and \"the fireplace\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_41159566/Validation_41159566_41159566_35267.416.png", "Validation_41159566/Validation_41159566_41159566_35269.415.png", "Validation_41159566/Validation_41159566_41159566_35271.414.png", "Validation_41159566/Validation_41159566_41159566_35273.413.png", "Validation_41159566/Validation_41159566_41159566_35275.412.png" ], "turn_id": 8, "scan_id": "Validation_41159566", "origin_question": "", "option": [], "answer": "", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_41159566/Validation_41159566_41159566_35277.412.png", "Validation_41159566/Validation_41159566_41159566_35279.411.png", "Validation_41159566/Validation_41159566_41159566_35281.410.png", "Validation_41159566/Validation_41159566_41159566_35283.426.png", "Validation_41159566/Validation_41159566_41159566_35285.425.png" ], "turn_id": 9, "scan_id": "Validation_41159566", "origin_question": "", "option": [], "answer": "", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_41159566/Validation_41159566_41159566_35287.424.png", "Validation_41159566/Validation_41159566_41159566_35289.423.png", "Validation_41159566/Validation_41159566_41159566_35291.423.png", "Validation_41159566/Validation_41159566_41159566_35293.422.png", "Validation_41159566/Validation_41159566_41159566_35295.421.png" ], "turn_id": 10, "scan_id": "Validation_41159566", "origin_question": "What will be the first-time appearance order of the following categories: cabinet, plant, table?", "option": [ "table, plant, cabinet", "cabinet, plant, table", "table, cabinet, plant", "plant, cabinet, table" ], "answer": "table, plant, cabinet", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: cabinet, plant, table? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['table, plant, cabinet', 'cabinet, plant, table', 'table, cabinet, plant', 'plant, cabinet, table'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47331646/Validation_47331646_47331646_21956.809.png", "Validation_47331646/Validation_47331646_47331646_21958.808.png", "Validation_47331646/Validation_47331646_47331646_21960.807.png", "Validation_47331646/Validation_47331646_47331646_21962.807.png", "Validation_47331646/Validation_47331646_47331646_21964.806.png" ], "turn_id": 1, "scan_id": "Validation_47331646", "origin_question": "Remember, have you seen any door(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any door(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_47331646/Validation_47331646_47331646_21966.805.png", "Validation_47331646/Validation_47331646_47331646_21968.804.png", "Validation_47331646/Validation_47331646_47331646_21970.804.png", "Validation_47331646/Validation_47331646_47331646_21972.803.png", "Validation_47331646/Validation_47331646_47331646_21974.802.png" ], "turn_id": 2, "scan_id": "Validation_47331646", "origin_question": "When did you first discover \"the traditional cream lamp with the glossy base\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the traditional cream lamp with the glossy base\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47331646/Validation_47331646_47331646_21976.801.png", "Validation_47331646/Validation_47331646_47331646_21978.817.png", "Validation_47331646/Validation_47331646_47331646_21980.816.png", "Validation_47331646/Validation_47331646_47331646_21982.815.png", "Validation_47331646/Validation_47331646_47331646_21984.815.png" ], "turn_id": 3, "scan_id": "Validation_47331646", "origin_question": "", "option": [], "answer": "", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "Validation_47331646/Validation_47331646_47331646_21986.814.png", "Validation_47331646/Validation_47331646_47331646_21988.813.png", "Validation_47331646/Validation_47331646_47331646_21990.812.png", "Validation_47331646/Validation_47331646_47331646_21992.811.png", "Validation_47331646/Validation_47331646_47331646_21994.810.png" ], "turn_id": 4, "scan_id": "Validation_47331646", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the mosaic-patterned couch in brown and gray tones\"; \"the sleek vertical white steel radiator\"; \"the traditional cream lamp with the glossy base\" ", "option": [ "the mosaic-patterned couch in brown and gray tones", "the sleek vertical white steel radiator", "the traditional cream lamp with the glossy base" ], "answer": "the sleek vertical white steel radiator", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the mosaic-patterned couch in brown and gray tones\"; \"the sleek vertical white steel radiator\"; \"the traditional cream lamp with the glossy base\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the mosaic-patterned couch in brown and gray tones', 'the sleek vertical white steel radiator', 'the traditional cream lamp with the glossy base'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_47331646/Validation_47331646_47331646_21996.810.png", "Validation_47331646/Validation_47331646_47331646_21998.809.png", "Validation_47331646/Validation_47331646_47331646_22000.808.png", "Validation_47331646/Validation_47331646_47331646_22002.807.png", "Validation_47331646/Validation_47331646_47331646_22004.806.png" ], "turn_id": 5, "scan_id": "Validation_47331646", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47331646/Validation_47331646_47331646_22006.805.png", "Validation_47331646/Validation_47331646_47331646_22008.805.png", "Validation_47331646/Validation_47331646_47331646_22010.804.png", "Validation_47331646/Validation_47331646_47331646_22012.803.png", "Validation_47331646/Validation_47331646_47331646_22014.802.png" ], "turn_id": 6, "scan_id": "Validation_47331646", "origin_question": "", "option": [], "answer": "", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "Validation_47331646/Validation_47331646_47331646_22016.802.png", "Validation_47331646/Validation_47331646_47331646_22018.801.png", "Validation_47331646/Validation_47331646_47331646_22020.800.png", "Validation_47331646/Validation_47331646_47331646_22022.816.png", "Validation_47331646/Validation_47331646_47331646_22024.815.png" ], "turn_id": 7, "scan_id": "Validation_47331646", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the sleek cylindrical transparent glass vase\"; \"the mosaic-patterned couch in brown and gray tones\"; \"the round, light gray upholstered stool with the wooden frame\" ", "option": [ "the sleek cylindrical transparent glass vase", "the mosaic-patterned couch in brown and gray tones", "the round, light gray upholstered stool with the wooden frame" ], "answer": "the sleek cylindrical transparent glass vase", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the sleek cylindrical transparent glass vase\"; \"the mosaic-patterned couch in brown and gray tones\"; \"the round, light gray upholstered stool with the wooden frame\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the sleek cylindrical transparent glass vase', 'the mosaic-patterned couch in brown and gray tones', 'the round, light gray upholstered stool with the wooden frame'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_47331646/Validation_47331646_47331646_22026.814.png", "Validation_47331646/Validation_47331646_47331646_22028.813.png", "Validation_47331646/Validation_47331646_47331646_22030.812.png", "Validation_47331646/Validation_47331646_47331646_22032.812.png", "Validation_47331646/Validation_47331646_47331646_22034.811.png" ], "turn_id": 8, "scan_id": "Validation_47331646", "origin_question": "By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the traditional cream lamp with the glossy base\", Y: \"the round, light gray upholstered stool with the wooden frame\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to Y and farther away from X", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the traditional cream lamp with the glossy base\", Y: \"the round, light gray upholstered stool with the wooden frame\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_47331646/Validation_47331646_47331646_22036.810.png", "Validation_47331646/Validation_47331646_47331646_22038.809.png", "Validation_47331646/Validation_47331646_47331646_22040.808.png", "Validation_47331646/Validation_47331646_47331646_22042.807.png", "Validation_47331646/Validation_47331646_47331646_22044.807.png" ], "turn_id": 9, "scan_id": "Validation_47331646", "origin_question": "By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the cream-colored wooden door with large glass panels\", Y: \"the natural beige wicker chair with the high backrest\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to Y and farther away from X", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the cream-colored wooden door with large glass panels\", Y: \"the natural beige wicker chair with the high backrest\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_47331646/Validation_47331646_47331646_22046.806.png", "Validation_47331646/Validation_47331646_47331646_22048.805.png", "Validation_47331646/Validation_47331646_47331646_22050.804.png", "Validation_47331646/Validation_47331646_47331646_22052.804.png", "Validation_47331646/Validation_47331646_47331646_22054.803.png" ], "turn_id": 10, "scan_id": "Validation_47331646", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the mosaic-patterned couch in brown and gray tones\" from your position?", "option": [], "answer": "159.36", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the mosaic-patterned couch in brown and gray tones\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0004_region6/550a4ce414334860aa94a9f9e1a3833c_i1_5.jpg", "1mp3d_0004_region6/550a4ce414334860aa94a9f9e1a3833c_i2_5.jpg", "1mp3d_0004_region6/3fa6056d3a874844ae03c2bb9ab1e78a_i1_0.jpg", "1mp3d_0004_region6/3fa6056d3a874844ae03c2bb9ab1e78a_i2_0.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0004_region6", "origin_question": "Remember, have you seen any purse(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any purse(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0004_region6/eacb6a103fbf44f4a7f5d339b11d3f6b_i1_1.jpg", "1mp3d_0004_region6/eacb6a103fbf44f4a7f5d339b11d3f6b_i2_1.jpg", "1mp3d_0004_region6/eacb6a103fbf44f4a7f5d339b11d3f6b_i1_2.jpg", "1mp3d_0004_region6/eacb6a103fbf44f4a7f5d339b11d3f6b_i2_2.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0004_region6", "origin_question": "Remember, have you seen any sink(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any sink(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "1mp3d_0004_region6/3fa6056d3a874844ae03c2bb9ab1e78a_i1_5.jpg", "1mp3d_0004_region6/3fa6056d3a874844ae03c2bb9ab1e78a_i2_5.jpg", "1mp3d_0004_region6/550a4ce414334860aa94a9f9e1a3833c_i1_4.jpg", "1mp3d_0004_region6/550a4ce414334860aa94a9f9e1a3833c_i2_4.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0004_region6", "origin_question": "Which direction is the fruit in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the fruit in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "1mp3d_0004_region6/3fa6056d3a874844ae03c2bb9ab1e78a_i1_4.jpg", "1mp3d_0004_region6/3fa6056d3a874844ae03c2bb9ab1e78a_i2_4.jpg", "1mp3d_0004_region6/3fa6056d3a874844ae03c2bb9ab1e78a_i1_3.jpg", "1mp3d_0004_region6/3fa6056d3a874844ae03c2bb9ab1e78a_i2_3.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0004_region6", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the round ceramic plate with black and white patterns\" from your position?", "option": [], "answer": "103.87", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the round ceramic plate with black and white patterns\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "1mp3d_0004_region6/eacb6a103fbf44f4a7f5d339b11d3f6b_i1_3.jpg", "1mp3d_0004_region6/eacb6a103fbf44f4a7f5d339b11d3f6b_i2_3.jpg", "1mp3d_0004_region6/eacb6a103fbf44f4a7f5d339b11d3f6b_i1_4.jpg", "1mp3d_0004_region6/eacb6a103fbf44f4a7f5d339b11d3f6b_i2_4.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0004_region6", "origin_question": "At the end of which turn were both of \"the large rectangular stainless steel stove with multiple burners\" and \"the fruit\" on your right side?", "option": [], "answer": "1", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the large rectangular stainless steel stove with multiple burners\" and \"the fruit\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "1mp3d_0004_region6/eacb6a103fbf44f4a7f5d339b11d3f6b_i1_5.jpg", "1mp3d_0004_region6/eacb6a103fbf44f4a7f5d339b11d3f6b_i2_5.jpg", "1mp3d_0004_region6/eacb6a103fbf44f4a7f5d339b11d3f6b_i1_0.jpg", "1mp3d_0004_region6/eacb6a103fbf44f4a7f5d339b11d3f6b_i2_0.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0004_region6", "origin_question": "Using your orientation at the end of turn 5 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "119.62", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 5 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "1mp3d_0078_region9/90440703eace49f3b867750beca05a25_i1_4.jpg", "1mp3d_0078_region9/90440703eace49f3b867750beca05a25_i2_4.jpg", "1mp3d_0078_region9/90440703eace49f3b867750beca05a25_i1_3.jpg", "1mp3d_0078_region9/90440703eace49f3b867750beca05a25_i2_3.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0078_region9", "origin_question": "Remember, how many plate(s) have you seen so far? ", "option": [], "answer": "4", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many plate(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "1mp3d_0078_region9/90440703eace49f3b867750beca05a25_i1_2.jpg", "1mp3d_0078_region9/90440703eace49f3b867750beca05a25_i2_2.jpg", "1mp3d_0078_region9/0eefd4bea0874786ab4946ca120bff73_i1_4.jpg", "1mp3d_0078_region9/0eefd4bea0874786ab4946ca120bff73_i2_4.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0078_region9", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular black and silver microwave appliance\", Y: \"the rectangular silver coffee maker with a shiny surface\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular black and silver microwave appliance\", Y: \"the rectangular silver coffee maker with a shiny surface\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "1mp3d_0078_region9/3c286cf678234163b7cb2ab3785fc498_i1_5.jpg", "1mp3d_0078_region9/3c286cf678234163b7cb2ab3785fc498_i2_5.jpg", "1mp3d_0078_region9/afd7d720b2f942a88f24b9d2e24a9581_i1_4.jpg", "1mp3d_0078_region9/afd7d720b2f942a88f24b9d2e24a9581_i2_4.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0078_region9", "origin_question": "Which two objects are on the same side of you now? \"the round wooden pan with a light beige color\", \"the counter\", \"the round silver stainless steel sink\". ", "option": [ "the round wooden pan with a light beige color, the counter", "the round wooden pan with a light beige color, the round silver stainless steel sink", "the counter, the round silver stainless steel sink" ], "answer": "the counter, the round silver stainless steel sink", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the round wooden pan with a light beige color\", \"the counter\", \"the round silver stainless steel sink\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the round wooden pan with a light beige color, the counter', 'the round wooden pan with a light beige color, the round silver stainless steel sink', 'the counter, the round silver stainless steel sink'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "1mp3d_0078_region9/3c286cf678234163b7cb2ab3785fc498_i1_3.jpg", "1mp3d_0078_region9/3c286cf678234163b7cb2ab3785fc498_i2_3.jpg", "1mp3d_0078_region9/0eefd4bea0874786ab4946ca120bff73_i1_3.jpg", "1mp3d_0078_region9/0eefd4bea0874786ab4946ca120bff73_i2_3.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0078_region9", "origin_question": "Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "35.1", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "1mp3d_0078_region9/f68715fa4e134a89958dad587b6cebbc_i1_2.jpg", "1mp3d_0078_region9/f68715fa4e134a89958dad587b6cebbc_i2_2.jpg", "1mp3d_0078_region9/f68715fa4e134a89958dad587b6cebbc_i1_1.jpg", "1mp3d_0078_region9/f68715fa4e134a89958dad587b6cebbc_i2_1.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0078_region9", "origin_question": "Please recall \"the rectangular silver coffee maker with a shiny surface\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.44", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the rectangular silver coffee maker with a shiny surface\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0078_region9/7ab6f8baf8bc41628a3b91fa4660ce43_i2_2.jpg", "1mp3d_0078_region9/7bbddf0c8df249a3b8a6da60b1b7fd57_i1_0.jpg", "1mp3d_0078_region9/7bbddf0c8df249a3b8a6da60b1b7fd57_i2_0.jpg", "1mp3d_0078_region9/32cd64f9c7d74e7185517bca189a00b4_i1_5.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0078_region9", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the rectangular black and silver microwave appliance\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the rectangular black and silver microwave appliance\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0078_region9/32cd64f9c7d74e7185517bca189a00b4_i2_5.jpg", "1mp3d_0078_region9/32cd64f9c7d74e7185517bca189a00b4_i1_0.jpg", "1mp3d_0078_region9/32cd64f9c7d74e7185517bca189a00b4_i2_0.jpg", "1mp3d_0078_region9/7bbddf0c8df249a3b8a6da60b1b7fd57_i2_3.jpg" ], "turn_id": 7, "scan_id": "1mp3d_0078_region9", "origin_question": "", "option": [], "answer": "", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "1mp3d_0078_region9/7bbddf0c8df249a3b8a6da60b1b7fd57_i1_1.jpg", "1mp3d_0078_region9/7bbddf0c8df249a3b8a6da60b1b7fd57_i2_1.jpg", "1mp3d_0078_region9/7ab6f8baf8bc41628a3b91fa4660ce43_i1_4.jpg", "1mp3d_0078_region9/7ab6f8baf8bc41628a3b91fa4660ce43_i2_4.jpg" ], "turn_id": 8, "scan_id": "1mp3d_0078_region9", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "1.66", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0559_02/00000.jpg", "scene0559_02/00030.jpg", "scene0559_02/00060.jpg", "scene0559_02/00090.jpg" ], "turn_id": 1, "scan_id": "scene0559_02", "origin_question": "Remember, have you seen any couch(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any couch(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0559_02/00120.jpg", "scene0559_02/00150.jpg", "scene0559_02/00180.jpg", "scene0559_02/00210.jpg" ], "turn_id": 2, "scan_id": "scene0559_02", "origin_question": "In which turn were you closest to \"the small round green ottoman made of soft velvet\"? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the small round green ottoman made of soft velvet\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0559_02/00240.jpg", "scene0559_02/00270.jpg", "scene0559_02/00300.jpg", "scene0559_02/00330.jpg" ], "turn_id": 3, "scan_id": "scene0559_02", "origin_question": "In which turn did you see both \"the small round green ottoman made of soft velvet\" and \"the gray sports shoe with white laces and a soft upper\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the small round green ottoman made of soft velvet\" and \"the gray sports shoe with white laces and a soft upper\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0559_02/00360.jpg", "scene0559_02/00390.jpg", "scene0559_02/00420.jpg", "scene0559_02/00450.jpg" ], "turn_id": 4, "scan_id": "scene0559_02", "origin_question": "What will be the first-time appearance order of the following categories: blackboard, clothes, ottoman?", "option": [ "clothes, blackboard, ottoman", "ottoman, blackboard, clothes", "ottoman, clothes, blackboard", "clothes, ottoman, blackboard" ], "answer": "ottoman, clothes, blackboard", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: blackboard, clothes, ottoman? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['clothes, blackboard, ottoman', 'ottoman, blackboard, clothes', 'ottoman, clothes, blackboard', 'clothes, ottoman, blackboard'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0559_02/00480.jpg", "scene0559_02/00510.jpg", "scene0559_02/00540.jpg", "scene0559_02/00570.jpg" ], "turn_id": 5, "scan_id": "scene0559_02", "origin_question": "What will be the first-time appearance order of the following categories: bench, blackboard, table?", "option": [ "blackboard, table, bench", "bench, blackboard, table", "table, bench, blackboard", "table, blackboard, bench" ], "answer": "table, bench, blackboard", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bench, blackboard, table? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['blackboard, table, bench', 'bench, blackboard, table', 'table, bench, blackboard', 'table, blackboard, bench'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0559_02/00600.jpg", "scene0559_02/00630.jpg", "scene0559_02/00660.jpg", "scene0559_02/00690.jpg" ], "turn_id": 6, "scan_id": "scene0559_02", "origin_question": "Please recall \"the small rectangular blackboard with a white background\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.91", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the small rectangular blackboard with a white background\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0204_00/00000.jpg", "scene0204_00/00030.jpg", "scene0204_00/00060.jpg", "scene0204_00/00090.jpg" ], "turn_id": 1, "scan_id": "scene0204_00", "origin_question": "Remember, have you seen any cup(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any cup(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0204_00/00120.jpg", "scene0204_00/00150.jpg", "scene0204_00/00180.jpg", "scene0204_00/00210.jpg" ], "turn_id": 2, "scan_id": "scene0204_00", "origin_question": "When did you first discover \"the large rectangular light brown wooden table\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the large rectangular light brown wooden table\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0204_00/00240.jpg", "scene0204_00/00270.jpg", "scene0204_00/00300.jpg", "scene0204_00/00330.jpg" ], "turn_id": 3, "scan_id": "scene0204_00", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "1.56", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0204_00/00360.jpg", "scene0204_00/00390.jpg", "scene0204_00/00420.jpg", "scene0204_00/00450.jpg" ], "turn_id": 4, "scan_id": "scene0204_00", "origin_question": "In which turn did you see both \"the large rectangular light brown wooden table\" and \"the rectangular white plastic switch of standard size\" simultaneously? ", "option": [], "answer": "3", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the large rectangular light brown wooden table\" and \"the rectangular white plastic switch of standard size\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0204_00/00480.jpg", "scene0204_00/00510.jpg", "scene0204_00/00540.jpg", "scene0204_00/00570.jpg" ], "turn_id": 5, "scan_id": "scene0204_00", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "70.57", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0204_00/00600.jpg", "scene0204_00/00630.jpg", "scene0204_00/00660.jpg", "scene0204_00/00690.jpg" ], "turn_id": 6, "scan_id": "scene0204_00", "origin_question": "At the end of which turn were both of \"the large rectangular light brown wooden table\" and \"the rectangular white plastic switch of standard size\" on your right side?", "option": [], "answer": "1", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the large rectangular light brown wooden table\" and \"the rectangular white plastic switch of standard size\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0204_00/00720.jpg", "scene0204_00/00750.jpg", "scene0204_00/00780.jpg", "scene0204_00/00810.jpg" ], "turn_id": 7, "scan_id": "scene0204_00", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the rectangular white plastic switch of standard size\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the rectangular white plastic switch of standard size\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0204_00/00840.jpg", "scene0204_00/00870.jpg", "scene0204_00/00900.jpg", "scene0204_00/00930.jpg" ], "turn_id": 8, "scan_id": "scene0204_00", "origin_question": "Compared to your position at the end of the 7 turn, are you now closer or farther away from \"the rectangular white plastic switch of standard size\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 7 turn, are you now closer or farther away from \"the rectangular white plastic switch of standard size\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0040_region49/3946fdd3e06a4701a588bd2e7b6e30da_i1_2.jpg", "1mp3d_0040_region49/3946fdd3e06a4701a588bd2e7b6e30da_i2_2.jpg", "1mp3d_0040_region49/3946fdd3e06a4701a588bd2e7b6e30da_i2_1.jpg", "1mp3d_0040_region49/5ca3a899d8b0431083b139d73ffdf6c9_i2_0.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0040_region49", "origin_question": "Remember, have you seen any chair(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any chair(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "1mp3d_0040_region49/5ca3a899d8b0431083b139d73ffdf6c9_i1_5.jpg", "1mp3d_0040_region49/5ca3a899d8b0431083b139d73ffdf6c9_i2_5.jpg", "1mp3d_0040_region49/a73871837379498eb16527c6b9c9aa20_i1_2.jpg", "1mp3d_0040_region49/a73871837379498eb16527c6b9c9aa20_i2_2.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0040_region49", "origin_question": "In which turn were you farthest from \"the medium round deep green flowerpot in ceramic\"? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the medium round deep green flowerpot in ceramic\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "1mp3d_0040_region49/a73871837379498eb16527c6b9c9aa20_i1_3.jpg", "1mp3d_0040_region49/a73871837379498eb16527c6b9c9aa20_i2_3.jpg", "1mp3d_0040_region49/a73871837379498eb16527c6b9c9aa20_i1_4.jpg", "1mp3d_0040_region49/a73871837379498eb16527c6b9c9aa20_i2_4.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0040_region49", "origin_question": "At the end of which turn were both of \"the large rectangular wooden table with dark wood grain\" and \"the medium round deep green flowerpot in ceramic\" on your right side?", "option": [], "answer": "2", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the large rectangular wooden table with dark wood grain\" and \"the medium round deep green flowerpot in ceramic\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0040_region49/5ca3a899d8b0431083b139d73ffdf6c9_i1_4.jpg", "1mp3d_0040_region49/5ca3a899d8b0431083b139d73ffdf6c9_i2_4.jpg", "1mp3d_0040_region49/5ca3a899d8b0431083b139d73ffdf6c9_i1_3.jpg", "1mp3d_0040_region49/5ca3a899d8b0431083b139d73ffdf6c9_i2_3.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0040_region49", "origin_question": "Remember, have you seen any plant(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any plant(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "1mp3d_0040_region49/3946fdd3e06a4701a588bd2e7b6e30da_i1_3.jpg", "1mp3d_0040_region49/3946fdd3e06a4701a588bd2e7b6e30da_i2_3.jpg", "1mp3d_0040_region49/b5145233e4ab457ba0c31f9eb20307bf_i1_3.jpg", "1mp3d_0040_region49/b5145233e4ab457ba0c31f9eb20307bf_i2_3.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0040_region49", "origin_question": "Assuming the direction you are facing at the end of the turn 3 is forward, did you move a certain distance forward or backward from that position?", "option": [ "Forward", "Backward" ], "answer": "Backward", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 3 is forward, did you move a certain distance forward or backward from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Forward', 'Backward'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "1mp3d_0030_region3/009755fbde8e43fe90b7d2721398dcec_i1_2.jpg", "1mp3d_0030_region3/009755fbde8e43fe90b7d2721398dcec_i2_2.jpg", "1mp3d_0030_region3/852eef595f4b4977b93b6a9019099932_i1_2.jpg", "1mp3d_0030_region3/852eef595f4b4977b93b6a9019099932_i2_2.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0030_region3", "origin_question": "Remember, how many stand(s) have you seen so far? ", "option": [], "answer": "2", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many stand(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "1mp3d_0030_region3/009755fbde8e43fe90b7d2721398dcec_i1_1.jpg", "1mp3d_0030_region3/009755fbde8e43fe90b7d2721398dcec_i2_1.jpg", "1mp3d_0030_region3/852eef595f4b4977b93b6a9019099932_i1_1.jpg", "1mp3d_0030_region3/852eef595f4b4977b93b6a9019099932_i2_1.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0030_region3", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the white rectangular wardrobe with vertical stripes\"; \"the rectangular, deep brown leather book\"; \"the bed\" ", "option": [ "the white rectangular wardrobe with vertical stripes", "the rectangular, deep brown leather book", "the bed" ], "answer": "the rectangular, deep brown leather book", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the white rectangular wardrobe with vertical stripes\"; \"the rectangular, deep brown leather book\"; \"the bed\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the white rectangular wardrobe with vertical stripes', 'the rectangular, deep brown leather book', 'the bed'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0030_region3/0d93e79881144102ac1ea0bc29b00e37_i1_2.jpg", "1mp3d_0030_region3/0d93e79881144102ac1ea0bc29b00e37_i2_2.jpg", "1mp3d_0030_region3/0d93e79881144102ac1ea0bc29b00e37_i1_1.jpg", "1mp3d_0030_region3/0d93e79881144102ac1ea0bc29b00e37_i2_1.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0030_region3", "origin_question": "", "option": [], "answer": "", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0030_region3/0d93e79881144102ac1ea0bc29b00e37_i1_0.jpg", "1mp3d_0030_region3/0d93e79881144102ac1ea0bc29b00e37_i2_0.jpg", "1mp3d_0030_region3/96facd619dc5415b836948e39487ebb2_i1_4.jpg", "1mp3d_0030_region3/96facd619dc5415b836948e39487ebb2_i2_4.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0030_region3", "origin_question": "", "option": [], "answer": "", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "1mp3d_0030_region3/2041c6faf91240e48969695c474dfbf2_i1_4.jpg", "1mp3d_0030_region3/2041c6faf91240e48969695c474dfbf2_i2_4.jpg", "1mp3d_0030_region3/96facd619dc5415b836948e39487ebb2_i1_3.jpg", "1mp3d_0030_region3/96facd619dc5415b836948e39487ebb2_i2_3.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0030_region3", "origin_question": "Assuming the direction you are facing at the end of the turn 3 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Right", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 3 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "1mp3d_0030_region3/96facd619dc5415b836948e39487ebb2_i1_2.jpg", "1mp3d_0030_region3/96facd619dc5415b836948e39487ebb2_i2_2.jpg", "1mp3d_0030_region3/1416ceed767d46a9beba0aede50504ed_i1_4.jpg", "1mp3d_0030_region3/1416ceed767d46a9beba0aede50504ed_i2_4.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0030_region3", "origin_question": "What will be the first-time appearance order of the following categories: bed, couch, vanity?", "option": [ "couch, bed, vanity", "vanity, couch, bed", "bed, vanity, couch", "couch, vanity, bed" ], "answer": "bed, vanity, couch", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bed, couch, vanity? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['couch, bed, vanity', 'vanity, couch, bed', 'bed, vanity, couch', 'couch, vanity, bed'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "1mp3d_0030_region3/1416ceed767d46a9beba0aede50504ed_i1_5.jpg", "1mp3d_0030_region3/1416ceed767d46a9beba0aede50504ed_i2_5.jpg", "1mp3d_0030_region3/164618f724c64b43934cb95b11c89593_i1_4.jpg", "1mp3d_0030_region3/164618f724c64b43934cb95b11c89593_i2_4.jpg" ], "turn_id": 7, "scan_id": "1mp3d_0030_region3", "origin_question": "How far is your current position from where you were at the end of the turn 5? (in meters)", "option": [], "answer": "1.62", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 5? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0143_00/00000.jpg", "scene0143_00/00030.jpg", "scene0143_00/00060.jpg", "scene0143_00/00090.jpg" ], "turn_id": 1, "scan_id": "scene0143_00", "origin_question": "Remember, how many lamp(s) have you seen so far? ", "option": [], "answer": "2", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many lamp(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 2 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0143_00/00120.jpg", "scene0143_00/00150.jpg", "scene0143_00/00180.jpg", "scene0143_00/00210.jpg" ], "turn_id": 2, "scan_id": "scene0143_00", "origin_question": "When did you first discover \"the rectangular wooden bed with a white bedsheet\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the rectangular wooden bed with a white bedsheet\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 2 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0067_region6/0bd07b7213b245f8a54ec4010f6ef1cc_i1_3.jpg", "1mp3d_0067_region6/0bd07b7213b245f8a54ec4010f6ef1cc_i2_3.jpg", "1mp3d_0067_region6/0bd07b7213b245f8a54ec4010f6ef1cc_i1_2.jpg", "1mp3d_0067_region6/0bd07b7213b245f8a54ec4010f6ef1cc_i2_2.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0067_region6", "origin_question": "Remember, have you seen any adhesive tape(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any adhesive tape(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 2 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "1mp3d_0067_region6/a4aaa9c679ba4a3dbdead1fb0bab0e02_i1_3.jpg", "1mp3d_0067_region6/a4aaa9c679ba4a3dbdead1fb0bab0e02_i2_3.jpg", "1mp3d_0067_region6/a4aaa9c679ba4a3dbdead1fb0bab0e02_i1_1.jpg", "1mp3d_0067_region6/a4aaa9c679ba4a3dbdead1fb0bab0e02_i2_1.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0067_region6", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular stainless steel microwave in shiny silver\", Y: \"the shelf\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "farther away from both X and Y", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular stainless steel microwave in shiny silver\", Y: \"the shelf\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 2 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0360_00/00000.jpg", "scene0360_00/00040.jpg", "scene0360_00/00080.jpg", "scene0360_00/00120.jpg", "scene0360_00/00160.jpg" ], "turn_id": 1, "scan_id": "scene0360_00", "origin_question": "Remember, have you seen any toothbrush(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any toothbrush(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0360_00/00200.jpg", "scene0360_00/00240.jpg", "scene0360_00/00280.jpg", "scene0360_00/00320.jpg", "scene0360_00/00360.jpg" ], "turn_id": 2, "scan_id": "scene0360_00", "origin_question": "Remember, have you seen any picture(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any picture(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0360_00/00400.jpg", "scene0360_00/00440.jpg", "scene0360_00/00480.jpg", "scene0360_00/00520.jpg", "scene0360_00/00560.jpg" ], "turn_id": 3, "scan_id": "scene0360_00", "origin_question": "In which turn were you closest to \"the rectangular, very light-colored sheet of paper\"? ", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the rectangular, very light-colored sheet of paper\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0360_00/00600.jpg", "scene0360_00/00640.jpg", "scene0360_00/00680.jpg", "scene0360_00/00720.jpg", "scene0360_00/00760.jpg" ], "turn_id": 4, "scan_id": "scene0360_00", "origin_question": "Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0360_00/00800.jpg", "scene0360_00/00850.jpg", "scene0360_00/00890.jpg", "scene0360_00/00930.jpg", "scene0360_00/00970.jpg" ], "turn_id": 5, "scan_id": "scene0360_00", "origin_question": "What will be the first-time appearance order of the following categories: bin, picture, socket?", "option": [ "picture, socket, bin", "bin, picture, socket", "picture, bin, socket", "bin, socket, picture" ], "answer": "bin, socket, picture", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bin, picture, socket? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['picture, socket, bin', 'bin, picture, socket', 'picture, bin, socket', 'bin, socket, picture'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0360_00/01010.jpg", "scene0360_00/01050.jpg", "scene0360_00/01090.jpg", "scene0360_00/01130.jpg", "scene0360_00/01170.jpg" ], "turn_id": 6, "scan_id": "scene0360_00", "origin_question": "When did you first discover \"the slender dark metal ladder\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the slender dark metal ladder\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0360_00/01210.jpg", "scene0360_00/01250.jpg", "scene0360_00/01290.jpg", "scene0360_00/01330.jpg", "scene0360_00/01370.jpg" ], "turn_id": 7, "scan_id": "scene0360_00", "origin_question": "Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "91.99", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0360_00/01410.jpg", "scene0360_00/01450.jpg", "scene0360_00/01490.jpg", "scene0360_00/01530.jpg", "scene0360_00/01570.jpg" ], "turn_id": 8, "scan_id": "scene0360_00", "origin_question": "Using your orientation at the end of turn 5 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 5 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0360_00/01610.jpg", "scene0360_00/01650.jpg", "scene0360_00/01690.jpg", "scene0360_00/01730.jpg", "scene0360_00/01770.jpg" ], "turn_id": 9, "scan_id": "scene0360_00", "origin_question": "What will be the first-time appearance order of the following categories: bin, door, picture?", "option": [ "picture, bin, door", "picture, door, bin", "bin, picture, door", "door, picture, bin" ], "answer": "bin, picture, door", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bin, door, picture? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['picture, bin, door', 'picture, door, bin', 'bin, picture, door', 'door, picture, bin'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0360_00/01810.jpg", "scene0360_00/01850.jpg", "scene0360_00/01890.jpg", "scene0360_00/01930.jpg", "scene0360_00/01970.jpg" ], "turn_id": 10, "scan_id": "scene0360_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0094_00/00000.jpg", "scene0094_00/00040.jpg", "scene0094_00/00080.jpg", "scene0094_00/00120.jpg" ], "turn_id": 1, "scan_id": "scene0094_00", "origin_question": "Remember, have you seen any sign(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any sign(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0094_00/00160.jpg", "scene0094_00/00200.jpg", "scene0094_00/00240.jpg", "scene0094_00/00280.jpg" ], "turn_id": 2, "scan_id": "scene0094_00", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the light brown rectangular wooden shelf\"; \"the small rectangular white plastic switch\"; \"the white ceramic single basin sink with a silver faucet\" ", "option": [ "the light brown rectangular wooden shelf", "the small rectangular white plastic switch", "the white ceramic single basin sink with a silver faucet" ], "answer": "the light brown rectangular wooden shelf", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the light brown rectangular wooden shelf\"; \"the small rectangular white plastic switch\"; \"the white ceramic single basin sink with a silver faucet\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the light brown rectangular wooden shelf', 'the small rectangular white plastic switch', 'the white ceramic single basin sink with a silver faucet'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "scene0094_00/00320.jpg", "scene0094_00/00360.jpg", "scene0094_00/00400.jpg", "scene0094_00/00440.jpg" ], "turn_id": 3, "scan_id": "scene0094_00", "origin_question": "Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Right", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0094_00/00480.jpg", "scene0094_00/00520.jpg", "scene0094_00/00560.jpg", "scene0094_00/00600.jpg" ], "turn_id": 4, "scan_id": "scene0094_00", "origin_question": "Which two objects are on the same side of you now? \"the small rectangular white plastic switch\", \"the white ceramic single basin sink with a silver faucet\", \"the silver metallic teapot with a black handle\". ", "option": [ "the small rectangular white plastic switch, the white ceramic single basin sink with a silver faucet", "the small rectangular white plastic switch, the silver metallic teapot with a black handle", "the white ceramic single basin sink with a silver faucet, the silver metallic teapot with a black handle" ], "answer": "the small rectangular white plastic switch, the white ceramic single basin sink with a silver faucet", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the small rectangular white plastic switch\", \"the white ceramic single basin sink with a silver faucet\", \"the silver metallic teapot with a black handle\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the small rectangular white plastic switch, the white ceramic single basin sink with a silver faucet', 'the small rectangular white plastic switch, the silver metallic teapot with a black handle', 'the white ceramic single basin sink with a silver faucet, the silver metallic teapot with a black handle'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0094_00/00640.jpg", "scene0094_00/00680.jpg", "scene0094_00/00720.jpg", "scene0094_00/00760.jpg" ], "turn_id": 5, "scan_id": "scene0094_00", "origin_question": "In which turn were you closest to \"the light brown rectangular wooden shelf\"? ", "option": [], "answer": "3", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the light brown rectangular wooden shelf\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0094_00/00800.jpg", "scene0094_00/00840.jpg", "scene0094_00/00880.jpg", "scene0094_00/00920.jpg" ], "turn_id": 6, "scan_id": "scene0094_00", "origin_question": "Which direction is the rectangular white framed glass window in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the rectangular white framed glass window in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "1mp3d_0044_region2/224642f8a0e0483eb94e594aa481c473_i1_1.jpg", "1mp3d_0044_region2/224642f8a0e0483eb94e594aa481c473_i2_1.jpg", "1mp3d_0044_region2/224642f8a0e0483eb94e594aa481c473_i1_2.jpg", "1mp3d_0044_region2/224642f8a0e0483eb94e594aa481c473_i2_2.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0044_region2", "origin_question": "Remember, how many switch(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many switch(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "1mp3d_0044_region2/3fa94e726f0c425aaba652db06f6f20b_i1_0.jpg", "1mp3d_0044_region2/3fa94e726f0c425aaba652db06f6f20b_i2_0.jpg", "1mp3d_0044_region2/2a8ee08b2da64479a42fdfe5c7562079_i1_1.jpg", "1mp3d_0044_region2/2a8ee08b2da64479a42fdfe5c7562079_i2_1.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0044_region2", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular multi-tiered marble countertop with swirling patterns\", Y: \"the small square white tissue box made of smooth material\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular multi-tiered marble countertop with swirling patterns\", Y: \"the small square white tissue box made of smooth material\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "1mp3d_0044_region2/01790b2fed1640a1bbb42f04b99f298a_i1_1.jpg", "1mp3d_0044_region2/01790b2fed1640a1bbb42f04b99f298a_i2_1.jpg", "1mp3d_0044_region2/01790b2fed1640a1bbb42f04b99f298a_i1_2.jpg", "1mp3d_0044_region2/01790b2fed1640a1bbb42f04b99f298a_i2_2.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0044_region2", "origin_question": "In which turn were you farthest from \"the small square white tissue box made of smooth material\"? ", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the small square white tissue box made of smooth material\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "1mp3d_0044_region2/2a8ee08b2da64479a42fdfe5c7562079_i1_2.jpg", "1mp3d_0044_region2/2a8ee08b2da64479a42fdfe5c7562079_i2_2.jpg", "1mp3d_0044_region2/5162c0936c8340239ef9d52e9d9ffc67_i1_5.jpg", "1mp3d_0044_region2/5162c0936c8340239ef9d52e9d9ffc67_i2_5.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0044_region2", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular multi-tiered marble countertop with swirling patterns\", Y: \"the long, dark patterned fabric curtain\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to X and farther away from Y", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular multi-tiered marble countertop with swirling patterns\", Y: \"the long, dark patterned fabric curtain\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "1mp3d_0044_region2/65672d1fd8784dda8f54ea18dab04d0e_i1_5.jpg", "1mp3d_0044_region2/65672d1fd8784dda8f54ea18dab04d0e_i2_5.jpg", "1mp3d_0044_region2/5162c0936c8340239ef9d52e9d9ffc67_i1_3.jpg", "1mp3d_0044_region2/5162c0936c8340239ef9d52e9d9ffc67_i2_3.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0044_region2", "origin_question": "Is the long, dark patterned fabric curtain to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the long, dark patterned fabric curtain to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "1mp3d_0044_region2/65672d1fd8784dda8f54ea18dab04d0e_i1_3.jpg", "1mp3d_0044_region2/65672d1fd8784dda8f54ea18dab04d0e_i2_3.jpg", "1mp3d_0044_region2/65672d1fd8784dda8f54ea18dab04d0e_i1_4.jpg", "1mp3d_0044_region2/65672d1fd8784dda8f54ea18dab04d0e_i2_4.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0044_region2", "origin_question": "In which turn did you see both \"the large white oval bathtub with a glossy finish\" and \"the microwave\" simultaneously? ", "option": [], "answer": "5", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the large white oval bathtub with a glossy finish\" and \"the microwave\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "1mp3d_0044_region2/e3a05e2cde9c4413ad372c30b1c50e66_i1_3.jpg", "1mp3d_0044_region2/e3a05e2cde9c4413ad372c30b1c50e66_i2_3.jpg", "1mp3d_0044_region2/b5c89126d17d47fea581e6db97107874_i1_2.jpg", "1mp3d_0044_region2/b5c89126d17d47fea581e6db97107874_i2_2.jpg" ], "turn_id": 7, "scan_id": "1mp3d_0044_region2", "origin_question": "Using your orientation at the end of turn 4 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "37.52", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 4 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "1mp3d_0044_region2/b5c89126d17d47fea581e6db97107874_i1_3.jpg", "1mp3d_0044_region2/b5c89126d17d47fea581e6db97107874_i2_3.jpg", "1mp3d_0044_region2/b5c89126d17d47fea581e6db97107874_i1_4.jpg", "1mp3d_0044_region2/b5c89126d17d47fea581e6db97107874_i2_4.jpg" ], "turn_id": 8, "scan_id": "1mp3d_0044_region2", "origin_question": "Among these three objects(\"the light brown cushioned chair with a curved back\",\"the small square white tissue box made of smooth material\",\"the microwave\" ), which one is the closest to you?", "option": [ "the light brown cushioned chair with a curved back", "the small square white tissue box made of smooth material", "the microwave" ], "answer": "the microwave", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects(\"the light brown cushioned chair with a curved back\",\"the small square white tissue box made of smooth material\",\"the microwave\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the light brown cushioned chair with a curved back', 'the small square white tissue box made of smooth material', 'the microwave'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "1mp3d_0044_region2/e3a05e2cde9c4413ad372c30b1c50e66_i1_4.jpg", "1mp3d_0044_region2/e3a05e2cde9c4413ad372c30b1c50e66_i2_4.jpg", "1mp3d_0044_region2/e3a05e2cde9c4413ad372c30b1c50e66_i1_5.jpg", "1mp3d_0044_region2/e3a05e2cde9c4413ad372c30b1c50e66_i2_5.jpg" ], "turn_id": 9, "scan_id": "1mp3d_0044_region2", "origin_question": "Please recall \"the rectangular multi-tiered marble countertop with swirling patterns\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.7", "user_message": "For the 9 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the rectangular multi-tiered marble countertop with swirling patterns\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0040_region11/9b5031bcbeba4456bcf32cbe3206f24c_i1_0.jpg", "1mp3d_0040_region11/9b5031bcbeba4456bcf32cbe3206f24c_i2_0.jpg", "1mp3d_0040_region11/d9349879995848549eda95af961040f7_i1_3.jpg", "1mp3d_0040_region11/d9349879995848549eda95af961040f7_i2_3.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0040_region11", "origin_question": "Remember, have you seen any window(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any window(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 2 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "1mp3d_0040_region11/9b5031bcbeba4456bcf32cbe3206f24c_i1_5.jpg", "1mp3d_0040_region11/9b5031bcbeba4456bcf32cbe3206f24c_i2_5.jpg", "1mp3d_0040_region11/9b5031bcbeba4456bcf32cbe3206f24c_i1_4.jpg", "1mp3d_0040_region11/9b5031bcbeba4456bcf32cbe3206f24c_i2_4.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0040_region11", "origin_question": "At the end of which turn were both of \"the lightweight striped curtain in white and blue\" and \"the blue cylindrical plastic bottle with a pump\" on your right side?", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the lightweight striped curtain in white and blue\" and \"the blue cylindrical plastic bottle with a pump\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 2 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0703_00/00000.jpg", "scene0703_00/00030.jpg", "scene0703_00/00060.jpg", "scene0703_00/00090.jpg" ], "turn_id": 1, "scan_id": "scene0703_00", "origin_question": "Remember, how many door(s) have you seen so far? ", "option": [], "answer": "2", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many door(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0703_00/00120.jpg", "scene0703_00/00150.jpg", "scene0703_00/00180.jpg", "scene0703_00/00210.jpg" ], "turn_id": 2, "scan_id": "scene0703_00", "origin_question": "When did you first discover \"the small rectangular light beige plastic switch\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the small rectangular light beige plastic switch\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0703_00/00240.jpg", "scene0703_00/00270.jpg", "scene0703_00/00300.jpg", "scene0703_00/00330.jpg" ], "turn_id": 3, "scan_id": "scene0703_00", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "47.77", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0703_00/00360.jpg", "scene0703_00/00390.jpg", "scene0703_00/00420.jpg", "scene0703_00/00450.jpg" ], "turn_id": 4, "scan_id": "scene0703_00", "origin_question": "Which direction is the small rectangular light beige plastic switch in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the small rectangular light beige plastic switch in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0703_00/00480.jpg", "scene0703_00/00510.jpg", "scene0703_00/00540.jpg", "scene0703_00/00570.jpg" ], "turn_id": 5, "scan_id": "scene0703_00", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0044_region39/ae7db65ba9064f66b142bba2ec5fc5d5_i1_0.jpg", "1mp3d_0044_region39/ae7db65ba9064f66b142bba2ec5fc5d5_i2_0.jpg", "1mp3d_0044_region39/ae7db65ba9064f66b142bba2ec5fc5d5_i1_1.jpg", "1mp3d_0044_region39/ae7db65ba9064f66b142bba2ec5fc5d5_i2_1.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0044_region39", "origin_question": "Remember, have you seen any toilet(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any toilet(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "1mp3d_0044_region39/ae7db65ba9064f66b142bba2ec5fc5d5_i1_2.jpg", "1mp3d_0044_region39/ae7db65ba9064f66b142bba2ec5fc5d5_i2_2.jpg", "1mp3d_0044_region39/128ef9515288432b8d4d0d259fe4060d_i1_2.jpg", "1mp3d_0044_region39/128ef9515288432b8d4d0d259fe4060d_i2_2.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0044_region39", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the white ceramic toilet with a closed lid\" from your position?", "option": [], "answer": "85.18", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the white ceramic toilet with a closed lid\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "1mp3d_0044_region39/8e7d323375344d50974984333252a3a1_i2_3.jpg", "1mp3d_0044_region39/8e7d323375344d50974984333252a3a1_i1_4.jpg", "1mp3d_0044_region39/8e7d323375344d50974984333252a3a1_i2_4.jpg", "1mp3d_0044_region39/128ef9515288432b8d4d0d259fe4060d_i1_3.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0044_region39", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "1mp3d_0044_region39/128ef9515288432b8d4d0d259fe4060d_i2_3.jpg", "1mp3d_0044_region39/8e7d323375344d50974984333252a3a1_i1_1.jpg", "1mp3d_0044_region39/8e7d323375344d50974984333252a3a1_i2_1.jpg", "1mp3d_0044_region39/8e7d323375344d50974984333252a3a1_i1_0.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0044_region39", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "1mp3d_0044_region39/8e7d323375344d50974984333252a3a1_i2_0.jpg", "1mp3d_0044_region39/8e7d323375344d50974984333252a3a1_i1_2.jpg", "1mp3d_0044_region39/8e7d323375344d50974984333252a3a1_i2_2.jpg", "1mp3d_0044_region39/128ef9515288432b8d4d0d259fe4060d_i1_1.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0044_region39", "origin_question": "Please recall \"the roll of white toilet paper made from standard paper\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.57", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the roll of white toilet paper made from standard paper\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0169_01/00000.jpg", "scene0169_01/00040.jpg", "scene0169_01/00080.jpg", "scene0169_01/00120.jpg" ], "turn_id": 1, "scan_id": "scene0169_01", "origin_question": "Remember, have you seen any chair(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any chair(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0169_01/00160.jpg", "scene0169_01/00200.jpg", "scene0169_01/00240.jpg", "scene0169_01/00280.jpg" ], "turn_id": 2, "scan_id": "scene0169_01", "origin_question": "At the end of which turn were both of \"the long rectangular light brown wooden table\" and \"the door\" on your right side?", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the long rectangular light brown wooden table\" and \"the door\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0169_01/00320.jpg", "scene0169_01/00360.jpg", "scene0169_01/00400.jpg", "scene0169_01/00440.jpg" ], "turn_id": 3, "scan_id": "scene0169_01", "origin_question": "Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "153.96", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "scene0169_01/00480.jpg", "scene0169_01/00520.jpg", "scene0169_01/00560.jpg", "scene0169_01/00600.jpg" ], "turn_id": 4, "scan_id": "scene0169_01", "origin_question": "Assuming the direction you are facing at the end of the turn 3 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Right", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 3 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0169_01/00640.jpg", "scene0169_01/00680.jpg", "scene0169_01/00720.jpg", "scene0169_01/00760.jpg" ], "turn_id": 5, "scan_id": "scene0169_01", "origin_question": "How far is your current position from where you were at the end of the turn 4? (in meters)", "option": [], "answer": "1.7", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 4? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0169_01/00800.jpg", "scene0169_01/00840.jpg", "scene0169_01/00880.jpg", "scene0169_01/00920.jpg" ], "turn_id": 6, "scan_id": "scene0169_01", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the door\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the door\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0169_01/00960.jpg", "scene0169_01/01000.jpg", "scene0169_01/01040.jpg", "scene0169_01/01080.jpg" ], "turn_id": 7, "scan_id": "scene0169_01", "origin_question": "By the end of the 4 turn, were you getting closer to or farther away from these two objects: X: \"the bin\", Y: \"the door\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 4 turn, were you getting closer to or farther away from these two objects: X: \"the bin\", Y: \"the door\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0169_01/01120.jpg", "scene0169_01/01160.jpg", "scene0169_01/01200.jpg", "scene0169_01/01240.jpg" ], "turn_id": 8, "scan_id": "scene0169_01", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "141.46", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0169_01/01280.jpg", "scene0169_01/01320.jpg", "scene0169_01/01360.jpg", "scene0169_01/01400.jpg" ], "turn_id": 9, "scan_id": "scene0169_01", "origin_question": "Please recall \"the bin\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "4.99", "user_message": "For the 9 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the bin\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0585_00/00000.jpg", "scene0585_00/00040.jpg", "scene0585_00/00080.jpg", "scene0585_00/00120.jpg", "scene0585_00/00160.jpg" ], "turn_id": 1, "scan_id": "scene0585_00", "origin_question": "Remember, have you seen any table(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any table(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0585_00/00200.jpg", "scene0585_00/00240.jpg", "scene0585_00/00280.jpg", "scene0585_00/00320.jpg", "scene0585_00/00360.jpg" ], "turn_id": 2, "scan_id": "scene0585_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0585_00/00400.jpg", "scene0585_00/00440.jpg", "scene0585_00/00480.jpg", "scene0585_00/00520.jpg", "scene0585_00/00560.jpg" ], "turn_id": 3, "scan_id": "scene0585_00", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the vertical black fabric rack with compartments\" from your position?", "option": [], "answer": "82.4", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the vertical black fabric rack with compartments\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0585_00/00600.jpg", "scene0585_00/00640.jpg", "scene0585_00/00680.jpg", "scene0585_00/00720.jpg", "scene0585_00/00760.jpg" ], "turn_id": 4, "scan_id": "scene0585_00", "origin_question": "In which turn did you see both \"the white rectangular sheet of paper\" and \"the rectangular light gray metal cabinet\" simultaneously? ", "option": [], "answer": "4", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the white rectangular sheet of paper\" and \"the rectangular light gray metal cabinet\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0585_00/00800.jpg", "scene0585_00/00840.jpg", "scene0585_00/00880.jpg", "scene0585_00/00920.jpg", "scene0585_00/00960.jpg" ], "turn_id": 5, "scan_id": "scene0585_00", "origin_question": "Which direction is the cylindrical beige plastic stool with a textured surface in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the cylindrical beige plastic stool with a textured surface in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0585_00/01000.jpg", "scene0585_00/01040.jpg", "scene0585_00/01080.jpg", "scene0585_00/01120.jpg", "scene0585_00/01160.jpg" ], "turn_id": 6, "scan_id": "scene0585_00", "origin_question": "In which turn did you see both \"the cylindrical beige plastic stool with a textured surface\" and \"the white rectangular sheet of paper\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the cylindrical beige plastic stool with a textured surface\" and \"the white rectangular sheet of paper\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0585_00/01200.jpg", "scene0585_00/01240.jpg", "scene0585_00/01280.jpg", "scene0585_00/01320.jpg", "scene0585_00/01360.jpg" ], "turn_id": 7, "scan_id": "scene0585_00", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the cylindrical beige plastic stool with a textured surface\" from your position?", "option": [], "answer": "178.64", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the cylindrical beige plastic stool with a textured surface\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0585_00/01400.jpg", "scene0585_00/01440.jpg", "scene0585_00/01480.jpg", "scene0585_00/01520.jpg", "scene0585_00/01560.jpg" ], "turn_id": 8, "scan_id": "scene0585_00", "origin_question": "What will be the first-time appearance order of the following categories: cart, paper cutter, stool?", "option": [ "paper cutter, cart, stool", "stool, cart, paper cutter", "stool, paper cutter, cart", "cart, stool, paper cutter" ], "answer": "stool, paper cutter, cart", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: cart, paper cutter, stool? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['paper cutter, cart, stool', 'stool, cart, paper cutter', 'stool, paper cutter, cart', 'cart, stool, paper cutter'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0585_00/01600.jpg", "scene0585_00/01640.jpg", "scene0585_00/01680.jpg", "scene0585_00/01720.jpg", "scene0585_00/01760.jpg" ], "turn_id": 9, "scan_id": "scene0585_00", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the cylindrical beige plastic stool with a textured surface\", Y: \"the rectangular light gray metal cabinet\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to Y and farther away from X", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the cylindrical beige plastic stool with a textured surface\", Y: \"the rectangular light gray metal cabinet\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "scene0585_00/01800.jpg", "scene0585_00/01840.jpg", "scene0585_00/01880.jpg", "scene0585_00/01920.jpg", "scene0585_00/01960.jpg" ], "turn_id": 10, "scan_id": "scene0585_00", "origin_question": "Among these three objects(\"the vertical metal mailbox with colorful compartments\",\"the vertical black fabric rack with compartments\",\"the sturdy rectangular paper cutter with a wooden base\" ), which one is the farthest from you?", "option": [ "the vertical metal mailbox with colorful compartments", "the vertical black fabric rack with compartments", "the sturdy rectangular paper cutter with a wooden base" ], "answer": "the sturdy rectangular paper cutter with a wooden base", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the vertical metal mailbox with colorful compartments\",\"the vertical black fabric rack with compartments\",\"the sturdy rectangular paper cutter with a wooden base\" ), which one is the farthest from you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the vertical metal mailbox with colorful compartments', 'the vertical black fabric rack with compartments', 'the sturdy rectangular paper cutter with a wooden base'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0004_region11/364f306248174f9cbeb6f3d612f50060_i1_4.jpg", "1mp3d_0004_region11/364f306248174f9cbeb6f3d612f50060_i2_4.jpg", "1mp3d_0004_region11/683b92657e2c4900b6e760760aac7057_i1_5.jpg", "1mp3d_0004_region11/683b92657e2c4900b6e760760aac7057_i2_5.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0004_region11", "origin_question": "Remember, have you seen any bench(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any bench(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0004_region11/364f306248174f9cbeb6f3d612f50060_i1_3.jpg", "1mp3d_0004_region11/364f306248174f9cbeb6f3d612f50060_i2_3.jpg", "1mp3d_0004_region11/683b92657e2c4900b6e760760aac7057_i1_4.jpg", "1mp3d_0004_region11/683b92657e2c4900b6e760760aac7057_i2_4.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0004_region11", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "1mp3d_0004_region11/364f306248174f9cbeb6f3d612f50060_i1_2.jpg", "1mp3d_0004_region11/364f306248174f9cbeb6f3d612f50060_i2_2.jpg", "1mp3d_0004_region11/13e891ec449b4cfcbee4c1302b55336a_i1_0.jpg", "1mp3d_0004_region11/13e891ec449b4cfcbee4c1302b55336a_i2_0.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0004_region11", "origin_question": "By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the cabinet\", Y: \"the window\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the cabinet\", Y: \"the window\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0004_region11/148ca4f631fe4c06af243a86e008fb67_i1_5.jpg", "1mp3d_0004_region11/148ca4f631fe4c06af243a86e008fb67_i2_5.jpg", "1mp3d_0004_region11/13e891ec449b4cfcbee4c1302b55336a_i1_2.jpg", "1mp3d_0004_region11/13e891ec449b4cfcbee4c1302b55336a_i2_2.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0004_region11", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the cabinet\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the cabinet\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "1mp3d_0004_region11/148ca4f631fe4c06af243a86e008fb67_i2_1.jpg", "1mp3d_0004_region11/13e891ec449b4cfcbee4c1302b55336a_i1_1.jpg", "1mp3d_0004_region11/13e891ec449b4cfcbee4c1302b55336a_i2_1.jpg", "1mp3d_0004_region11/148ca4f631fe4c06af243a86e008fb67_i1_0.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0004_region11", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "67.63", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0498_02/00000.jpg", "scene0498_02/00030.jpg", "scene0498_02/00060.jpg", "scene0498_02/00090.jpg" ], "turn_id": 1, "scan_id": "scene0498_02", "origin_question": "Remember, have you seen any light(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any light(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0498_02/00120.jpg", "scene0498_02/00150.jpg", "scene0498_02/00180.jpg", "scene0498_02/00210.jpg" ], "turn_id": 2, "scan_id": "scene0498_02", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the rectangular stainless steel microwave in silver color\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the rectangular stainless steel microwave in silver color\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0498_02/00240.jpg", "scene0498_02/00270.jpg", "scene0498_02/00300.jpg", "scene0498_02/00330.jpg" ], "turn_id": 3, "scan_id": "scene0498_02", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the rectangular silver stainless steel bin\"; \"the door\"; \"the round, light-colored ceramic plate\" ", "option": [ "the rectangular silver stainless steel bin", "the door", "the round, light-colored ceramic plate" ], "answer": "the rectangular silver stainless steel bin", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the rectangular silver stainless steel bin\"; \"the door\"; \"the round, light-colored ceramic plate\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular silver stainless steel bin', 'the door', 'the round, light-colored ceramic plate'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0498_02/00360.jpg", "scene0498_02/00390.jpg", "scene0498_02/00420.jpg", "scene0498_02/00450.jpg" ], "turn_id": 4, "scan_id": "scene0498_02", "origin_question": "At the end of which turn were both of \"the rectangular brown wooden table of medium size\" and \"the window\" on your right side?", "option": [], "answer": "1", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the rectangular brown wooden table of medium size\" and \"the window\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0498_02/00480.jpg", "scene0498_02/00510.jpg", "scene0498_02/00540.jpg", "scene0498_02/00570.jpg" ], "turn_id": 5, "scan_id": "scene0498_02", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the rectangular stainless steel sink\" from your position?", "option": [], "answer": "58.01", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the rectangular stainless steel sink\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0498_02/00600.jpg", "scene0498_02/00630.jpg", "scene0498_02/00660.jpg", "scene0498_02/00690.jpg" ], "turn_id": 6, "scan_id": "scene0498_02", "origin_question": "Please recall \"the rectangular brown wooden table of medium size\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.46", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the rectangular brown wooden table of medium size\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0498_02/00720.jpg", "scene0498_02/00750.jpg", "scene0498_02/00780.jpg", "scene0498_02/00810.jpg" ], "turn_id": 7, "scan_id": "scene0498_02", "origin_question": "What will be the first-time appearance order of the following categories: bin, bottle, table?", "option": [ "bin, bottle, table", "table, bin, bottle", "bin, table, bottle", "table, bottle, bin" ], "answer": "table, bin, bottle", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bin, bottle, table? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['bin, bottle, table', 'table, bin, bottle', 'bin, table, bottle', 'table, bottle, bin'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0040_region27/bfe100fb6ef24661831ea5d65e4e62ce_i1_4.jpg", "1mp3d_0040_region27/bfe100fb6ef24661831ea5d65e4e62ce_i2_4.jpg", "1mp3d_0040_region27/a1d210d6a26b45528017394d195a7019_i1_4.jpg", "1mp3d_0040_region27/a1d210d6a26b45528017394d195a7019_i2_4.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0040_region27", "origin_question": "", "option": [], "answer": "", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 3 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0040_region27/fd7bc314137242a498ccec4264834f8a_i1_0.jpg", "1mp3d_0040_region27/fd7bc314137242a498ccec4264834f8a_i2_0.jpg", "1mp3d_0040_region27/9186192bc71b4e868af5b35244643ca4_i1_5.jpg", "1mp3d_0040_region27/9186192bc71b4e868af5b35244643ca4_i2_5.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0040_region27", "origin_question": "Remember, have you seen any thermostat(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any thermostat(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 3 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "1mp3d_0040_region27/9186192bc71b4e868af5b35244643ca4_i1_4.jpg", "1mp3d_0040_region27/9186192bc71b4e868af5b35244643ca4_i2_4.jpg", "1mp3d_0040_region27/9186192bc71b4e868af5b35244643ca4_i1_3.jpg", "1mp3d_0040_region27/9186192bc71b4e868af5b35244643ca4_i2_3.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0040_region27", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "119.33", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 3 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "1mp3d_0057_region20/a9bdea807d534025a937b7c715bc974c_i1_4.jpg", "1mp3d_0057_region20/a9bdea807d534025a937b7c715bc974c_i2_4.jpg", "1mp3d_0057_region20/f5a31e39d8fa40e89bbbc110bf25de2c_i1_3.jpg", "1mp3d_0057_region20/f5a31e39d8fa40e89bbbc110bf25de2c_i2_3.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0057_region20", "origin_question": "Remember, how many letter(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many letter(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "1mp3d_0057_region20/f5a31e39d8fa40e89bbbc110bf25de2c_i1_4.jpg", "1mp3d_0057_region20/f5a31e39d8fa40e89bbbc110bf25de2c_i2_4.jpg", "1mp3d_0057_region20/a9bdea807d534025a937b7c715bc974c_i1_3.jpg", "1mp3d_0057_region20/a9bdea807d534025a937b7c715bc974c_i2_3.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0057_region20", "origin_question": "Which two objects are on the same side of you now? \"the sleek black flat screen television set\", \"the light-colored wooden cabinet with three drawers\", \"the rectangular wooden bed with a natural wood headboard\". ", "option": [ "the sleek black flat screen television set, the light-colored wooden cabinet with three drawers", "the sleek black flat screen television set, the rectangular wooden bed with a natural wood headboard", "the light-colored wooden cabinet with three drawers, the rectangular wooden bed with a natural wood headboard" ], "answer": "the sleek black flat screen television set, the light-colored wooden cabinet with three drawers", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the sleek black flat screen television set\", \"the light-colored wooden cabinet with three drawers\", \"the rectangular wooden bed with a natural wood headboard\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the sleek black flat screen television set, the light-colored wooden cabinet with three drawers', 'the sleek black flat screen television set, the rectangular wooden bed with a natural wood headboard', 'the light-colored wooden cabinet with three drawers, the rectangular wooden bed with a natural wood headboard'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "1mp3d_0057_region20/a9bdea807d534025a937b7c715bc974c_i1_2.jpg", "1mp3d_0057_region20/a9bdea807d534025a937b7c715bc974c_i2_2.jpg", "1mp3d_0057_region20/64685718397043c08a14614bb91e36bc_i1_3.jpg", "1mp3d_0057_region20/64685718397043c08a14614bb91e36bc_i2_3.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0057_region20", "origin_question": "When did you first discover \"the sleek black flat screen television set\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the sleek black flat screen television set\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "1mp3d_0057_region20/64685718397043c08a14614bb91e36bc_i1_2.jpg", "1mp3d_0057_region20/64685718397043c08a14614bb91e36bc_i2_2.jpg", "1mp3d_0057_region20/0ce7a1fe811a44d98744e968d9b44dc4_i1_2.jpg", "1mp3d_0057_region20/0ce7a1fe811a44d98744e968d9b44dc4_i2_2.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0057_region20", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "1mp3d_0057_region20/2dc819c5a5864cc8a953a8af8030acf4_i2_3.jpg", "1mp3d_0057_region20/2dc819c5a5864cc8a953a8af8030acf4_i1_4.jpg", "1mp3d_0057_region20/2dc819c5a5864cc8a953a8af8030acf4_i2_4.jpg", "1mp3d_0057_region20/2dc819c5a5864cc8a953a8af8030acf4_i1_5.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0057_region20", "origin_question": "At the end of which turn were both of \"the sleek black flat screen television set\" and \"the rectangular wooden bed with a natural wood headboard\" on your left side?", "option": [], "answer": "4", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the sleek black flat screen television set\" and \"the rectangular wooden bed with a natural wood headboard\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "1mp3d_0057_region20/2dc819c5a5864cc8a953a8af8030acf4_i2_5.jpg", "1mp3d_0057_region20/2dc819c5a5864cc8a953a8af8030acf4_i1_0.jpg", "1mp3d_0057_region20/2dc819c5a5864cc8a953a8af8030acf4_i2_0.jpg", "1mp3d_0057_region20/0ce7a1fe811a44d98744e968d9b44dc4_i1_0.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0057_region20", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "144.47", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "1mp3d_0057_region20/0ce7a1fe811a44d98744e968d9b44dc4_i2_0.jpg", "1mp3d_0057_region20/0ce7a1fe811a44d98744e968d9b44dc4_i1_5.jpg", "1mp3d_0057_region20/0ce7a1fe811a44d98744e968d9b44dc4_i2_5.jpg", "1mp3d_0057_region20/c84fffdd336f4820a82c2063c2eed7b9_i2_5.jpg" ], "turn_id": 7, "scan_id": "1mp3d_0057_region20", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the sleek black flat screen television set\" from your position?", "option": [], "answer": "141.26", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the sleek black flat screen television set\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0019_01/00000.jpg", "scene0019_01/00030.jpg", "scene0019_01/00060.jpg", "scene0019_01/00090.jpg" ], "turn_id": 1, "scan_id": "scene0019_01", "origin_question": "Remember, how many sack(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many sack(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0019_01/00120.jpg", "scene0019_01/00150.jpg", "scene0019_01/00180.jpg", "scene0019_01/00210.jpg" ], "turn_id": 2, "scan_id": "scene0019_01", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the medium-sized white cylindrical bin\"; \"the black tall rectangular machine\"; \"the rectangular white lampshade on a brown metal lamp\" ", "option": [ "the medium-sized white cylindrical bin", "the black tall rectangular machine", "the rectangular white lampshade on a brown metal lamp" ], "answer": "the medium-sized white cylindrical bin", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the medium-sized white cylindrical bin\"; \"the black tall rectangular machine\"; \"the rectangular white lampshade on a brown metal lamp\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the medium-sized white cylindrical bin', 'the black tall rectangular machine', 'the rectangular white lampshade on a brown metal lamp'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0019_01/00240.jpg", "scene0019_01/00270.jpg", "scene0019_01/00300.jpg", "scene0019_01/00330.jpg" ], "turn_id": 3, "scan_id": "scene0019_01", "origin_question": "Please recall \"the rectangular white lampshade on a brown metal lamp\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.93", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the rectangular white lampshade on a brown metal lamp\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0019_01/00360.jpg", "scene0019_01/00390.jpg", "scene0019_01/00420.jpg", "scene0019_01/00450.jpg" ], "turn_id": 4, "scan_id": "scene0019_01", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the rectangular white lampshade on a brown metal lamp\" from your position?", "option": [], "answer": "143.62", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the rectangular white lampshade on a brown metal lamp\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0019_01/00480.jpg", "scene0019_01/00510.jpg", "scene0019_01/00540.jpg", "scene0019_01/00570.jpg" ], "turn_id": 5, "scan_id": "scene0019_01", "origin_question": "Which direction is the medium-sized white cylindrical bin in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the medium-sized white cylindrical bin in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0231_02/00000.jpg", "scene0231_02/00040.jpg", "scene0231_02/00080.jpg", "scene0231_02/00120.jpg", "scene0231_02/00160.jpg" ], "turn_id": 1, "scan_id": "scene0231_02", "origin_question": "Remember, have you seen any headboard(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any headboard(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0231_02/00200.jpg", "scene0231_02/00240.jpg", "scene0231_02/00280.jpg", "scene0231_02/00320.jpg", "scene0231_02/00360.jpg" ], "turn_id": 2, "scan_id": "scene0231_02", "origin_question": "Which direction is the backpack in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the backpack in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0231_02/00400.jpg", "scene0231_02/00440.jpg", "scene0231_02/00480.jpg", "scene0231_02/00520.jpg", "scene0231_02/00560.jpg" ], "turn_id": 3, "scan_id": "scene0231_02", "origin_question": "Which direction is the rectangular pink or light brown pillow in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the rectangular pink or light brown pillow in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0231_02/00600.jpg", "scene0231_02/00640.jpg", "scene0231_02/00680.jpg", "scene0231_02/00720.jpg", "scene0231_02/00760.jpg" ], "turn_id": 4, "scan_id": "scene0231_02", "origin_question": "What will be the first-time appearance order of the following categories: bin, dishwasher, poster?", "option": [ "bin, poster, dishwasher", "poster, dishwasher, bin", "dishwasher, poster, bin", "bin, dishwasher, poster" ], "answer": "bin, poster, dishwasher", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bin, dishwasher, poster? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['bin, poster, dishwasher', 'poster, dishwasher, bin', 'dishwasher, poster, bin', 'bin, dishwasher, poster'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0231_02/00800.jpg", "scene0231_02/00840.jpg", "scene0231_02/00880.jpg", "scene0231_02/00920.jpg", "scene0231_02/00960.jpg" ], "turn_id": 5, "scan_id": "scene0231_02", "origin_question": "Which direction is the round wooden table in soft brown color in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the round wooden table in soft brown color in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0231_02/01000.jpg", "scene0231_02/01040.jpg", "scene0231_02/01080.jpg", "scene0231_02/01120.jpg", "scene0231_02/01160.jpg" ], "turn_id": 6, "scan_id": "scene0231_02", "origin_question": "At the end of which turn were both of \"the round wooden table in soft brown color\" and \"the poster\" on your right side?", "option": [], "answer": "2", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the round wooden table in soft brown color\" and \"the poster\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0231_02/01200.jpg", "scene0231_02/01240.jpg", "scene0231_02/01280.jpg", "scene0231_02/01320.jpg", "scene0231_02/01360.jpg" ], "turn_id": 7, "scan_id": "scene0231_02", "origin_question": "Is the tall black steel support with a round base to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the tall black steel support with a round base to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0231_02/01400.jpg", "scene0231_02/01440.jpg", "scene0231_02/01480.jpg", "scene0231_02/01520.jpg", "scene0231_02/01560.jpg" ], "turn_id": 8, "scan_id": "scene0231_02", "origin_question": "In which turn did you see both \"the medium-sized lamp with a white cylindrical shade\" and \"the backpack\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the medium-sized lamp with a white cylindrical shade\" and \"the backpack\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0231_02/01600.jpg", "scene0231_02/01640.jpg", "scene0231_02/01680.jpg", "scene0231_02/01720.jpg", "scene0231_02/01760.jpg" ], "turn_id": 9, "scan_id": "scene0231_02", "origin_question": "How far is your current position from where you were at the end of the turn 6? (in meters)", "option": [], "answer": "1.92", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 6? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0231_02/01800.jpg", "scene0231_02/01840.jpg", "scene0231_02/01880.jpg", "scene0231_02/01920.jpg", "scene0231_02/01960.jpg" ], "turn_id": 10, "scan_id": "scene0231_02", "origin_question": "Please recall \"the medium-sized lamp with a white cylindrical shade\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "3.69", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the medium-sized lamp with a white cylindrical shade\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0221_00/00000.jpg", "scene0221_00/00030.jpg", "scene0221_00/00060.jpg", "scene0221_00/00090.jpg" ], "turn_id": 1, "scan_id": "scene0221_00", "origin_question": "Remember, have you seen any radio(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any radio(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0221_00/00120.jpg", "scene0221_00/00150.jpg", "scene0221_00/00180.jpg", "scene0221_00/00210.jpg" ], "turn_id": 2, "scan_id": "scene0221_00", "origin_question": "Is the mirror to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the mirror to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0221_00/00240.jpg", "scene0221_00/00270.jpg", "scene0221_00/00300.jpg", "scene0221_00/00330.jpg" ], "turn_id": 3, "scan_id": "scene0221_00", "origin_question": "Is the white standard-sized sheet of plain paper to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the white standard-sized sheet of plain paper to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0221_00/00360.jpg", "scene0221_00/00390.jpg", "scene0221_00/00420.jpg", "scene0221_00/00450.jpg" ], "turn_id": 4, "scan_id": "scene0221_00", "origin_question": "Which two objects are on the same side of you now? \"the white standard-sized sheet of plain paper\", \"the rectangular brown wooden stand\", \"the radiator\". ", "option": [ "the white standard-sized sheet of plain paper, the rectangular brown wooden stand", "the white standard-sized sheet of plain paper, the radiator", "the rectangular brown wooden stand, the radiator" ], "answer": "the white standard-sized sheet of plain paper, the rectangular brown wooden stand", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the white standard-sized sheet of plain paper\", \"the rectangular brown wooden stand\", \"the radiator\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the white standard-sized sheet of plain paper, the rectangular brown wooden stand', 'the white standard-sized sheet of plain paper, the radiator', 'the rectangular brown wooden stand, the radiator'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0221_00/00480.jpg", "scene0221_00/00510.jpg", "scene0221_00/00540.jpg", "scene0221_00/00570.jpg" ], "turn_id": 5, "scan_id": "scene0221_00", "origin_question": "What will be the first-time appearance order of the following categories: door, mirror, notebook?", "option": [ "notebook, mirror, door", "notebook, door, mirror", "door, mirror, notebook", "mirror, door, notebook" ], "answer": "door, mirror, notebook", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: door, mirror, notebook? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['notebook, mirror, door', 'notebook, door, mirror', 'door, mirror, notebook', 'mirror, door, notebook'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0221_00/00600.jpg", "scene0221_00/00630.jpg", "scene0221_00/00660.jpg", "scene0221_00/00690.jpg" ], "turn_id": 6, "scan_id": "scene0221_00", "origin_question": "When did you first discover \"the radiator\"? (index of the turn)", "option": [], "answer": "3", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the radiator\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0221_00/00720.jpg", "scene0221_00/00750.jpg", "scene0221_00/00780.jpg", "scene0221_00/00810.jpg" ], "turn_id": 7, "scan_id": "scene0221_00", "origin_question": "Which two objects are on the same side of you now? \"the rectangular light beige wooden cabinet\", \"the modern lamp with a white cylindrical lampshade\", \"the rectangular, soft brown wooden desk\". ", "option": [ "the rectangular light beige wooden cabinet, the modern lamp with a white cylindrical lampshade", "the rectangular light beige wooden cabinet, the rectangular, soft brown wooden desk", "the modern lamp with a white cylindrical lampshade, the rectangular, soft brown wooden desk" ], "answer": "the rectangular light beige wooden cabinet, the modern lamp with a white cylindrical lampshade", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the rectangular light beige wooden cabinet\", \"the modern lamp with a white cylindrical lampshade\", \"the rectangular, soft brown wooden desk\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular light beige wooden cabinet, the modern lamp with a white cylindrical lampshade', 'the rectangular light beige wooden cabinet, the rectangular, soft brown wooden desk', 'the modern lamp with a white cylindrical lampshade, the rectangular, soft brown wooden desk'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0221_00/00840.jpg", "scene0221_00/00870.jpg", "scene0221_00/00900.jpg", "scene0221_00/00930.jpg" ], "turn_id": 8, "scan_id": "scene0221_00", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the mirror\" from your position?", "option": [], "answer": "115.71", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the mirror\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0351_00/00000.jpg", "scene0351_00/00040.jpg", "scene0351_00/00080.jpg", "scene0351_00/00120.jpg" ], "turn_id": 1, "scan_id": "scene0351_00", "origin_question": "Remember, have you seen any window(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any window(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0351_00/00160.jpg", "scene0351_00/00200.jpg", "scene0351_00/00240.jpg", "scene0351_00/00280.jpg" ], "turn_id": 2, "scan_id": "scene0351_00", "origin_question": "Which direction is the rectangular window with a white frame and blinds in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the rectangular window with a white frame and blinds in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0351_00/00320.jpg", "scene0351_00/00360.jpg", "scene0351_00/00400.jpg", "scene0351_00/00440.jpg" ], "turn_id": 3, "scan_id": "scene0351_00", "origin_question": "Which direction is the small plant with shiny green leaves in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the small plant with shiny green leaves in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0351_00/00480.jpg", "scene0351_00/00520.jpg", "scene0351_00/00560.jpg", "scene0351_00/00600.jpg" ], "turn_id": 4, "scan_id": "scene0351_00", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the towel\" from your position?", "option": [], "answer": "117.71", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the towel\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0351_00/00640.jpg", "scene0351_00/00680.jpg", "scene0351_00/00720.jpg", "scene0351_00/00760.jpg" ], "turn_id": 5, "scan_id": "scene0351_00", "origin_question": "Is the rectangular white plastic blackboard to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the rectangular white plastic blackboard to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0351_00/00800.jpg", "scene0351_00/00840.jpg", "scene0351_00/00880.jpg", "scene0351_00/00920.jpg" ], "turn_id": 6, "scan_id": "scene0351_00", "origin_question": "By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular window with a white frame and blinds\", Y: \"the radiator\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular window with a white frame and blinds\", Y: \"the radiator\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0351_00/00960.jpg", "scene0351_00/01000.jpg", "scene0351_00/01040.jpg", "scene0351_00/01080.jpg" ], "turn_id": 7, "scan_id": "scene0351_00", "origin_question": "Which direction is the towel in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the towel in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0351_00/01120.jpg", "scene0351_00/01160.jpg", "scene0351_00/01200.jpg", "scene0351_00/01240.jpg" ], "turn_id": 8, "scan_id": "scene0351_00", "origin_question": "Please recall \"the towel\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "3.25", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the towel\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0351_00/01280.jpg", "scene0351_00/01320.jpg", "scene0351_00/01360.jpg", "scene0351_00/01400.jpg" ], "turn_id": 9, "scan_id": "scene0351_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 9 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0552_01/00000.jpg", "scene0552_01/00030.jpg", "scene0552_01/00060.jpg", "scene0552_01/00090.jpg" ], "turn_id": 1, "scan_id": "scene0552_01", "origin_question": "Remember, have you seen any bucket(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any bucket(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0552_01/00120.jpg", "scene0552_01/00150.jpg", "scene0552_01/00180.jpg", "scene0552_01/00210.jpg" ], "turn_id": 2, "scan_id": "scene0552_01", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0552_01/00240.jpg", "scene0552_01/00270.jpg", "scene0552_01/00300.jpg", "scene0552_01/00330.jpg" ], "turn_id": 3, "scan_id": "scene0552_01", "origin_question": "Which two objects are on the same side of you now? \"the rectangular green metal paper cutter\", \"the standard white plastic bucket\", \"the white rectangular wall-mounted soap dispenser\". ", "option": [ "the rectangular green metal paper cutter, the standard white plastic bucket", "the rectangular green metal paper cutter, the white rectangular wall-mounted soap dispenser", "the standard white plastic bucket, the white rectangular wall-mounted soap dispenser" ], "answer": "the rectangular green metal paper cutter, the standard white plastic bucket", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the rectangular green metal paper cutter\", \"the standard white plastic bucket\", \"the white rectangular wall-mounted soap dispenser\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular green metal paper cutter, the standard white plastic bucket', 'the rectangular green metal paper cutter, the white rectangular wall-mounted soap dispenser', 'the standard white plastic bucket, the white rectangular wall-mounted soap dispenser'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0552_01/00360.jpg", "scene0552_01/00390.jpg", "scene0552_01/00420.jpg", "scene0552_01/00450.jpg" ], "turn_id": 4, "scan_id": "scene0552_01", "origin_question": "Which direction is the rectangular, shiny dispenser with a dark color in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the rectangular, shiny dispenser with a dark color in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0552_01/00480.jpg", "scene0552_01/00510.jpg", "scene0552_01/00540.jpg", "scene0552_01/00570.jpg" ], "turn_id": 5, "scan_id": "scene0552_01", "origin_question": "Which two objects are on the same side of you now? \"the multi-functional gray plastic copier with a sturdy structure\", \"the rectangular green metal paper cutter\", \"the sleek white rectangular microwave with a digital display\". ", "option": [ "the multi-functional gray plastic copier with a sturdy structure, the rectangular green metal paper cutter", "the multi-functional gray plastic copier with a sturdy structure, the sleek white rectangular microwave with a digital display", "the rectangular green metal paper cutter, the sleek white rectangular microwave with a digital display" ], "answer": "the multi-functional gray plastic copier with a sturdy structure, the rectangular green metal paper cutter", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the multi-functional gray plastic copier with a sturdy structure\", \"the rectangular green metal paper cutter\", \"the sleek white rectangular microwave with a digital display\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the multi-functional gray plastic copier with a sturdy structure, the rectangular green metal paper cutter', 'the multi-functional gray plastic copier with a sturdy structure, the sleek white rectangular microwave with a digital display', 'the rectangular green metal paper cutter, the sleek white rectangular microwave with a digital display'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0338_00/00000.jpg", "scene0338_00/00030.jpg", "scene0338_00/00060.jpg", "scene0338_00/00090.jpg" ], "turn_id": 1, "scan_id": "scene0338_00", "origin_question": "Remember, have you seen any mailbox(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any mailbox(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0338_00/00120.jpg", "scene0338_00/00150.jpg", "scene0338_00/00180.jpg", "scene0338_00/00210.jpg" ], "turn_id": 2, "scan_id": "scene0338_00", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the rectangular black bin with a flat top\"; \"the large modern gray plastic copier\"; \"the standard-sized rectangular white paper envelope\" ", "option": [ "the rectangular black bin with a flat top", "the large modern gray plastic copier", "the standard-sized rectangular white paper envelope" ], "answer": "the rectangular black bin with a flat top", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the rectangular black bin with a flat top\"; \"the large modern gray plastic copier\"; \"the standard-sized rectangular white paper envelope\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular black bin with a flat top', 'the large modern gray plastic copier', 'the standard-sized rectangular white paper envelope'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0338_00/00240.jpg", "scene0338_00/00270.jpg", "scene0338_00/00300.jpg", "scene0338_00/00330.jpg" ], "turn_id": 3, "scan_id": "scene0338_00", "origin_question": "Please recall \"the white rectangular mailbox made of metal or plastic\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.88", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the white rectangular mailbox made of metal or plastic\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0338_00/00360.jpg", "scene0338_00/00390.jpg", "scene0338_00/00420.jpg", "scene0338_00/00450.jpg" ], "turn_id": 4, "scan_id": "scene0338_00", "origin_question": "When did you first discover \"the rectangular deep gray counter with wood and metal\"? (index of the turn)", "option": [], "answer": "3", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the rectangular deep gray counter with wood and metal\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0338_00/00480.jpg", "scene0338_00/00510.jpg", "scene0338_00/00540.jpg", "scene0338_00/00570.jpg" ], "turn_id": 5, "scan_id": "scene0338_00", "origin_question": "Remember, how many detergent(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many detergent(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0338_00/00600.jpg", "scene0338_00/00630.jpg", "scene0338_00/00660.jpg", "scene0338_00/00690.jpg" ], "turn_id": 6, "scan_id": "scene0338_00", "origin_question": "In which turn did you see both \"the large modern gray plastic copier\" and \"the door\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the large modern gray plastic copier\" and \"the door\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "1mp3d_0040_region34/ded0eb84b03d4b3a86147a6b84b9a1cd_i1_3.jpg", "1mp3d_0040_region34/ded0eb84b03d4b3a86147a6b84b9a1cd_i2_3.jpg", "1mp3d_0040_region34/399345689db24e29a7e8993c629aa09f_i1_4.jpg", "1mp3d_0040_region34/399345689db24e29a7e8993c629aa09f_i2_4.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0040_region34", "origin_question": "Remember, how many cap(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many cap(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "1mp3d_0040_region34/af5b78be15074119b26c2cc68cc6c5ac_i1_4.jpg", "1mp3d_0040_region34/af5b78be15074119b26c2cc68cc6c5ac_i2_4.jpg", "1mp3d_0040_region34/dab8cb8e76294cfba05d7f8d0b5165be_i1_3.jpg", "1mp3d_0040_region34/dab8cb8e76294cfba05d7f8d0b5165be_i2_3.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0040_region34", "origin_question": "Remember, how many globe(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many globe(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "1mp3d_0040_region34/dab8cb8e76294cfba05d7f8d0b5165be_i1_5.jpg", "1mp3d_0040_region34/dab8cb8e76294cfba05d7f8d0b5165be_i2_5.jpg", "1mp3d_0040_region34/af5b78be15074119b26c2cc68cc6c5ac_i1_3.jpg", "1mp3d_0040_region34/af5b78be15074119b26c2cc68cc6c5ac_i2_3.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0040_region34", "origin_question": "Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance forward or backward from that position?", "option": [ "Forward", "Backward" ], "answer": "Backward", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance forward or backward from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Forward', 'Backward'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0040_region34/399345689db24e29a7e8993c629aa09f_i1_2.jpg", "1mp3d_0040_region34/399345689db24e29a7e8993c629aa09f_i2_2.jpg", "1mp3d_0040_region34/399345689db24e29a7e8993c629aa09f_i1_3.jpg", "1mp3d_0040_region34/399345689db24e29a7e8993c629aa09f_i2_3.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0040_region34", "origin_question": "", "option": [], "answer": "", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "1mp3d_0040_region34/eea2e371b6ac43c4a72edf8546be9388_i1_1.jpg", "1mp3d_0040_region34/eea2e371b6ac43c4a72edf8546be9388_i2_1.jpg", "1mp3d_0040_region34/eea2e371b6ac43c4a72edf8546be9388_i1_0.jpg", "1mp3d_0040_region34/eea2e371b6ac43c4a72edf8546be9388_i2_0.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0040_region34", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "2.67", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "1mp3d_0040_region34/3e1701bd29544e47b96d79d370933af5_i1_1.jpg", "1mp3d_0040_region34/3e1701bd29544e47b96d79d370933af5_i2_1.jpg", "1mp3d_0040_region34/a4c2f102b23b4d4b87b21d3cb5d0cfa2_i1_1.jpg", "1mp3d_0040_region34/a4c2f102b23b4d4b87b21d3cb5d0cfa2_i2_1.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0040_region34", "origin_question": "Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "1mp3d_0040_region34/77fa4b2db2cf45a0abff5833de1bc079_i1_1.jpg", "1mp3d_0040_region34/77fa4b2db2cf45a0abff5833de1bc079_i2_1.jpg", "1mp3d_0040_region34/b6169f1b5e0a42489ed145131e9ae486_i1_0.jpg", "1mp3d_0040_region34/b6169f1b5e0a42489ed145131e9ae486_i2_0.jpg" ], "turn_id": 7, "scan_id": "1mp3d_0040_region34", "origin_question": "Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance forward or backward from that position?", "option": [ "Forward", "Backward" ], "answer": "Forward", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance forward or backward from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Forward', 'Backward'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "1mp3d_0040_region34/77fa4b2db2cf45a0abff5833de1bc079_i1_0.jpg", "1mp3d_0040_region34/77fa4b2db2cf45a0abff5833de1bc079_i2_0.jpg", "1mp3d_0040_region34/a4c2f102b23b4d4b87b21d3cb5d0cfa2_i1_0.jpg", "1mp3d_0040_region34/a4c2f102b23b4d4b87b21d3cb5d0cfa2_i2_0.jpg" ], "turn_id": 8, "scan_id": "1mp3d_0040_region34", "origin_question": "How far is your current position from where you were at the end of the turn 5? (in meters)", "option": [], "answer": "2.61", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 5? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0040_region34/3e1701bd29544e47b96d79d370933af5_i1_3.jpg", "1mp3d_0040_region34/3e1701bd29544e47b96d79d370933af5_i2_3.jpg", "1mp3d_0040_region34/3e1701bd29544e47b96d79d370933af5_i1_4.jpg", "1mp3d_0040_region34/3e1701bd29544e47b96d79d370933af5_i2_4.jpg" ], "turn_id": 9, "scan_id": "1mp3d_0040_region34", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the large flat-panel television with a black frame\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 9 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the large flat-panel television with a black frame\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0549_00/00000.jpg", "scene0549_00/00030.jpg", "scene0549_00/00060.jpg", "scene0549_00/00090.jpg" ], "turn_id": 1, "scan_id": "scene0549_00", "origin_question": "Remember, have you seen any camera(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any camera(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0549_00/00120.jpg", "scene0549_00/00150.jpg", "scene0549_00/00180.jpg", "scene0549_00/00210.jpg" ], "turn_id": 2, "scan_id": "scene0549_00", "origin_question": "In which turn did you see both \"the rectangular three-seater glossy green leather couch\" and \"the door\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the rectangular three-seater glossy green leather couch\" and \"the door\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0549_00/00240.jpg", "scene0549_00/00270.jpg", "scene0549_00/00300.jpg", "scene0549_00/00330.jpg" ], "turn_id": 3, "scan_id": "scene0549_00", "origin_question": "How far is your current position from where you were at the end of the turn 1? (in meters)", "option": [], "answer": "3.3", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 1? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0549_00/00360.jpg", "scene0549_00/00390.jpg", "scene0549_00/00420.jpg", "scene0549_00/00450.jpg" ], "turn_id": 4, "scan_id": "scene0549_00", "origin_question": "What will be the first-time appearance order of the following categories: couch, plant, shelf?", "option": [ "couch, shelf, plant", "shelf, couch, plant", "plant, couch, shelf", "couch, plant, shelf" ], "answer": "couch, shelf, plant", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: couch, plant, shelf? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['couch, shelf, plant', 'shelf, couch, plant', 'plant, couch, shelf', 'couch, plant, shelf'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0549_00/00480.jpg", "scene0549_00/00510.jpg", "scene0549_00/00540.jpg", "scene0549_00/00570.jpg" ], "turn_id": 5, "scan_id": "scene0549_00", "origin_question": "What will be the first-time appearance order of the following categories: plant, shelf, table?", "option": [ "plant, table, shelf", "table, plant, shelf", "shelf, table, plant", "table, shelf, plant" ], "answer": "table, shelf, plant", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: plant, shelf, table? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['plant, table, shelf', 'table, plant, shelf', 'shelf, table, plant', 'table, shelf, plant'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0549_00/00600.jpg", "scene0549_00/00630.jpg", "scene0549_00/00660.jpg", "scene0549_00/00690.jpg" ], "turn_id": 6, "scan_id": "scene0549_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0583_02/00000.jpg", "scene0583_02/00040.jpg", "scene0583_02/00080.jpg", "scene0583_02/00120.jpg", "scene0583_02/00160.jpg" ], "turn_id": 1, "scan_id": "scene0583_02", "origin_question": "Remember, have you seen any bottle(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any bottle(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0583_02/00200.jpg", "scene0583_02/00240.jpg", "scene0583_02/00280.jpg", "scene0583_02/00320.jpg", "scene0583_02/00360.jpg" ], "turn_id": 2, "scan_id": "scene0583_02", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the white rectangular plastic bottle with a pump top\"; \"the rectangular picture with blue, green, and white floral patterns\"; \"the well-maintained rectangular wooden door with a brown surface\" ", "option": [ "the white rectangular plastic bottle with a pump top", "the rectangular picture with blue, green, and white floral patterns", "the well-maintained rectangular wooden door with a brown surface" ], "answer": "the rectangular picture with blue, green, and white floral patterns", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the white rectangular plastic bottle with a pump top\"; \"the rectangular picture with blue, green, and white floral patterns\"; \"the well-maintained rectangular wooden door with a brown surface\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the white rectangular plastic bottle with a pump top', 'the rectangular picture with blue, green, and white floral patterns', 'the well-maintained rectangular wooden door with a brown surface'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0583_02/00400.jpg", "scene0583_02/00440.jpg", "scene0583_02/00480.jpg", "scene0583_02/00520.jpg", "scene0583_02/00560.jpg" ], "turn_id": 3, "scan_id": "scene0583_02", "origin_question": "When did you first discover \"the standard rectangular ceramic bathtub in a light color\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the standard rectangular ceramic bathtub in a light color\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0583_02/00600.jpg", "scene0583_02/00640.jpg", "scene0583_02/00680.jpg", "scene0583_02/00720.jpg", "scene0583_02/00760.jpg" ], "turn_id": 4, "scan_id": "scene0583_02", "origin_question": "How far is your current position from where you were at the end of the turn 1? (in meters)", "option": [], "answer": "1.4", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 1? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0583_02/00800.jpg", "scene0583_02/00840.jpg", "scene0583_02/00880.jpg", "scene0583_02/00920.jpg", "scene0583_02/00960.jpg" ], "turn_id": 5, "scan_id": "scene0583_02", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the window\" from your position?", "option": [], "answer": "144.8", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the window\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0583_02/01000.jpg", "scene0583_02/01040.jpg", "scene0583_02/01080.jpg", "scene0583_02/01120.jpg", "scene0583_02/01160.jpg" ], "turn_id": 6, "scan_id": "scene0583_02", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the well-maintained rectangular wooden door with a brown surface\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the well-maintained rectangular wooden door with a brown surface\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0583_02/01200.jpg", "scene0583_02/01240.jpg", "scene0583_02/01280.jpg", "scene0583_02/01320.jpg", "scene0583_02/01360.jpg" ], "turn_id": 7, "scan_id": "scene0583_02", "origin_question": "Using your orientation at the end of turn 6 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "145.52", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 6 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0583_02/01400.jpg", "scene0583_02/01440.jpg", "scene0583_02/01480.jpg", "scene0583_02/01520.jpg", "scene0583_02/01560.jpg" ], "turn_id": 8, "scan_id": "scene0583_02", "origin_question": "What will be the first-time appearance order of the following categories: bathtub, picture, toilet paper?", "option": [ "toilet paper, bathtub, picture", "picture, bathtub, toilet paper", "picture, toilet paper, bathtub", "bathtub, picture, toilet paper" ], "answer": "bathtub, picture, toilet paper", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bathtub, picture, toilet paper? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['toilet paper, bathtub, picture', 'picture, bathtub, toilet paper', 'picture, toilet paper, bathtub', 'bathtub, picture, toilet paper'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0478_01/00000.jpg", "scene0478_01/00030.jpg", "scene0478_01/00060.jpg", "scene0478_01/00090.jpg" ], "turn_id": 1, "scan_id": "scene0478_01", "origin_question": "Remember, have you seen any printer(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any printer(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0478_01/00120.jpg", "scene0478_01/00150.jpg", "scene0478_01/00180.jpg", "scene0478_01/00210.jpg" ], "turn_id": 2, "scan_id": "scene0478_01", "origin_question": "In which turn did you see both \"the large white plastic copier with a control panel\" and \"the bin\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the large white plastic copier with a control panel\" and \"the bin\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0478_01/00240.jpg", "scene0478_01/00270.jpg", "scene0478_01/00300.jpg", "scene0478_01/00330.jpg" ], "turn_id": 3, "scan_id": "scene0478_01", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the rectangular light gray multi-level tray\" from your position?", "option": [], "answer": "150.64", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the rectangular light gray multi-level tray\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0478_01/00360.jpg", "scene0478_01/00390.jpg", "scene0478_01/00420.jpg", "scene0478_01/00450.jpg" ], "turn_id": 4, "scan_id": "scene0478_01", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the small, vibrant book with a colorful cover\"; \"the modern glass door with a black metal frame\"; \"the large black counter with a flat top surface\" ", "option": [ "the small, vibrant book with a colorful cover", "the modern glass door with a black metal frame", "the large black counter with a flat top surface" ], "answer": "the modern glass door with a black metal frame", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the small, vibrant book with a colorful cover\"; \"the modern glass door with a black metal frame\"; \"the large black counter with a flat top surface\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the small, vibrant book with a colorful cover', 'the modern glass door with a black metal frame', 'the large black counter with a flat top surface'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0478_01/00480.jpg", "scene0478_01/00510.jpg", "scene0478_01/00540.jpg", "scene0478_01/00570.jpg" ], "turn_id": 5, "scan_id": "scene0478_01", "origin_question": "Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0478_01/00600.jpg", "scene0478_01/00630.jpg", "scene0478_01/00660.jpg", "scene0478_01/00690.jpg" ], "turn_id": 6, "scan_id": "scene0478_01", "origin_question": "What will be the first-time appearance order of the following categories: book, cabinet, door?", "option": [ "cabinet, book, door", "cabinet, door, book", "book, cabinet, door", "door, book, cabinet" ], "answer": "cabinet, book, door", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: book, cabinet, door? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['cabinet, book, door', 'cabinet, door, book', 'book, cabinet, door', 'door, book, cabinet'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0044_region44/697500b5e2f84c7a9884dd4388e26263_i1_0.jpg", "1mp3d_0044_region44/697500b5e2f84c7a9884dd4388e26263_i2_0.jpg", "1mp3d_0044_region44/768b72a0c6ac4aaa8295d6366f4edded_i1_2.jpg", "1mp3d_0044_region44/697500b5e2f84c7a9884dd4388e26263_i1_5.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0044_region44", "origin_question": "", "option": [], "answer": "", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 3 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "1mp3d_0044_region44/697500b5e2f84c7a9884dd4388e26263_i2_5.jpg", "1mp3d_0044_region44/768b72a0c6ac4aaa8295d6366f4edded_i1_5.jpg", "1mp3d_0044_region44/768b72a0c6ac4aaa8295d6366f4edded_i2_5.jpg", "1mp3d_0044_region44/697500b5e2f84c7a9884dd4388e26263_i1_4.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0044_region44", "origin_question": "When did you first discover \"the square outdoor table with a mosaic tile top\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the square outdoor table with a mosaic tile top\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 3 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0044_region44/697500b5e2f84c7a9884dd4388e26263_i2_4.jpg", "1mp3d_0044_region44/768b72a0c6ac4aaa8295d6366f4edded_i1_1.jpg", "1mp3d_0044_region44/768b72a0c6ac4aaa8295d6366f4edded_i2_1.jpg", "1mp3d_0044_region44/c2750dd86e7642caaaf189f59f82b843_i1_3.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0044_region44", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the square outdoor table with a mosaic tile top\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the square outdoor table with a mosaic tile top\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 3 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0067_region23/ecff9ecf0cfe4d8bb83260fc092f3b00_i1_1.jpg", "1mp3d_0067_region23/ecff9ecf0cfe4d8bb83260fc092f3b00_i2_1.jpg", "1mp3d_0067_region23/28c09c307b11487c999f88e1e9ec3231_i1_2.jpg", "1mp3d_0067_region23/28c09c307b11487c999f88e1e9ec3231_i2_2.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0067_region23", "origin_question": "Remember, have you seen any water cooler(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any water cooler(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0067_region23/47bec259eaac4e00a4db928751889af6_i1_0.jpg", "1mp3d_0067_region23/47bec259eaac4e00a4db928751889af6_i2_0.jpg", "1mp3d_0067_region23/47bec259eaac4e00a4db928751889af6_i1_5.jpg", "1mp3d_0067_region23/47bec259eaac4e00a4db928751889af6_i2_5.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0067_region23", "origin_question": "Remember, have you seen any door(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any door(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0067_region23/28c09c307b11487c999f88e1e9ec3231_i1_1.jpg", "1mp3d_0067_region23/28c09c307b11487c999f88e1e9ec3231_i2_1.jpg", "1mp3d_0067_region23/ea5cf8f62e1e46bd805e248b8286228b_i1_2.jpg", "1mp3d_0067_region23/ea5cf8f62e1e46bd805e248b8286228b_i2_2.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0067_region23", "origin_question": "", "option": [], "answer": "", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "1mp3d_0067_region23/ea5cf8f62e1e46bd805e248b8286228b_i1_3.jpg", "1mp3d_0067_region23/ea5cf8f62e1e46bd805e248b8286228b_i2_3.jpg", "1mp3d_0067_region23/ea5cf8f62e1e46bd805e248b8286228b_i1_4.jpg", "1mp3d_0067_region23/ea5cf8f62e1e46bd805e248b8286228b_i2_4.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0067_region23", "origin_question": "By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the stove\", Y: \"the small, round black wooden table with decorative patterns\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "farther away from both X and Y", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the stove\", Y: \"the small, round black wooden table with decorative patterns\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "1mp3d_0067_region23/28c09c307b11487c999f88e1e9ec3231_i1_5.jpg", "1mp3d_0067_region23/28c09c307b11487c999f88e1e9ec3231_i2_5.jpg", "1mp3d_0067_region23/28c09c307b11487c999f88e1e9ec3231_i1_4.jpg", "1mp3d_0067_region23/28c09c307b11487c999f88e1e9ec3231_i2_4.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0067_region23", "origin_question": "Remember, how many paper cutter(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many paper cutter(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "1mp3d_0067_region23/ecff9ecf0cfe4d8bb83260fc092f3b00_i1_4.jpg", "1mp3d_0067_region23/ecff9ecf0cfe4d8bb83260fc092f3b00_i2_4.jpg", "1mp3d_0067_region23/ecff9ecf0cfe4d8bb83260fc092f3b00_i1_5.jpg", "1mp3d_0067_region23/ecff9ecf0cfe4d8bb83260fc092f3b00_i2_5.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0067_region23", "origin_question": "Using your orientation at the end of turn 4 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "58.08", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 4 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "1mp3d_0040_region18/6fb3acd268e9467cb3d47045fd8a1f2d_i1_3.jpg", "1mp3d_0040_region18/6fb3acd268e9467cb3d47045fd8a1f2d_i2_3.jpg", "1mp3d_0040_region18/ecb4c9d8de6f43f2ab5e642f3e6049a6_i1_2.jpg", "1mp3d_0040_region18/ecb4c9d8de6f43f2ab5e642f3e6049a6_i2_2.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0040_region18", "origin_question": "Remember, how many glass(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many glass(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0040_region18/6fb3acd268e9467cb3d47045fd8a1f2d_i1_2.jpg", "1mp3d_0040_region18/6fb3acd268e9467cb3d47045fd8a1f2d_i2_2.jpg", "1mp3d_0040_region18/592ac8be752f45e889a5170e4e3c2060_i1_2.jpg", "1mp3d_0040_region18/592ac8be752f45e889a5170e4e3c2060_i2_2.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0040_region18", "origin_question": "Remember, have you seen any chair(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any chair(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "1mp3d_0040_region18/32ae1b4f7fb4435fa1311bcc9031b9c3_i1_4.jpg", "1mp3d_0040_region18/32ae1b4f7fb4435fa1311bcc9031b9c3_i2_4.jpg", "1mp3d_0040_region18/32ae1b4f7fb4435fa1311bcc9031b9c3_i1_5.jpg", "1mp3d_0040_region18/32ae1b4f7fb4435fa1311bcc9031b9c3_i2_5.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0040_region18", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "1.92", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "1mp3d_0040_region18/592ac8be752f45e889a5170e4e3c2060_i1_3.jpg", "1mp3d_0040_region18/592ac8be752f45e889a5170e4e3c2060_i2_3.jpg", "1mp3d_0040_region18/592ac8be752f45e889a5170e4e3c2060_i1_1.jpg", "1mp3d_0040_region18/592ac8be752f45e889a5170e4e3c2060_i2_1.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0040_region18", "origin_question": "Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "64.81", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "1mp3d_0040_region18/965df3e44811465e92a56c189c024b88_i1_3.jpg", "1mp3d_0040_region18/965df3e44811465e92a56c189c024b88_i2_3.jpg", "1mp3d_0040_region18/3c2fad423ba64923b42b7d9453bcf000_i1_2.jpg", "1mp3d_0040_region18/3c2fad423ba64923b42b7d9453bcf000_i2_2.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0040_region18", "origin_question": "How far is your current position from where you were at the end of the turn 1? (in meters)", "option": [], "answer": "6.83", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 1? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "1mp3d_0040_region18/001499e78f734a4eaf14e727140baf21_i1_5.jpg", "1mp3d_0040_region18/001499e78f734a4eaf14e727140baf21_i2_5.jpg", "1mp3d_0040_region18/001499e78f734a4eaf14e727140baf21_i1_0.jpg", "1mp3d_0040_region18/001499e78f734a4eaf14e727140baf21_i2_0.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0040_region18", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "1mp3d_0040_region18/3c2fad423ba64923b42b7d9453bcf000_i1_1.jpg", "1mp3d_0040_region18/3c2fad423ba64923b42b7d9453bcf000_i2_1.jpg", "1mp3d_0040_region18/965df3e44811465e92a56c189c024b88_i1_1.jpg", "1mp3d_0040_region18/965df3e44811465e92a56c189c024b88_i2_1.jpg" ], "turn_id": 7, "scan_id": "1mp3d_0040_region18", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "118.17", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "1mp3d_0040_region18/965df3e44811465e92a56c189c024b88_i1_2.jpg", "1mp3d_0040_region18/965df3e44811465e92a56c189c024b88_i2_2.jpg", "1mp3d_0040_region18/365e81eb660e4cdea4634560b0b166f8_i1_5.jpg", "1mp3d_0040_region18/365e81eb660e4cdea4634560b0b166f8_i2_5.jpg" ], "turn_id": 8, "scan_id": "1mp3d_0040_region18", "origin_question": "At the end of which turn were both of \"the medium-sized white rectangular projector\" and \"the tall dark metal rack with multiple layers\" on your right side?", "option": [], "answer": "7", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the medium-sized white rectangular projector\" and \"the tall dark metal rack with multiple layers\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "1mp3d_0040_region18/365e81eb660e4cdea4634560b0b166f8_i1_0.jpg", "1mp3d_0040_region18/365e81eb660e4cdea4634560b0b166f8_i2_0.jpg", "1mp3d_0040_region18/365e81eb660e4cdea4634560b0b166f8_i1_1.jpg", "1mp3d_0040_region18/365e81eb660e4cdea4634560b0b166f8_i2_1.jpg" ], "turn_id": 9, "scan_id": "1mp3d_0040_region18", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the large rectangular white screen with black borders\" from your position?", "option": [], "answer": "104.91", "user_message": "For the 9 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the large rectangular white screen with black borders\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0074_region8/1227344eaa1f424b83db585fd665fb2a_i1_2.jpg", "1mp3d_0074_region8/1227344eaa1f424b83db585fd665fb2a_i2_2.jpg", "1mp3d_0074_region8/db0e4e7b77c04948b581a11f633df145_i1_3.jpg", "1mp3d_0074_region8/db0e4e7b77c04948b581a11f633df145_i2_3.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0074_region8", "origin_question": "Remember, have you seen any curtain(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any curtain(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "1mp3d_0074_region8/c9b7a323a72141628f0c48aa7eb3430e_i1_3.jpg", "1mp3d_0074_region8/c9b7a323a72141628f0c48aa7eb3430e_i2_3.jpg", "1mp3d_0074_region8/db0e4e7b77c04948b581a11f633df145_i1_4.jpg", "1mp3d_0074_region8/db0e4e7b77c04948b581a11f633df145_i2_4.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0074_region8", "origin_question": "When did you first discover \"the long, gray curtain made of soft fabric\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the long, gray curtain made of soft fabric\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "1mp3d_0074_region8/68e4c4477aa442cbbb9a9f1e54b9b831_i2_0.jpg", "1mp3d_0074_region8/96490c4da78240058d1f76a07ed95c9d_i1_4.jpg", "1mp3d_0074_region8/96490c4da78240058d1f76a07ed95c9d_i2_4.jpg", "1mp3d_0074_region8/96490c4da78240058d1f76a07ed95c9d_i2_0.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0074_region8", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the large rectangular glass window with a black frame\", Y: \"the long, gray curtain made of soft fabric\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "farther away from both X and Y", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the large rectangular glass window with a black frame\", Y: \"the long, gray curtain made of soft fabric\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0074_region8/68e4c4477aa442cbbb9a9f1e54b9b831_i1_5.jpg", "1mp3d_0074_region8/68e4c4477aa442cbbb9a9f1e54b9b831_i2_5.jpg", "1mp3d_0074_region8/96490c4da78240058d1f76a07ed95c9d_i1_5.jpg", "1mp3d_0074_region8/96490c4da78240058d1f76a07ed95c9d_i2_5.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0074_region8", "origin_question": "", "option": [], "answer": "", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0280_02/00000.jpg", "scene0280_02/00040.jpg", "scene0280_02/00080.jpg", "scene0280_02/00120.jpg" ], "turn_id": 1, "scan_id": "scene0280_02", "origin_question": "Remember, have you seen any blackboard(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any blackboard(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0280_02/00160.jpg", "scene0280_02/00200.jpg", "scene0280_02/00240.jpg", "scene0280_02/00280.jpg" ], "turn_id": 2, "scan_id": "scene0280_02", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the standard chair with a dark fabric covering\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the standard chair with a dark fabric covering\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0280_02/00320.jpg", "scene0280_02/00360.jpg", "scene0280_02/00400.jpg", "scene0280_02/00440.jpg" ], "turn_id": 3, "scan_id": "scene0280_02", "origin_question": "Please recall \"the standard chair with a dark fabric covering\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.38", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the standard chair with a dark fabric covering\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0280_02/00480.jpg", "scene0280_02/00520.jpg", "scene0280_02/00560.jpg", "scene0280_02/00600.jpg" ], "turn_id": 4, "scan_id": "scene0280_02", "origin_question": "When did you first discover \"the medium-sized rectangular dark brown wood dresser\"? (index of the turn)", "option": [], "answer": "3", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the medium-sized rectangular dark brown wood dresser\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0280_02/00640.jpg", "scene0280_02/00680.jpg", "scene0280_02/00720.jpg", "scene0280_02/00760.jpg" ], "turn_id": 5, "scan_id": "scene0280_02", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0280_02/00800.jpg", "scene0280_02/00840.jpg", "scene0280_02/00880.jpg", "scene0280_02/00920.jpg" ], "turn_id": 6, "scan_id": "scene0280_02", "origin_question": "What will be the first-time appearance order of the following categories: chair, stand, towel?", "option": [ "chair, towel, stand", "towel, chair, stand", "towel, stand, chair", "stand, chair, towel" ], "answer": "stand, chair, towel", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: chair, stand, towel? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['chair, towel, stand', 'towel, chair, stand', 'towel, stand, chair', 'stand, chair, towel'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0280_02/00960.jpg", "scene0280_02/01000.jpg", "scene0280_02/01040.jpg", "scene0280_02/01080.jpg" ], "turn_id": 7, "scan_id": "scene0280_02", "origin_question": "What will be the first-time appearance order of the following categories: ball, chair, vanity?", "option": [ "vanity, chair, ball", "ball, chair, vanity", "ball, vanity, chair", "chair, vanity, ball" ], "answer": "ball, chair, vanity", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: ball, chair, vanity? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['vanity, chair, ball', 'ball, chair, vanity', 'ball, vanity, chair', 'chair, vanity, ball'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0280_02/01120.jpg", "scene0280_02/01160.jpg", "scene0280_02/01200.jpg", "scene0280_02/01240.jpg" ], "turn_id": 8, "scan_id": "scene0280_02", "origin_question": "Which direction is the medium rectangular glossy black laptop in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the medium rectangular glossy black laptop in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0280_02/01280.jpg", "scene0280_02/01320.jpg", "scene0280_02/01360.jpg", "scene0280_02/01400.jpg" ], "turn_id": 9, "scan_id": "scene0280_02", "origin_question": "At the end of which turn were both of \"the medium-sized, round, green ball\" and \"the medium-sized rectangular dark brown wood dresser\" on your left side?", "option": [], "answer": "2", "user_message": "For the 9 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the medium-sized, round, green ball\" and \"the medium-sized rectangular dark brown wood dresser\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0389_00/00000.jpg", "scene0389_00/00040.jpg", "scene0389_00/00080.jpg", "scene0389_00/00120.jpg" ], "turn_id": 1, "scan_id": "scene0389_00", "origin_question": "Remember, have you seen any clock(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any clock(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0389_00/00160.jpg", "scene0389_00/00200.jpg", "scene0389_00/00240.jpg", "scene0389_00/00280.jpg" ], "turn_id": 2, "scan_id": "scene0389_00", "origin_question": "When did you first discover \"the double bed with a wooden headboard and white bedspread\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the double bed with a wooden headboard and white bedspread\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0389_00/00320.jpg", "scene0389_00/00360.jpg", "scene0389_00/00400.jpg", "scene0389_00/00440.jpg" ], "turn_id": 3, "scan_id": "scene0389_00", "origin_question": "Which two objects are on the same side of you now? \"the rectangular board with a blue fabric surface\", \"the rectangular blue plastic box without a lid\", \"the horizontal, white metal radiator with rectangular shape\". ", "option": [ "the rectangular board with a blue fabric surface, the rectangular blue plastic box without a lid", "the rectangular board with a blue fabric surface, the horizontal, white metal radiator with rectangular shape", "the rectangular blue plastic box without a lid, the horizontal, white metal radiator with rectangular shape" ], "answer": "the rectangular board with a blue fabric surface, the rectangular blue plastic box without a lid", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the rectangular board with a blue fabric surface\", \"the rectangular blue plastic box without a lid\", \"the horizontal, white metal radiator with rectangular shape\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular board with a blue fabric surface, the rectangular blue plastic box without a lid', 'the rectangular board with a blue fabric surface, the horizontal, white metal radiator with rectangular shape', 'the rectangular blue plastic box without a lid, the horizontal, white metal radiator with rectangular shape'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0389_00/00480.jpg", "scene0389_00/00520.jpg", "scene0389_00/00560.jpg", "scene0389_00/00600.jpg" ], "turn_id": 4, "scan_id": "scene0389_00", "origin_question": "Is the rectangular board with a blue fabric surface to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the rectangular board with a blue fabric surface to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0389_00/00640.jpg", "scene0389_00/00680.jpg", "scene0389_00/00720.jpg", "scene0389_00/00760.jpg" ], "turn_id": 5, "scan_id": "scene0389_00", "origin_question": "Please recall \"the medium-sized rectangular dark plastic telephone\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.41", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the medium-sized rectangular dark plastic telephone\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0389_00/00800.jpg", "scene0389_00/00840.jpg", "scene0389_00/00880.jpg", "scene0389_00/00920.jpg" ], "turn_id": 6, "scan_id": "scene0389_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0389_00/00960.jpg", "scene0389_00/01000.jpg", "scene0389_00/01040.jpg", "scene0389_00/01080.jpg" ], "turn_id": 7, "scan_id": "scene0389_00", "origin_question": "When did you first discover \"the rectangular, rich brown wooden table\"? (index of the turn)", "option": [], "answer": "5", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the rectangular, rich brown wooden table\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0389_00/01120.jpg", "scene0389_00/01160.jpg", "scene0389_00/01200.jpg", "scene0389_00/01240.jpg" ], "turn_id": 8, "scan_id": "scene0389_00", "origin_question": "In which turn did you see both \"the light-colored fabric curtain with a subtle pattern\" and \"the rectangular, rich brown wooden table\" simultaneously? ", "option": [], "answer": "5", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the light-colored fabric curtain with a subtle pattern\" and \"the rectangular, rich brown wooden table\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0047_region16/67b8535df7ae45809c807744186d7a1f_i1_4.jpg", "1mp3d_0047_region16/67b8535df7ae45809c807744186d7a1f_i2_4.jpg", "1mp3d_0047_region16/b11577607584497581c9433f5406c55b_i1_2.jpg", "1mp3d_0047_region16/b11577607584497581c9433f5406c55b_i2_2.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0047_region16", "origin_question": "Remember, have you seen any toilet(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any toilet(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 2 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "1mp3d_0047_region16/67b8535df7ae45809c807744186d7a1f_i1_3.jpg", "1mp3d_0047_region16/67b8535df7ae45809c807744186d7a1f_i2_3.jpg", "1mp3d_0047_region16/67b8535df7ae45809c807744186d7a1f_i1_2.jpg", "1mp3d_0047_region16/67b8535df7ae45809c807744186d7a1f_i2_2.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0047_region16", "origin_question": "In which turn did you see both \"the large white cabinet with a black countertop\" and \"the modern white ceramic toilet with a smooth bowl\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the large white cabinet with a black countertop\" and \"the modern white ceramic toilet with a smooth bowl\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 2 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0057_region5/6a289d9b67cd43f4b59aab2037af221b_i1_2.jpg", "1mp3d_0057_region5/6a289d9b67cd43f4b59aab2037af221b_i2_2.jpg", "1mp3d_0057_region5/2112dc2d19404cbdbc41dae916377e19_i1_3.jpg", "1mp3d_0057_region5/2112dc2d19404cbdbc41dae916377e19_i2_3.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0057_region5", "origin_question": "Remember, have you seen any stairs(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any stairs(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "1mp3d_0057_region5/2112dc2d19404cbdbc41dae916377e19_i1_0.jpg", "1mp3d_0057_region5/2112dc2d19404cbdbc41dae916377e19_i2_0.jpg", "1mp3d_0057_region5/c06443aebf13409da70a8184e17663dd_i1_0.jpg", "1mp3d_0057_region5/c06443aebf13409da70a8184e17663dd_i2_0.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0057_region5", "origin_question": "In which turn were you closest to \"the stairs\"? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the stairs\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0057_region5/b1eaadc1fb0b4bb7b5f4e7afeb0fc9da_i2_3.jpg", "1mp3d_0057_region5/c06443aebf13409da70a8184e17663dd_i1_3.jpg", "1mp3d_0057_region5/c06443aebf13409da70a8184e17663dd_i2_3.jpg", "1mp3d_0057_region5/b1eaadc1fb0b4bb7b5f4e7afeb0fc9da_i1_0.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0057_region5", "origin_question": "", "option": [], "answer": "", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "1mp3d_0057_region5/b1eaadc1fb0b4bb7b5f4e7afeb0fc9da_i2_0.jpg", "1mp3d_0057_region5/b1eaadc1fb0b4bb7b5f4e7afeb0fc9da_i2_2.jpg", "1mp3d_0057_region5/b1eaadc1fb0b4bb7b5f4e7afeb0fc9da_i1_5.jpg", "1mp3d_0057_region5/b1eaadc1fb0b4bb7b5f4e7afeb0fc9da_i2_5.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0057_region5", "origin_question": "When did you first discover \"the stairs\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the stairs\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0274_01/00000.jpg", "scene0274_01/00040.jpg", "scene0274_01/00080.jpg", "scene0274_01/00120.jpg" ], "turn_id": 1, "scan_id": "scene0274_01", "origin_question": "Remember, have you seen any door(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any door(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0274_01/00160.jpg", "scene0274_01/00200.jpg", "scene0274_01/00240.jpg", "scene0274_01/00280.jpg" ], "turn_id": 2, "scan_id": "scene0274_01", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0274_01/00320.jpg", "scene0274_01/00360.jpg", "scene0274_01/00400.jpg", "scene0274_01/00440.jpg" ], "turn_id": 3, "scan_id": "scene0274_01", "origin_question": "At the end of which turn were both of \"the small rectangular green bin with a black lid\" and \"the rectangular white plastic telephone\" on your right side?", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the small rectangular green bin with a black lid\" and \"the rectangular white plastic telephone\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0274_01/00480.jpg", "scene0274_01/00520.jpg", "scene0274_01/00560.jpg", "scene0274_01/00600.jpg" ], "turn_id": 4, "scan_id": "scene0274_01", "origin_question": "Which two objects are on the same side of you now? \"the small rectangular green bin with a black lid\", \"the rectangular white plastic telephone\", \"the rectangular grey and red rubber eraser\". ", "option": [ "the small rectangular green bin with a black lid, the rectangular white plastic telephone", "the small rectangular green bin with a black lid, the rectangular grey and red rubber eraser", "the rectangular white plastic telephone, the rectangular grey and red rubber eraser" ], "answer": "the small rectangular green bin with a black lid, the rectangular white plastic telephone", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the small rectangular green bin with a black lid\", \"the rectangular white plastic telephone\", \"the rectangular grey and red rubber eraser\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the small rectangular green bin with a black lid, the rectangular white plastic telephone', 'the small rectangular green bin with a black lid, the rectangular grey and red rubber eraser', 'the rectangular white plastic telephone, the rectangular grey and red rubber eraser'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0274_01/00640.jpg", "scene0274_01/00680.jpg", "scene0274_01/00720.jpg", "scene0274_01/00760.jpg" ], "turn_id": 5, "scan_id": "scene0274_01", "origin_question": "In which turn did you see both \"the rectangular grey and red rubber eraser\" and \"the cylindrical white bottle with a green label\" simultaneously? ", "option": [], "answer": "4", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the rectangular grey and red rubber eraser\" and \"the cylindrical white bottle with a green label\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0274_01/00800.jpg", "scene0274_01/00840.jpg", "scene0274_01/00880.jpg", "scene0274_01/00920.jpg" ], "turn_id": 6, "scan_id": "scene0274_01", "origin_question": "Which direction is the small rectangular green bin with a black lid in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the small rectangular green bin with a black lid in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0274_01/00960.jpg", "scene0274_01/01000.jpg", "scene0274_01/01040.jpg", "scene0274_01/01080.jpg" ], "turn_id": 7, "scan_id": "scene0274_01", "origin_question": "What will be the first-time appearance order of the following categories: bin, chair, eraser?", "option": [ "eraser, chair, bin", "chair, eraser, bin", "bin, eraser, chair", "bin, chair, eraser" ], "answer": "bin, chair, eraser", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bin, chair, eraser? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['eraser, chair, bin', 'chair, eraser, bin', 'bin, eraser, chair', 'bin, chair, eraser'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0274_01/01120.jpg", "scene0274_01/01160.jpg", "scene0274_01/01200.jpg", "scene0274_01/01240.jpg" ], "turn_id": 8, "scan_id": "scene0274_01", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the cylindrical white bottle with a green label\" from your position?", "option": [], "answer": "73.53", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the cylindrical white bottle with a green label\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0274_01/01280.jpg", "scene0274_01/01320.jpg", "scene0274_01/01360.jpg", "scene0274_01/01400.jpg" ], "turn_id": 9, "scan_id": "scene0274_01", "origin_question": "Which direction is the rectangular white plastic telephone in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 9 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the rectangular white plastic telephone in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0131_01/00000.jpg", "scene0131_01/00040.jpg", "scene0131_01/00080.jpg", "scene0131_01/00120.jpg" ], "turn_id": 1, "scan_id": "scene0131_01", "origin_question": "Remember, have you seen any jar(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any jar(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0131_01/00160.jpg", "scene0131_01/00200.jpg", "scene0131_01/00240.jpg", "scene0131_01/00280.jpg" ], "turn_id": 2, "scan_id": "scene0131_01", "origin_question": "At the end of which turn were both of \"the round soccer ball with a black and white pattern\" and \"the medium-sized cylindrical light-colored tube\" on your left side?", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the round soccer ball with a black and white pattern\" and \"the medium-sized cylindrical light-colored tube\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0131_01/00320.jpg", "scene0131_01/00360.jpg", "scene0131_01/00400.jpg", "scene0131_01/00440.jpg" ], "turn_id": 3, "scan_id": "scene0131_01", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the compact vertical black stainless steel refrigerator\" from your position?", "option": [], "answer": "114.58", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the compact vertical black stainless steel refrigerator\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0131_01/00480.jpg", "scene0131_01/00520.jpg", "scene0131_01/00560.jpg", "scene0131_01/00600.jpg" ], "turn_id": 4, "scan_id": "scene0131_01", "origin_question": "Remember, have you seen any fan(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any fan(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "scene0131_01/00640.jpg", "scene0131_01/00680.jpg", "scene0131_01/00720.jpg", "scene0131_01/00760.jpg" ], "turn_id": 5, "scan_id": "scene0131_01", "origin_question": "Among these three objects(\"the round soccer ball with a black and white pattern\",\"the medium-sized cylindrical light-colored tube\",\"the white cylindrical cup\" ), which one is the closest to you?", "option": [ "the round soccer ball with a black and white pattern", "the medium-sized cylindrical light-colored tube", "the white cylindrical cup" ], "answer": "the white cylindrical cup", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects(\"the round soccer ball with a black and white pattern\",\"the medium-sized cylindrical light-colored tube\",\"the white cylindrical cup\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the round soccer ball with a black and white pattern', 'the medium-sized cylindrical light-colored tube', 'the white cylindrical cup'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0131_01/00800.jpg", "scene0131_01/00840.jpg", "scene0131_01/00880.jpg", "scene0131_01/00920.jpg" ], "turn_id": 6, "scan_id": "scene0131_01", "origin_question": "In which turn did you see both \"the modern rectangular orange couch with a metal frame\" and \"the rectangular wooden shelf with a simple design\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the modern rectangular orange couch with a metal frame\" and \"the rectangular wooden shelf with a simple design\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0131_01/00960.jpg", "scene0131_01/01000.jpg", "scene0131_01/01040.jpg", "scene0131_01/01080.jpg" ], "turn_id": 7, "scan_id": "scene0131_01", "origin_question": "Please recall \"the round black fan with a basic design\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.42", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the round black fan with a basic design\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0030_region16/164618f724c64b43934cb95b11c89593_i1_3.jpg", "1mp3d_0030_region16/164618f724c64b43934cb95b11c89593_i2_3.jpg", "1mp3d_0030_region16/d89bee33427d45a5bd0cd5e1b4f8dd34_i1_5.jpg", "1mp3d_0030_region16/d89bee33427d45a5bd0cd5e1b4f8dd34_i2_5.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0030_region16", "origin_question": "Remember, have you seen any chair(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any chair(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "1mp3d_0030_region16/d89bee33427d45a5bd0cd5e1b4f8dd34_i1_4.jpg", "1mp3d_0030_region16/d89bee33427d45a5bd0cd5e1b4f8dd34_i2_4.jpg", "1mp3d_0030_region16/164618f724c64b43934cb95b11c89593_i1_2.jpg", "1mp3d_0030_region16/164618f724c64b43934cb95b11c89593_i2_2.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0030_region16", "origin_question": "Is the vivid rectangular picture depicting a seascape to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the vivid rectangular picture depicting a seascape to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "1mp3d_0030_region16/e4753c26daf64c8492bb3fad5d959fb3_i1_2.jpg", "1mp3d_0030_region16/e4753c26daf64c8492bb3fad5d959fb3_i2_2.jpg", "1mp3d_0030_region16/76dc52af56e842c5b53f25d93e938299_i1_4.jpg", "1mp3d_0030_region16/76dc52af56e842c5b53f25d93e938299_i2_4.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0030_region16", "origin_question": "Which direction is the vivid rectangular picture depicting a seascape in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the vivid rectangular picture depicting a seascape in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "1mp3d_0030_region16/be1a38f2bec74de79395d0a938234d38_i1_0.jpg", "1mp3d_0030_region16/be1a38f2bec74de79395d0a938234d38_i2_0.jpg", "1mp3d_0030_region16/2b4c37304fd54440a3d1e7fd223e26d5_i1_0.jpg", "1mp3d_0030_region16/2b4c37304fd54440a3d1e7fd223e26d5_i2_0.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0030_region16", "origin_question": "Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Left", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "1mp3d_0030_region16/448d573c3160494c8715d3e7314f4afb_i1_5.jpg", "1mp3d_0030_region16/448d573c3160494c8715d3e7314f4afb_i2_5.jpg", "1mp3d_0030_region16/448d573c3160494c8715d3e7314f4afb_i1_1.jpg", "1mp3d_0030_region16/448d573c3160494c8715d3e7314f4afb_i2_1.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0030_region16", "origin_question": "In which turn did you see both \"the vivid rectangular picture depicting a seascape\" and \"the rectangular wooden cabinet with a natural gloss finish\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the vivid rectangular picture depicting a seascape\" and \"the rectangular wooden cabinet with a natural gloss finish\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "1mp3d_0030_region16/32c98b5cfe804dd185a8cb6ecb91a63f_i1_4.jpg", "1mp3d_0030_region16/32c98b5cfe804dd185a8cb6ecb91a63f_i2_4.jpg", "1mp3d_0030_region16/0fa0a4dc5aea4285b225cda527388ecb_i1_3.jpg", "1mp3d_0030_region16/0fa0a4dc5aea4285b225cda527388ecb_i2_3.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0030_region16", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the colorful woven carpet with geometric patterns\" from your position?", "option": [], "answer": "78.94", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the colorful woven carpet with geometric patterns\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "1mp3d_0030_region16/0fa0a4dc5aea4285b225cda527388ecb_i1_4.jpg", "1mp3d_0030_region16/0fa0a4dc5aea4285b225cda527388ecb_i2_4.jpg", "1mp3d_0030_region16/0fa0a4dc5aea4285b225cda527388ecb_i1_2.jpg", "1mp3d_0030_region16/0fa0a4dc5aea4285b225cda527388ecb_i2_2.jpg" ], "turn_id": 7, "scan_id": "1mp3d_0030_region16", "origin_question": "When did you first discover \"the rail\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the rail\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "1mp3d_0030_region16/32c98b5cfe804dd185a8cb6ecb91a63f_i1_1.jpg", "1mp3d_0030_region16/32c98b5cfe804dd185a8cb6ecb91a63f_i2_1.jpg", "1mp3d_0030_region16/2709f0b41bf14286b7f7c3a40b91d921_i1_5.jpg", "1mp3d_0030_region16/2709f0b41bf14286b7f7c3a40b91d921_i2_5.jpg" ], "turn_id": 8, "scan_id": "1mp3d_0030_region16", "origin_question": "How far is your current position from where you were at the end of the turn 3? (in meters)", "option": [], "answer": "1.93", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 3? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0353_00/00000.jpg", "scene0353_00/00040.jpg", "scene0353_00/00080.jpg", "scene0353_00/00120.jpg", "scene0353_00/00160.jpg" ], "turn_id": 1, "scan_id": "scene0353_00", "origin_question": "Remember, how many dresser(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many dresser(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0353_00/00200.jpg", "scene0353_00/00240.jpg", "scene0353_00/00280.jpg", "scene0353_00/00320.jpg", "scene0353_00/00360.jpg" ], "turn_id": 2, "scan_id": "scene0353_00", "origin_question": "Which two objects are on the same side of you now? \"the blackboard\", \"the bin\", \"the rectangular wall calendar with a white background\". ", "option": [ "the blackboard, the bin", "the blackboard, the rectangular wall calendar with a white background", "the bin, the rectangular wall calendar with a white background" ], "answer": "the blackboard, the bin", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the blackboard\", \"the bin\", \"the rectangular wall calendar with a white background\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the blackboard, the bin', 'the blackboard, the rectangular wall calendar with a white background', 'the bin, the rectangular wall calendar with a white background'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0353_00/00400.jpg", "scene0353_00/00440.jpg", "scene0353_00/00480.jpg", "scene0353_00/00520.jpg", "scene0353_00/00560.jpg" ], "turn_id": 3, "scan_id": "scene0353_00", "origin_question": "At the end of which turn were both of \"the blackboard\" and \"the rectangular wall calendar with a white background\" on your left side?", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the blackboard\" and \"the rectangular wall calendar with a white background\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0353_00/00600.jpg", "scene0353_00/00640.jpg", "scene0353_00/00680.jpg", "scene0353_00/00720.jpg", "scene0353_00/00760.jpg" ], "turn_id": 4, "scan_id": "scene0353_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "scene0353_00/00800.jpg", "scene0353_00/00840.jpg", "scene0353_00/00880.jpg", "scene0353_00/00920.jpg", "scene0353_00/00960.jpg" ], "turn_id": 5, "scan_id": "scene0353_00", "origin_question": "Among these three objects(\"the blackboard\",\"the bin\",\"the medium, soft, colorful jacket\" ), which one is the closest to you?", "option": [ "the blackboard", "the bin", "the medium, soft, colorful jacket" ], "answer": "the medium, soft, colorful jacket", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the blackboard\",\"the bin\",\"the medium, soft, colorful jacket\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the blackboard', 'the bin', 'the medium, soft, colorful jacket'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0353_00/01000.jpg", "scene0353_00/01040.jpg", "scene0353_00/01080.jpg", "scene0353_00/01120.jpg", "scene0353_00/01160.jpg" ], "turn_id": 6, "scan_id": "scene0353_00", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the rectangular canvas picture in a white frame\" from your position?", "option": [], "answer": "107.83", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the rectangular canvas picture in a white frame\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0353_00/01200.jpg", "scene0353_00/01240.jpg", "scene0353_00/01280.jpg", "scene0353_00/01320.jpg", "scene0353_00/01360.jpg" ], "turn_id": 7, "scan_id": "scene0353_00", "origin_question": "Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "76.35", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0353_00/01400.jpg", "scene0353_00/01440.jpg", "scene0353_00/01480.jpg", "scene0353_00/01520.jpg", "scene0353_00/01560.jpg" ], "turn_id": 8, "scan_id": "scene0353_00", "origin_question": "Which direction is the rectangular wall calendar with a white background in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the rectangular wall calendar with a white background in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0353_00/01600.jpg", "scene0353_00/01640.jpg", "scene0353_00/01680.jpg", "scene0353_00/01720.jpg", "scene0353_00/01760.jpg" ], "turn_id": 9, "scan_id": "scene0353_00", "origin_question": "Please recall \"the curtain\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.92", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the curtain\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0353_00/01800.jpg", "scene0353_00/01840.jpg", "scene0353_00/01880.jpg", "scene0353_00/01920.jpg", "scene0353_00/01960.jpg" ], "turn_id": 10, "scan_id": "scene0353_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0025_00/00000.jpg", "scene0025_00/00040.jpg", "scene0025_00/00080.jpg", "scene0025_00/00120.jpg", "scene0025_00/00160.jpg" ], "turn_id": 1, "scan_id": "scene0025_00", "origin_question": "Remember, have you seen any basket(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any basket(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0025_00/00200.jpg", "scene0025_00/00240.jpg", "scene0025_00/00280.jpg", "scene0025_00/00320.jpg", "scene0025_00/00360.jpg" ], "turn_id": 2, "scan_id": "scene0025_00", "origin_question": "In which turn did you see both \"the standard white keyboard made of plastic\" and \"the rectangular blackboard made of plastic\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the standard white keyboard made of plastic\" and \"the rectangular blackboard made of plastic\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0025_00/00400.jpg", "scene0025_00/00440.jpg", "scene0025_00/00480.jpg", "scene0025_00/00520.jpg", "scene0025_00/00560.jpg" ], "turn_id": 3, "scan_id": "scene0025_00", "origin_question": "Which direction is the standard white keyboard made of plastic in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the standard white keyboard made of plastic in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0025_00/00600.jpg", "scene0025_00/00640.jpg", "scene0025_00/00680.jpg", "scene0025_00/00720.jpg", "scene0025_00/00760.jpg" ], "turn_id": 4, "scan_id": "scene0025_00", "origin_question": "Please recall \"the cylindrical plastic bottle with a red cap\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.29", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the cylindrical plastic bottle with a red cap\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0025_00/00800.jpg", "scene0025_00/00840.jpg", "scene0025_00/00880.jpg", "scene0025_00/00920.jpg", "scene0025_00/00960.jpg" ], "turn_id": 5, "scan_id": "scene0025_00", "origin_question": "What will be the first-time appearance order of the following categories: desk, door, fan?", "option": [ "fan, desk, door", "desk, fan, door", "desk, door, fan", "door, fan, desk" ], "answer": "desk, door, fan", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: desk, door, fan? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['fan, desk, door', 'desk, fan, door', 'desk, door, fan', 'door, fan, desk'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0025_00/01000.jpg", "scene0025_00/01040.jpg", "scene0025_00/01080.jpg", "scene0025_00/01120.jpg", "scene0025_00/01160.jpg" ], "turn_id": 6, "scan_id": "scene0025_00", "origin_question": "Which direction is the rectangular, striped cotton pillow in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the rectangular, striped cotton pillow in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0025_00/01200.jpg", "scene0025_00/01240.jpg", "scene0025_00/01280.jpg", "scene0025_00/01320.jpg", "scene0025_00/01360.jpg" ], "turn_id": 7, "scan_id": "scene0025_00", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the rectangular, striped cotton pillow\" from your position?", "option": [], "answer": "109.66", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the rectangular, striped cotton pillow\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0025_00/01400.jpg", "scene0025_00/01440.jpg", "scene0025_00/01480.jpg", "scene0025_00/01520.jpg", "scene0025_00/01560.jpg" ], "turn_id": 8, "scan_id": "scene0025_00", "origin_question": "Which two objects are on the same side of you now? \"the blinds\", \"the cylindrical metal pen with a green cap\", \"the rectangular, striped cotton pillow\". ", "option": [ "the blinds, the cylindrical metal pen with a green cap", "the blinds, the rectangular, striped cotton pillow", "the cylindrical metal pen with a green cap, the rectangular, striped cotton pillow" ], "answer": "the blinds, the cylindrical metal pen with a green cap", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the blinds\", \"the cylindrical metal pen with a green cap\", \"the rectangular, striped cotton pillow\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the blinds, the cylindrical metal pen with a green cap', 'the blinds, the rectangular, striped cotton pillow', 'the cylindrical metal pen with a green cap, the rectangular, striped cotton pillow'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0025_00/01600.jpg", "scene0025_00/01640.jpg", "scene0025_00/01680.jpg", "scene0025_00/01720.jpg", "scene0025_00/01760.jpg" ], "turn_id": 9, "scan_id": "scene0025_00", "origin_question": "Using your orientation at the end of turn 8 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "156.48", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 8 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "1mp3d_0072_region5/11df17e75c974cec81b1b76aeab73399_i1_3.jpg", "1mp3d_0072_region5/11df17e75c974cec81b1b76aeab73399_i2_3.jpg", "1mp3d_0072_region5/9f14e40b3bca40c292a9c10af54164ff_i1_5.jpg", "1mp3d_0072_region5/9f14e40b3bca40c292a9c10af54164ff_i2_5.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0072_region5", "origin_question": "Remember, how many chair(s) have you seen so far? ", "option": [], "answer": "4", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many chair(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "1mp3d_0072_region5/8064037d7d254b2a9802647b9a31cfcb_i1_2.jpg", "1mp3d_0072_region5/8064037d7d254b2a9802647b9a31cfcb_i2_2.jpg", "1mp3d_0072_region5/48a174fdebfc429c9d5cca4e74d7fbdd_i1_1.jpg", "1mp3d_0072_region5/48a174fdebfc429c9d5cca4e74d7fbdd_i2_1.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0072_region5", "origin_question": "Which direction is the rectangular dark wooden footstool in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the rectangular dark wooden footstool in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "1mp3d_0072_region5/48a174fdebfc429c9d5cca4e74d7fbdd_i1_0.jpg", "1mp3d_0072_region5/48a174fdebfc429c9d5cca4e74d7fbdd_i2_0.jpg", "1mp3d_0072_region5/8064037d7d254b2a9802647b9a31cfcb_i1_0.jpg", "1mp3d_0072_region5/8064037d7d254b2a9802647b9a31cfcb_i2_0.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0072_region5", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular white couch with smooth fabric\", Y: \"the bin\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to X and farther away from Y", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular white couch with smooth fabric\", Y: \"the bin\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0072_region5/4cc573783ab84a5f9f191ce255a13b97_i1_2.jpg", "1mp3d_0072_region5/4cc573783ab84a5f9f191ce255a13b97_i2_2.jpg", "1mp3d_0072_region5/33296641c793430a893601e4ecbddcba_i1_1.jpg", "1mp3d_0072_region5/33296641c793430a893601e4ecbddcba_i2_1.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0072_region5", "origin_question": "", "option": [], "answer": "", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "1mp3d_0072_region5/33296641c793430a893601e4ecbddcba_i1_2.jpg", "1mp3d_0072_region5/33296641c793430a893601e4ecbddcba_i2_2.jpg", "1mp3d_0072_region5/33296641c793430a893601e4ecbddcba_i1_3.jpg", "1mp3d_0072_region5/33296641c793430a893601e4ecbddcba_i2_3.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0072_region5", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the rectangular black box with a smooth exterior\" from your position?", "option": [], "answer": "165.13", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the rectangular black box with a smooth exterior\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "1mp3d_0072_region5/33296641c793430a893601e4ecbddcba_i1_4.jpg", "1mp3d_0072_region5/33296641c793430a893601e4ecbddcba_i2_4.jpg", "1mp3d_0072_region5/33296641c793430a893601e4ecbddcba_i1_5.jpg", "1mp3d_0072_region5/33296641c793430a893601e4ecbddcba_i2_5.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0072_region5", "origin_question": "Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Left", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0006_02/00000.jpg", "scene0006_02/00040.jpg", "scene0006_02/00080.jpg", "scene0006_02/00120.jpg", "scene0006_02/00160.jpg" ], "turn_id": 1, "scan_id": "scene0006_02", "origin_question": "Remember, how many picture(s) have you seen so far? ", "option": [], "answer": "4", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many picture(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0006_02/00200.jpg", "scene0006_02/00240.jpg", "scene0006_02/00280.jpg", "scene0006_02/00320.jpg", "scene0006_02/00360.jpg" ], "turn_id": 2, "scan_id": "scene0006_02", "origin_question": "At the end of which turn were both of \"the dark round wooden stand with a smooth surface\" and \"the cylindrical white lamp made of plastic and glass\" on your right side?", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the dark round wooden stand with a smooth surface\" and \"the cylindrical white lamp made of plastic and glass\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0006_02/00400.jpg", "scene0006_02/00440.jpg", "scene0006_02/00480.jpg", "scene0006_02/00520.jpg", "scene0006_02/00560.jpg" ], "turn_id": 3, "scan_id": "scene0006_02", "origin_question": "In which turn did you see both \"the rectangular, dark wooden table suitable for writing\" and \"the rectangular black or dark grey telephone\" simultaneously? ", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the rectangular, dark wooden table suitable for writing\" and \"the rectangular black or dark grey telephone\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0006_02/00600.jpg", "scene0006_02/00640.jpg", "scene0006_02/00680.jpg", "scene0006_02/00720.jpg", "scene0006_02/00760.jpg" ], "turn_id": 4, "scan_id": "scene0006_02", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the flat-panel TV with a dark matte surface\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the flat-panel TV with a dark matte surface\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0006_02/00800.jpg", "scene0006_02/00840.jpg", "scene0006_02/00880.jpg", "scene0006_02/00920.jpg", "scene0006_02/00960.jpg" ], "turn_id": 5, "scan_id": "scene0006_02", "origin_question": "Is the cylindrical dark bin made of sturdy material to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the cylindrical dark bin made of sturdy material to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0006_02/01000.jpg", "scene0006_02/01040.jpg", "scene0006_02/01080.jpg", "scene0006_02/01120.jpg", "scene0006_02/01160.jpg" ], "turn_id": 6, "scan_id": "scene0006_02", "origin_question": "Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "103.85", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0006_02/01200.jpg", "scene0006_02/01240.jpg", "scene0006_02/01280.jpg", "scene0006_02/01320.jpg", "scene0006_02/01360.jpg" ], "turn_id": 7, "scan_id": "scene0006_02", "origin_question": "What will be the first-time appearance order of the following categories: backpack, door, table?", "option": [ "table, door, backpack", "backpack, table, door", "table, backpack, door", "backpack, door, table" ], "answer": "backpack, door, table", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: backpack, door, table? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['table, door, backpack', 'backpack, table, door', 'table, backpack, door', 'backpack, door, table'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0006_02/01400.jpg", "scene0006_02/01440.jpg", "scene0006_02/01480.jpg", "scene0006_02/01520.jpg", "scene0006_02/01560.jpg" ], "turn_id": 8, "scan_id": "scene0006_02", "origin_question": "By the end of the 3 turn, were you getting closer to or farther away from these two objects: X: \"the small cylindrical metal bottle with a checkered pattern\", Y: \"the black leather rectangular shoe of medium size\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 3 turn, were you getting closer to or farther away from these two objects: X: \"the small cylindrical metal bottle with a checkered pattern\", Y: \"the black leather rectangular shoe of medium size\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "scene0006_02/01600.jpg", "scene0006_02/01640.jpg", "scene0006_02/01680.jpg", "scene0006_02/01720.jpg", "scene0006_02/01760.jpg" ], "turn_id": 9, "scan_id": "scene0006_02", "origin_question": "Among these three objects(\"the flat-panel TV with a dark matte surface\",\"the rectangular black or dark grey telephone\",\"the small cylindrical metal bottle with a checkered pattern\" ), which one is the closest to you?", "option": [ "the flat-panel TV with a dark matte surface", "the rectangular black or dark grey telephone", "the small cylindrical metal bottle with a checkered pattern" ], "answer": "the flat-panel TV with a dark matte surface", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the flat-panel TV with a dark matte surface\",\"the rectangular black or dark grey telephone\",\"the small cylindrical metal bottle with a checkered pattern\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the flat-panel TV with a dark matte surface', 'the rectangular black or dark grey telephone', 'the small cylindrical metal bottle with a checkered pattern'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0006_02/01800.jpg", "scene0006_02/01840.jpg", "scene0006_02/01880.jpg", "scene0006_02/01920.jpg", "scene0006_02/01960.jpg" ], "turn_id": 10, "scan_id": "scene0006_02", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the small cylindrical metal bottle with a checkered pattern\" from your position?", "option": [], "answer": "161.29", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the small cylindrical metal bottle with a checkered pattern\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0082_00/00000.jpg", "scene0082_00/00030.jpg", "scene0082_00/00060.jpg", "scene0082_00/00090.jpg" ], "turn_id": 1, "scan_id": "scene0082_00", "origin_question": "Remember, have you seen any chair(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any chair(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 3 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0082_00/00120.jpg", "scene0082_00/00150.jpg", "scene0082_00/00180.jpg", "scene0082_00/00210.jpg" ], "turn_id": 2, "scan_id": "scene0082_00", "origin_question": "Remember, how many hamper(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many hamper(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 3 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0082_00/00240.jpg", "scene0082_00/00270.jpg", "scene0082_00/00300.jpg", "scene0082_00/00330.jpg" ], "turn_id": 3, "scan_id": "scene0082_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 3 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0044_region30/5b797c9f28c44c4585f07395b16ff9a5_i1_1.jpg", "1mp3d_0044_region30/5b797c9f28c44c4585f07395b16ff9a5_i2_1.jpg", "1mp3d_0044_region30/5b797c9f28c44c4585f07395b16ff9a5_i1_5.jpg", "1mp3d_0044_region30/5b797c9f28c44c4585f07395b16ff9a5_i2_5.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0044_region30", "origin_question": "Remember, have you seen any headphones(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any headphones(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0044_region30/7dbc584f79f44cf586ffc70dc3f2ff2a_i1_0.jpg", "1mp3d_0044_region30/7dbc584f79f44cf586ffc70dc3f2ff2a_i2_0.jpg", "1mp3d_0044_region30/74d3b3736c6047a6a904bc68c6aa9f36_i1_0.jpg", "1mp3d_0044_region30/74d3b3736c6047a6a904bc68c6aa9f36_i2_0.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0044_region30", "origin_question": "Remember, have you seen any light(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any light(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "1mp3d_0044_region30/7dbc584f79f44cf586ffc70dc3f2ff2a_i2_3.jpg", "1mp3d_0044_region30/74d3b3736c6047a6a904bc68c6aa9f36_i1_1.jpg", "1mp3d_0044_region30/74d3b3736c6047a6a904bc68c6aa9f36_i2_1.jpg", "1mp3d_0044_region30/9dbe2b1ec9484d4da2cd10f502449c72_i1_5.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0044_region30", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the tall, glossy black wooden door\", Y: \"the small light fixture with a cylindrical white lampshade\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "farther away from both X and Y", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the tall, glossy black wooden door\", Y: \"the small light fixture with a cylindrical white lampshade\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "1mp3d_0044_region30/9dbe2b1ec9484d4da2cd10f502449c72_i2_5.jpg", "1mp3d_0044_region30/978ffb12f601402c894982880f8eefa6_i1_3.jpg", "1mp3d_0044_region30/978ffb12f601402c894982880f8eefa6_i2_3.jpg", "1mp3d_0044_region30/b1eb72881a38479a9e5ea4dac5b415f6_i1_5.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0044_region30", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the tall, elegant yellow vase\" from your position?", "option": [], "answer": "79.32", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the tall, elegant yellow vase\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "1mp3d_0044_region30/b1eb72881a38479a9e5ea4dac5b415f6_i2_5.jpg", "1mp3d_0044_region30/b1b145e1a343455fb1529d804e2cd36b_i1_2.jpg", "1mp3d_0044_region30/b1b145e1a343455fb1529d804e2cd36b_i2_2.jpg", "1mp3d_0044_region30/b1eb72881a38479a9e5ea4dac5b415f6_i1_0.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0044_region30", "origin_question": "How far is your current position from where you were at the end of the turn 3? (in meters)", "option": [], "answer": "2.84", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 3? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "1mp3d_0044_region30/b1eb72881a38479a9e5ea4dac5b415f6_i2_0.jpg", "1mp3d_0044_region30/b1eb72881a38479a9e5ea4dac5b415f6_i1_1.jpg", "1mp3d_0044_region30/b1eb72881a38479a9e5ea4dac5b415f6_i2_1.jpg", "1mp3d_0044_region30/b1eb72881a38479a9e5ea4dac5b415f6_i1_2.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0044_region30", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "2.99", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "1mp3d_0044_region30/b1eb72881a38479a9e5ea4dac5b415f6_i2_2.jpg", "1mp3d_0044_region30/b1b145e1a343455fb1529d804e2cd36b_i1_0.jpg", "1mp3d_0044_region30/b1b145e1a343455fb1529d804e2cd36b_i2_0.jpg", "1mp3d_0044_region30/b1b145e1a343455fb1529d804e2cd36b_i1_1.jpg" ], "turn_id": 7, "scan_id": "1mp3d_0044_region30", "origin_question": "How far is your current position from where you were at the end of the turn 6? (in meters)", "option": [], "answer": "2.15", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 6? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0044_region30/b1b145e1a343455fb1529d804e2cd36b_i2_1.jpg", "1mp3d_0044_region30/bff9c16c112048898cedad2b5d3cbe32_i1_2.jpg", "1mp3d_0044_region30/bff9c16c112048898cedad2b5d3cbe32_i2_2.jpg", "1mp3d_0044_region30/bff9c16c112048898cedad2b5d3cbe32_i1_1.jpg" ], "turn_id": 8, "scan_id": "1mp3d_0044_region30", "origin_question": "Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the large rectangular white projection screen\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the large rectangular white projection screen\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0078_region0/37165e7f48cc40bd864b8f8817a04c63_i1_0.jpg", "1mp3d_0078_region0/37165e7f48cc40bd864b8f8817a04c63_i2_0.jpg", "1mp3d_0078_region0/dc00a03cb9f847fea498d765a31ec2b8_i1_0.jpg", "1mp3d_0078_region0/dc00a03cb9f847fea498d765a31ec2b8_i2_0.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0078_region0", "origin_question": "Remember, have you seen any dresser(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any dresser(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "1mp3d_0078_region0/f8f72f32f6594620b17161a621b9c872_i1_3.jpg", "1mp3d_0078_region0/f8f72f32f6594620b17161a621b9c872_i2_3.jpg", "1mp3d_0078_region0/f8f72f32f6594620b17161a621b9c872_i1_4.jpg", "1mp3d_0078_region0/f8f72f32f6594620b17161a621b9c872_i2_4.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0078_region0", "origin_question": "When did you first discover \"the square full-size bed with a dark wooden frame\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the square full-size bed with a dark wooden frame\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0078_region0/b46b1193e47449ffb1e6977555cb7222_i1_3.jpg", "1mp3d_0078_region0/b46b1193e47449ffb1e6977555cb7222_i2_3.jpg", "1mp3d_0078_region0/f8f72f32f6594620b17161a621b9c872_i1_5.jpg", "1mp3d_0078_region0/f8f72f32f6594620b17161a621b9c872_i2_5.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0078_region0", "origin_question": "Remember, have you seen any door(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any door(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "1mp3d_0078_region0/dc00a03cb9f847fea498d765a31ec2b8_i1_5.jpg", "1mp3d_0078_region0/dc00a03cb9f847fea498d765a31ec2b8_i2_5.jpg", "1mp3d_0078_region0/37165e7f48cc40bd864b8f8817a04c63_i1_5.jpg", "1mp3d_0078_region0/37165e7f48cc40bd864b8f8817a04c63_i2_5.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0078_region0", "origin_question": "When did you first discover \"the natural light brown rectangular wooden door\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the natural light brown rectangular wooden door\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "1mp3d_0078_region0/7f8730408c084c1da6a42059767adee3_i1_5.jpg", "1mp3d_0078_region0/7f8730408c084c1da6a42059767adee3_i2_5.jpg", "1mp3d_0078_region0/7f8730408c084c1da6a42059767adee3_i1_4.jpg", "1mp3d_0078_region0/7f8730408c084c1da6a42059767adee3_i2_4.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0078_region0", "origin_question": "By the end of the 3 turn, were you getting closer to or farther away from these two objects: X: \"the square full-size bed with a dark wooden frame\", Y: \"the natural light brown rectangular wooden door\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to Y and farther away from X", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 3 turn, were you getting closer to or farther away from these two objects: X: \"the square full-size bed with a dark wooden frame\", Y: \"the natural light brown rectangular wooden door\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0078_region0/d64f111089894930b35dd4663eb9aa34_i1_3.jpg", "1mp3d_0078_region0/d64f111089894930b35dd4663eb9aa34_i2_3.jpg", "1mp3d_0078_region0/39298a3e5bff4258ad4e8dc6031f4f1a_i1_5.jpg", "1mp3d_0078_region0/39298a3e5bff4258ad4e8dc6031f4f1a_i2_5.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0078_region0", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the natural light brown rectangular wooden door\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the natural light brown rectangular wooden door\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0064_01/00000.jpg", "scene0064_01/00030.jpg", "scene0064_01/00060.jpg", "scene0064_01/00090.jpg" ], "turn_id": 1, "scan_id": "scene0064_01", "origin_question": "Remember, have you seen any range hood(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any range hood(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0064_01/00120.jpg", "scene0064_01/00150.jpg", "scene0064_01/00180.jpg", "scene0064_01/00210.jpg" ], "turn_id": 2, "scan_id": "scene0064_01", "origin_question": "In which turn were you farthest from \"the medium-sized red and black rectangular backpack\"? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the medium-sized red and black rectangular backpack\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "scene0064_01/00240.jpg", "scene0064_01/00270.jpg", "scene0064_01/00300.jpg", "scene0064_01/00330.jpg" ], "turn_id": 3, "scan_id": "scene0064_01", "origin_question": "Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Left", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0064_01/00360.jpg", "scene0064_01/00390.jpg", "scene0064_01/00420.jpg", "scene0064_01/00450.jpg" ], "turn_id": 4, "scan_id": "scene0064_01", "origin_question": "How far is your current position from where you were at the end of the turn 1? (in meters)", "option": [], "answer": "3.29", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 1? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0064_01/00480.jpg", "scene0064_01/00510.jpg", "scene0064_01/00540.jpg", "scene0064_01/00570.jpg" ], "turn_id": 5, "scan_id": "scene0064_01", "origin_question": "Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the medium-sized red and black rectangular backpack\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the medium-sized red and black rectangular backpack\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0064_01/00600.jpg", "scene0064_01/00630.jpg", "scene0064_01/00660.jpg", "scene0064_01/00690.jpg" ], "turn_id": 6, "scan_id": "scene0064_01", "origin_question": "When did you first discover \"the door\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the door\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0392_01/00000.jpg", "scene0392_01/00040.jpg", "scene0392_01/00080.jpg", "scene0392_01/00120.jpg", "scene0392_01/00160.jpg" ], "turn_id": 1, "scan_id": "scene0392_01", "origin_question": "Remember, have you seen any laptop(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any laptop(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0392_01/00200.jpg", "scene0392_01/00240.jpg", "scene0392_01/00280.jpg", "scene0392_01/00320.jpg", "scene0392_01/00360.jpg" ], "turn_id": 2, "scan_id": "scene0392_01", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the rectangular light-colored wooden desk with drawers\"; \"the sleek black modern chair with a curved backrest\"; \"the tall rectangular mirror with a reflective glass surface\" ", "option": [ "the rectangular light-colored wooden desk with drawers", "the sleek black modern chair with a curved backrest", "the tall rectangular mirror with a reflective glass surface" ], "answer": "the tall rectangular mirror with a reflective glass surface", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the rectangular light-colored wooden desk with drawers\"; \"the sleek black modern chair with a curved backrest\"; \"the tall rectangular mirror with a reflective glass surface\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular light-colored wooden desk with drawers', 'the sleek black modern chair with a curved backrest', 'the tall rectangular mirror with a reflective glass surface'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0392_01/00400.jpg", "scene0392_01/00440.jpg", "scene0392_01/00480.jpg", "scene0392_01/00520.jpg", "scene0392_01/00560.jpg" ], "turn_id": 3, "scan_id": "scene0392_01", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "110.19", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0392_01/00600.jpg", "scene0392_01/00640.jpg", "scene0392_01/00680.jpg", "scene0392_01/00720.jpg", "scene0392_01/00760.jpg" ], "turn_id": 4, "scan_id": "scene0392_01", "origin_question": "Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "157.74", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0392_01/00800.jpg", "scene0392_01/00840.jpg", "scene0392_01/00880.jpg", "scene0392_01/00920.jpg", "scene0392_01/00960.jpg" ], "turn_id": 5, "scan_id": "scene0392_01", "origin_question": "Which direction is the radiator in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the radiator in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0392_01/01000.jpg", "scene0392_01/01040.jpg", "scene0392_01/01080.jpg", "scene0392_01/01120.jpg", "scene0392_01/01160.jpg" ], "turn_id": 6, "scan_id": "scene0392_01", "origin_question": "Which two objects are on the same side of you now? \"the bed\", \"the tall rectangular mirror with a reflective glass surface\", \"the radiator\". ", "option": [ "the bed, the tall rectangular mirror with a reflective glass surface", "the bed, the radiator", "the tall rectangular mirror with a reflective glass surface, the radiator" ], "answer": "the bed, the tall rectangular mirror with a reflective glass surface", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the bed\", \"the tall rectangular mirror with a reflective glass surface\", \"the radiator\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the bed, the tall rectangular mirror with a reflective glass surface', 'the bed, the radiator', 'the tall rectangular mirror with a reflective glass surface, the radiator'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0392_01/01200.jpg", "scene0392_01/01240.jpg", "scene0392_01/01280.jpg", "scene0392_01/01320.jpg", "scene0392_01/01380.jpg" ], "turn_id": 7, "scan_id": "scene0392_01", "origin_question": "Please recall \"the tall rectangular mirror with a reflective glass surface\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.43", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the tall rectangular mirror with a reflective glass surface\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0392_01/01420.jpg", "scene0392_01/01460.jpg", "scene0392_01/01500.jpg", "scene0392_01/01540.jpg", "scene0392_01/01580.jpg" ], "turn_id": 8, "scan_id": "scene0392_01", "origin_question": "Is the light brown wooden rectangular dresser with five drawers to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the light brown wooden rectangular dresser with five drawers to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0392_01/01620.jpg", "scene0392_01/01660.jpg", "scene0392_01/01700.jpg", "scene0392_01/01740.jpg", "scene0392_01/01780.jpg" ], "turn_id": 9, "scan_id": "scene0392_01", "origin_question": "Using your orientation at the end of turn 6 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 6 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0392_01/01820.jpg", "scene0392_01/01860.jpg", "scene0392_01/01900.jpg", "scene0392_01/01940.jpg", "scene0392_01/01980.jpg" ], "turn_id": 10, "scan_id": "scene0392_01", "origin_question": "Which direction is the radiator in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the radiator in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0100_01/00000.jpg", "scene0100_01/00040.jpg", "scene0100_01/00080.jpg", "scene0100_01/00120.jpg" ], "turn_id": 1, "scan_id": "scene0100_01", "origin_question": "Remember, have you seen any bathtub(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any bathtub(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0100_01/00160.jpg", "scene0100_01/00200.jpg", "scene0100_01/00240.jpg", "scene0100_01/00280.jpg" ], "turn_id": 2, "scan_id": "scene0100_01", "origin_question": "In which turn were you closest to \"the standard white ceramic toilet with a closed seat\"? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the standard white ceramic toilet with a closed seat\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0100_01/00320.jpg", "scene0100_01/00360.jpg", "scene0100_01/00400.jpg", "scene0100_01/00440.jpg" ], "turn_id": 3, "scan_id": "scene0100_01", "origin_question": "At the end of which turn were both of \"the roll of white, soft, absorbent toilet paper\" and \"the deep blue rectangular mat of soft material\" on your left side?", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the roll of white, soft, absorbent toilet paper\" and \"the deep blue rectangular mat of soft material\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0100_01/00480.jpg", "scene0100_01/00520.jpg", "scene0100_01/00560.jpg", "scene0100_01/00600.jpg" ], "turn_id": 4, "scan_id": "scene0100_01", "origin_question": "Please recall \"the circular, metallic single-handle switch\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.91", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the circular, metallic single-handle switch\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0100_01/00640.jpg", "scene0100_01/00680.jpg", "scene0100_01/00720.jpg", "scene0100_01/00760.jpg" ], "turn_id": 5, "scan_id": "scene0100_01", "origin_question": "Is the cylindrical, white plastic bin to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the cylindrical, white plastic bin to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0100_01/00800.jpg", "scene0100_01/00840.jpg", "scene0100_01/00880.jpg", "scene0100_01/00920.jpg" ], "turn_id": 6, "scan_id": "scene0100_01", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the circular, metallic single-handle switch\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the circular, metallic single-handle switch\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0100_01/00960.jpg", "scene0100_01/01000.jpg", "scene0100_01/01040.jpg", "scene0100_01/01080.jpg" ], "turn_id": 7, "scan_id": "scene0100_01", "origin_question": "Which two objects are on the same side of you now? \"the roll of white, soft, absorbent toilet paper\", \"the bathtub\", \"the round, light brown ceramic sink\". ", "option": [ "the roll of white, soft, absorbent toilet paper, the bathtub", "the roll of white, soft, absorbent toilet paper, the round, light brown ceramic sink", "the bathtub, the round, light brown ceramic sink" ], "answer": "the roll of white, soft, absorbent toilet paper, the bathtub", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the roll of white, soft, absorbent toilet paper\", \"the bathtub\", \"the round, light brown ceramic sink\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the roll of white, soft, absorbent toilet paper, the bathtub', 'the roll of white, soft, absorbent toilet paper, the round, light brown ceramic sink', 'the bathtub, the round, light brown ceramic sink'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0701_00/00000.jpg", "scene0701_00/00040.jpg", "scene0701_00/00080.jpg", "scene0701_00/00120.jpg" ], "turn_id": 1, "scan_id": "scene0701_00", "origin_question": "Remember, have you seen any table(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any table(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0701_00/00160.jpg", "scene0701_00/00200.jpg", "scene0701_00/00240.jpg", "scene0701_00/00280.jpg" ], "turn_id": 2, "scan_id": "scene0701_00", "origin_question": "Remember, how many couch(s) have you seen so far? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many couch(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0701_00/00320.jpg", "scene0701_00/00360.jpg", "scene0701_00/00400.jpg", "scene0701_00/00440.jpg" ], "turn_id": 3, "scan_id": "scene0701_00", "origin_question": "When did you first discover \"the bright red medium-sized backpack with black straps\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the bright red medium-sized backpack with black straps\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0701_00/00480.jpg", "scene0701_00/00520.jpg", "scene0701_00/00560.jpg", "scene0701_00/00600.jpg" ], "turn_id": 4, "scan_id": "scene0701_00", "origin_question": "Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "43.49", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0701_00/00640.jpg", "scene0701_00/00680.jpg", "scene0701_00/00720.jpg", "scene0701_00/00760.jpg" ], "turn_id": 5, "scan_id": "scene0701_00", "origin_question": "Using your orientation at the end of turn 4 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "80.33", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 4 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0701_00/00800.jpg", "scene0701_00/00840.jpg", "scene0701_00/00880.jpg", "scene0701_00/00920.jpg" ], "turn_id": 6, "scan_id": "scene0701_00", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the round brown leather table with a smooth surface\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the round brown leather table with a smooth surface\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0067_region30/527045377f974a7d8853d919218b98bd_i1_4.jpg", "1mp3d_0067_region30/527045377f974a7d8853d919218b98bd_i2_4.jpg", "1mp3d_0067_region30/92d6818b71734d2c9267bc430f4f6490_i1_2.jpg", "1mp3d_0067_region30/92d6818b71734d2c9267bc430f4f6490_i2_2.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0067_region30", "origin_question": "Remember, have you seen any picture(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any picture(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "1mp3d_0067_region30/97dfbaecdc3f417385083e8302b2b694_i1_1.jpg", "1mp3d_0067_region30/97dfbaecdc3f417385083e8302b2b694_i2_1.jpg", "1mp3d_0067_region30/0cf943705a6f4cf798dbcc605c811a50_i1_1.jpg", "1mp3d_0067_region30/0cf943705a6f4cf798dbcc605c811a50_i2_1.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0067_region30", "origin_question": "When did you first discover \"the set of three rectangular sliding doors\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the set of three rectangular sliding doors\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "1mp3d_0067_region30/735cc8eae07743a1aa5d41ba878b0eb4_i1_2.jpg", "1mp3d_0067_region30/735cc8eae07743a1aa5d41ba878b0eb4_i2_2.jpg", "1mp3d_0067_region30/735cc8eae07743a1aa5d41ba878b0eb4_i1_3.jpg", "1mp3d_0067_region30/735cc8eae07743a1aa5d41ba878b0eb4_i2_3.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0067_region30", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "1mp3d_0067_region30/0cf943705a6f4cf798dbcc605c811a50_i1_3.jpg", "1mp3d_0067_region30/0cf943705a6f4cf798dbcc605c811a50_i2_3.jpg", "1mp3d_0067_region30/97dfbaecdc3f417385083e8302b2b694_i1_3.jpg", "1mp3d_0067_region30/97dfbaecdc3f417385083e8302b2b694_i2_3.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0067_region30", "origin_question": "Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "147.54", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "1mp3d_0067_region30/92d6818b71734d2c9267bc430f4f6490_i1_3.jpg", "1mp3d_0067_region30/92d6818b71734d2c9267bc430f4f6490_i2_3.jpg", "1mp3d_0067_region30/527045377f974a7d8853d919218b98bd_i1_5.jpg", "1mp3d_0067_region30/527045377f974a7d8853d919218b98bd_i2_5.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0067_region30", "origin_question": "Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "86.6", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "1mp3d_0067_region30/527045377f974a7d8853d919218b98bd_i1_0.jpg", "1mp3d_0067_region30/527045377f974a7d8853d919218b98bd_i2_0.jpg", "1mp3d_0067_region30/92d6818b71734d2c9267bc430f4f6490_i1_0.jpg", "1mp3d_0067_region30/92d6818b71734d2c9267bc430f4f6490_i2_0.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0067_region30", "origin_question": "Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "1mp3d_0067_region30/92d6818b71734d2c9267bc430f4f6490_i1_1.jpg", "1mp3d_0067_region30/92d6818b71734d2c9267bc430f4f6490_i2_1.jpg", "1mp3d_0067_region30/97dfbaecdc3f417385083e8302b2b694_i1_4.jpg", "1mp3d_0067_region30/97dfbaecdc3f417385083e8302b2b694_i2_4.jpg" ], "turn_id": 7, "scan_id": "1mp3d_0067_region30", "origin_question": "Using your orientation at the end of turn 6 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "32.83", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 6 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0067_region30/0cf943705a6f4cf798dbcc605c811a50_i1_4.jpg", "1mp3d_0067_region30/0cf943705a6f4cf798dbcc605c811a50_i2_4.jpg", "1mp3d_0067_region30/735cc8eae07743a1aa5d41ba878b0eb4_i1_4.jpg", "1mp3d_0067_region30/735cc8eae07743a1aa5d41ba878b0eb4_i2_4.jpg" ], "turn_id": 8, "scan_id": "1mp3d_0067_region30", "origin_question": "Compared to your position at the end of the 7 turn, are you now closer or farther away from \"the set of three rectangular sliding doors\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 7 turn, are you now closer or farther away from \"the set of three rectangular sliding doors\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0513_00/00000.jpg", "scene0513_00/00040.jpg", "scene0513_00/00080.jpg", "scene0513_00/00120.jpg", "scene0513_00/00160.jpg" ], "turn_id": 1, "scan_id": "scene0513_00", "origin_question": "Remember, have you seen any window(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any window(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0513_00/00200.jpg", "scene0513_00/00240.jpg", "scene0513_00/00280.jpg", "scene0513_00/00320.jpg", "scene0513_00/00360.jpg" ], "turn_id": 2, "scan_id": "scene0513_00", "origin_question": "Remember, how many table(s) have you seen so far? ", "option": [], "answer": "4", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many table(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0513_00/00400.jpg", "scene0513_00/00440.jpg", "scene0513_00/00480.jpg", "scene0513_00/00520.jpg", "scene0513_00/00560.jpg" ], "turn_id": 3, "scan_id": "scene0513_00", "origin_question": "Remember, have you seen any chair(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any chair(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0513_00/00600.jpg", "scene0513_00/00640.jpg", "scene0513_00/00680.jpg", "scene0513_00/00720.jpg", "scene0513_00/00760.jpg" ], "turn_id": 4, "scan_id": "scene0513_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0513_00/00800.jpg", "scene0513_00/00840.jpg", "scene0513_00/00880.jpg", "scene0513_00/00920.jpg", "scene0513_00/00960.jpg" ], "turn_id": 5, "scan_id": "scene0513_00", "origin_question": "Remember, how many ottoman(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many ottoman(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0513_00/01000.jpg", "scene0513_00/01040.jpg", "scene0513_00/01080.jpg", "scene0513_00/01120.jpg", "scene0513_00/01160.jpg" ], "turn_id": 6, "scan_id": "scene0513_00", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "2.58", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0513_00/01200.jpg", "scene0513_00/01240.jpg", "scene0513_00/01280.jpg", "scene0513_00/01320.jpg", "scene0513_00/01360.jpg" ], "turn_id": 7, "scan_id": "scene0513_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0513_00/01400.jpg", "scene0513_00/01440.jpg", "scene0513_00/01480.jpg", "scene0513_00/01520.jpg", "scene0513_00/01560.jpg" ], "turn_id": 8, "scan_id": "scene0513_00", "origin_question": "Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0047_region20/d573afade1c74167ac0d0b1a2e37ddc4_i1_1.jpg", "1mp3d_0047_region20/d573afade1c74167ac0d0b1a2e37ddc4_i2_1.jpg", "1mp3d_0047_region20/4890b24fbb59414892ca854895896126_i1_0.jpg", "1mp3d_0047_region20/4890b24fbb59414892ca854895896126_i2_0.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0047_region20", "origin_question": "Remember, have you seen any chair(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any chair(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "1mp3d_0047_region20/d573afade1c74167ac0d0b1a2e37ddc4_i1_2.jpg", "1mp3d_0047_region20/d573afade1c74167ac0d0b1a2e37ddc4_i2_2.jpg", "1mp3d_0047_region20/98f2fab16dc44b82aef16383127ee674_i1_2.jpg", "1mp3d_0047_region20/98f2fab16dc44b82aef16383127ee674_i2_2.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0047_region20", "origin_question": "Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Right", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0047_region20/791d70f94a124a9e8013f173ce51ff08_i1_5.jpg", "1mp3d_0047_region20/791d70f94a124a9e8013f173ce51ff08_i2_5.jpg", "1mp3d_0047_region20/98f2fab16dc44b82aef16383127ee674_i1_1.jpg", "1mp3d_0047_region20/98f2fab16dc44b82aef16383127ee674_i2_1.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0047_region20", "origin_question": "", "option": [], "answer": "", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "1mp3d_0047_region20/98f2fab16dc44b82aef16383127ee674_i1_0.jpg", "1mp3d_0047_region20/98f2fab16dc44b82aef16383127ee674_i2_0.jpg", "1mp3d_0047_region20/d573afade1c74167ac0d0b1a2e37ddc4_i1_0.jpg", "1mp3d_0047_region20/d573afade1c74167ac0d0b1a2e37ddc4_i2_0.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0047_region20", "origin_question": "Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "126.87", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "1mp3d_0047_region20/4890b24fbb59414892ca854895896126_i1_2.jpg", "1mp3d_0047_region20/4890b24fbb59414892ca854895896126_i2_2.jpg", "1mp3d_0047_region20/4890b24fbb59414892ca854895896126_i1_3.jpg", "1mp3d_0047_region20/4890b24fbb59414892ca854895896126_i2_3.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0047_region20", "origin_question": "How far is your current position from where you were at the end of the turn 3? (in meters)", "option": [], "answer": "2.46", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 3? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0656_00/00000.jpg", "scene0656_00/00040.jpg", "scene0656_00/00080.jpg", "scene0656_00/00120.jpg", "scene0656_00/00160.jpg" ], "turn_id": 1, "scan_id": "scene0656_00", "origin_question": "Remember, have you seen any table(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any table(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0656_00/00200.jpg", "scene0656_00/00240.jpg", "scene0656_00/00280.jpg", "scene0656_00/00320.jpg", "scene0656_00/00360.jpg" ], "turn_id": 2, "scan_id": "scene0656_00", "origin_question": "Please recall \"the rectangular wooden stand with two curved drawers\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.4", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the rectangular wooden stand with two curved drawers\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0656_00/00400.jpg", "scene0656_00/00440.jpg", "scene0656_00/00480.jpg", "scene0656_00/00520.jpg", "scene0656_00/00560.jpg" ], "turn_id": 3, "scan_id": "scene0656_00", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the standard double or queen-size bed with a wooden frame\"; \"the adhesive tape\"; \"the dark brown wooden dresser with multiple drawers\" ", "option": [ "the standard double or queen-size bed with a wooden frame", "the adhesive tape", "the dark brown wooden dresser with multiple drawers" ], "answer": "the dark brown wooden dresser with multiple drawers", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the standard double or queen-size bed with a wooden frame\"; \"the adhesive tape\"; \"the dark brown wooden dresser with multiple drawers\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the standard double or queen-size bed with a wooden frame', 'the adhesive tape', 'the dark brown wooden dresser with multiple drawers'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0656_00/00600.jpg", "scene0656_00/00640.jpg", "scene0656_00/00680.jpg", "scene0656_00/00720.jpg", "scene0656_00/00760.jpg" ], "turn_id": 4, "scan_id": "scene0656_00", "origin_question": "Please recall \"the rectangular wooden writing table with two gold-handled drawers\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.62", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the rectangular wooden writing table with two gold-handled drawers\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "scene0656_00/00800.jpg", "scene0656_00/00840.jpg", "scene0656_00/00880.jpg", "scene0656_00/00920.jpg", "scene0656_00/00960.jpg" ], "turn_id": 5, "scan_id": "scene0656_00", "origin_question": "Among these three objects(\"the rectangular wooden writing table with two gold-handled drawers\",\"the small, cylindrical pink bin\",\"the rectangular wooden shelf with three layers\" ), which one is the closest to you?", "option": [ "the rectangular wooden writing table with two gold-handled drawers", "the small, cylindrical pink bin", "the rectangular wooden shelf with three layers" ], "answer": "the rectangular wooden writing table with two gold-handled drawers", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the rectangular wooden writing table with two gold-handled drawers\",\"the small, cylindrical pink bin\",\"the rectangular wooden shelf with three layers\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular wooden writing table with two gold-handled drawers', 'the small, cylindrical pink bin', 'the rectangular wooden shelf with three layers'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0656_00/01000.jpg", "scene0656_00/01040.jpg", "scene0656_00/01080.jpg", "scene0656_00/01120.jpg", "scene0656_00/01160.jpg" ], "turn_id": 6, "scan_id": "scene0656_00", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0656_00/01200.jpg", "scene0656_00/01240.jpg", "scene0656_00/01280.jpg", "scene0656_00/01320.jpg", "scene0656_00/01360.jpg" ], "turn_id": 7, "scan_id": "scene0656_00", "origin_question": "Which direction is the medium-sized chair with a dark cushioned seat in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the medium-sized chair with a dark cushioned seat in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0656_00/01400.jpg", "scene0656_00/01440.jpg", "scene0656_00/01480.jpg", "scene0656_00/01520.jpg", "scene0656_00/01560.jpg" ], "turn_id": 8, "scan_id": "scene0656_00", "origin_question": "Remember, how many door knob(s) have you seen so far? ", "option": [], "answer": "1", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many door knob(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0260_01/00000.jpg", "scene0260_01/00040.jpg", "scene0260_01/00080.jpg", "scene0260_01/00120.jpg", "scene0260_01/00160.jpg" ], "turn_id": 1, "scan_id": "scene0260_01", "origin_question": "Remember, have you seen any fan(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any fan(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0260_01/00200.jpg", "scene0260_01/00240.jpg", "scene0260_01/00280.jpg", "scene0260_01/00320.jpg", "scene0260_01/00360.jpg" ], "turn_id": 2, "scan_id": "scene0260_01", "origin_question": "In which turn did you see both \"the small, cylindrical silver metal bin\" and \"the bottle\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the small, cylindrical silver metal bin\" and \"the bottle\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0260_01/00400.jpg", "scene0260_01/00440.jpg", "scene0260_01/00480.jpg", "scene0260_01/00520.jpg", "scene0260_01/00560.jpg" ], "turn_id": 3, "scan_id": "scene0260_01", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the black, streamlined hair dryer with a rounded body\"; \"the vanity\"; \"the mirror\" ", "option": [ "the black, streamlined hair dryer with a rounded body", "the vanity", "the mirror" ], "answer": "the mirror", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the black, streamlined hair dryer with a rounded body\"; \"the vanity\"; \"the mirror\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the black, streamlined hair dryer with a rounded body', 'the vanity', 'the mirror'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0260_01/00600.jpg", "scene0260_01/00640.jpg", "scene0260_01/00680.jpg", "scene0260_01/00720.jpg", "scene0260_01/00760.jpg" ], "turn_id": 4, "scan_id": "scene0260_01", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the rectangular, framed canvas picture with colorful artwork\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the rectangular, framed canvas picture with colorful artwork\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0260_01/00800.jpg", "scene0260_01/00840.jpg", "scene0260_01/00880.jpg", "scene0260_01/00920.jpg", "scene0260_01/00960.jpg" ], "turn_id": 5, "scan_id": "scene0260_01", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the rectangular, light beige wooden door\"; \"the light-colored striped fabric curtain\"; \"the switch\" ", "option": [ "the rectangular, light beige wooden door", "the light-colored striped fabric curtain", "the switch" ], "answer": "the light-colored striped fabric curtain", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the rectangular, light beige wooden door\"; \"the light-colored striped fabric curtain\"; \"the switch\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular, light beige wooden door', 'the light-colored striped fabric curtain', 'the switch'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0260_01/01000.jpg", "scene0260_01/01040.jpg", "scene0260_01/01080.jpg", "scene0260_01/01120.jpg", "scene0260_01/01160.jpg" ], "turn_id": 6, "scan_id": "scene0260_01", "origin_question": "Is the small, cylindrical silver metal bin to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the small, cylindrical silver metal bin to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0260_01/01200.jpg", "scene0260_01/01240.jpg", "scene0260_01/01280.jpg", "scene0260_01/01320.jpg", "scene0260_01/01360.jpg" ], "turn_id": 7, "scan_id": "scene0260_01", "origin_question": "Please recall \"the single white roll of toilet paper\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.82", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the single white roll of toilet paper\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0260_01/01400.jpg", "scene0260_01/01440.jpg", "scene0260_01/01480.jpg", "scene0260_01/01520.jpg", "scene0260_01/01560.jpg" ], "turn_id": 8, "scan_id": "scene0260_01", "origin_question": "Which two objects are on the same side of you now? \"the standard white ceramic toilet with a round bowl\", \"the single white roll of toilet paper\", \"the switch\". ", "option": [ "the standard white ceramic toilet with a round bowl, the single white roll of toilet paper", "the standard white ceramic toilet with a round bowl, the switch", "the single white roll of toilet paper, the switch" ], "answer": "the standard white ceramic toilet with a round bowl, the single white roll of toilet paper", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the standard white ceramic toilet with a round bowl\", \"the single white roll of toilet paper\", \"the switch\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the standard white ceramic toilet with a round bowl, the single white roll of toilet paper', 'the standard white ceramic toilet with a round bowl, the switch', 'the single white roll of toilet paper, the switch'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0260_01/01600.jpg", "scene0260_01/01640.jpg", "scene0260_01/01680.jpg", "scene0260_01/01720.jpg", "scene0260_01/01760.jpg" ], "turn_id": 9, "scan_id": "scene0260_01", "origin_question": "Compared to your position at the end of the 6 turn, are you now closer or farther away from \"the rectangular, framed canvas picture with colorful artwork\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 6 turn, are you now closer or farther away from \"the rectangular, framed canvas picture with colorful artwork\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "scene0260_01/01800.jpg", "scene0260_01/01840.jpg", "scene0260_01/01880.jpg", "scene0260_01/01930.jpg", "scene0260_01/01970.jpg" ], "turn_id": 10, "scan_id": "scene0260_01", "origin_question": "Among these three objects(\"the small, cylindrical silver metal bin\",\"the single white roll of toilet paper\",\"the vanity\" ), which one is the closest to you?", "option": [ "the small, cylindrical silver metal bin", "the single white roll of toilet paper", "the vanity" ], "answer": "the vanity", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the small, cylindrical silver metal bin\",\"the single white roll of toilet paper\",\"the vanity\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the small, cylindrical silver metal bin', 'the single white roll of toilet paper', 'the vanity'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0067_region15/6266c02b09394f7e895f45aea7a1d010_i1_4.jpg", "1mp3d_0067_region15/6266c02b09394f7e895f45aea7a1d010_i2_4.jpg", "1mp3d_0067_region15/cf01932242e74d7fb73e9abb0666d71b_i1_3.jpg", "1mp3d_0067_region15/cf01932242e74d7fb73e9abb0666d71b_i2_3.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0067_region15", "origin_question": "Remember, have you seen any toilet(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any toilet(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "1mp3d_0067_region15/0376844787ff4d4bb442bb2446b9f3af_i1_5.jpg", "1mp3d_0067_region15/0376844787ff4d4bb442bb2446b9f3af_i2_5.jpg", "1mp3d_0067_region15/0376844787ff4d4bb442bb2446b9f3af_i2_4.jpg", "1mp3d_0067_region15/0376844787ff4d4bb442bb2446b9f3af_i1_0.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0067_region15", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the toilet paper\"; \"the toilet\"; \"the rectangular, deep brown wooden door with six panels\" ", "option": [ "the toilet paper", "the toilet", "the rectangular, deep brown wooden door with six panels" ], "answer": "the rectangular, deep brown wooden door with six panels", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the toilet paper\"; \"the toilet\"; \"the rectangular, deep brown wooden door with six panels\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the toilet paper', 'the toilet', 'the rectangular, deep brown wooden door with six panels'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0067_region15/0376844787ff4d4bb442bb2446b9f3af_i2_0.jpg", "1mp3d_0067_region15/cf01932242e74d7fb73e9abb0666d71b_i1_5.jpg", "1mp3d_0067_region15/cf01932242e74d7fb73e9abb0666d71b_i2_5.jpg", "1mp3d_0067_region15/e72479914a95422abb9a01ce4bdf0ab6_i2_1.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0067_region15", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the toilet paper\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the toilet paper\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "1mp3d_0067_region15/e72479914a95422abb9a01ce4bdf0ab6_i1_0.jpg", "1mp3d_0067_region15/e72479914a95422abb9a01ce4bdf0ab6_i2_0.jpg", "1mp3d_0067_region15/cf01932242e74d7fb73e9abb0666d71b_i1_4.jpg", "1mp3d_0067_region15/cf01932242e74d7fb73e9abb0666d71b_i2_4.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0067_region15", "origin_question": "In which turn did you see both \"the toilet paper\" and \"the rectangular, deep brown wooden door with six panels\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the toilet paper\" and \"the rectangular, deep brown wooden door with six panels\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "1mp3d_0067_region15/ae8304134312434bb91c6bb7ec16d7b4_i1_5.jpg", "1mp3d_0067_region15/ae8304134312434bb91c6bb7ec16d7b4_i2_5.jpg", "1mp3d_0067_region15/ae8304134312434bb91c6bb7ec16d7b4_i1_0.jpg", "1mp3d_0067_region15/ae8304134312434bb91c6bb7ec16d7b4_i2_0.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0067_region15", "origin_question": "When did you first discover \"the rectangular, deep brown wooden door with six panels\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the rectangular, deep brown wooden door with six panels\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0449_01/00000.jpg", "scene0449_01/00040.jpg", "scene0449_01/00080.jpg", "scene0449_01/00120.jpg", "scene0449_01/00160.jpg" ], "turn_id": 1, "scan_id": "scene0449_01", "origin_question": "Remember, have you seen any toilet(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any toilet(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0449_01/00200.jpg", "scene0449_01/00240.jpg", "scene0449_01/00280.jpg", "scene0449_01/00320.jpg", "scene0449_01/00360.jpg" ], "turn_id": 2, "scan_id": "scene0449_01", "origin_question": "When did you first discover \"the mirror\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the mirror\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0449_01/00400.jpg", "scene0449_01/00440.jpg", "scene0449_01/00480.jpg", "scene0449_01/00520.jpg", "scene0449_01/00560.jpg" ], "turn_id": 3, "scan_id": "scene0449_01", "origin_question": "Which direction is the toilet in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the toilet in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "scene0449_01/00600.jpg", "scene0449_01/00640.jpg", "scene0449_01/00680.jpg", "scene0449_01/00720.jpg", "scene0449_01/00760.jpg" ], "turn_id": 4, "scan_id": "scene0449_01", "origin_question": "Among these three objects(\"the toilet\",\"the bin\",\"the sink\" ), which one is the farthest from you?", "option": [ "the toilet", "the bin", "the sink" ], "answer": "the bin", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the toilet\",\"the bin\",\"the sink\" ), which one is the farthest from you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the toilet', 'the bin', 'the sink'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0449_01/00800.jpg", "scene0449_01/00840.jpg", "scene0449_01/00880.jpg", "scene0449_01/00920.jpg", "scene0449_01/00960.jpg" ], "turn_id": 5, "scan_id": "scene0449_01", "origin_question": "What will be the first-time appearance order of the following categories: mirror, sink, toilet?", "option": [ "toilet, mirror, sink", "toilet, sink, mirror", "mirror, sink, toilet", "mirror, toilet, sink" ], "answer": "toilet, mirror, sink", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: mirror, sink, toilet? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['toilet, mirror, sink', 'toilet, sink, mirror', 'mirror, sink, toilet', 'mirror, toilet, sink'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0449_01/01000.jpg", "scene0449_01/01040.jpg", "scene0449_01/01080.jpg", "scene0449_01/01120.jpg", "scene0449_01/01160.jpg" ], "turn_id": 6, "scan_id": "scene0449_01", "origin_question": "Which direction is the standard rectangular plastic switch in neutral color in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the standard rectangular plastic switch in neutral color in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0449_01/01200.jpg", "scene0449_01/01240.jpg", "scene0449_01/01280.jpg", "scene0449_01/01320.jpg", "scene0449_01/01360.jpg" ], "turn_id": 7, "scan_id": "scene0449_01", "origin_question": "At the end of which turn were both of \"the rectangular, soft silver tissue box with a smooth surface\" and \"the bin\" on your right side?", "option": [], "answer": "3", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the rectangular, soft silver tissue box with a smooth surface\" and \"the bin\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0449_01/01400.jpg", "scene0449_01/01440.jpg", "scene0449_01/01480.jpg", "scene0449_01/01520.jpg", "scene0449_01/01560.jpg" ], "turn_id": 8, "scan_id": "scene0449_01", "origin_question": "Please recall \"the simple white drawstring bag made of woven fabric\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.83", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the simple white drawstring bag made of woven fabric\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0449_01/01600.jpg", "scene0449_01/01640.jpg", "scene0449_01/01680.jpg", "scene0449_01/01720.jpg", "scene0449_01/01760.jpg" ], "turn_id": 9, "scan_id": "scene0449_01", "origin_question": "By the end of the 7 turn, were you getting closer to or farther away from these two objects: X: \"the toilet\", Y: \"the rectangular, soft silver tissue box with a smooth surface\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 7 turn, were you getting closer to or farther away from these two objects: X: \"the toilet\", Y: \"the rectangular, soft silver tissue box with a smooth surface\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0044_region23/6906646ef3a5423184e1d7938ba4f148_i2_5.jpg", "1mp3d_0044_region23/6906646ef3a5423184e1d7938ba4f148_i1_4.jpg", "1mp3d_0044_region23/6906646ef3a5423184e1d7938ba4f148_i2_4.jpg", "1mp3d_0044_region23/611e887f347a4bd4afbf09666ef657ac_i1_2.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0044_region23", "origin_question": "Remember, have you seen any faucet(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any faucet(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "1mp3d_0044_region23/611e887f347a4bd4afbf09666ef657ac_i2_2.jpg", "1mp3d_0044_region23/d70f2efc18f34b57a642332228e899d7_i1_2.jpg", "1mp3d_0044_region23/d70f2efc18f34b57a642332228e899d7_i2_2.jpg", "1mp3d_0044_region23/5e1ef15505c04411ad4e540be80b0f0b_i1_1.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0044_region23", "origin_question": "Is the tall, slender faucet with a dark metallic surface to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the tall, slender faucet with a dark metallic surface to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "1mp3d_0044_region23/5e1ef15505c04411ad4e540be80b0f0b_i2_1.jpg", "1mp3d_0044_region23/5e1ef15505c04411ad4e540be80b0f0b_i1_0.jpg", "1mp3d_0044_region23/5e1ef15505c04411ad4e540be80b0f0b_i2_0.jpg", "1mp3d_0044_region23/5e1ef15505c04411ad4e540be80b0f0b_i1_5.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0044_region23", "origin_question": "What will be the first-time appearance order of the following categories: bottle, table, washing machine?", "option": [ "washing machine, bottle, table", "table, bottle, washing machine", "bottle, table, washing machine", "bottle, washing machine, table" ], "answer": "bottle, washing machine, table", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bottle, table, washing machine? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['washing machine, bottle, table', 'table, bottle, washing machine', 'bottle, table, washing machine', 'bottle, washing machine, table'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "1mp3d_0044_region23/5e1ef15505c04411ad4e540be80b0f0b_i2_5.jpg", "1mp3d_0044_region23/d70f2efc18f34b57a642332228e899d7_i1_0.jpg", "1mp3d_0044_region23/d70f2efc18f34b57a642332228e899d7_i2_0.jpg", "1mp3d_0044_region23/611e887f347a4bd4afbf09666ef657ac_i1_1.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0044_region23", "origin_question": "Among these three objects(\"the large rectangular white sink with rounded corners\",\"the small rectangular wooden picture frame in deep brown\",\"the washing machine\" ), which one is the farthest from you?", "option": [ "the large rectangular white sink with rounded corners", "the small rectangular wooden picture frame in deep brown", "the washing machine" ], "answer": "the small rectangular wooden picture frame in deep brown", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects(\"the large rectangular white sink with rounded corners\",\"the small rectangular wooden picture frame in deep brown\",\"the washing machine\" ), which one is the farthest from you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the large rectangular white sink with rounded corners', 'the small rectangular wooden picture frame in deep brown', 'the washing machine'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "1mp3d_0044_region23/611e887f347a4bd4afbf09666ef657ac_i2_1.jpg", "1mp3d_0044_region23/611e887f347a4bd4afbf09666ef657ac_i1_0.jpg", "1mp3d_0044_region23/611e887f347a4bd4afbf09666ef657ac_i2_0.jpg", "1mp3d_0044_region23/0df37f2d82bd48fb963df1839e33fd32_i1_3.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0044_region23", "origin_question": "What will be the first-time appearance order of the following categories: bottle, door, washing machine?", "option": [ "door, washing machine, bottle", "door, bottle, washing machine", "bottle, washing machine, door", "bottle, door, washing machine" ], "answer": "bottle, washing machine, door", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bottle, door, washing machine? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['door, washing machine, bottle', 'door, bottle, washing machine', 'bottle, washing machine, door', 'bottle, door, washing machine'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0629_00/00000.jpg", "scene0629_00/00040.jpg", "scene0629_00/00080.jpg", "scene0629_00/00120.jpg", "scene0629_00/00160.jpg" ], "turn_id": 1, "scan_id": "scene0629_00", "origin_question": "Remember, have you seen any plant(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any plant(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0629_00/00200.jpg", "scene0629_00/00240.jpg", "scene0629_00/00280.jpg", "scene0629_00/00320.jpg", "scene0629_00/00360.jpg" ], "turn_id": 2, "scan_id": "scene0629_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0629_00/00400.jpg", "scene0629_00/00440.jpg", "scene0629_00/00480.jpg", "scene0629_00/00520.jpg", "scene0629_00/00560.jpg" ], "turn_id": 3, "scan_id": "scene0629_00", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the tall potted plant with white flowers and green leaves\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the tall potted plant with white flowers and green leaves\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0629_00/00600.jpg", "scene0629_00/00640.jpg", "scene0629_00/00680.jpg", "scene0629_00/00720.jpg", "scene0629_00/00760.jpg" ], "turn_id": 4, "scan_id": "scene0629_00", "origin_question": "In which turn were you closest to \"the white wooden door with six rectangular panels\"? ", "option": [], "answer": "3", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the white wooden door with six rectangular panels\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0629_00/00800.jpg", "scene0629_00/00840.jpg", "scene0629_00/00880.jpg", "scene0629_00/00920.jpg", "scene0629_00/00960.jpg" ], "turn_id": 5, "scan_id": "scene0629_00", "origin_question": "Which direction is the black, soft, velvety mattress in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the black, soft, velvety mattress in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "scene0629_00/01000.jpg", "scene0629_00/01040.jpg", "scene0629_00/01080.jpg", "scene0629_00/01120.jpg", "scene0629_00/01160.jpg" ], "turn_id": 6, "scan_id": "scene0629_00", "origin_question": "Among these three objects(\"the tall potted plant with white flowers and green leaves\",\"the black, soft, velvety mattress\",\"the table\" ), which one is the closest to you?", "option": [ "the tall potted plant with white flowers and green leaves", "the black, soft, velvety mattress", "the table" ], "answer": "the black, soft, velvety mattress", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the tall potted plant with white flowers and green leaves\",\"the black, soft, velvety mattress\",\"the table\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the tall potted plant with white flowers and green leaves', 'the black, soft, velvety mattress', 'the table'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0629_00/01200.jpg", "scene0629_00/01240.jpg", "scene0629_00/01280.jpg", "scene0629_00/01320.jpg", "scene0629_00/01360.jpg" ], "turn_id": 7, "scan_id": "scene0629_00", "origin_question": "Using your orientation at the end of turn 5 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "76.98", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 5 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0629_00/01400.jpg", "scene0629_00/01440.jpg", "scene0629_00/01480.jpg", "scene0629_00/01520.jpg", "scene0629_00/01560.jpg" ], "turn_id": 8, "scan_id": "scene0629_00", "origin_question": "By the end of the 4 turn, were you getting closer to or farther away from these two objects: X: \"the tall potted plant with white flowers and green leaves\", Y: \"the black, soft, velvety mattress\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "farther away from both X and Y", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 4 turn, were you getting closer to or farther away from these two objects: X: \"the tall potted plant with white flowers and green leaves\", Y: \"the black, soft, velvety mattress\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0629_00/01600.jpg", "scene0629_00/01640.jpg", "scene0629_00/01680.jpg", "scene0629_00/01720.jpg", "scene0629_00/01760.jpg" ], "turn_id": 9, "scan_id": "scene0629_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0629_00/01800.jpg", "scene0629_00/01840.jpg", "scene0629_00/01880.jpg", "scene0629_00/01920.jpg", "scene0629_00/01960.jpg" ], "turn_id": 10, "scan_id": "scene0629_00", "origin_question": "Which two objects are on the same side of you now? \"the tall potted plant with white flowers and green leaves\", \"the table\", \"the cylindrical plastic bottle with a blue cap\". ", "option": [ "the tall potted plant with white flowers and green leaves, the table", "the tall potted plant with white flowers and green leaves, the cylindrical plastic bottle with a blue cap", "the table, the cylindrical plastic bottle with a blue cap" ], "answer": "the tall potted plant with white flowers and green leaves, the table", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the tall potted plant with white flowers and green leaves\", \"the table\", \"the cylindrical plastic bottle with a blue cap\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the tall potted plant with white flowers and green leaves, the table', 'the tall potted plant with white flowers and green leaves, the cylindrical plastic bottle with a blue cap', 'the table, the cylindrical plastic bottle with a blue cap'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0451_04/00000.jpg", "scene0451_04/00040.jpg", "scene0451_04/00080.jpg", "scene0451_04/00120.jpg", "scene0451_04/00160.jpg" ], "turn_id": 1, "scan_id": "scene0451_04", "origin_question": "Remember, have you seen any pitcher(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any pitcher(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0451_04/00200.jpg", "scene0451_04/00240.jpg", "scene0451_04/00280.jpg", "scene0451_04/00320.jpg", "scene0451_04/00360.jpg" ], "turn_id": 2, "scan_id": "scene0451_04", "origin_question": "In which turn did you see both \"the large rectangular wooden dining table with a white tablecloth\" and \"the fruit\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the large rectangular wooden dining table with a white tablecloth\" and \"the fruit\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0451_04/00400.jpg", "scene0451_04/00440.jpg", "scene0451_04/00480.jpg", "scene0451_04/00520.jpg", "scene0451_04/00560.jpg" ], "turn_id": 3, "scan_id": "scene0451_04", "origin_question": "Please recall \"the large rectangular window with a white frame\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.52", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the large rectangular window with a white frame\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0451_04/00600.jpg", "scene0451_04/00640.jpg", "scene0451_04/00680.jpg", "scene0451_04/00720.jpg", "scene0451_04/00760.jpg" ], "turn_id": 4, "scan_id": "scene0451_04", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "156.34", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0451_04/00800.jpg", "scene0451_04/00840.jpg", "scene0451_04/00880.jpg", "scene0451_04/00920.jpg", "scene0451_04/00960.jpg" ], "turn_id": 5, "scan_id": "scene0451_04", "origin_question": "", "option": [], "answer": "", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0451_04/01000.jpg", "scene0451_04/01040.jpg", "scene0451_04/01080.jpg", "scene0451_04/01120.jpg", "scene0451_04/01160.jpg" ], "turn_id": 6, "scan_id": "scene0451_04", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "1.63", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0451_04/01200.jpg", "scene0451_04/01240.jpg", "scene0451_04/01280.jpg", "scene0451_04/01320.jpg", "scene0451_04/01360.jpg" ], "turn_id": 7, "scan_id": "scene0451_04", "origin_question": "Which direction is the large rectangular window with a white frame in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the large rectangular window with a white frame in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0451_04/01400.jpg", "scene0451_04/01440.jpg", "scene0451_04/01480.jpg", "scene0451_04/01520.jpg", "scene0451_04/01560.jpg" ], "turn_id": 8, "scan_id": "scene0451_04", "origin_question": "What will be the first-time appearance order of the following categories: curtain, fruit, switch?", "option": [ "curtain, switch, fruit", "curtain, fruit, switch", "fruit, curtain, switch", "switch, curtain, fruit" ], "answer": "curtain, fruit, switch", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: curtain, fruit, switch? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['curtain, switch, fruit', 'curtain, fruit, switch', 'fruit, curtain, switch', 'switch, curtain, fruit'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0591_00/00000.jpg", "scene0591_00/00040.jpg", "scene0591_00/00080.jpg", "scene0591_00/00120.jpg", "scene0591_00/00160.jpg" ], "turn_id": 1, "scan_id": "scene0591_00", "origin_question": "Remember, how many chair(s) have you seen so far? ", "option": [], "answer": "2", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many chair(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0591_00/00200.jpg", "scene0591_00/00240.jpg", "scene0591_00/00280.jpg", "scene0591_00/00320.jpg", "scene0591_00/00360.jpg" ], "turn_id": 2, "scan_id": "scene0591_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0591_00/00400.jpg", "scene0591_00/00440.jpg", "scene0591_00/00480.jpg", "scene0591_00/00520.jpg", "scene0591_00/00560.jpg" ], "turn_id": 3, "scan_id": "scene0591_00", "origin_question": "Is the sleek black medium-sized backpack made of nylon to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the sleek black medium-sized backpack made of nylon to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0591_00/00600.jpg", "scene0591_00/00640.jpg", "scene0591_00/00680.jpg", "scene0591_00/00720.jpg", "scene0591_00/00760.jpg" ], "turn_id": 4, "scan_id": "scene0591_00", "origin_question": "What will be the first-time appearance order of the following categories: cabinet, keyboard, pillow?", "option": [ "pillow, cabinet, keyboard", "cabinet, pillow, keyboard", "pillow, keyboard, cabinet", "cabinet, keyboard, pillow" ], "answer": "cabinet, keyboard, pillow", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: cabinet, keyboard, pillow? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['pillow, cabinet, keyboard', 'cabinet, pillow, keyboard', 'pillow, keyboard, cabinet', 'cabinet, keyboard, pillow'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0591_00/00800.jpg", "scene0591_00/00840.jpg", "scene0591_00/00880.jpg", "scene0591_00/00920.jpg", "scene0591_00/00960.jpg" ], "turn_id": 5, "scan_id": "scene0591_00", "origin_question": "Which direction is the sleek black medium-sized backpack made of nylon in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the sleek black medium-sized backpack made of nylon in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0591_00/01000.jpg", "scene0591_00/01040.jpg", "scene0591_00/01080.jpg", "scene0591_00/01120.jpg", "scene0591_00/01160.jpg" ], "turn_id": 6, "scan_id": "scene0591_00", "origin_question": "Using your orientation at the end of turn 5 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "144.16", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 5 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "scene0591_00/01200.jpg", "scene0591_00/01240.jpg", "scene0591_00/01280.jpg", "scene0591_00/01320.jpg", "scene0591_00/01360.jpg" ], "turn_id": 7, "scan_id": "scene0591_00", "origin_question": "Among these three objects(\"the window\",\"the sleek oval black mouse made of plastic\",\"the brown rectangular wooden door with a smooth surface\" ), which one is the closest to you?", "option": [ "the window", "the sleek oval black mouse made of plastic", "the brown rectangular wooden door with a smooth surface" ], "answer": "the brown rectangular wooden door with a smooth surface", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the window\",\"the sleek oval black mouse made of plastic\",\"the brown rectangular wooden door with a smooth surface\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the window', 'the sleek oval black mouse made of plastic', 'the brown rectangular wooden door with a smooth surface'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0591_00/01400.jpg", "scene0591_00/01440.jpg", "scene0591_00/01480.jpg", "scene0591_00/01520.jpg", "scene0591_00/01560.jpg" ], "turn_id": 8, "scan_id": "scene0591_00", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the modern black rectangular keyboard with white markings\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the modern black rectangular keyboard with white markings\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0040_region53/e06ec7ada8304843a634f6aa91c9165d_i1_0.jpg", "1mp3d_0040_region53/e06ec7ada8304843a634f6aa91c9165d_i2_0.jpg", "1mp3d_0040_region53/527864a208f246dbb52fb55adcb4751f_i1_1.jpg", "1mp3d_0040_region53/527864a208f246dbb52fb55adcb4751f_i2_1.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0040_region53", "origin_question": "Remember, have you seen any footstool(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any footstool(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0040_region53/527864a208f246dbb52fb55adcb4751f_i1_2.jpg", "1mp3d_0040_region53/527864a208f246dbb52fb55adcb4751f_i2_2.jpg", "1mp3d_0040_region53/e06ec7ada8304843a634f6aa91c9165d_i1_1.jpg", "1mp3d_0040_region53/e06ec7ada8304843a634f6aa91c9165d_i2_1.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0040_region53", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "1mp3d_0040_region53/d7e72b80531c417095989e69398cf1d8_i1_0.jpg", "1mp3d_0040_region53/d7e72b80531c417095989e69398cf1d8_i2_0.jpg", "1mp3d_0040_region53/3fe15c0ca3734249b98f3465e612b8db_i1_0.jpg", "1mp3d_0040_region53/3fe15c0ca3734249b98f3465e612b8db_i2_0.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0040_region53", "origin_question": "Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Right", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "1mp3d_0040_region53/3fe15c0ca3734249b98f3465e612b8db_i1_5.jpg", "1mp3d_0040_region53/3fe15c0ca3734249b98f3465e612b8db_i2_5.jpg", "1mp3d_0040_region53/d7e72b80531c417095989e69398cf1d8_i1_4.jpg", "1mp3d_0040_region53/d7e72b80531c417095989e69398cf1d8_i2_4.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0040_region53", "origin_question": "Assuming the direction you are facing at the end of the turn 3 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Left", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 3 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0040_region53/e06ec7ada8304843a634f6aa91c9165d_i1_3.jpg", "1mp3d_0040_region53/e06ec7ada8304843a634f6aa91c9165d_i2_3.jpg", "1mp3d_0040_region53/e06ec7ada8304843a634f6aa91c9165d_i1_2.jpg", "1mp3d_0040_region53/e06ec7ada8304843a634f6aa91c9165d_i2_2.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0040_region53", "origin_question": "", "option": [], "answer": "", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "1mp3d_0040_region53/d7e72b80531c417095989e69398cf1d8_i1_2.jpg", "1mp3d_0040_region53/d7e72b80531c417095989e69398cf1d8_i2_2.jpg", "1mp3d_0040_region53/3fe15c0ca3734249b98f3465e612b8db_i1_1.jpg", "1mp3d_0040_region53/3fe15c0ca3734249b98f3465e612b8db_i2_1.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0040_region53", "origin_question": "Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0029_region15/180b0d7146704223a12bcef6699d0121_i1_0.jpg", "1mp3d_0029_region15/180b0d7146704223a12bcef6699d0121_i2_0.jpg", "1mp3d_0029_region15/554ecce763be4bd49e89c1bfb89676f3_i1_5.jpg", "1mp3d_0029_region15/554ecce763be4bd49e89c1bfb89676f3_i2_5.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0029_region15", "origin_question": "Remember, have you seen any picture(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any picture(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "1mp3d_0029_region15/554ecce763be4bd49e89c1bfb89676f3_i1_0.jpg", "1mp3d_0029_region15/554ecce763be4bd49e89c1bfb89676f3_i2_0.jpg", "1mp3d_0029_region15/180b0d7146704223a12bcef6699d0121_i1_1.jpg", "1mp3d_0029_region15/180b0d7146704223a12bcef6699d0121_i2_1.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0029_region15", "origin_question": "When did you first discover \"the vent\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the vent\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0029_region15/2cbca953eeaf437799479c9111c1ee8a_i1_4.jpg", "1mp3d_0029_region15/2cbca953eeaf437799479c9111c1ee8a_i2_4.jpg", "1mp3d_0029_region15/aaa9cd95e66c4564b2065763e8dfe013_i1_0.jpg", "1mp3d_0029_region15/aaa9cd95e66c4564b2065763e8dfe013_i2_0.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0029_region15", "origin_question": "", "option": [], "answer": "", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "1mp3d_0029_region15/aaa9cd95e66c4564b2065763e8dfe013_i1_5.jpg", "1mp3d_0029_region15/aaa9cd95e66c4564b2065763e8dfe013_i2_5.jpg", "1mp3d_0029_region15/2cbca953eeaf437799479c9111c1ee8a_i1_3.jpg", "1mp3d_0029_region15/2cbca953eeaf437799479c9111c1ee8a_i2_3.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0029_region15", "origin_question": "When did you first discover \"the vase\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the vase\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "1mp3d_0029_region15/2cbca953eeaf437799479c9111c1ee8a_i1_2.jpg", "1mp3d_0029_region15/2cbca953eeaf437799479c9111c1ee8a_i2_2.jpg", "1mp3d_0029_region15/3ef76f35c6ae4ee6a69122e2413900e5_i1_1.jpg", "1mp3d_0029_region15/3ef76f35c6ae4ee6a69122e2413900e5_i2_1.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0029_region15", "origin_question": "Is the light to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the light to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0029_region15/3ef76f35c6ae4ee6a69122e2413900e5_i1_2.jpg", "1mp3d_0029_region15/3ef76f35c6ae4ee6a69122e2413900e5_i2_2.jpg", "1mp3d_0029_region15/3ef76f35c6ae4ee6a69122e2413900e5_i1_3.jpg", "1mp3d_0029_region15/3ef76f35c6ae4ee6a69122e2413900e5_i2_3.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0029_region15", "origin_question": "Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the vase\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the vase\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0461_00/00000.jpg", "scene0461_00/00030.jpg", "scene0461_00/00060.jpg", "scene0461_00/00090.jpg" ], "turn_id": 1, "scan_id": "scene0461_00", "origin_question": "Remember, how many candlestick(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many candlestick(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0461_00/00120.jpg", "scene0461_00/00150.jpg", "scene0461_00/00180.jpg", "scene0461_00/00210.jpg" ], "turn_id": 2, "scan_id": "scene0461_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0461_00/00240.jpg", "scene0461_00/00270.jpg", "scene0461_00/00300.jpg", "scene0461_00/00330.jpg" ], "turn_id": 3, "scan_id": "scene0461_00", "origin_question": "Remember, have you seen any tissue(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any tissue(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0461_00/00360.jpg", "scene0461_00/00390.jpg", "scene0461_00/00420.jpg", "scene0461_00/00450.jpg" ], "turn_id": 4, "scan_id": "scene0461_00", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0247_01/00000.jpg", "scene0247_01/00040.jpg", "scene0247_01/00080.jpg", "scene0247_01/00120.jpg", "scene0247_01/00160.jpg" ], "turn_id": 1, "scan_id": "scene0247_01", "origin_question": "Remember, have you seen any grab bar(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any grab bar(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0247_01/00200.jpg", "scene0247_01/00240.jpg", "scene0247_01/00280.jpg", "scene0247_01/00320.jpg", "scene0247_01/00360.jpg" ], "turn_id": 2, "scan_id": "scene0247_01", "origin_question": "Is the black rectangular plastic bin to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the black rectangular plastic bin to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0247_01/00400.jpg", "scene0247_01/00440.jpg", "scene0247_01/00480.jpg", "scene0247_01/00520.jpg", "scene0247_01/00560.jpg" ], "turn_id": 3, "scan_id": "scene0247_01", "origin_question": "Which direction is the cylindrical roll of white paper in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the cylindrical roll of white paper in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0247_01/00600.jpg", "scene0247_01/00640.jpg", "scene0247_01/00680.jpg", "scene0247_01/00720.jpg", "scene0247_01/00760.jpg" ], "turn_id": 4, "scan_id": "scene0247_01", "origin_question": "Which direction is the stainless steel teapot with a black handle in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the stainless steel teapot with a black handle in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0247_01/00800.jpg", "scene0247_01/00840.jpg", "scene0247_01/00880.jpg", "scene0247_01/00920.jpg", "scene0247_01/00960.jpg" ], "turn_id": 5, "scan_id": "scene0247_01", "origin_question": "Please recall \"the dishwasher\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "4.07", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the dishwasher\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0247_01/01000.jpg", "scene0247_01/01040.jpg", "scene0247_01/01080.jpg", "scene0247_01/01120.jpg", "scene0247_01/01160.jpg" ], "turn_id": 6, "scan_id": "scene0247_01", "origin_question": "Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "115.95", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0247_01/01200.jpg", "scene0247_01/01240.jpg", "scene0247_01/01280.jpg", "scene0247_01/01320.jpg", "scene0247_01/01360.jpg" ], "turn_id": 7, "scan_id": "scene0247_01", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the cylindrical roll of white paper\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the cylindrical roll of white paper\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0247_01/01400.jpg", "scene0247_01/01440.jpg", "scene0247_01/01480.jpg", "scene0247_01/01520.jpg", "scene0247_01/01560.jpg" ], "turn_id": 8, "scan_id": "scene0247_01", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the stainless steel teapot with a black handle\" from your position?", "option": [], "answer": "131.19", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the stainless steel teapot with a black handle\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "scene0247_01/01600.jpg", "scene0247_01/01640.jpg", "scene0247_01/01680.jpg", "scene0247_01/01720.jpg", "scene0247_01/01760.jpg" ], "turn_id": 9, "scan_id": "scene0247_01", "origin_question": "Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance forward or backward from that position?", "option": [ "Forward", "Backward" ], "answer": "Backward", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance forward or backward from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Forward', 'Backward'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0247_01/01800.jpg", "scene0247_01/01840.jpg", "scene0247_01/01880.jpg", "scene0247_01/01920.jpg", "scene0247_01/01960.jpg" ], "turn_id": 10, "scan_id": "scene0247_01", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the standard stainless steel upright refrigerator with a freezer\" from your position?", "option": [], "answer": "110.14", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the standard stainless steel upright refrigerator with a freezer\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0426_03/00000.jpg", "scene0426_03/00040.jpg", "scene0426_03/00080.jpg", "scene0426_03/00120.jpg" ], "turn_id": 1, "scan_id": "scene0426_03", "origin_question": "Remember, have you seen any washbasin(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any washbasin(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0426_03/00160.jpg", "scene0426_03/00200.jpg", "scene0426_03/00240.jpg", "scene0426_03/00280.jpg" ], "turn_id": 2, "scan_id": "scene0426_03", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the door\"; \"the black leather chair with a shiny surface\"; \"the bed\" ", "option": [ "the door", "the black leather chair with a shiny surface", "the bed" ], "answer": "the bed", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the door\"; \"the black leather chair with a shiny surface\"; \"the bed\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the door', 'the black leather chair with a shiny surface', 'the bed'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0426_03/00320.jpg", "scene0426_03/00360.jpg", "scene0426_03/00400.jpg", "scene0426_03/00440.jpg" ], "turn_id": 3, "scan_id": "scene0426_03", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0426_03/00480.jpg", "scene0426_03/00520.jpg", "scene0426_03/00560.jpg", "scene0426_03/00600.jpg" ], "turn_id": 4, "scan_id": "scene0426_03", "origin_question": "What will be the first-time appearance order of the following categories: chair, desk, remote control?", "option": [ "desk, chair, remote control", "chair, desk, remote control", "chair, remote control, desk", "remote control, desk, chair" ], "answer": "chair, desk, remote control", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: chair, desk, remote control? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['desk, chair, remote control', 'chair, desk, remote control', 'chair, remote control, desk', 'remote control, desk, chair'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0426_03/00640.jpg", "scene0426_03/00680.jpg", "scene0426_03/00720.jpg", "scene0426_03/00760.jpg" ], "turn_id": 5, "scan_id": "scene0426_03", "origin_question": "Which two objects are on the same side of you now? \"the standard the4 white sheet of paper\", \"the bed\", \"the bulky, flat monitor in dark colors\". ", "option": [ "the standard the4 white sheet of paper, the bed", "the standard the4 white sheet of paper, the bulky, flat monitor in dark colors", "the bed, the bulky, flat monitor in dark colors" ], "answer": "the bed, the bulky, flat monitor in dark colors", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the standard the4 white sheet of paper\", \"the bed\", \"the bulky, flat monitor in dark colors\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the standard the4 white sheet of paper, the bed', 'the standard the4 white sheet of paper, the bulky, flat monitor in dark colors', 'the bed, the bulky, flat monitor in dark colors'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0426_03/00800.jpg", "scene0426_03/00840.jpg", "scene0426_03/00880.jpg", "scene0426_03/00920.jpg" ], "turn_id": 6, "scan_id": "scene0426_03", "origin_question": "In which turn did you see both \"the sleek black flat-screen TV with a plastic frame\" and \"the bed\" simultaneously? ", "option": [], "answer": "6", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the sleek black flat-screen TV with a plastic frame\" and \"the bed\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0426_03/00960.jpg", "scene0426_03/01000.jpg", "scene0426_03/01040.jpg", "scene0426_03/01080.jpg" ], "turn_id": 7, "scan_id": "scene0426_03", "origin_question": "Which direction is the blanket in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the blanket in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0029_region23/16c937ffd817478e8ea187ffbd13eba3_i1_5.jpg", "1mp3d_0029_region23/16c937ffd817478e8ea187ffbd13eba3_i2_5.jpg", "1mp3d_0029_region23/f2fbf6ec3aae4a1e863435d270f5de83_i1_2.jpg", "1mp3d_0029_region23/f2fbf6ec3aae4a1e863435d270f5de83_i2_2.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0029_region23", "origin_question": "Remember, have you seen any window(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any window(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "1mp3d_0029_region23/16c937ffd817478e8ea187ffbd13eba3_i1_2.jpg", "1mp3d_0029_region23/16c937ffd817478e8ea187ffbd13eba3_i2_2.jpg", "1mp3d_0029_region23/8b7e19a595cb492b8f0f03fbaaabc9da_i1_1.jpg", "1mp3d_0029_region23/8b7e19a595cb492b8f0f03fbaaabc9da_i2_1.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0029_region23", "origin_question": "Which direction is the neatly arranged white couch with three cushions in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the neatly arranged white couch with three cushions in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "1mp3d_0029_region23/94bf8ede95e44311a9e8e629a4e1a898_i1_1.jpg", "1mp3d_0029_region23/94bf8ede95e44311a9e8e629a4e1a898_i2_1.jpg", "1mp3d_0029_region23/8b7e19a595cb492b8f0f03fbaaabc9da_i1_3.jpg", "1mp3d_0029_region23/8b7e19a595cb492b8f0f03fbaaabc9da_i2_3.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0029_region23", "origin_question": "Which two objects are on the same side of you now? \"the neatly arranged white couch with three cushions\", \"the large rectangular window with a white frame\", \"the rectangular wooden bed with a white bedsheet\". ", "option": [ "the neatly arranged white couch with three cushions, the large rectangular window with a white frame", "the neatly arranged white couch with three cushions, the rectangular wooden bed with a white bedsheet", "the large rectangular window with a white frame, the rectangular wooden bed with a white bedsheet" ], "answer": "the neatly arranged white couch with three cushions, the large rectangular window with a white frame", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the neatly arranged white couch with three cushions\", \"the large rectangular window with a white frame\", \"the rectangular wooden bed with a white bedsheet\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the neatly arranged white couch with three cushions, the large rectangular window with a white frame', 'the neatly arranged white couch with three cushions, the rectangular wooden bed with a white bedsheet', 'the large rectangular window with a white frame, the rectangular wooden bed with a white bedsheet'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "1mp3d_0029_region23/8b7e19a595cb492b8f0f03fbaaabc9da_i1_2.jpg", "1mp3d_0029_region23/8b7e19a595cb492b8f0f03fbaaabc9da_i2_2.jpg", "1mp3d_0029_region23/8b7e19a595cb492b8f0f03fbaaabc9da_i1_0.jpg", "1mp3d_0029_region23/8b7e19a595cb492b8f0f03fbaaabc9da_i2_0.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0029_region23", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the mirror\" from your position?", "option": [], "answer": "163.27", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the mirror\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "1mp3d_0029_region23/8b7e19a595cb492b8f0f03fbaaabc9da_i1_5.jpg", "1mp3d_0029_region23/8b7e19a595cb492b8f0f03fbaaabc9da_i2_5.jpg", "1mp3d_0029_region23/94bf8ede95e44311a9e8e629a4e1a898_i1_4.jpg", "1mp3d_0029_region23/94bf8ede95e44311a9e8e629a4e1a898_i2_4.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0029_region23", "origin_question": "Please recall \"the rectangular wooden bed with a white bedsheet\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.51", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the rectangular wooden bed with a white bedsheet\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "1mp3d_0029_region23/3d21c0cbc98b4c3da6585fabef9c68c3_i1_0.jpg", "1mp3d_0029_region23/3d21c0cbc98b4c3da6585fabef9c68c3_i2_0.jpg", "1mp3d_0029_region23/64b5e1a6f2c64d779bd3a7286b1bf7d3_i1_0.jpg", "1mp3d_0029_region23/64b5e1a6f2c64d779bd3a7286b1bf7d3_i2_0.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0029_region23", "origin_question": "In which turn were you farthest from \"the rectangular white book made of paper\"? ", "option": [], "answer": "6", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the rectangular white book made of paper\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0164_00/00000.jpg", "scene0164_00/00040.jpg", "scene0164_00/00080.jpg", "scene0164_00/00120.jpg", "scene0164_00/00160.jpg" ], "turn_id": 1, "scan_id": "scene0164_00", "origin_question": "Remember, have you seen any coffee maker(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any coffee maker(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0164_00/00200.jpg", "scene0164_00/00240.jpg", "scene0164_00/00280.jpg", "scene0164_00/00320.jpg", "scene0164_00/00360.jpg" ], "turn_id": 2, "scan_id": "scene0164_00", "origin_question": "In which turn did you see both \"the stainless steel electric stove with a white body\" and \"the medium black and gray rectangular backpack\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the stainless steel electric stove with a white body\" and \"the medium black and gray rectangular backpack\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0164_00/00400.jpg", "scene0164_00/00440.jpg", "scene0164_00/00480.jpg", "scene0164_00/00520.jpg", "scene0164_00/00560.jpg" ], "turn_id": 3, "scan_id": "scene0164_00", "origin_question": "Which direction is the stylish silver and black coffee maker in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the stylish silver and black coffee maker in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0164_00/00600.jpg", "scene0164_00/00640.jpg", "scene0164_00/00680.jpg", "scene0164_00/00720.jpg", "scene0164_00/00760.jpg" ], "turn_id": 4, "scan_id": "scene0164_00", "origin_question": "Please recall \"the microwave\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.1", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the microwave\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0164_00/00800.jpg", "scene0164_00/00840.jpg", "scene0164_00/00880.jpg", "scene0164_00/00920.jpg", "scene0164_00/00960.jpg" ], "turn_id": 5, "scan_id": "scene0164_00", "origin_question": "Which direction is the oven in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the oven in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0164_00/01000.jpg", "scene0164_00/01040.jpg", "scene0164_00/01080.jpg", "scene0164_00/01120.jpg", "scene0164_00/01160.jpg" ], "turn_id": 6, "scan_id": "scene0164_00", "origin_question": "Please recall \"the stainless steel kettle with a black handle and lid\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.97", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the stainless steel kettle with a black handle and lid\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0164_00/01200.jpg", "scene0164_00/01240.jpg", "scene0164_00/01280.jpg", "scene0164_00/01320.jpg", "scene0164_00/01360.jpg" ], "turn_id": 7, "scan_id": "scene0164_00", "origin_question": "Which direction is the small rectangular white plastic dish rack in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the small rectangular white plastic dish rack in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0164_00/01400.jpg", "scene0164_00/01440.jpg", "scene0164_00/01480.jpg", "scene0164_00/01520.jpg", "scene0164_00/01560.jpg" ], "turn_id": 8, "scan_id": "scene0164_00", "origin_question": "Which two objects are on the same side of you now? \"the microwave\", \"the black and silver rectangular metal toaster\", \"the refrigerator\". ", "option": [ "the microwave, the black and silver rectangular metal toaster", "the microwave, the refrigerator", "the black and silver rectangular metal toaster, the refrigerator" ], "answer": "the microwave, the black and silver rectangular metal toaster", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the microwave\", \"the black and silver rectangular metal toaster\", \"the refrigerator\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the microwave, the black and silver rectangular metal toaster', 'the microwave, the refrigerator', 'the black and silver rectangular metal toaster, the refrigerator'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0164_00/01600.jpg", "scene0164_00/01640.jpg", "scene0164_00/01680.jpg", "scene0164_00/01720.jpg", "scene0164_00/01760.jpg" ], "turn_id": 9, "scan_id": "scene0164_00", "origin_question": "Which two objects are on the same side of you now? \"the small rectangular white plastic dish rack\", \"the stainless steel electric stove with a white body\", \"the oven\". ", "option": [ "the small rectangular white plastic dish rack, the stainless steel electric stove with a white body", "the small rectangular white plastic dish rack, the oven", "the stainless steel electric stove with a white body, the oven" ], "answer": "the stainless steel electric stove with a white body, the oven", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the small rectangular white plastic dish rack\", \"the stainless steel electric stove with a white body\", \"the oven\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the small rectangular white plastic dish rack, the stainless steel electric stove with a white body', 'the small rectangular white plastic dish rack, the oven', 'the stainless steel electric stove with a white body, the oven'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0362_00/00000.jpg", "scene0362_00/00040.jpg", "scene0362_00/00080.jpg", "scene0362_00/00120.jpg", "scene0362_00/00160.jpg" ], "turn_id": 1, "scan_id": "scene0362_00", "origin_question": "Remember, have you seen any plunger(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any plunger(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0362_00/00200.jpg", "scene0362_00/00240.jpg", "scene0362_00/00280.jpg", "scene0362_00/00320.jpg", "scene0362_00/00360.jpg" ], "turn_id": 2, "scan_id": "scene0362_00", "origin_question": "When did you first discover \"the curtain\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the curtain\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0362_00/00400.jpg", "scene0362_00/00440.jpg", "scene0362_00/00480.jpg", "scene0362_00/00520.jpg", "scene0362_00/00560.jpg" ], "turn_id": 3, "scan_id": "scene0362_00", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "1.14", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0362_00/00600.jpg", "scene0362_00/00640.jpg", "scene0362_00/00680.jpg", "scene0362_00/00720.jpg", "scene0362_00/00760.jpg" ], "turn_id": 4, "scan_id": "scene0362_00", "origin_question": "When did you first discover \"the sleek black flat-panel TV with a modern design\"? (index of the turn)", "option": [], "answer": "4", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the sleek black flat-panel TV with a modern design\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0362_00/00800.jpg", "scene0362_00/00840.jpg", "scene0362_00/00880.jpg", "scene0362_00/00920.jpg", "scene0362_00/00960.jpg" ], "turn_id": 5, "scan_id": "scene0362_00", "origin_question": "Which two objects are on the same side of you now? \"the framed picture with abstract shapes in red, yellow, and beige\", \"the beige horizontal radiator with rows of bars\", \"the modern chair with a black mesh backrest\". ", "option": [ "the framed picture with abstract shapes in red, yellow, and beige, the beige horizontal radiator with rows of bars", "the framed picture with abstract shapes in red, yellow, and beige, the modern chair with a black mesh backrest", "the beige horizontal radiator with rows of bars, the modern chair with a black mesh backrest" ], "answer": "the framed picture with abstract shapes in red, yellow, and beige, the beige horizontal radiator with rows of bars", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the framed picture with abstract shapes in red, yellow, and beige\", \"the beige horizontal radiator with rows of bars\", \"the modern chair with a black mesh backrest\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the framed picture with abstract shapes in red, yellow, and beige, the beige horizontal radiator with rows of bars', 'the framed picture with abstract shapes in red, yellow, and beige, the modern chair with a black mesh backrest', 'the beige horizontal radiator with rows of bars, the modern chair with a black mesh backrest'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0362_00/01000.jpg", "scene0362_00/01040.jpg", "scene0362_00/01080.jpg", "scene0362_00/01120.jpg", "scene0362_00/01160.jpg" ], "turn_id": 6, "scan_id": "scene0362_00", "origin_question": "Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the rectangular, beige wooden desk with a simple appearance\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the rectangular, beige wooden desk with a simple appearance\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "scene0362_00/01200.jpg", "scene0362_00/01240.jpg", "scene0362_00/01280.jpg", "scene0362_00/01320.jpg", "scene0362_00/01360.jpg" ], "turn_id": 7, "scan_id": "scene0362_00", "origin_question": "Among these three objects(\"the rectangular two-seater couch with deep red fabric\",\"the beige horizontal radiator with rows of bars\",\"the sleek black flat-panel TV with a modern design\" ), which one is the closest to you?", "option": [ "the rectangular two-seater couch with deep red fabric", "the beige horizontal radiator with rows of bars", "the sleek black flat-panel TV with a modern design" ], "answer": "the sleek black flat-panel TV with a modern design", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the rectangular two-seater couch with deep red fabric\",\"the beige horizontal radiator with rows of bars\",\"the sleek black flat-panel TV with a modern design\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular two-seater couch with deep red fabric', 'the beige horizontal radiator with rows of bars', 'the sleek black flat-panel TV with a modern design'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0362_00/01400.jpg", "scene0362_00/01440.jpg", "scene0362_00/01480.jpg", "scene0362_00/01520.jpg", "scene0362_00/01560.jpg" ], "turn_id": 8, "scan_id": "scene0362_00", "origin_question": "Which two objects are on the same side of you now? \"the rectangular two-seater couch with deep red fabric\", \"the modern chair with a black mesh backrest\", \"the medium-sized hardcover book with a dark cover\". ", "option": [ "the rectangular two-seater couch with deep red fabric, the modern chair with a black mesh backrest", "the rectangular two-seater couch with deep red fabric, the medium-sized hardcover book with a dark cover", "the modern chair with a black mesh backrest, the medium-sized hardcover book with a dark cover" ], "answer": "the modern chair with a black mesh backrest, the medium-sized hardcover book with a dark cover", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the rectangular two-seater couch with deep red fabric\", \"the modern chair with a black mesh backrest\", \"the medium-sized hardcover book with a dark cover\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular two-seater couch with deep red fabric, the modern chair with a black mesh backrest', 'the rectangular two-seater couch with deep red fabric, the medium-sized hardcover book with a dark cover', 'the modern chair with a black mesh backrest, the medium-sized hardcover book with a dark cover'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0593_00/00000.jpg", "scene0593_00/00040.jpg", "scene0593_00/00080.jpg", "scene0593_00/00120.jpg", "scene0593_00/00160.jpg" ], "turn_id": 1, "scan_id": "scene0593_00", "origin_question": "Remember, have you seen any bin(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any bin(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0593_00/00200.jpg", "scene0593_00/00240.jpg", "scene0593_00/00280.jpg", "scene0593_00/00320.jpg", "scene0593_00/00360.jpg" ], "turn_id": 2, "scan_id": "scene0593_00", "origin_question": "At the end of which turn were both of \"the small rectangular black plastic bin\" and \"the white, standard-sized copier made of plastic\" on your left side?", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the small rectangular black plastic bin\" and \"the white, standard-sized copier made of plastic\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0593_00/00400.jpg", "scene0593_00/00440.jpg", "scene0593_00/00480.jpg", "scene0593_00/00520.jpg", "scene0593_00/00560.jpg" ], "turn_id": 3, "scan_id": "scene0593_00", "origin_question": "At the end of which turn were both of \"the modern paper cutter with a black handle and silver blade\" and \"the medium-sized white printer with a blue cover\" on your left side?", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the modern paper cutter with a black handle and silver blade\" and \"the medium-sized white printer with a blue cover\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0593_00/00600.jpg", "scene0593_00/00640.jpg", "scene0593_00/00680.jpg", "scene0593_00/00720.jpg", "scene0593_00/00760.jpg" ], "turn_id": 4, "scan_id": "scene0593_00", "origin_question": "Which direction is the medium-sized white printer with a blue cover in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the medium-sized white printer with a blue cover in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0593_00/00800.jpg", "scene0593_00/00840.jpg", "scene0593_00/00880.jpg", "scene0593_00/00920.jpg", "scene0593_00/00960.jpg" ], "turn_id": 5, "scan_id": "scene0593_00", "origin_question": "By the end of the 4 turn, were you getting closer to or farther away from these two objects: X: \"the white, standard-sized copier made of plastic\", Y: \"the medium-sized white printer with a blue cover\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "farther away from both X and Y", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 4 turn, were you getting closer to or farther away from these two objects: X: \"the white, standard-sized copier made of plastic\", Y: \"the medium-sized white printer with a blue cover\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "scene0593_00/01000.jpg", "scene0593_00/01040.jpg", "scene0593_00/01080.jpg", "scene0593_00/01120.jpg", "scene0593_00/01160.jpg" ], "turn_id": 6, "scan_id": "scene0593_00", "origin_question": "Among these three objects(\"the white, standard-sized copier made of plastic\",\"the modern paper cutter with a black handle and silver blade\",\"the door\" ), which one is the farthest from you?", "option": [ "the white, standard-sized copier made of plastic", "the modern paper cutter with a black handle and silver blade", "the door" ], "answer": "the modern paper cutter with a black handle and silver blade", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the white, standard-sized copier made of plastic\",\"the modern paper cutter with a black handle and silver blade\",\"the door\" ), which one is the farthest from you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the white, standard-sized copier made of plastic', 'the modern paper cutter with a black handle and silver blade', 'the door'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0593_00/01200.jpg", "scene0593_00/01240.jpg", "scene0593_00/01280.jpg", "scene0593_00/01320.jpg", "scene0593_00/01360.jpg" ], "turn_id": 7, "scan_id": "scene0593_00", "origin_question": "Is the couch to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the couch to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0593_00/01400.jpg", "scene0593_00/01440.jpg", "scene0593_00/01480.jpg", "scene0593_00/01520.jpg", "scene0593_00/01560.jpg" ], "turn_id": 8, "scan_id": "scene0593_00", "origin_question": "In which turn did you see both \"the white, standard-sized copier made of plastic\" and \"the modern paper cutter with a black handle and silver blade\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the white, standard-sized copier made of plastic\" and \"the modern paper cutter with a black handle and silver blade\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0593_00/01600.jpg", "scene0593_00/01640.jpg", "scene0593_00/01680.jpg", "scene0593_00/01720.jpg", "scene0593_00/01760.jpg" ], "turn_id": 9, "scan_id": "scene0593_00", "origin_question": "Is the modern paper cutter with a black handle and silver blade to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the modern paper cutter with a black handle and silver blade to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0593_00/01800.jpg", "scene0593_00/01840.jpg", "scene0593_00/01880.jpg", "scene0593_00/01920.jpg", "scene0593_00/01960.jpg" ], "turn_id": 10, "scan_id": "scene0593_00", "origin_question": "At the end of which turn were both of \"the couch\" and \"the rectangular light brown board made of soft wood\" on your right side?", "option": [], "answer": "3", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the couch\" and \"the rectangular light brown board made of soft wood\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0114_01/00000.jpg", "scene0114_01/00040.jpg", "scene0114_01/00080.jpg", "scene0114_01/00120.jpg", "scene0114_01/00160.jpg" ], "turn_id": 1, "scan_id": "scene0114_01", "origin_question": "Remember, have you seen any food(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any food(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0114_01/00200.jpg", "scene0114_01/00240.jpg", "scene0114_01/00280.jpg", "scene0114_01/00320.jpg", "scene0114_01/00360.jpg" ], "turn_id": 2, "scan_id": "scene0114_01", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the single rectangular white sheet of paper\" from your position?", "option": [], "answer": "108.09", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the single rectangular white sheet of paper\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0114_01/00400.jpg", "scene0114_01/00440.jpg", "scene0114_01/00480.jpg", "scene0114_01/00520.jpg", "scene0114_01/00560.jpg" ], "turn_id": 3, "scan_id": "scene0114_01", "origin_question": "At the end of which turn were both of \"the lifelike toy model of Homer Simpson\" and \"the medium-sized brown wooden door\" on your right side?", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the lifelike toy model of Homer Simpson\" and \"the medium-sized brown wooden door\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0114_01/00600.jpg", "scene0114_01/00640.jpg", "scene0114_01/00680.jpg", "scene0114_01/00720.jpg", "scene0114_01/00760.jpg" ], "turn_id": 4, "scan_id": "scene0114_01", "origin_question": "", "option": [], "answer": "", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0114_01/00800.jpg", "scene0114_01/00840.jpg", "scene0114_01/00880.jpg", "scene0114_01/00920.jpg", "scene0114_01/00960.jpg" ], "turn_id": 5, "scan_id": "scene0114_01", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the single rectangular white sheet of paper\" from your position?", "option": [], "answer": "56.32", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the single rectangular white sheet of paper\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0114_01/01000.jpg", "scene0114_01/01040.jpg", "scene0114_01/01080.jpg", "scene0114_01/01120.jpg", "scene0114_01/01160.jpg" ], "turn_id": 6, "scan_id": "scene0114_01", "origin_question": "Is the white corded rectangular plastic telephone to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the white corded rectangular plastic telephone to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0114_01/01200.jpg", "scene0114_01/01240.jpg", "scene0114_01/01280.jpg", "scene0114_01/01320.jpg", "scene0114_01/01360.jpg" ], "turn_id": 7, "scan_id": "scene0114_01", "origin_question": "Compared to your position at the end of the 6 turn, are you now closer or farther away from \"the medium-sized brown wooden door\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 6 turn, are you now closer or farther away from \"the medium-sized brown wooden door\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0114_01/01400.jpg", "scene0114_01/01440.jpg", "scene0114_01/01480.jpg", "scene0114_01/01520.jpg", "scene0114_01/01560.jpg" ], "turn_id": 8, "scan_id": "scene0114_01", "origin_question": "Using your orientation at the end of turn 7 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 7 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0114_01/01600.jpg", "scene0114_01/01640.jpg", "scene0114_01/01680.jpg", "scene0114_01/01720.jpg", "scene0114_01/01760.jpg" ], "turn_id": 9, "scan_id": "scene0114_01", "origin_question": "When did you first discover \"the medium-sized black ergonomic mouse made of plastic\"? (index of the turn)", "option": [], "answer": "4", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the medium-sized black ergonomic mouse made of plastic\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0474_04/00000.jpg", "scene0474_04/00030.jpg", "scene0474_04/00060.jpg", "scene0474_04/00090.jpg" ], "turn_id": 1, "scan_id": "scene0474_04", "origin_question": "Remember, have you seen any bin(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any bin(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0474_04/00120.jpg", "scene0474_04/00150.jpg", "scene0474_04/00180.jpg", "scene0474_04/00210.jpg" ], "turn_id": 2, "scan_id": "scene0474_04", "origin_question": "Which direction is the rectangular blue bin with a label in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the rectangular blue bin with a label in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0474_04/00240.jpg", "scene0474_04/00270.jpg", "scene0474_04/00300.jpg", "scene0474_04/00330.jpg" ], "turn_id": 3, "scan_id": "scene0474_04", "origin_question": "Is the rectangular wooden door with a smooth, rich brown surface to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the rectangular wooden door with a smooth, rich brown surface to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0474_04/00360.jpg", "scene0474_04/00390.jpg", "scene0474_04/00420.jpg", "scene0474_04/00450.jpg" ], "turn_id": 4, "scan_id": "scene0474_04", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the rectangular blue bin with a label\" from your position?", "option": [], "answer": "48.88", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the rectangular blue bin with a label\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0474_04/00480.jpg", "scene0474_04/00510.jpg", "scene0474_04/00540.jpg", "scene0474_04/00570.jpg" ], "turn_id": 5, "scan_id": "scene0474_04", "origin_question": "In which turn did you see both \"the rectangular wooden door with a smooth, rich brown surface\" and \"the rectangular wooden desk with a built-in bookshelf\" simultaneously? ", "option": [], "answer": "4", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the rectangular wooden door with a smooth, rich brown surface\" and \"the rectangular wooden desk with a built-in bookshelf\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0474_04/00610.jpg", "scene0474_04/00640.jpg", "scene0474_04/00670.jpg", "scene0474_04/00700.jpg" ], "turn_id": 6, "scan_id": "scene0474_04", "origin_question": "What will be the first-time appearance order of the following categories: bin, couch, desk?", "option": [ "desk, bin, couch", "couch, bin, desk", "bin, couch, desk", "couch, desk, bin" ], "answer": "bin, couch, desk", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bin, couch, desk? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['desk, bin, couch', 'couch, bin, desk', 'bin, couch, desk', 'couch, desk, bin'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0474_04/00730.jpg", "scene0474_04/00760.jpg", "scene0474_04/00790.jpg", "scene0474_04/00820.jpg" ], "turn_id": 7, "scan_id": "scene0474_04", "origin_question": "Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the medium-sized black backpack with textured fabric\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the medium-sized black backpack with textured fabric\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0626_01/00000.jpg", "scene0626_01/00030.jpg", "scene0626_01/00060.jpg", "scene0626_01/00090.jpg" ], "turn_id": 1, "scan_id": "scene0626_01", "origin_question": "Remember, have you seen any keyboard(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any keyboard(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0626_01/00120.jpg", "scene0626_01/00150.jpg", "scene0626_01/00180.jpg", "scene0626_01/00210.jpg" ], "turn_id": 2, "scan_id": "scene0626_01", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0626_01/00240.jpg", "scene0626_01/00270.jpg", "scene0626_01/00300.jpg", "scene0626_01/00330.jpg" ], "turn_id": 3, "scan_id": "scene0626_01", "origin_question": "In which turn did you see both \"the rectangular blackboard with a white background\" and \"the rectangular book with a white cover and image\" simultaneously? ", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the rectangular blackboard with a white background\" and \"the rectangular book with a white cover and image\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0626_01/00360.jpg", "scene0626_01/00390.jpg", "scene0626_01/00420.jpg", "scene0626_01/00450.jpg" ], "turn_id": 4, "scan_id": "scene0626_01", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the medium-sized purple umbrella made of lightweight material\" from your position?", "option": [], "answer": "82.29", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the medium-sized purple umbrella made of lightweight material\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0626_01/00480.jpg", "scene0626_01/00510.jpg", "scene0626_01/00540.jpg", "scene0626_01/00570.jpg" ], "turn_id": 5, "scan_id": "scene0626_01", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the person\"; \"the cozy, light brown cotton couch\"; \"the medium-sized purple umbrella made of lightweight material\" ", "option": [ "the person", "the cozy, light brown cotton couch", "the medium-sized purple umbrella made of lightweight material" ], "answer": "the person", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the person\"; \"the cozy, light brown cotton couch\"; \"the medium-sized purple umbrella made of lightweight material\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the person', 'the cozy, light brown cotton couch', 'the medium-sized purple umbrella made of lightweight material'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0626_01/00600.jpg", "scene0626_01/00630.jpg", "scene0626_01/00660.jpg", "scene0626_01/00690.jpg" ], "turn_id": 6, "scan_id": "scene0626_01", "origin_question": "At the end of which turn were both of \"the rectangular black keyboard made of plastic\" and \"the cozy, light brown cotton couch\" on your right side?", "option": [], "answer": "5", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the rectangular black keyboard made of plastic\" and \"the cozy, light brown cotton couch\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0044_region15/5ae7f77201964f5386990b3ce719df94_i1_1.jpg", "1mp3d_0044_region15/5ae7f77201964f5386990b3ce719df94_i2_1.jpg", "1mp3d_0044_region15/6167a976f60f4baebdd66477d83f2fd1_i1_3.jpg", "1mp3d_0044_region15/6167a976f60f4baebdd66477d83f2fd1_i2_3.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0044_region15", "origin_question": "Remember, have you seen any door(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any door(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0044_region15/5ae7f77201964f5386990b3ce719df94_i1_3.jpg", "1mp3d_0044_region15/5ae7f77201964f5386990b3ce719df94_i2_3.jpg", "1mp3d_0044_region15/6167a976f60f4baebdd66477d83f2fd1_i1_2.jpg", "1mp3d_0044_region15/6167a976f60f4baebdd66477d83f2fd1_i2_2.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0044_region15", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "1mp3d_0044_region15/d1f4a03375d4471d8b80de83620c52e6_i1_3.jpg", "1mp3d_0044_region15/d1f4a03375d4471d8b80de83620c52e6_i2_3.jpg", "1mp3d_0044_region15/d1f4a03375d4471d8b80de83620c52e6_i1_4.jpg", "1mp3d_0044_region15/d1f4a03375d4471d8b80de83620c52e6_i2_4.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0044_region15", "origin_question": "Which two objects are on the same side of you now? \"the dark wood vanity with exquisite carvings\", \"the beige hand towel made of soft cotton\", \"the toilet paper\". ", "option": [ "the dark wood vanity with exquisite carvings, the beige hand towel made of soft cotton", "the dark wood vanity with exquisite carvings, the toilet paper", "the beige hand towel made of soft cotton, the toilet paper" ], "answer": "the dark wood vanity with exquisite carvings, the beige hand towel made of soft cotton", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the dark wood vanity with exquisite carvings\", \"the beige hand towel made of soft cotton\", \"the toilet paper\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the dark wood vanity with exquisite carvings, the beige hand towel made of soft cotton', 'the dark wood vanity with exquisite carvings, the toilet paper', 'the beige hand towel made of soft cotton, the toilet paper'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "1mp3d_0044_region15/d1f4a03375d4471d8b80de83620c52e6_i1_5.jpg", "1mp3d_0044_region15/d1f4a03375d4471d8b80de83620c52e6_i2_5.jpg", "1mp3d_0044_region15/6167a976f60f4baebdd66477d83f2fd1_i1_0.jpg", "1mp3d_0044_region15/6167a976f60f4baebdd66477d83f2fd1_i2_0.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0044_region15", "origin_question": "Which direction is the dark brown rectangular wooden door in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the dark brown rectangular wooden door in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "1mp3d_0044_region15/6167a976f60f4baebdd66477d83f2fd1_i1_5.jpg", "1mp3d_0044_region15/6167a976f60f4baebdd66477d83f2fd1_i2_5.jpg", "1mp3d_0044_region15/6167a976f60f4baebdd66477d83f2fd1_i1_4.jpg", "1mp3d_0044_region15/6167a976f60f4baebdd66477d83f2fd1_i2_4.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0044_region15", "origin_question": "What will be the first-time appearance order of the following categories: light, toilet paper, towel?", "option": [ "light, towel, toilet paper", "light, toilet paper, towel", "towel, toilet paper, light", "toilet paper, towel, light" ], "answer": "toilet paper, towel, light", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: light, toilet paper, towel? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['light, towel, toilet paper', 'light, toilet paper, towel', 'towel, toilet paper, light', 'toilet paper, towel, light'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "1mp3d_0044_region15/5ae7f77201964f5386990b3ce719df94_i1_5.jpg", "1mp3d_0044_region15/5ae7f77201964f5386990b3ce719df94_i2_5.jpg", "1mp3d_0044_region15/5ae7f77201964f5386990b3ce719df94_i1_0.jpg", "1mp3d_0044_region15/5ae7f77201964f5386990b3ce719df94_i2_0.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0044_region15", "origin_question": "Which two objects are on the same side of you now? \"the toilet paper\", \"the rectangular glass window with dark wooden frame\", \"the beige hand towel made of soft cotton\". ", "option": [ "the toilet paper, the rectangular glass window with dark wooden frame", "the toilet paper, the beige hand towel made of soft cotton", "the rectangular glass window with dark wooden frame, the beige hand towel made of soft cotton" ], "answer": "the toilet paper, the rectangular glass window with dark wooden frame", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the toilet paper\", \"the rectangular glass window with dark wooden frame\", \"the beige hand towel made of soft cotton\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the toilet paper, the rectangular glass window with dark wooden frame', 'the toilet paper, the beige hand towel made of soft cotton', 'the rectangular glass window with dark wooden frame, the beige hand towel made of soft cotton'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0084_02/00000.jpg", "scene0084_02/00040.jpg", "scene0084_02/00080.jpg", "scene0084_02/00120.jpg" ], "turn_id": 1, "scan_id": "scene0084_02", "origin_question": "Remember, have you seen any book(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any book(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0084_02/00160.jpg", "scene0084_02/00200.jpg", "scene0084_02/00240.jpg", "scene0084_02/00280.jpg" ], "turn_id": 2, "scan_id": "scene0084_02", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the dark rectangular cotton-padded jacket\" from your position?", "option": [], "answer": "54.04", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the dark rectangular cotton-padded jacket\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0084_02/00320.jpg", "scene0084_02/00360.jpg", "scene0084_02/00400.jpg", "scene0084_02/00440.jpg" ], "turn_id": 3, "scan_id": "scene0084_02", "origin_question": "Is the dark rectangular cotton-padded jacket to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the dark rectangular cotton-padded jacket to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0084_02/00480.jpg", "scene0084_02/00520.jpg", "scene0084_02/00560.jpg", "scene0084_02/00600.jpg" ], "turn_id": 4, "scan_id": "scene0084_02", "origin_question": "At the end of which turn were both of \"the modern beige or light gray shower with a vertical head\" and \"the round, metallic knob-style switch with a silver surface\" on your right side?", "option": [], "answer": "4", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the modern beige or light gray shower with a vertical head\" and \"the round, metallic knob-style switch with a silver surface\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0084_02/00640.jpg", "scene0084_02/00680.jpg", "scene0084_02/00720.jpg", "scene0084_02/00760.jpg" ], "turn_id": 5, "scan_id": "scene0084_02", "origin_question": "What will be the first-time appearance order of the following categories: door, switch, toilet?", "option": [ "switch, toilet, door", "toilet, door, switch", "door, switch, toilet", "toilet, switch, door" ], "answer": "door, switch, toilet", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: door, switch, toilet? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['switch, toilet, door', 'toilet, door, switch', 'door, switch, toilet', 'toilet, switch, door'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0084_02/00800.jpg", "scene0084_02/00840.jpg", "scene0084_02/00880.jpg", "scene0084_02/00920.jpg" ], "turn_id": 6, "scan_id": "scene0084_02", "origin_question": "Which two objects are on the same side of you now? \"the dark rectangular cotton-padded jacket\", \"the rectangular metal holder with a minimalist design\", \"the large white rectangular metal radiator\". ", "option": [ "the dark rectangular cotton-padded jacket, the rectangular metal holder with a minimalist design", "the dark rectangular cotton-padded jacket, the large white rectangular metal radiator", "the rectangular metal holder with a minimalist design, the large white rectangular metal radiator" ], "answer": "the rectangular metal holder with a minimalist design, the large white rectangular metal radiator", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the dark rectangular cotton-padded jacket\", \"the rectangular metal holder with a minimalist design\", \"the large white rectangular metal radiator\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the dark rectangular cotton-padded jacket, the rectangular metal holder with a minimalist design', 'the dark rectangular cotton-padded jacket, the large white rectangular metal radiator', 'the rectangular metal holder with a minimalist design, the large white rectangular metal radiator'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0084_02/00960.jpg", "scene0084_02/01000.jpg", "scene0084_02/01040.jpg", "scene0084_02/01080.jpg" ], "turn_id": 7, "scan_id": "scene0084_02", "origin_question": "How far is your current position from where you were at the end of the turn 1? (in meters)", "option": [], "answer": "1.91", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 1? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0040_region4/e71839ebe96b481783835db43fec5cbd_i1_5.jpg", "1mp3d_0040_region4/e71839ebe96b481783835db43fec5cbd_i2_5.jpg", "1mp3d_0040_region4/ea1a6578bc034b31aa0c9844a987ff7d_i1_2.jpg", "1mp3d_0040_region4/ea1a6578bc034b31aa0c9844a987ff7d_i2_2.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0040_region4", "origin_question": "Remember, have you seen any window(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any window(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0040_region4/e71839ebe96b481783835db43fec5cbd_i1_0.jpg", "1mp3d_0040_region4/e71839ebe96b481783835db43fec5cbd_i2_0.jpg", "1mp3d_0040_region4/ea1a6578bc034b31aa0c9844a987ff7d_i2_1.jpg", "1mp3d_0040_region4/21060452879b403cae060c2a21503889_i1_2.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0040_region4", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0040_region4/21060452879b403cae060c2a21503889_i2_2.jpg", "1mp3d_0040_region4/1f169ea3d73c4cc8b7f653c4d4ff15cb_i1_2.jpg", "1mp3d_0040_region4/1f169ea3d73c4cc8b7f653c4d4ff15cb_i2_2.jpg", "1mp3d_0040_region4/21060452879b403cae060c2a21503889_i1_1.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0040_region4", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the rectangular, warm amber textured window\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the rectangular, warm amber textured window\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "1mp3d_0040_region4/21060452879b403cae060c2a21503889_i2_1.jpg", "1mp3d_0040_region4/ea1a6578bc034b31aa0c9844a987ff7d_i1_0.jpg", "1mp3d_0040_region4/ea1a6578bc034b31aa0c9844a987ff7d_i2_0.jpg", "1mp3d_0040_region4/ea1a6578bc034b31aa0c9844a987ff7d_i1_5.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0040_region4", "origin_question": "How far is your current position from where you were at the end of the turn 3? (in meters)", "option": [], "answer": "1.45", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 3? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "1mp3d_0040_region4/ea1a6578bc034b31aa0c9844a987ff7d_i2_5.jpg", "1mp3d_0040_region4/21060452879b403cae060c2a21503889_i1_5.jpg", "1mp3d_0040_region4/21060452879b403cae060c2a21503889_i2_5.jpg", "1mp3d_0040_region4/1f169ea3d73c4cc8b7f653c4d4ff15cb_i1_4.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0040_region4", "origin_question": "By the end of the 4 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular, warm amber textured window\", Y: \"Off-white rectangular fabric blinds of standard size\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "farther away from both X and Y", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 4 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular, warm amber textured window\", Y: \"Off-white rectangular fabric blinds of standard size\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0665_00/00000.jpg", "scene0665_00/00040.jpg", "scene0665_00/00080.jpg", "scene0665_00/00120.jpg" ], "turn_id": 1, "scan_id": "scene0665_00", "origin_question": "Remember, how many chair(s) have you seen so far? ", "option": [], "answer": "8", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many chair(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0665_00/00160.jpg", "scene0665_00/00200.jpg", "scene0665_00/00240.jpg", "scene0665_00/00280.jpg" ], "turn_id": 2, "scan_id": "scene0665_00", "origin_question": "Remember, have you seen any table(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any table(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0665_00/00320.jpg", "scene0665_00/00360.jpg", "scene0665_00/00400.jpg", "scene0665_00/00440.jpg" ], "turn_id": 3, "scan_id": "scene0665_00", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "74.56", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0665_00/00480.jpg", "scene0665_00/00520.jpg", "scene0665_00/00560.jpg", "scene0665_00/00600.jpg" ], "turn_id": 4, "scan_id": "scene0665_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0665_00/00640.jpg", "scene0665_00/00680.jpg", "scene0665_00/00720.jpg", "scene0665_00/00760.jpg" ], "turn_id": 5, "scan_id": "scene0665_00", "origin_question": "How far is your current position from where you were at the end of the turn 4? (in meters)", "option": [], "answer": "2.66", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 4? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0665_00/00800.jpg", "scene0665_00/00850.jpg", "scene0665_00/00890.jpg", "scene0665_00/00930.jpg" ], "turn_id": 6, "scan_id": "scene0665_00", "origin_question": "Please recall \"the dark, soft cotton wrap without patterns\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.29", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the dark, soft cotton wrap without patterns\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0618_00/00000.jpg", "scene0618_00/00030.jpg", "scene0618_00/00060.jpg", "scene0618_00/00090.jpg" ], "turn_id": 1, "scan_id": "scene0618_00", "origin_question": "Remember, have you seen any chair(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any chair(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0618_00/00120.jpg", "scene0618_00/00150.jpg", "scene0618_00/00180.jpg", "scene0618_00/00210.jpg" ], "turn_id": 2, "scan_id": "scene0618_00", "origin_question": "Remember, have you seen any pool table(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any pool table(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0618_00/00240.jpg", "scene0618_00/00270.jpg", "scene0618_00/00300.jpg", "scene0618_00/00330.jpg" ], "turn_id": 3, "scan_id": "scene0618_00", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0618_00/00360.jpg", "scene0618_00/00390.jpg", "scene0618_00/00420.jpg", "scene0618_00/00450.jpg" ], "turn_id": 4, "scan_id": "scene0618_00", "origin_question": "Remember, have you seen any chair(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any chair(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0618_00/00480.jpg", "scene0618_00/00510.jpg", "scene0618_00/00540.jpg", "scene0618_00/00570.jpg" ], "turn_id": 5, "scan_id": "scene0618_00", "origin_question": "How far is your current position from where you were at the end of the turn 4? (in meters)", "option": [], "answer": "1.5", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 4? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0618_00/00600.jpg", "scene0618_00/00630.jpg", "scene0618_00/00660.jpg", "scene0618_00/00690.jpg" ], "turn_id": 6, "scan_id": "scene0618_00", "origin_question": "Using your orientation at the end of turn 4 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "98.39", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 4 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0618_00/00720.jpg", "scene0618_00/00750.jpg", "scene0618_00/00780.jpg", "scene0618_00/00810.jpg" ], "turn_id": 7, "scan_id": "scene0618_00", "origin_question": "What will be the first-time appearance order of the following categories: chair, pool table, window?", "option": [ "window, chair, pool table", "pool table, window, chair", "chair, pool table, window", "window, pool table, chair" ], "answer": "chair, pool table, window", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: chair, pool table, window? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['window, chair, pool table', 'pool table, window, chair', 'chair, pool table, window', 'window, pool table, chair'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0004_region18/805d7e3629234abbbc3de4c5eb7598ab_i1_1.jpg", "1mp3d_0004_region18/805d7e3629234abbbc3de4c5eb7598ab_i2_1.jpg", "1mp3d_0004_region18/805d7e3629234abbbc3de4c5eb7598ab_i1_2.jpg", "1mp3d_0004_region18/805d7e3629234abbbc3de4c5eb7598ab_i2_2.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0004_region18", "origin_question": "Remember, have you seen any table(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any table(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "1mp3d_0004_region18/805d7e3629234abbbc3de4c5eb7598ab_i1_3.jpg", "1mp3d_0004_region18/805d7e3629234abbbc3de4c5eb7598ab_i2_3.jpg", "1mp3d_0004_region18/1b0d7310dda2480d840cb78d86ec7236_i1_3.jpg", "1mp3d_0004_region18/1b0d7310dda2480d840cb78d86ec7236_i2_3.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0004_region18", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the bowl of round, green fruit\" from your position?", "option": [], "answer": "63.5", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the bowl of round, green fruit\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "1mp3d_0004_region18/1b0d7310dda2480d840cb78d86ec7236_i1_2.jpg", "1mp3d_0004_region18/1b0d7310dda2480d840cb78d86ec7236_i2_2.jpg", "1mp3d_0004_region18/1b0d7310dda2480d840cb78d86ec7236_i1_1.jpg", "1mp3d_0004_region18/1b0d7310dda2480d840cb78d86ec7236_i2_1.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0004_region18", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "1mp3d_0004_region18/1b0d7310dda2480d840cb78d86ec7236_i1_0.jpg", "1mp3d_0004_region18/1b0d7310dda2480d840cb78d86ec7236_i2_0.jpg", "1mp3d_0004_region18/dd567b91bf7147b3869277c4d0691452_i1_4.jpg", "1mp3d_0004_region18/dd567b91bf7147b3869277c4d0691452_i2_4.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0004_region18", "origin_question": "Which direction is the statue in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the statue in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0004_region18/dd567b91bf7147b3869277c4d0691452_i1_5.jpg", "1mp3d_0004_region18/dd567b91bf7147b3869277c4d0691452_i2_5.jpg", "1mp3d_0004_region18/dd567b91bf7147b3869277c4d0691452_i1_0.jpg", "1mp3d_0004_region18/dd567b91bf7147b3869277c4d0691452_i2_0.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0004_region18", "origin_question": "", "option": [], "answer": "", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "1mp3d_0004_region18/dd567b91bf7147b3869277c4d0691452_i1_1.jpg", "1mp3d_0004_region18/dd567b91bf7147b3869277c4d0691452_i2_1.jpg", "1mp3d_0004_region18/dd567b91bf7147b3869277c4d0691452_i1_2.jpg", "1mp3d_0004_region18/dd567b91bf7147b3869277c4d0691452_i2_2.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0004_region18", "origin_question": "Which direction is the statue in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the statue in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0084_region13/5125c4a0a99f45788e694d5a73ba99c7_i1_5.jpg", "1mp3d_0084_region13/5125c4a0a99f45788e694d5a73ba99c7_i2_5.jpg", "1mp3d_0084_region13/7743b72f7c3d40749363873307eef9c9_i1_1.jpg", "1mp3d_0084_region13/7743b72f7c3d40749363873307eef9c9_i2_1.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0084_region13", "origin_question": "", "option": [], "answer": "", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 3 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0084_region13/7743b72f7c3d40749363873307eef9c9_i1_2.jpg", "1mp3d_0084_region13/7743b72f7c3d40749363873307eef9c9_i2_2.jpg", "1mp3d_0084_region13/7743b72f7c3d40749363873307eef9c9_i1_3.jpg", "1mp3d_0084_region13/7743b72f7c3d40749363873307eef9c9_i2_3.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0084_region13", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 3 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "1mp3d_0084_region13/5125c4a0a99f45788e694d5a73ba99c7_i2_0.jpg", "1mp3d_0084_region13/20f70ab4efec496c86529075648caef2_i1_5.jpg", "1mp3d_0084_region13/20f70ab4efec496c86529075648caef2_i2_5.jpg", "1mp3d_0084_region13/1b321779a4374c2b952c51820daa9e6c_i1_5.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0084_region13", "origin_question": "Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "27.52", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 3 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0575_01/00000.jpg", "scene0575_01/00040.jpg", "scene0575_01/00080.jpg", "scene0575_01/00120.jpg", "scene0575_01/00160.jpg" ], "turn_id": 1, "scan_id": "scene0575_01", "origin_question": "Remember, have you seen any chair(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any chair(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0575_01/00200.jpg", "scene0575_01/00240.jpg", "scene0575_01/00280.jpg", "scene0575_01/00320.jpg", "scene0575_01/00360.jpg" ], "turn_id": 2, "scan_id": "scene0575_01", "origin_question": "In which turn did you see both \"the rectangular frosted glass window with clean lines\" and \"the medium rectangular light brown wooden table\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the rectangular frosted glass window with clean lines\" and \"the medium rectangular light brown wooden table\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0575_01/00400.jpg", "scene0575_01/00440.jpg", "scene0575_01/00480.jpg", "scene0575_01/00520.jpg", "scene0575_01/00560.jpg" ], "turn_id": 3, "scan_id": "scene0575_01", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "1.21", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0575_01/00600.jpg", "scene0575_01/00640.jpg", "scene0575_01/00680.jpg", "scene0575_01/00720.jpg", "scene0575_01/00760.jpg" ], "turn_id": 4, "scan_id": "scene0575_01", "origin_question": "In which turn were you closest to \"the medium rectangular light brown wooden table\"? ", "option": [], "answer": "2", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the medium rectangular light brown wooden table\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0575_01/00800.jpg", "scene0575_01/00840.jpg", "scene0575_01/00880.jpg", "scene0575_01/00920.jpg", "scene0575_01/00960.jpg" ], "turn_id": 5, "scan_id": "scene0575_01", "origin_question": "How far is your current position from where you were at the end of the turn 1? (in meters)", "option": [], "answer": "1.98", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 1? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0575_01/01000.jpg", "scene0575_01/01040.jpg", "scene0575_01/01080.jpg", "scene0575_01/01120.jpg", "scene0575_01/01160.jpg" ], "turn_id": 6, "scan_id": "scene0575_01", "origin_question": "Is the cylindrical black bin with a glossy surface to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the cylindrical black bin with a glossy surface to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0575_01/01200.jpg", "scene0575_01/01240.jpg", "scene0575_01/01280.jpg", "scene0575_01/01320.jpg", "scene0575_01/01360.jpg" ], "turn_id": 7, "scan_id": "scene0575_01", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "1.28", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0575_01/01400.jpg", "scene0575_01/01440.jpg", "scene0575_01/01480.jpg", "scene0575_01/01520.jpg", "scene0575_01/01560.jpg" ], "turn_id": 8, "scan_id": "scene0575_01", "origin_question": "How far is your current position from where you were at the end of the turn 3? (in meters)", "option": [], "answer": "2.68", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 3? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0575_01/01600.jpg", "scene0575_01/01640.jpg", "scene0575_01/01680.jpg", "scene0575_01/01720.jpg", "scene0575_01/01760.jpg" ], "turn_id": 9, "scan_id": "scene0575_01", "origin_question": "By the end of the 6 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular frosted glass window with clean lines\", Y: \"the cylindrical black bin with a glossy surface\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to Y and farther away from X", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 6 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular frosted glass window with clean lines\", Y: \"the cylindrical black bin with a glossy surface\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0575_01/01800.jpg", "scene0575_01/01840.jpg", "scene0575_01/01880.jpg", "scene0575_01/01920.jpg", "scene0575_01/01960.jpg" ], "turn_id": 10, "scan_id": "scene0575_01", "origin_question": "Compared to your position at the end of the 5 turn, are you now closer or farther away from \"the cylindrical black bin with a glossy surface\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 5 turn, are you now closer or farther away from \"the cylindrical black bin with a glossy surface\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0074_region1/84055072c9154370bc2a76c86daf011c_i1_3.jpg", "1mp3d_0074_region1/84055072c9154370bc2a76c86daf011c_i2_3.jpg", "1mp3d_0074_region1/84055072c9154370bc2a76c86daf011c_i1_4.jpg", "1mp3d_0074_region1/84055072c9154370bc2a76c86daf011c_i2_4.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0074_region1", "origin_question": "Remember, have you seen any plant(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any plant(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "1mp3d_0074_region1/3e8f012538474392bf311dcec2e2153a_i1_5.jpg", "1mp3d_0074_region1/3e8f012538474392bf311dcec2e2153a_i2_5.jpg", "1mp3d_0074_region1/3e8f012538474392bf311dcec2e2153a_i1_0.jpg", "1mp3d_0074_region1/3e8f012538474392bf311dcec2e2153a_i2_0.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0074_region1", "origin_question": "When did you first discover \"the tall chair with a black metal frame and beige fabric seat\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the tall chair with a black metal frame and beige fabric seat\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "1mp3d_0074_region1/84055072c9154370bc2a76c86daf011c_i1_2.jpg", "1mp3d_0074_region1/84055072c9154370bc2a76c86daf011c_i2_2.jpg", "1mp3d_0074_region1/84055072c9154370bc2a76c86daf011c_i1_1.jpg", "1mp3d_0074_region1/84055072c9154370bc2a76c86daf011c_i2_1.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0074_region1", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the large rectangular white screen with blue and gray decorations\" from your position?", "option": [], "answer": "55.11", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the large rectangular white screen with blue and gray decorations\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "1mp3d_0074_region1/84055072c9154370bc2a76c86daf011c_i1_0.jpg", "1mp3d_0074_region1/84055072c9154370bc2a76c86daf011c_i2_0.jpg", "1mp3d_0074_region1/6ca981e907be49749ed4aead19864515_i1_0.jpg", "1mp3d_0074_region1/6ca981e907be49749ed4aead19864515_i2_0.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0074_region1", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "1mp3d_0074_region1/6ca981e907be49749ed4aead19864515_i1_2.jpg", "1mp3d_0074_region1/6ca981e907be49749ed4aead19864515_i2_2.jpg", "1mp3d_0074_region1/6ca981e907be49749ed4aead19864515_i1_1.jpg", "1mp3d_0074_region1/6ca981e907be49749ed4aead19864515_i2_1.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0074_region1", "origin_question": "Please recall \"the tall chair with a black metal frame and beige fabric seat\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.02", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the tall chair with a black metal frame and beige fabric seat\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "1mp3d_0074_region1/6ca981e907be49749ed4aead19864515_i1_3.jpg", "1mp3d_0074_region1/6ca981e907be49749ed4aead19864515_i2_3.jpg", "1mp3d_0074_region1/6ca981e907be49749ed4aead19864515_i1_5.jpg", "1mp3d_0074_region1/6ca981e907be49749ed4aead19864515_i2_5.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0074_region1", "origin_question": "Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0568_02/00000.jpg", "scene0568_02/00040.jpg", "scene0568_02/00080.jpg", "scene0568_02/00120.jpg" ], "turn_id": 1, "scan_id": "scene0568_02", "origin_question": "Remember, have you seen any couch(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any couch(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0568_02/00160.jpg", "scene0568_02/00200.jpg", "scene0568_02/00240.jpg", "scene0568_02/00280.jpg" ], "turn_id": 2, "scan_id": "scene0568_02", "origin_question": "When did you first discover \"the rectangular coffee table with a glass top\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the rectangular coffee table with a glass top\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0568_02/00320.jpg", "scene0568_02/00360.jpg", "scene0568_02/00400.jpg", "scene0568_02/00440.jpg" ], "turn_id": 3, "scan_id": "scene0568_02", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0568_02/00480.jpg", "scene0568_02/00520.jpg", "scene0568_02/00560.jpg", "scene0568_02/00600.jpg" ], "turn_id": 4, "scan_id": "scene0568_02", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the rectangular coffee table with a glass top\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the rectangular coffee table with a glass top\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0568_02/00640.jpg", "scene0568_02/00680.jpg", "scene0568_02/00720.jpg", "scene0568_02/00760.jpg" ], "turn_id": 5, "scan_id": "scene0568_02", "origin_question": "Which direction is the rectangular coffee table with a glass top in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the rectangular coffee table with a glass top in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0568_02/00800.jpg", "scene0568_02/00840.jpg", "scene0568_02/00880.jpg", "scene0568_02/00920.jpg" ], "turn_id": 6, "scan_id": "scene0568_02", "origin_question": "Which two objects are on the same side of you now? \"the bright red rectangular couch made of soft fabric\", \"the small rectangular white air conditioner with a modern design\", \"the vibrant rectangular soft carpet with intricate patterns\". ", "option": [ "the bright red rectangular couch made of soft fabric, the small rectangular white air conditioner with a modern design", "the bright red rectangular couch made of soft fabric, the vibrant rectangular soft carpet with intricate patterns", "the small rectangular white air conditioner with a modern design, the vibrant rectangular soft carpet with intricate patterns" ], "answer": "the bright red rectangular couch made of soft fabric, the small rectangular white air conditioner with a modern design", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the bright red rectangular couch made of soft fabric\", \"the small rectangular white air conditioner with a modern design\", \"the vibrant rectangular soft carpet with intricate patterns\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the bright red rectangular couch made of soft fabric, the small rectangular white air conditioner with a modern design', 'the bright red rectangular couch made of soft fabric, the vibrant rectangular soft carpet with intricate patterns', 'the small rectangular white air conditioner with a modern design, the vibrant rectangular soft carpet with intricate patterns'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0568_02/00960.jpg", "scene0568_02/01000.jpg", "scene0568_02/01040.jpg", "scene0568_02/01080.jpg" ], "turn_id": 7, "scan_id": "scene0568_02", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the gray rectangular metal stand with three shelves\"; \"the rectangular white plastic radiator\"; \"the white curtain with colorful edges made of cotton\" ", "option": [ "the gray rectangular metal stand with three shelves", "the rectangular white plastic radiator", "the white curtain with colorful edges made of cotton" ], "answer": "the gray rectangular metal stand with three shelves", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the gray rectangular metal stand with three shelves\"; \"the rectangular white plastic radiator\"; \"the white curtain with colorful edges made of cotton\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the gray rectangular metal stand with three shelves', 'the rectangular white plastic radiator', 'the white curtain with colorful edges made of cotton'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0568_02/01120.jpg", "scene0568_02/01160.jpg", "scene0568_02/01200.jpg", "scene0568_02/01240.jpg" ], "turn_id": 8, "scan_id": "scene0568_02", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the vibrant rectangular soft carpet with intricate patterns\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the vibrant rectangular soft carpet with intricate patterns\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0078_region11/c9030e71fc29447599a5c21d90261ca5_i2_0.jpg", "1mp3d_0078_region11/f461fdc9b5c54386bcb1350bf4072015_i1_5.jpg", "1mp3d_0078_region11/f461fdc9b5c54386bcb1350bf4072015_i2_5.jpg", "1mp3d_0078_region11/0fc2263dc29f4a9183320eddd33edb58_i1_0.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0078_region11", "origin_question": "Remember, have you seen any plant(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any plant(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0078_region11/0fc2263dc29f4a9183320eddd33edb58_i2_0.jpg", "1mp3d_0078_region11/f461fdc9b5c54386bcb1350bf4072015_i1_4.jpg", "1mp3d_0078_region11/f461fdc9b5c54386bcb1350bf4072015_i2_4.jpg", "1mp3d_0078_region11/f461fdc9b5c54386bcb1350bf4072015_i1_3.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0078_region11", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the tall, slender plant with green leaves\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the tall, slender plant with green leaves\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "1mp3d_0078_region11/f461fdc9b5c54386bcb1350bf4072015_i2_3.jpg", "1mp3d_0078_region11/2287f7e926a7402a8723eaf0db683d09_i1_2.jpg", "1mp3d_0078_region11/2287f7e926a7402a8723eaf0db683d09_i2_2.jpg", "1mp3d_0078_region11/a26c4e73940a44c89caa6788fcf4ca97_i1_2.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0078_region11", "origin_question": "Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "80.43", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "1mp3d_0078_region11/a26c4e73940a44c89caa6788fcf4ca97_i2_2.jpg", "1mp3d_0078_region11/e47436300f4f4b90a2415ea87eee0903_i1_3.jpg", "1mp3d_0078_region11/e47436300f4f4b90a2415ea87eee0903_i2_3.jpg", "1mp3d_0078_region11/e52749304010410ba3a1e058fe627b6d_i1_2.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0078_region11", "origin_question": "Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "108.51", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "1mp3d_0078_region11/e52749304010410ba3a1e058fe627b6d_i2_2.jpg", "1mp3d_0078_region11/7ab6f8baf8bc41628a3b91fa4660ce43_i1_0.jpg", "1mp3d_0078_region11/7ab6f8baf8bc41628a3b91fa4660ce43_i2_0.jpg", "1mp3d_0078_region11/99efd71d29484d2c837ab4a49750fab0_i1_0.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0078_region11", "origin_question": "By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the washbasin\", Y: \"the round metal light with a reflective surface\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "farther away from both X and Y", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the washbasin\", Y: \"the round metal light with a reflective surface\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0078_region11/99efd71d29484d2c837ab4a49750fab0_i2_0.jpg", "1mp3d_0078_region11/99efd71d29484d2c837ab4a49750fab0_i2_1.jpg", "1mp3d_0078_region11/7ab6f8baf8bc41628a3b91fa4660ce43_i1_1.jpg", "1mp3d_0078_region11/7ab6f8baf8bc41628a3b91fa4660ce43_i2_1.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0078_region11", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the round metal light with a reflective surface\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the round metal light with a reflective surface\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "1mp3d_0078_region11/7bbddf0c8df249a3b8a6da60b1b7fd57_i1_3.jpg", "1mp3d_0078_region11/7bbddf0c8df249a3b8a6da60b1b7fd57_i2_3.jpg", "1mp3d_0078_region11/1576d62e7bbb45e8a5ef9e7bb37b1839_i1_3.jpg", "1mp3d_0078_region11/1576d62e7bbb45e8a5ef9e7bb37b1839_i2_3.jpg" ], "turn_id": 7, "scan_id": "1mp3d_0078_region11", "origin_question": "Using your orientation at the end of turn 5 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 5 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0078_region11/1576d62e7bbb45e8a5ef9e7bb37b1839_i1_4.jpg", "1mp3d_0078_region11/1576d62e7bbb45e8a5ef9e7bb37b1839_i2_4.jpg", "1mp3d_0078_region11/32cd64f9c7d74e7185517bca189a00b4_i1_2.jpg", "1mp3d_0078_region11/32cd64f9c7d74e7185517bca189a00b4_i1_1.jpg" ], "turn_id": 8, "scan_id": "1mp3d_0078_region11", "origin_question": "Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the round metal light with a reflective surface\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the round metal light with a reflective surface\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0578_00/00000.jpg", "scene0578_00/00040.jpg", "scene0578_00/00080.jpg", "scene0578_00/00120.jpg" ], "turn_id": 1, "scan_id": "scene0578_00", "origin_question": "Remember, have you seen any table(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any table(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0578_00/00160.jpg", "scene0578_00/00200.jpg", "scene0578_00/00240.jpg", "scene0578_00/00280.jpg" ], "turn_id": 2, "scan_id": "scene0578_00", "origin_question": "In which turn were you closest to \"the rectangular wooden window with clean glass panes\"? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the rectangular wooden window with clean glass panes\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0578_00/00320.jpg", "scene0578_00/00360.jpg", "scene0578_00/00400.jpg", "scene0578_00/00440.jpg" ], "turn_id": 3, "scan_id": "scene0578_00", "origin_question": "What will be the first-time appearance order of the following categories: monitor, speaker, table?", "option": [ "table, monitor, speaker", "monitor, speaker, table", "monitor, table, speaker", "table, speaker, monitor" ], "answer": "table, monitor, speaker", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: monitor, speaker, table? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['table, monitor, speaker', 'monitor, speaker, table', 'monitor, table, speaker', 'table, speaker, monitor'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0578_00/00480.jpg", "scene0578_00/00520.jpg", "scene0578_00/00560.jpg", "scene0578_00/00600.jpg" ], "turn_id": 4, "scan_id": "scene0578_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0578_00/00640.jpg", "scene0578_00/00680.jpg", "scene0578_00/00720.jpg", "scene0578_00/00760.jpg" ], "turn_id": 5, "scan_id": "scene0578_00", "origin_question": "Is the flat monitor with black borders and a silver base to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the flat monitor with black borders and a silver base to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0578_00/00800.jpg", "scene0578_00/00840.jpg", "scene0578_00/00880.jpg", "scene0578_00/00920.jpg" ], "turn_id": 6, "scan_id": "scene0578_00", "origin_question": "At the end of which turn were both of \"the medium-sized, deep navy blue backpack made of nylon\" and \"the flat monitor with black borders and a silver base\" on your left side?", "option": [], "answer": "5", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the medium-sized, deep navy blue backpack made of nylon\" and \"the flat monitor with black borders and a silver base\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0578_00/00960.jpg", "scene0578_00/01000.jpg", "scene0578_00/01040.jpg", "scene0578_00/01080.jpg" ], "turn_id": 7, "scan_id": "scene0578_00", "origin_question": "Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the medium-sized, deep navy blue backpack made of nylon\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the medium-sized, deep navy blue backpack made of nylon\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0578_00/01120.jpg", "scene0578_00/01160.jpg", "scene0578_00/01200.jpg", "scene0578_00/01240.jpg" ], "turn_id": 8, "scan_id": "scene0578_00", "origin_question": "By the end of the 4 turn, were you getting closer to or farther away from these two objects: X: \"the flat monitor with black borders and a silver base\", Y: \"the cylindrical black plastic bin\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to Y and farther away from X", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 4 turn, were you getting closer to or farther away from these two objects: X: \"the flat monitor with black borders and a silver base\", Y: \"the cylindrical black plastic bin\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0578_00/01280.jpg", "scene0578_00/01320.jpg", "scene0578_00/01360.jpg", "scene0578_00/01400.jpg" ], "turn_id": 9, "scan_id": "scene0578_00", "origin_question": "Is the medium-sized, deep navy blue backpack made of nylon to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 9 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the medium-sized, deep navy blue backpack made of nylon to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0269_02/00000.jpg", "scene0269_02/00030.jpg", "scene0269_02/00060.jpg", "scene0269_02/00090.jpg" ], "turn_id": 1, "scan_id": "scene0269_02", "origin_question": "Remember, have you seen any ladder(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any ladder(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0269_02/00120.jpg", "scene0269_02/00150.jpg", "scene0269_02/00180.jpg", "scene0269_02/00210.jpg" ], "turn_id": 2, "scan_id": "scene0269_02", "origin_question": "Remember, have you seen any package(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any package(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0269_02/00240.jpg", "scene0269_02/00270.jpg", "scene0269_02/00300.jpg", "scene0269_02/00330.jpg" ], "turn_id": 3, "scan_id": "scene0269_02", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0269_02/00360.jpg", "scene0269_02/00390.jpg", "scene0269_02/00420.jpg", "scene0269_02/00450.jpg" ], "turn_id": 4, "scan_id": "scene0269_02", "origin_question": "When did you first discover \"the door\"? (index of the turn)", "option": [], "answer": "3", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the door\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0269_02/00480.jpg", "scene0269_02/00510.jpg", "scene0269_02/00540.jpg", "scene0269_02/00570.jpg" ], "turn_id": 5, "scan_id": "scene0269_02", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0269_02/00600.jpg", "scene0269_02/00630.jpg", "scene0269_02/00660.jpg", "scene0269_02/00690.jpg" ], "turn_id": 6, "scan_id": "scene0269_02", "origin_question": "Using your orientation at the end of turn 4 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "150.75", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 4 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0269_02/00720.jpg", "scene0269_02/00750.jpg", "scene0269_02/00780.jpg", "scene0269_02/00810.jpg" ], "turn_id": 7, "scan_id": "scene0269_02", "origin_question": "Using your orientation at the end of turn 6 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "95.11", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 6 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0040_region40/b3fa9ba954b442449ed078b050cd2a94_i1_0.jpg", "1mp3d_0040_region40/b3fa9ba954b442449ed078b050cd2a94_i2_0.jpg", "1mp3d_0040_region40/3398c144f24745bda02fd6895b58361e_i1_0.jpg", "1mp3d_0040_region40/3398c144f24745bda02fd6895b58361e_i2_0.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0040_region40", "origin_question": "Remember, have you seen any picture(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any picture(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0040_region40/aa833bdadf00424eaf72f5245556ad77_i1_5.jpg", "1mp3d_0040_region40/aa833bdadf00424eaf72f5245556ad77_i2_5.jpg", "1mp3d_0040_region40/3193cd5f6d7a4b2bbf85d5043cbc6352_i1_4.jpg", "1mp3d_0040_region40/3193cd5f6d7a4b2bbf85d5043cbc6352_i2_4.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0040_region40", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "1mp3d_0040_region40/421a3c211c2c4abdb53c55ab1c2e21eb_i1_4.jpg", "1mp3d_0040_region40/421a3c211c2c4abdb53c55ab1c2e21eb_i2_4.jpg", "1mp3d_0040_region40/3193cd5f6d7a4b2bbf85d5043cbc6352_i1_5.jpg", "1mp3d_0040_region40/3193cd5f6d7a4b2bbf85d5043cbc6352_i2_5.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0040_region40", "origin_question": "Please recall \"the concrete arch with a rounded top and beige color\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "5.09", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the concrete arch with a rounded top and beige color\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0040_region40/aa833bdadf00424eaf72f5245556ad77_i1_2.jpg", "1mp3d_0040_region40/aa833bdadf00424eaf72f5245556ad77_i2_2.jpg", "1mp3d_0040_region40/3398c144f24745bda02fd6895b58361e_i1_5.jpg", "1mp3d_0040_region40/3398c144f24745bda02fd6895b58361e_i2_5.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0040_region40", "origin_question": "", "option": [], "answer": "", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "1mp3d_0040_region40/b3fa9ba954b442449ed078b050cd2a94_i1_5.jpg", "1mp3d_0040_region40/b3fa9ba954b442449ed078b050cd2a94_i2_5.jpg", "1mp3d_0040_region40/3398c144f24745bda02fd6895b58361e_i1_4.jpg", "1mp3d_0040_region40/3398c144f24745bda02fd6895b58361e_i2_4.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0040_region40", "origin_question": "Please recall \"the grand, deep brown decoration\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.16", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the grand, deep brown decoration\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "1mp3d_0040_region40/aa833bdadf00424eaf72f5245556ad77_i1_1.jpg", "1mp3d_0040_region40/aa833bdadf00424eaf72f5245556ad77_i2_1.jpg", "1mp3d_0040_region40/2ad15b07cbe24771b6217553df356bea_i1_2.jpg", "1mp3d_0040_region40/2ad15b07cbe24771b6217553df356bea_i2_2.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0040_region40", "origin_question": "What will be the first-time appearance order of the following categories: arch, picture, window?", "option": [ "arch, picture, window", "arch, window, picture", "picture, arch, window", "window, picture, arch" ], "answer": "picture, arch, window", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: arch, picture, window? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['arch, picture, window', 'arch, window, picture', 'picture, arch, window', 'window, picture, arch'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0697_01/00000.jpg", "scene0697_01/00040.jpg", "scene0697_01/00080.jpg", "scene0697_01/00120.jpg", "scene0697_01/00160.jpg" ], "turn_id": 1, "scan_id": "scene0697_01", "origin_question": "Remember, have you seen any bed(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any bed(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0697_01/00200.jpg", "scene0697_01/00240.jpg", "scene0697_01/00280.jpg", "scene0697_01/00320.jpg", "scene0697_01/00360.jpg" ], "turn_id": 2, "scan_id": "scene0697_01", "origin_question": "In which turn did you see both \"the standard queen-size bed with a floral bedsheet\" and \"the silver round metal clock with a clear dial\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the standard queen-size bed with a floral bedsheet\" and \"the silver round metal clock with a clear dial\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0697_01/00400.jpg", "scene0697_01/00440.jpg", "scene0697_01/00480.jpg", "scene0697_01/00520.jpg", "scene0697_01/00560.jpg" ], "turn_id": 3, "scan_id": "scene0697_01", "origin_question": "", "option": [], "answer": "", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0697_01/00600.jpg", "scene0697_01/00640.jpg", "scene0697_01/00680.jpg", "scene0697_01/00720.jpg", "scene0697_01/00760.jpg" ], "turn_id": 4, "scan_id": "scene0697_01", "origin_question": "In which turn did you see both \"the standard queen-size bed with a floral bedsheet\" and \"the rectangular tissue box with purple and white patterns\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the standard queen-size bed with a floral bedsheet\" and \"the rectangular tissue box with purple and white patterns\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0697_01/00800.jpg", "scene0697_01/00840.jpg", "scene0697_01/00880.jpg", "scene0697_01/00920.jpg", "scene0697_01/00960.jpg" ], "turn_id": 5, "scan_id": "scene0697_01", "origin_question": "Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the rectangular plastic bottle with a blue pump head\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the rectangular plastic bottle with a blue pump head\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0697_01/01000.jpg", "scene0697_01/01040.jpg", "scene0697_01/01080.jpg", "scene0697_01/01120.jpg", "scene0697_01/01160.jpg" ], "turn_id": 6, "scan_id": "scene0697_01", "origin_question": "What will be the first-time appearance order of the following categories: tissue box, vanity, window?", "option": [ "vanity, window, tissue box", "window, tissue box, vanity", "window, vanity, tissue box", "tissue box, window, vanity" ], "answer": "tissue box, window, vanity", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: tissue box, vanity, window? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['vanity, window, tissue box', 'window, tissue box, vanity', 'window, vanity, tissue box', 'tissue box, window, vanity'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "scene0697_01/01200.jpg", "scene0697_01/01240.jpg", "scene0697_01/01280.jpg", "scene0697_01/01320.jpg", "scene0697_01/01360.jpg" ], "turn_id": 7, "scan_id": "scene0697_01", "origin_question": "Among these three objects(\"the black-framed picture with five family photos\",\"the switch\",\"the bin\" ), which one is the farthest from you?", "option": [ "the black-framed picture with five family photos", "the switch", "the bin" ], "answer": "the black-framed picture with five family photos", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the black-framed picture with five family photos\",\"the switch\",\"the bin\" ), which one is the farthest from you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the black-framed picture with five family photos', 'the switch', 'the bin'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0697_01/01400.jpg", "scene0697_01/01440.jpg", "scene0697_01/01480.jpg", "scene0697_01/01520.jpg", "scene0697_01/01560.jpg" ], "turn_id": 8, "scan_id": "scene0697_01", "origin_question": "Is the bin to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the bin to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "scene0697_01/01600.jpg", "scene0697_01/01640.jpg", "scene0697_01/01680.jpg", "scene0697_01/01720.jpg", "scene0697_01/01760.jpg" ], "turn_id": 9, "scan_id": "scene0697_01", "origin_question": "Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Left", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0697_01/01800.jpg", "scene0697_01/01840.jpg", "scene0697_01/01880.jpg", "scene0697_01/01920.jpg", "scene0697_01/01960.jpg" ], "turn_id": 10, "scan_id": "scene0697_01", "origin_question": "Which direction is the silver round metal clock with a clear dial in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the silver round metal clock with a clear dial in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0695_01/00000.jpg", "scene0695_01/00040.jpg", "scene0695_01/00080.jpg", "scene0695_01/00490.jpg", "scene0695_01/00530.jpg" ], "turn_id": 1, "scan_id": "scene0695_01", "origin_question": "Remember, have you seen any backpack(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any backpack(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0695_01/00570.jpg", "scene0695_01/00610.jpg", "scene0695_01/00650.jpg", "scene0695_01/00690.jpg", "scene0695_01/00730.jpg" ], "turn_id": 2, "scan_id": "scene0695_01", "origin_question": "When did you first discover \"the cup\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the cup\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0695_01/00770.jpg", "scene0695_01/00810.jpg", "scene0695_01/00850.jpg", "scene0695_01/00890.jpg", "scene0695_01/00930.jpg" ], "turn_id": 3, "scan_id": "scene0695_01", "origin_question": "Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "79.26", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0695_01/00970.jpg", "scene0695_01/01010.jpg", "scene0695_01/01050.jpg", "scene0695_01/01090.jpg", "scene0695_01/01130.jpg" ], "turn_id": 4, "scan_id": "scene0695_01", "origin_question": "In which turn were you farthest from \"the bed\"? ", "option": [], "answer": "3", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the bed\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0695_01/01170.jpg", "scene0695_01/01210.jpg", "scene0695_01/01250.jpg", "scene0695_01/01290.jpg", "scene0695_01/01330.jpg" ], "turn_id": 5, "scan_id": "scene0695_01", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the cylindrical silver mesh basket made of plastic\"; \"the rectangular light-colored mat with subtle patterns\"; \"the book\" ", "option": [ "the cylindrical silver mesh basket made of plastic", "the rectangular light-colored mat with subtle patterns", "the book" ], "answer": "the cylindrical silver mesh basket made of plastic", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the cylindrical silver mesh basket made of plastic\"; \"the rectangular light-colored mat with subtle patterns\"; \"the book\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the cylindrical silver mesh basket made of plastic', 'the rectangular light-colored mat with subtle patterns', 'the book'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0695_01/01370.jpg", "scene0695_01/01410.jpg", "scene0695_01/01450.jpg", "scene0695_01/01490.jpg", "scene0695_01/01530.jpg" ], "turn_id": 6, "scan_id": "scene0695_01", "origin_question": "Which two objects are on the same side of you now? \"the bed\", \"the rectangular wooden shelf of moderate size\", \"the white rectangular remote control with several buttons\". ", "option": [ "the bed, the rectangular wooden shelf of moderate size", "the bed, the white rectangular remote control with several buttons", "the rectangular wooden shelf of moderate size, the white rectangular remote control with several buttons" ], "answer": "the rectangular wooden shelf of moderate size, the white rectangular remote control with several buttons", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the bed\", \"the rectangular wooden shelf of moderate size\", \"the white rectangular remote control with several buttons\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the bed, the rectangular wooden shelf of moderate size', 'the bed, the white rectangular remote control with several buttons', 'the rectangular wooden shelf of moderate size, the white rectangular remote control with several buttons'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0695_01/01570.jpg", "scene0695_01/01610.jpg", "scene0695_01/01650.jpg", "scene0695_01/01690.jpg", "scene0695_01/01730.jpg" ], "turn_id": 7, "scan_id": "scene0695_01", "origin_question": "Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0695_01/01770.jpg", "scene0695_01/01810.jpg", "scene0695_01/01850.jpg", "scene0695_01/01890.jpg", "scene0695_01/01930.jpg" ], "turn_id": 8, "scan_id": "scene0695_01", "origin_question": "Which two objects are on the same side of you now? \"the slim, silver laptop with a turquoise decoration\", \"the cup\", \"the cylindrical silver mesh basket made of plastic\". ", "option": [ "the slim, silver laptop with a turquoise decoration, the cup", "the slim, silver laptop with a turquoise decoration, the cylindrical silver mesh basket made of plastic", "the cup, the cylindrical silver mesh basket made of plastic" ], "answer": "the slim, silver laptop with a turquoise decoration, the cup", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the slim, silver laptop with a turquoise decoration\", \"the cup\", \"the cylindrical silver mesh basket made of plastic\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the slim, silver laptop with a turquoise decoration, the cup', 'the slim, silver laptop with a turquoise decoration, the cylindrical silver mesh basket made of plastic', 'the cup, the cylindrical silver mesh basket made of plastic'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0072_01/00000.jpg", "scene0072_01/00040.jpg", "scene0072_01/00080.jpg", "scene0072_01/00120.jpg" ], "turn_id": 1, "scan_id": "scene0072_01", "origin_question": "Remember, how many dustpan(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many dustpan(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0072_01/00160.jpg", "scene0072_01/00200.jpg", "scene0072_01/00240.jpg", "scene0072_01/00280.jpg" ], "turn_id": 2, "scan_id": "scene0072_01", "origin_question": "When did you first discover \"the rustic cylindrical light made of wood and metal\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the rustic cylindrical light made of wood and metal\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0072_01/00320.jpg", "scene0072_01/00360.jpg", "scene0072_01/00400.jpg", "scene0072_01/00440.jpg" ], "turn_id": 3, "scan_id": "scene0072_01", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the medium-sized rectangular light-colored shelf\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the medium-sized rectangular light-colored shelf\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0072_01/00480.jpg", "scene0072_01/00520.jpg", "scene0072_01/00560.jpg", "scene0072_01/00600.jpg" ], "turn_id": 4, "scan_id": "scene0072_01", "origin_question": "Which two objects are on the same side of you now? \"the wardrobe\", \"the simple wooden stool with a blue cushion seat\", \"the simple single bed with a yellow bedsheet\". ", "option": [ "the wardrobe, the simple wooden stool with a blue cushion seat", "the wardrobe, the simple single bed with a yellow bedsheet", "the simple wooden stool with a blue cushion seat, the simple single bed with a yellow bedsheet" ], "answer": "the simple wooden stool with a blue cushion seat, the simple single bed with a yellow bedsheet", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the wardrobe\", \"the simple wooden stool with a blue cushion seat\", \"the simple single bed with a yellow bedsheet\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the wardrobe, the simple wooden stool with a blue cushion seat', 'the wardrobe, the simple single bed with a yellow bedsheet', 'the simple wooden stool with a blue cushion seat, the simple single bed with a yellow bedsheet'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0072_01/00640.jpg", "scene0072_01/00680.jpg", "scene0072_01/00720.jpg", "scene0072_01/00760.jpg" ], "turn_id": 5, "scan_id": "scene0072_01", "origin_question": "In which turn were you closest to \"the large rectangular natural wood cabinet with transparent doors\"? ", "option": [], "answer": "5", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the large rectangular natural wood cabinet with transparent doors\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0072_01/00800.jpg", "scene0072_01/00840.jpg", "scene0072_01/00880.jpg", "scene0072_01/00920.jpg" ], "turn_id": 6, "scan_id": "scene0072_01", "origin_question": "At the end of which turn were both of \"the rustic cylindrical light made of wood and metal\" and \"the large rectangular natural wood cabinet with transparent doors\" on your left side?", "option": [], "answer": "2", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the rustic cylindrical light made of wood and metal\" and \"the large rectangular natural wood cabinet with transparent doors\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "scene0072_01/00960.jpg", "scene0072_01/01000.jpg", "scene0072_01/01040.jpg", "scene0072_01/01080.jpg" ], "turn_id": 7, "scan_id": "scene0072_01", "origin_question": "Among these three objects(\"the rectangular white cast iron radiator\",\"the large rectangular natural wood cabinet with transparent doors\",\"the round, natural brown woven basket\" ), which one is the closest to you?", "option": [ "the rectangular white cast iron radiator", "the large rectangular natural wood cabinet with transparent doors", "the round, natural brown woven basket" ], "answer": "the large rectangular natural wood cabinet with transparent doors", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects(\"the rectangular white cast iron radiator\",\"the large rectangular natural wood cabinet with transparent doors\",\"the round, natural brown woven basket\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular white cast iron radiator', 'the large rectangular natural wood cabinet with transparent doors', 'the round, natural brown woven basket'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0057_region16/f658e9b6b19e471c8c98f9175b03d8a1_i1_5.jpg", "1mp3d_0057_region16/f658e9b6b19e471c8c98f9175b03d8a1_i2_5.jpg", "1mp3d_0057_region16/f658e9b6b19e471c8c98f9175b03d8a1_i1_0.jpg", "1mp3d_0057_region16/f658e9b6b19e471c8c98f9175b03d8a1_i2_0.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0057_region16", "origin_question": "Remember, have you seen any towel(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any towel(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0057_region16/3fd138436a1c483593a8730fddff5187_i1_4.jpg", "1mp3d_0057_region16/3fd138436a1c483593a8730fddff5187_i2_4.jpg", "1mp3d_0057_region16/d4507886bc484283a312687dc32d3740_i1_5.jpg", "1mp3d_0057_region16/d4507886bc484283a312687dc32d3740_i2_5.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0057_region16", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0057_region16/0c17fd1d27164de4ad2274e6e0d64a89_i1_5.jpg", "1mp3d_0057_region16/0c17fd1d27164de4ad2274e6e0d64a89_i2_5.jpg", "1mp3d_0057_region16/d4507886bc484283a312687dc32d3740_i1_0.jpg", "1mp3d_0057_region16/d4507886bc484283a312687dc32d3740_i2_0.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0057_region16", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the rectangular mirror with a brown border\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the rectangular mirror with a brown border\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "1mp3d_0057_region16/3fd138436a1c483593a8730fddff5187_i1_0.jpg", "1mp3d_0057_region16/3fd138436a1c483593a8730fddff5187_i2_0.jpg", "1mp3d_0057_region16/3fd138436a1c483593a8730fddff5187_i1_1.jpg", "1mp3d_0057_region16/3fd138436a1c483593a8730fddff5187_i2_1.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0057_region16", "origin_question": "What will be the first-time appearance order of the following categories: mirror, toilet, towel?", "option": [ "towel, mirror, toilet", "mirror, toilet, towel", "toilet, mirror, towel", "toilet, towel, mirror" ], "answer": "towel, mirror, toilet", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: mirror, toilet, towel? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['towel, mirror, toilet', 'mirror, toilet, towel', 'toilet, mirror, towel', 'toilet, towel, mirror'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "1mp3d_0057_region16/d4507886bc484283a312687dc32d3740_i1_2.jpg", "1mp3d_0057_region16/d4507886bc484283a312687dc32d3740_i2_2.jpg", "1mp3d_0057_region16/d4507886bc484283a312687dc32d3740_i2_3.jpg", "1mp3d_0057_region16/0c17fd1d27164de4ad2274e6e0d64a89_i1_4.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0057_region16", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the fabric curtain with a brown and beige grid pattern\", Y: \"the sink\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to Y and farther away from X", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the fabric curtain with a brown and beige grid pattern\", Y: \"the sink\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0471_02/00000.jpg", "scene0471_02/00030.jpg", "scene0471_02/00060.jpg", "scene0471_02/00090.jpg" ], "turn_id": 1, "scan_id": "scene0471_02", "origin_question": "Remember, how many piano(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many piano(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0471_02/00120.jpg", "scene0471_02/00150.jpg", "scene0471_02/00180.jpg", "scene0471_02/00210.jpg" ], "turn_id": 2, "scan_id": "scene0471_02", "origin_question": "In which turn were you closest to \"the dispenser\"? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the dispenser\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0471_02/00240.jpg", "scene0471_02/00270.jpg", "scene0471_02/00300.jpg", "scene0471_02/00330.jpg" ], "turn_id": 3, "scan_id": "scene0471_02", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the dispenser\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the dispenser\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0471_02/00360.jpg", "scene0471_02/00390.jpg", "scene0471_02/00420.jpg", "scene0471_02/00450.jpg" ], "turn_id": 4, "scan_id": "scene0471_02", "origin_question": "Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0471_02/00480.jpg", "scene0471_02/00510.jpg", "scene0471_02/00540.jpg", "scene0471_02/00570.jpg" ], "turn_id": 5, "scan_id": "scene0471_02", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0262_01/00000.jpg", "scene0262_01/00040.jpg", "scene0262_01/00080.jpg", "scene0262_01/00120.jpg", "scene0262_01/00160.jpg" ], "turn_id": 1, "scan_id": "scene0262_01", "origin_question": "Remember, how many stove(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many stove(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0262_01/00200.jpg", "scene0262_01/00240.jpg", "scene0262_01/00280.jpg", "scene0262_01/00320.jpg", "scene0262_01/00360.jpg" ], "turn_id": 2, "scan_id": "scene0262_01", "origin_question": "In which turn were you farthest from \"the lamp\"? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the lamp\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0262_01/00400.jpg", "scene0262_01/00440.jpg", "scene0262_01/00480.jpg", "scene0262_01/00520.jpg", "scene0262_01/00560.jpg" ], "turn_id": 3, "scan_id": "scene0262_01", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "1.4", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0262_01/00600.jpg", "scene0262_01/00640.jpg", "scene0262_01/00680.jpg", "scene0262_01/00720.jpg", "scene0262_01/00760.jpg" ], "turn_id": 4, "scan_id": "scene0262_01", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the small, grey wooden rectangular cabinet\"; \"the blue hardcover book with paper material\"; \"the small transparent glass bottle with a black cap\" ", "option": [ "the small, grey wooden rectangular cabinet", "the blue hardcover book with paper material", "the small transparent glass bottle with a black cap" ], "answer": "the small transparent glass bottle with a black cap", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the small, grey wooden rectangular cabinet\"; \"the blue hardcover book with paper material\"; \"the small transparent glass bottle with a black cap\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the small, grey wooden rectangular cabinet', 'the blue hardcover book with paper material', 'the small transparent glass bottle with a black cap'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0262_01/00800.jpg", "scene0262_01/00840.jpg", "scene0262_01/00880.jpg", "scene0262_01/00920.jpg", "scene0262_01/00960.jpg" ], "turn_id": 5, "scan_id": "scene0262_01", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0262_01/01000.jpg", "scene0262_01/01040.jpg", "scene0262_01/01080.jpg", "scene0262_01/01120.jpg", "scene0262_01/01160.jpg" ], "turn_id": 6, "scan_id": "scene0262_01", "origin_question": "Which direction is the small rectangular picture with a white shiny frame in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the small rectangular picture with a white shiny frame in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0262_01/01200.jpg", "scene0262_01/01240.jpg", "scene0262_01/01280.jpg", "scene0262_01/01320.jpg", "scene0262_01/01360.jpg" ], "turn_id": 7, "scan_id": "scene0262_01", "origin_question": "Which two objects are on the same side of you now? \"the black and silver rectangular telephone\", \"the rectangular white plastic keyboard with black keycaps\", \"the gray, single-person couch made of soft fabric\". ", "option": [ "the black and silver rectangular telephone, the rectangular white plastic keyboard with black keycaps", "the black and silver rectangular telephone, the gray, single-person couch made of soft fabric", "the rectangular white plastic keyboard with black keycaps, the gray, single-person couch made of soft fabric" ], "answer": "the black and silver rectangular telephone, the rectangular white plastic keyboard with black keycaps", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the black and silver rectangular telephone\", \"the rectangular white plastic keyboard with black keycaps\", \"the gray, single-person couch made of soft fabric\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the black and silver rectangular telephone, the rectangular white plastic keyboard with black keycaps', 'the black and silver rectangular telephone, the gray, single-person couch made of soft fabric', 'the rectangular white plastic keyboard with black keycaps, the gray, single-person couch made of soft fabric'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0262_01/01400.jpg", "scene0262_01/01440.jpg", "scene0262_01/01480.jpg", "scene0262_01/01520.jpg", "scene0262_01/01560.jpg" ], "turn_id": 8, "scan_id": "scene0262_01", "origin_question": "", "option": [], "answer": "", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0262_01/01600.jpg", "scene0262_01/01640.jpg", "scene0262_01/01680.jpg", "scene0262_01/01720.jpg", "scene0262_01/01760.jpg" ], "turn_id": 9, "scan_id": "scene0262_01", "origin_question": "In which turn did you see both \"the lamp\" and \"the white rectangular desk made of sturdy materials\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the lamp\" and \"the white rectangular desk made of sturdy materials\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0262_01/01800.jpg", "scene0262_01/01840.jpg", "scene0262_01/01880.jpg", "scene0262_01/01920.jpg", "scene0262_01/01960.jpg" ], "turn_id": 10, "scan_id": "scene0262_01", "origin_question": "Which direction is the gray, single-person couch made of soft fabric in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the gray, single-person couch made of soft fabric in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0608_02/00000.jpg", "scene0608_02/00040.jpg", "scene0608_02/00080.jpg", "scene0608_02/00120.jpg", "scene0608_02/00160.jpg" ], "turn_id": 1, "scan_id": "scene0608_02", "origin_question": "Remember, have you seen any plant(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any plant(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0608_02/00200.jpg", "scene0608_02/00240.jpg", "scene0608_02/00280.jpg", "scene0608_02/00320.jpg", "scene0608_02/00360.jpg" ], "turn_id": 2, "scan_id": "scene0608_02", "origin_question": "In which turn did you see both \"the window\" and \"the tv\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the window\" and \"the tv\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0608_02/00400.jpg", "scene0608_02/00440.jpg", "scene0608_02/00480.jpg", "scene0608_02/00520.jpg", "scene0608_02/00560.jpg" ], "turn_id": 3, "scan_id": "scene0608_02", "origin_question": "Which two objects are on the same side of you now? \"the window\", \"the stylish black rectangular stand with closed doors\", \"the three-seater green fabric couch with rolled armrests\". ", "option": [ "the window, the stylish black rectangular stand with closed doors", "the window, the three-seater green fabric couch with rolled armrests", "the stylish black rectangular stand with closed doors, the three-seater green fabric couch with rolled armrests" ], "answer": "the window, the stylish black rectangular stand with closed doors", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the window\", \"the stylish black rectangular stand with closed doors\", \"the three-seater green fabric couch with rolled armrests\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the window, the stylish black rectangular stand with closed doors', 'the window, the three-seater green fabric couch with rolled armrests', 'the stylish black rectangular stand with closed doors, the three-seater green fabric couch with rolled armrests'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0608_02/00600.jpg", "scene0608_02/00640.jpg", "scene0608_02/00680.jpg", "scene0608_02/00720.jpg", "scene0608_02/00760.jpg" ], "turn_id": 4, "scan_id": "scene0608_02", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "31.91", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0608_02/00800.jpg", "scene0608_02/00840.jpg", "scene0608_02/00880.jpg", "scene0608_02/00920.jpg", "scene0608_02/00960.jpg" ], "turn_id": 5, "scan_id": "scene0608_02", "origin_question": "Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the classic black and white piano\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the classic black and white piano\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0608_02/01000.jpg", "scene0608_02/01040.jpg", "scene0608_02/01080.jpg", "scene0608_02/01120.jpg", "scene0608_02/01160.jpg" ], "turn_id": 6, "scan_id": "scene0608_02", "origin_question": "Please recall \"the window\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.87", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the window\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0608_02/01200.jpg", "scene0608_02/01240.jpg", "scene0608_02/01280.jpg", "scene0608_02/01320.jpg", "scene0608_02/01360.jpg" ], "turn_id": 7, "scan_id": "scene0608_02", "origin_question": "Is the window to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the window to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0608_02/01400.jpg", "scene0608_02/01440.jpg", "scene0608_02/01480.jpg", "scene0608_02/01520.jpg", "scene0608_02/01560.jpg" ], "turn_id": 8, "scan_id": "scene0608_02", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0608_02/01600.jpg", "scene0608_02/01640.jpg", "scene0608_02/01680.jpg", "scene0608_02/01720.jpg", "scene0608_02/01760.jpg" ], "turn_id": 9, "scan_id": "scene0608_02", "origin_question": "Is the tv to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the tv to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0608_02/01800.jpg", "scene0608_02/01840.jpg", "scene0608_02/01880.jpg", "scene0608_02/01920.jpg", "scene0608_02/01960.jpg" ], "turn_id": 10, "scan_id": "scene0608_02", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0488_00/00000.jpg", "scene0488_00/00030.jpg", "scene0488_00/00060.jpg", "scene0488_00/00090.jpg" ], "turn_id": 1, "scan_id": "scene0488_00", "origin_question": "Remember, have you seen any counter(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any counter(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0488_00/00120.jpg", "scene0488_00/00150.jpg", "scene0488_00/00180.jpg", "scene0488_00/00210.jpg" ], "turn_id": 2, "scan_id": "scene0488_00", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the white rectangular electric stove with four burners\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the white rectangular electric stove with four burners\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0488_00/00240.jpg", "scene0488_00/00270.jpg", "scene0488_00/00300.jpg", "scene0488_00/00330.jpg" ], "turn_id": 3, "scan_id": "scene0488_00", "origin_question": "Please recall \"the white rectangular microwave with a digital display\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.83", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the white rectangular microwave with a digital display\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0488_00/00360.jpg", "scene0488_00/00390.jpg", "scene0488_00/00420.jpg", "scene0488_00/00450.jpg" ], "turn_id": 4, "scan_id": "scene0488_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0488_00/00480.jpg", "scene0488_00/00510.jpg", "scene0488_00/00540.jpg", "scene0488_00/00570.jpg" ], "turn_id": 5, "scan_id": "scene0488_00", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the white rectangular microwave with a digital display\" from your position?", "option": [], "answer": "69.61", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the white rectangular microwave with a digital display\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0671_00/00000.jpg", "scene0671_00/00030.jpg", "scene0671_00/00060.jpg", "scene0671_00/00090.jpg" ], "turn_id": 1, "scan_id": "scene0671_00", "origin_question": "Remember, how many backpack(s) have you seen so far? ", "option": [], "answer": "2", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many backpack(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0671_00/00120.jpg", "scene0671_00/00150.jpg", "scene0671_00/00180.jpg", "scene0671_00/00210.jpg" ], "turn_id": 2, "scan_id": "scene0671_00", "origin_question": "When did you first discover \"the bottle\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the bottle\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0671_00/00240.jpg", "scene0671_00/00270.jpg", "scene0671_00/00300.jpg", "scene0671_00/00330.jpg" ], "turn_id": 3, "scan_id": "scene0671_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0671_00/00360.jpg", "scene0671_00/00390.jpg", "scene0671_00/00420.jpg", "scene0671_00/00450.jpg" ], "turn_id": 4, "scan_id": "scene0671_00", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "107.44", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0671_00/00480.jpg", "scene0671_00/00510.jpg", "scene0671_00/00540.jpg", "scene0671_00/00570.jpg" ], "turn_id": 5, "scan_id": "scene0671_00", "origin_question": "When did you first discover \"the green leafy plant in a transparent glass vase\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the green leafy plant in a transparent glass vase\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0671_00/00600.jpg", "scene0671_00/00630.jpg", "scene0671_00/00660.jpg", "scene0671_00/00690.jpg" ], "turn_id": 6, "scan_id": "scene0671_00", "origin_question": "Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "70.83", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0086_02/00000.jpg", "scene0086_02/00040.jpg", "scene0086_02/00080.jpg", "scene0086_02/00120.jpg" ], "turn_id": 1, "scan_id": "scene0086_02", "origin_question": "Remember, have you seen any shelf(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any shelf(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0086_02/00160.jpg", "scene0086_02/00200.jpg", "scene0086_02/00240.jpg", "scene0086_02/00280.jpg" ], "turn_id": 2, "scan_id": "scene0086_02", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the medium-sized silver metal rail\"; \"the tall, green plant with long leaves\"; \"the small silver metal bin with a plastic liner\" ", "option": [ "the medium-sized silver metal rail", "the tall, green plant with long leaves", "the small silver metal bin with a plastic liner" ], "answer": "the medium-sized silver metal rail", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the medium-sized silver metal rail\"; \"the tall, green plant with long leaves\"; \"the small silver metal bin with a plastic liner\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the medium-sized silver metal rail', 'the tall, green plant with long leaves', 'the small silver metal bin with a plastic liner'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0086_02/00320.jpg", "scene0086_02/00360.jpg", "scene0086_02/00400.jpg", "scene0086_02/00440.jpg" ], "turn_id": 3, "scan_id": "scene0086_02", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0086_02/00480.jpg", "scene0086_02/00520.jpg", "scene0086_02/00560.jpg", "scene0086_02/00600.jpg" ], "turn_id": 4, "scan_id": "scene0086_02", "origin_question": "", "option": [], "answer": "", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0086_02/00640.jpg", "scene0086_02/00680.jpg", "scene0086_02/00720.jpg", "scene0086_02/00760.jpg" ], "turn_id": 5, "scan_id": "scene0086_02", "origin_question": "In which turn did you see both \"the medium-sized plastic bottle with a white body and blue cap\" and \"the tall, green plant with long leaves\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the medium-sized plastic bottle with a white body and blue cap\" and \"the tall, green plant with long leaves\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0086_02/00800.jpg", "scene0086_02/00840.jpg", "scene0086_02/00880.jpg", "scene0086_02/00920.jpg" ], "turn_id": 6, "scan_id": "scene0086_02", "origin_question": "At the end of which turn were both of \"the black oval-shaped plastic soap dispenser\" and \"the tall, green plant with long leaves\" on your left side?", "option": [], "answer": "6", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the black oval-shaped plastic soap dispenser\" and \"the tall, green plant with long leaves\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0086_02/00960.jpg", "scene0086_02/01000.jpg", "scene0086_02/01040.jpg", "scene0086_02/01080.jpg" ], "turn_id": 7, "scan_id": "scene0086_02", "origin_question": "What will be the first-time appearance order of the following categories: bottle, door, mirror?", "option": [ "mirror, bottle, door", "mirror, door, bottle", "bottle, door, mirror", "door, mirror, bottle" ], "answer": "bottle, door, mirror", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bottle, door, mirror? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['mirror, bottle, door', 'mirror, door, bottle', 'bottle, door, mirror', 'door, mirror, bottle'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0084_region5/e5d83bdf3c66439aacf480b2f04a778f_i1_1.jpg", "1mp3d_0084_region5/e5d83bdf3c66439aacf480b2f04a778f_i2_1.jpg", "1mp3d_0084_region5/951adb0311204b58a0280f0cbcbcfa19_i1_5.jpg", "1mp3d_0084_region5/951adb0311204b58a0280f0cbcbcfa19_i2_5.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0084_region5", "origin_question": "Remember, have you seen any chair(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any chair(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "1mp3d_0084_region5/e5d83bdf3c66439aacf480b2f04a778f_i1_2.jpg", "1mp3d_0084_region5/e5d83bdf3c66439aacf480b2f04a778f_i2_2.jpg", "1mp3d_0084_region5/c8e2b36198cb423d8d0803e41d615b0d_i1_1.jpg", "1mp3d_0084_region5/c8e2b36198cb423d8d0803e41d615b0d_i2_1.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0084_region5", "origin_question": "In which turn did you see both \"the rectangular white desk with a minimalist design\" and \"the medium-sized, rectangular partition in beige\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the rectangular white desk with a minimalist design\" and \"the medium-sized, rectangular partition in beige\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0084_region5/fc6a063f71534f2ba7595fdaac55854c_i1_0.jpg", "1mp3d_0084_region5/fc6a063f71534f2ba7595fdaac55854c_i2_0.jpg", "1mp3d_0084_region5/cfa52adccaae4c59879a267d3c8d44e9_i1_0.jpg", "1mp3d_0084_region5/cfa52adccaae4c59879a267d3c8d44e9_i2_0.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0084_region5", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the large white wooden shelf with three layers\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the large white wooden shelf with three layers\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "1mp3d_0084_region5/fc6a063f71534f2ba7595fdaac55854c_i1_1.jpg", "1mp3d_0084_region5/fc6a063f71534f2ba7595fdaac55854c_i2_1.jpg", "1mp3d_0084_region5/8ad5a339add04ce887b4eb94d40914c6_i1_4.jpg", "1mp3d_0084_region5/8ad5a339add04ce887b4eb94d40914c6_i2_4.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0084_region5", "origin_question": "At the end of which turn were both of \"the large rectangular black metal window frame\" and \"the lively green potted plant in a brown pot\" on your right side?", "option": [], "answer": "2", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the large rectangular black metal window frame\" and \"the lively green potted plant in a brown pot\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0084_region5/a70fcde73fc544d48b247edcc4dc0490_i1_1.jpg", "1mp3d_0084_region5/a70fcde73fc544d48b247edcc4dc0490_i2_1.jpg", "1mp3d_0084_region5/a70fcde73fc544d48b247edcc4dc0490_i1_0.jpg", "1mp3d_0084_region5/a70fcde73fc544d48b247edcc4dc0490_i2_0.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0084_region5", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the large, circular white modern fireplace\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the large, circular white modern fireplace\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "1mp3d_0084_region5/effa4354d216404ab0f9ea64cbec3144_i1_1.jpg", "1mp3d_0084_region5/effa4354d216404ab0f9ea64cbec3144_i2_1.jpg", "1mp3d_0084_region5/effa4354d216404ab0f9ea64cbec3144_i1_2.jpg", "1mp3d_0084_region5/effa4354d216404ab0f9ea64cbec3144_i2_2.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0084_region5", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the stylish black and silver printer with a digital display\"; \"the rectangular white desk with a minimalist design\"; \"the greenish-yellow statue with a metal base\" ", "option": [ "the stylish black and silver printer with a digital display", "the rectangular white desk with a minimalist design", "the greenish-yellow statue with a metal base" ], "answer": "the greenish-yellow statue with a metal base", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the stylish black and silver printer with a digital display\"; \"the rectangular white desk with a minimalist design\"; \"the greenish-yellow statue with a metal base\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the stylish black and silver printer with a digital display', 'the rectangular white desk with a minimalist design', 'the greenish-yellow statue with a metal base'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "1mp3d_0084_region5/6dcd567836f24a25b8be08ab110e3ed2_i1_2.jpg", "1mp3d_0084_region5/6dcd567836f24a25b8be08ab110e3ed2_i2_2.jpg", "1mp3d_0084_region5/e2cff4dba01a4a1aa9296bf39db26f20_i1_2.jpg", "1mp3d_0084_region5/e2cff4dba01a4a1aa9296bf39db26f20_i2_2.jpg" ], "turn_id": 7, "scan_id": "1mp3d_0084_region5", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the large rectangular black metal window frame\" from your position?", "option": [], "answer": "132.96", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the large rectangular black metal window frame\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0084_region5/e32c0ea5881d40e29ca6f58ff24e435c_i1_2.jpg", "1mp3d_0084_region5/e32c0ea5881d40e29ca6f58ff24e435c_i2_2.jpg", "1mp3d_0084_region5/05d8dd39b5de44b58287a6de59cdba37_i1_5.jpg", "1mp3d_0084_region5/05d8dd39b5de44b58287a6de59cdba37_i2_5.jpg" ], "turn_id": 8, "scan_id": "1mp3d_0084_region5", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the lively green potted plant in a brown pot\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the lively green potted plant in a brown pot\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "1mp3d_0084_region5/e32c0ea5881d40e29ca6f58ff24e435c_i1_3.jpg", "1mp3d_0084_region5/e32c0ea5881d40e29ca6f58ff24e435c_i2_3.jpg", "1mp3d_0084_region5/e2cff4dba01a4a1aa9296bf39db26f20_i1_3.jpg", "1mp3d_0084_region5/e2cff4dba01a4a1aa9296bf39db26f20_i2_3.jpg" ], "turn_id": 9, "scan_id": "1mp3d_0084_region5", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the large rectangular black metal window frame\" from your position?", "option": [], "answer": "72.29", "user_message": "For the 9 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the large rectangular black metal window frame\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0663_02/00000.jpg", "scene0663_02/00040.jpg", "scene0663_02/00080.jpg", "scene0663_02/00120.jpg", "scene0663_02/00160.jpg" ], "turn_id": 1, "scan_id": "scene0663_02", "origin_question": "Remember, have you seen any table(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any table(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0663_02/00200.jpg", "scene0663_02/00240.jpg", "scene0663_02/00280.jpg", "scene0663_02/00320.jpg", "scene0663_02/00360.jpg" ], "turn_id": 2, "scan_id": "scene0663_02", "origin_question": "Which two objects are on the same side of you now? \"the small green rectangular bin\", \"the black ergonomic wired optical mouse made of plastic\", \"the blue cylindrical ceramic cup with a handle\". ", "option": [ "the small green rectangular bin, the black ergonomic wired optical mouse made of plastic", "the small green rectangular bin, the blue cylindrical ceramic cup with a handle", "the black ergonomic wired optical mouse made of plastic, the blue cylindrical ceramic cup with a handle" ], "answer": "the black ergonomic wired optical mouse made of plastic, the blue cylindrical ceramic cup with a handle", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the small green rectangular bin\", \"the black ergonomic wired optical mouse made of plastic\", \"the blue cylindrical ceramic cup with a handle\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the small green rectangular bin, the black ergonomic wired optical mouse made of plastic', 'the small green rectangular bin, the blue cylindrical ceramic cup with a handle', 'the black ergonomic wired optical mouse made of plastic, the blue cylindrical ceramic cup with a handle'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0663_02/00400.jpg", "scene0663_02/00440.jpg", "scene0663_02/00480.jpg", "scene0663_02/00520.jpg", "scene0663_02/00560.jpg" ], "turn_id": 3, "scan_id": "scene0663_02", "origin_question": "Please recall \"the dark rectangular plastic telephone\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.63", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the dark rectangular plastic telephone\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0663_02/00600.jpg", "scene0663_02/00640.jpg", "scene0663_02/00680.jpg", "scene0663_02/00720.jpg", "scene0663_02/00760.jpg" ], "turn_id": 4, "scan_id": "scene0663_02", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the window\"; \"the dark rectangular plastic telephone\"; \"the blue roll of adhesive tape with a white stripe\" ", "option": [ "the window", "the dark rectangular plastic telephone", "the blue roll of adhesive tape with a white stripe" ], "answer": "the window", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the window\"; \"the dark rectangular plastic telephone\"; \"the blue roll of adhesive tape with a white stripe\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the window', 'the dark rectangular plastic telephone', 'the blue roll of adhesive tape with a white stripe'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0663_02/00800.jpg", "scene0663_02/00840.jpg", "scene0663_02/00880.jpg", "scene0663_02/00920.jpg", "scene0663_02/00960.jpg" ], "turn_id": 5, "scan_id": "scene0663_02", "origin_question": "Remember, have you seen any mouse(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any mouse(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0663_02/01000.jpg", "scene0663_02/01040.jpg", "scene0663_02/01080.jpg", "scene0663_02/01120.jpg", "scene0663_02/01160.jpg" ], "turn_id": 6, "scan_id": "scene0663_02", "origin_question": "Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "68.69", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0663_02/01200.jpg", "scene0663_02/01240.jpg", "scene0663_02/01280.jpg", "scene0663_02/01320.jpg", "scene0663_02/01360.jpg" ], "turn_id": 7, "scan_id": "scene0663_02", "origin_question": "Which two objects are on the same side of you now? \"the small green rectangular bin\", \"the dark brown rectangular wooden desk\", \"the pair of dark clothes and a shiny jacket\". ", "option": [ "the small green rectangular bin, the dark brown rectangular wooden desk", "the small green rectangular bin, the pair of dark clothes and a shiny jacket", "the dark brown rectangular wooden desk, the pair of dark clothes and a shiny jacket" ], "answer": "the small green rectangular bin, the dark brown rectangular wooden desk", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the small green rectangular bin\", \"the dark brown rectangular wooden desk\", \"the pair of dark clothes and a shiny jacket\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the small green rectangular bin, the dark brown rectangular wooden desk', 'the small green rectangular bin, the pair of dark clothes and a shiny jacket', 'the dark brown rectangular wooden desk, the pair of dark clothes and a shiny jacket'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0663_02/01400.jpg", "scene0663_02/01440.jpg", "scene0663_02/01480.jpg", "scene0663_02/01520.jpg", "scene0663_02/01560.jpg" ], "turn_id": 8, "scan_id": "scene0663_02", "origin_question": "Please recall \"the blue roll of adhesive tape with a white stripe\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.18", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the blue roll of adhesive tape with a white stripe\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0663_02/01600.jpg", "scene0663_02/01640.jpg", "scene0663_02/01680.jpg", "scene0663_02/01720.jpg", "scene0663_02/01760.jpg" ], "turn_id": 9, "scan_id": "scene0663_02", "origin_question": "Which two objects are on the same side of you now? \"the rectangular brown wooden table\", \"the black ergonomic wired optical mouse made of plastic\", \"the dark brown rectangular wooden desk\". ", "option": [ "the rectangular brown wooden table, the black ergonomic wired optical mouse made of plastic", "the rectangular brown wooden table, the dark brown rectangular wooden desk", "the black ergonomic wired optical mouse made of plastic, the dark brown rectangular wooden desk" ], "answer": "the rectangular brown wooden table, the black ergonomic wired optical mouse made of plastic", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the rectangular brown wooden table\", \"the black ergonomic wired optical mouse made of plastic\", \"the dark brown rectangular wooden desk\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular brown wooden table, the black ergonomic wired optical mouse made of plastic', 'the rectangular brown wooden table, the dark brown rectangular wooden desk', 'the black ergonomic wired optical mouse made of plastic, the dark brown rectangular wooden desk'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0663_02/01800.jpg", "scene0663_02/01840.jpg", "scene0663_02/01880.jpg", "scene0663_02/01920.jpg", "scene0663_02/01960.jpg" ], "turn_id": 10, "scan_id": "scene0663_02", "origin_question": "", "option": [], "answer": "", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0029_region30/1838a8fb8e9f48cebcd079d5373cba02_i1_2.jpg", "1mp3d_0029_region30/1838a8fb8e9f48cebcd079d5373cba02_i2_2.jpg", "1mp3d_0029_region30/d2f529abd1404526bb735dc0b5588df3_i1_1.jpg", "1mp3d_0029_region30/d2f529abd1404526bb735dc0b5588df3_i2_1.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0029_region30", "origin_question": "Remember, have you seen any light(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any light(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "1mp3d_0029_region30/d2f529abd1404526bb735dc0b5588df3_i1_2.jpg", "1mp3d_0029_region30/d2f529abd1404526bb735dc0b5588df3_i2_2.jpg", "1mp3d_0029_region30/6376b741b50a4418b3dc3fde791c3c09_i1_0.jpg", "1mp3d_0029_region30/6376b741b50a4418b3dc3fde791c3c09_i2_0.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0029_region30", "origin_question": "When did you first discover \"the small rectangular white metal vent with a grid pattern\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the small rectangular white metal vent with a grid pattern\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "1mp3d_0029_region30/9e4c92fd7eb74504baecf55a3264716e_i1_0.jpg", "1mp3d_0029_region30/9e4c92fd7eb74504baecf55a3264716e_i2_0.jpg", "1mp3d_0029_region30/b564162b2c7d4033bfe6ef3dfb959c9e_i1_4.jpg", "1mp3d_0029_region30/b564162b2c7d4033bfe6ef3dfb959c9e_i2_4.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0029_region30", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "2.42", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0029_region30/4291cbfde1024467b1548934653f07ce_i1_5.jpg", "1mp3d_0029_region30/4291cbfde1024467b1548934653f07ce_i2_5.jpg", "1mp3d_0029_region30/4291cbfde1024467b1548934653f07ce_i1_4.jpg", "1mp3d_0029_region30/b6c62c6714d44a7080c83ec374607bcb_i1_1.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0029_region30", "origin_question": "Remember, have you seen any globe(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any globe(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "1mp3d_0029_region30/b6c62c6714d44a7080c83ec374607bcb_i2_1.jpg", "1mp3d_0029_region30/92fb09a83f8949619b9dc5bda2855456_i1_1.jpg", "1mp3d_0029_region30/92fb09a83f8949619b9dc5bda2855456_i2_1.jpg", "1mp3d_0029_region30/1679b5de39e548d38ba240f2fd99cae9_i1_0.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0029_region30", "origin_question": "How far is your current position from where you were at the end of the turn 4? (in meters)", "option": [], "answer": "2.42", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 4? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0029_region30/1679b5de39e548d38ba240f2fd99cae9_i2_0.jpg", "1mp3d_0029_region30/1679b5de39e548d38ba240f2fd99cae9_i1_1.jpg", "1mp3d_0029_region30/1679b5de39e548d38ba240f2fd99cae9_i2_1.jpg", "1mp3d_0029_region30/92fb09a83f8949619b9dc5bda2855456_i1_2.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0029_region30", "origin_question": "", "option": [], "answer": "", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "1mp3d_0029_region30/92fb09a83f8949619b9dc5bda2855456_i2_2.jpg", "1mp3d_0029_region30/3429ecc6a320475fada43d43848b9aae_i1_1.jpg", "1mp3d_0029_region30/3429ecc6a320475fada43d43848b9aae_i2_1.jpg", "1mp3d_0029_region30/3429ecc6a320475fada43d43848b9aae_i1_2.jpg" ], "turn_id": 7, "scan_id": "1mp3d_0029_region30", "origin_question": "Is the small rectangular white metal vent with a grid pattern to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the small rectangular white metal vent with a grid pattern to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_41069051/Validation_41069051_41069051_5084.234.png", "Validation_41069051/Validation_41069051_41069051_5086.217.png", "Validation_41069051/Validation_41069051_41069051_5088.216.png", "Validation_41069051/Validation_41069051_41069051_5090.215.png" ], "turn_id": 1, "scan_id": "Validation_41069051", "origin_question": "Remember, have you seen any backpack(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any backpack(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_41069051/Validation_41069051_41069051_5092.214.png", "Validation_41069051/Validation_41069051_41069051_5094.213.png", "Validation_41069051/Validation_41069051_41069051_5096.213.png", "Validation_41069051/Validation_41069051_41069051_5098.212.png" ], "turn_id": 2, "scan_id": "Validation_41069051", "origin_question": "When did you first discover \"the glossy white bathtub with integrated handles\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the glossy white bathtub with integrated handles\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_41069051/Validation_41069051_41069051_5100.211.png", "Validation_41069051/Validation_41069051_41069051_5102.210.png", "Validation_41069051/Validation_41069051_41069051_5104.209.png", "Validation_41069051/Validation_41069051_41069051_5106.208.png" ], "turn_id": 3, "scan_id": "Validation_41069051", "origin_question": "What will be the first-time appearance order of the following categories: bathtub, bin, toilet?", "option": [ "toilet, bathtub, bin", "bin, toilet, bathtub", "toilet, bin, bathtub", "bathtub, toilet, bin" ], "answer": "bathtub, toilet, bin", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bathtub, bin, toilet? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['toilet, bathtub, bin', 'bin, toilet, bathtub', 'toilet, bin, bathtub', 'bathtub, toilet, bin'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_41069051/Validation_41069051_41069051_5108.208.png", "Validation_41069051/Validation_41069051_41069051_5110.207.png", "Validation_41069051/Validation_41069051_41069051_5112.206.png", "Validation_41069051/Validation_41069051_41069051_5114.205.png" ], "turn_id": 4, "scan_id": "Validation_41069051", "origin_question": "At the end of which turn were both of \"the glossy white ceramic oval sink on the pedestal\" and \"the modern oval-shaped white ceramic toilet\" on your left side?", "option": [], "answer": "2", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the glossy white ceramic oval sink on the pedestal\" and \"the modern oval-shaped white ceramic toilet\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_41069051/Validation_41069051_41069051_5116.204.png", "Validation_41069051/Validation_41069051_41069051_5118.204.png", "Validation_41069051/Validation_41069051_41069051_5120.203.png", "Validation_41069051/Validation_41069051_41069051_5122.219.png" ], "turn_id": 5, "scan_id": "Validation_41069051", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the compact, cylindrical silver metal bin\" from your position?", "option": [], "answer": "178.72", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the compact, cylindrical silver metal bin\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "Validation_41069051/Validation_41069051_41069051_5124.218.png", "Validation_41069051/Validation_41069051_41069051_5126.217.png", "Validation_41069051/Validation_41069051_41069051_5128.216.png", "Validation_41069051/Validation_41069051_41069051_5130.215.png" ], "turn_id": 6, "scan_id": "Validation_41069051", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the glossy white ceramic oval sink on the pedestal\"; \"the sleek rectangular silver mirror with the minimalistic frame\"; \"the door\" ", "option": [ "the glossy white ceramic oval sink on the pedestal", "the sleek rectangular silver mirror with the minimalistic frame", "the door" ], "answer": "the sleek rectangular silver mirror with the minimalistic frame", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the glossy white ceramic oval sink on the pedestal\"; \"the sleek rectangular silver mirror with the minimalistic frame\"; \"the door\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the glossy white ceramic oval sink on the pedestal', 'the sleek rectangular silver mirror with the minimalistic frame', 'the door'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_41069051/Validation_41069051_41069051_5132.214.png", "Validation_41069051/Validation_41069051_41069051_5134.214.png", "Validation_41069051/Validation_41069051_41069051_5136.213.png", "Validation_41069051/Validation_41069051_41069051_5138.212.png" ], "turn_id": 7, "scan_id": "Validation_41069051", "origin_question": "Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "41.71", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_41069051/Validation_41069051_41069051_5140.211.png", "Validation_41069051/Validation_41069051_41069051_5142.210.png", "Validation_41069051/Validation_41069051_41069051_5144.209.png", "Validation_41069051/Validation_41069051_41069051_5146.209.png" ], "turn_id": 8, "scan_id": "Validation_41069051", "origin_question": "Please recall \"the creamy white plush towel with the raised pattern\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.41", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the creamy white plush towel with the raised pattern\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42445026/Validation_42445026_42445026_48936.087.png", "Validation_42445026/Validation_42445026_42445026_48938.070.png", "Validation_42445026/Validation_42445026_42445026_48940.069.png", "Validation_42445026/Validation_42445026_42445026_48942.068.png" ], "turn_id": 1, "scan_id": "Validation_42445026", "origin_question": "Remember, have you seen any bucket(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any bucket(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_42445026/Validation_42445026_42445026_48944.068.png", "Validation_42445026/Validation_42445026_42445026_48946.067.png", "Validation_42445026/Validation_42445026_42445026_48948.066.png", "Validation_42445026/Validation_42445026_42445026_48950.065.png" ], "turn_id": 2, "scan_id": "Validation_42445026", "origin_question": "Please recall \"the tall cream-colored cabinet with wood-toned trims\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.54", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the tall cream-colored cabinet with wood-toned trims\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_42445026/Validation_42445026_42445026_48952.064.png", "Validation_42445026/Validation_42445026_42445026_48954.064.png", "Validation_42445026/Validation_42445026_42445026_48956.063.png", "Validation_42445026/Validation_42445026_42445026_48958.062.png" ], "turn_id": 3, "scan_id": "Validation_42445026", "origin_question": "Which direction is the tall cream-colored cabinet with wood-toned trims in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the tall cream-colored cabinet with wood-toned trims in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "Validation_42445026/Validation_42445026_42445026_48960.061.png", "Validation_42445026/Validation_42445026_42445026_48962.060.png", "Validation_42445026/Validation_42445026_42445026_48964.060.png", "Validation_42445026/Validation_42445026_42445026_48966.059.png" ], "turn_id": 4, "scan_id": "Validation_42445026", "origin_question": "Among these three objects(\"the table\",\"the contemporary white plastic chair with the curved seat\",\"the door\" ), which one is the closest to you?", "option": [ "the table", "the contemporary white plastic chair with the curved seat", "the door" ], "answer": "the door", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects(\"the table\",\"the contemporary white plastic chair with the curved seat\",\"the door\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the table', 'the contemporary white plastic chair with the curved seat', 'the door'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_42445026/Validation_42445026_42445026_48968.058.png", "Validation_42445026/Validation_42445026_42445026_48970.057.png", "Validation_42445026/Validation_42445026_42445026_48972.056.png", "Validation_42445026/Validation_42445026_42445026_48974.056.png" ], "turn_id": 5, "scan_id": "Validation_42445026", "origin_question": "Using your orientation at the end of turn 4 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "71.98", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 4 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_42445026/Validation_42445026_42445026_48976.055.png", "Validation_42445026/Validation_42445026_42445026_48978.071.png", "Validation_42445026/Validation_42445026_42445026_48980.070.png", "Validation_42445026/Validation_42445026_42445026_48982.069.png" ], "turn_id": 6, "scan_id": "Validation_42445026", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the rectangular window with the white frame\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the rectangular window with the white frame\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_42445026/Validation_42445026_42445026_48984.068.png", "Validation_42445026/Validation_42445026_42445026_48986.068.png", "Validation_42445026/Validation_42445026_42445026_48988.067.png", "Validation_42445026/Validation_42445026_42445026_48990.066.png" ], "turn_id": 7, "scan_id": "Validation_42445026", "origin_question": "Which two objects are on the same side of you now? \"the compact, off-white curved lamp\", \"the door\", \"the table\". ", "option": [ "the compact, off-white curved lamp, the door", "the compact, off-white curved lamp, the table", "the door, the table" ], "answer": "the compact, off-white curved lamp, the door", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the compact, off-white curved lamp\", \"the door\", \"the table\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the compact, off-white curved lamp, the door', 'the compact, off-white curved lamp, the table', 'the door, the table'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_42445026/Validation_42445026_42445026_48992.065.png", "Validation_42445026/Validation_42445026_42445026_48994.065.png", "Validation_42445026/Validation_42445026_42445026_48996.064.png", "Validation_42445026/Validation_42445026_42445026_48998.063.png" ], "turn_id": 8, "scan_id": "Validation_42445026", "origin_question": "Compared to your position at the end of the 7 turn, are you now closer or farther away from \"the tall cream-colored cabinet with wood-toned trims\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 7 turn, are you now closer or farther away from \"the tall cream-colored cabinet with wood-toned trims\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_48458652/Validation_48458652_48458652_7120.213.png", "Validation_48458652/Validation_48458652_48458652_7122.213.png", "Validation_48458652/Validation_48458652_48458652_7124.212.png", "Validation_48458652/Validation_48458652_48458652_7126.211.png" ], "turn_id": 1, "scan_id": "Validation_48458652", "origin_question": "Remember, have you seen any mirror(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any mirror(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_48458652/Validation_48458652_48458652_7128.210.png", "Validation_48458652/Validation_48458652_48458652_7130.209.png", "Validation_48458652/Validation_48458652_48458652_7132.208.png", "Validation_48458652/Validation_48458652_48458652_7134.208.png" ], "turn_id": 2, "scan_id": "Validation_48458652", "origin_question": "Which direction is the sleek white vertical radiator with horizontal fins in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the sleek white vertical radiator with horizontal fins in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_48458652/Validation_48458652_48458652_7136.207.png", "Validation_48458652/Validation_48458652_48458652_7138.206.png", "Validation_48458652/Validation_48458652_48458652_7140.205.png", "Validation_48458652/Validation_48458652_48458652_7142.204.png" ], "turn_id": 3, "scan_id": "Validation_48458652", "origin_question": "Is the sleek white vertical radiator with horizontal fins to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the sleek white vertical radiator with horizontal fins to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_48458652/Validation_48458652_48458652_7144.203.png", "Validation_48458652/Validation_48458652_48458652_7146.203.png", "Validation_48458652/Validation_48458652_48458652_7148.202.png", "Validation_48458652/Validation_48458652_48458652_7150.201.png" ], "turn_id": 4, "scan_id": "Validation_48458652", "origin_question": "What will be the first-time appearance order of the following categories: backpack, bed, chair?", "option": [ "bed, chair, backpack", "bed, backpack, chair", "backpack, chair, bed", "chair, bed, backpack" ], "answer": "chair, bed, backpack", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: backpack, bed, chair? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['bed, chair, backpack', 'bed, backpack, chair', 'backpack, chair, bed', 'chair, bed, backpack'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_48458652/Validation_48458652_48458652_7152.217.png", "Validation_48458652/Validation_48458652_48458652_7154.216.png", "Validation_48458652/Validation_48458652_48458652_7156.215.png", "Validation_48458652/Validation_48458652_48458652_7158.214.png" ], "turn_id": 5, "scan_id": "Validation_48458652", "origin_question": "By the end of the 4 turn, were you getting closer to or farther away from these two objects: X: \"the sleek white vertical radiator with horizontal fins\", Y: \"the cylindrical terracottthe flowerpot made of clay\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to X and farther away from Y", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 4 turn, were you getting closer to or farther away from these two objects: X: \"the sleek white vertical radiator with horizontal fins\", Y: \"the cylindrical terracottthe flowerpot made of clay\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_48458652/Validation_48458652_48458652_7160.213.png", "Validation_48458652/Validation_48458652_48458652_7162.213.png", "Validation_48458652/Validation_48458652_48458652_7164.212.png", "Validation_48458652/Validation_48458652_48458652_7166.211.png" ], "turn_id": 6, "scan_id": "Validation_48458652", "origin_question": "By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the cylindrical fabric bucket with the white mesh design\", Y: \"the sleek white vertical radiator with horizontal fins\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "farther away from both X and Y", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the cylindrical fabric bucket with the white mesh design\", Y: \"the sleek white vertical radiator with horizontal fins\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47334368/Validation_47334368_47334368_63274.204.png", "Validation_47334368/Validation_47334368_47334368_63276.186.png", "Validation_47334368/Validation_47334368_47334368_63278.185.png", "Validation_47334368/Validation_47334368_47334368_63280.185.png", "Validation_47334368/Validation_47334368_47334368_63282.184.png" ], "turn_id": 1, "scan_id": "Validation_47334368", "origin_question": "Remember, have you seen any washing machine(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any washing machine(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_47334368/Validation_47334368_47334368_63284.183.png", "Validation_47334368/Validation_47334368_47334368_63286.182.png", "Validation_47334368/Validation_47334368_47334368_63288.181.png", "Validation_47334368/Validation_47334368_47334368_63290.197.png", "Validation_47334368/Validation_47334368_47334368_63292.196.png" ], "turn_id": 2, "scan_id": "Validation_47334368", "origin_question": "Is the sleek, white rectangular ceramic sink to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the sleek, white rectangular ceramic sink to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_47334368/Validation_47334368_47334368_63294.196.png", "Validation_47334368/Validation_47334368_47334368_63296.195.png", "Validation_47334368/Validation_47334368_47334368_63298.194.png", "Validation_47334368/Validation_47334368_47334368_63300.193.png", "Validation_47334368/Validation_47334368_47334368_63302.192.png" ], "turn_id": 3, "scan_id": "Validation_47334368", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the compact, sleek light gray cabinet\" from your position?", "option": [], "answer": "92.6", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the compact, sleek light gray cabinet\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47334368/Validation_47334368_47334368_63304.191.png", "Validation_47334368/Validation_47334368_47334368_63306.191.png", "Validation_47334368/Validation_47334368_47334368_63308.190.png", "Validation_47334368/Validation_47334368_47334368_63310.189.png", "Validation_47334368/Validation_47334368_47334368_63312.188.png" ], "turn_id": 4, "scan_id": "Validation_47334368", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the colorful array of cylindrical and oval toiletries\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the colorful array of cylindrical and oval toiletries\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_47334368/Validation_47334368_47334368_63314.187.png", "Validation_47334368/Validation_47334368_47334368_63316.187.png", "Validation_47334368/Validation_47334368_47334368_63318.186.png", "Validation_47334368/Validation_47334368_47334368_63320.185.png", "Validation_47334368/Validation_47334368_47334368_63322.184.png" ], "turn_id": 5, "scan_id": "Validation_47334368", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the sleek, white rectangular ceramic sink\" from your position?", "option": [], "answer": "74.64", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the sleek, white rectangular ceramic sink\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_47334368/Validation_47334368_47334368_63324.183.png", "Validation_47334368/Validation_47334368_47334368_63326.183.png", "Validation_47334368/Validation_47334368_47334368_63328.182.png", "Validation_47334368/Validation_47334368_47334368_63330.181.png", "Validation_47334368/Validation_47334368_47334368_63332.197.png" ], "turn_id": 6, "scan_id": "Validation_47334368", "origin_question": "Is the sleek white front-loading washing machine to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the sleek white front-loading washing machine to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_47334368/Validation_47334368_47334368_63334.196.png", "Validation_47334368/Validation_47334368_47334368_63336.195.png", "Validation_47334368/Validation_47334368_47334368_63338.194.png", "Validation_47334368/Validation_47334368_47334368_63340.193.png", "Validation_47334368/Validation_47334368_47334368_63342.193.png" ], "turn_id": 7, "scan_id": "Validation_47334368", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the vibrant orange plastic basin with the round shape\" from your position?", "option": [], "answer": "66.59", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the vibrant orange plastic basin with the round shape\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47334368/Validation_47334368_47334368_63344.192.png", "Validation_47334368/Validation_47334368_47334368_63346.191.png", "Validation_47334368/Validation_47334368_47334368_63348.190.png", "Validation_47334368/Validation_47334368_47334368_63350.189.png", "Validation_47334368/Validation_47334368_47334368_63352.188.png" ], "turn_id": 8, "scan_id": "Validation_47334368", "origin_question": "Which direction is the sleek white front-loading washing machine in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the sleek white front-loading washing machine in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_47334368/Validation_47334368_47334368_63354.188.png", "Validation_47334368/Validation_47334368_47334368_63356.187.png", "Validation_47334368/Validation_47334368_47334368_63358.186.png", "Validation_47334368/Validation_47334368_47334368_63360.185.png", "Validation_47334368/Validation_47334368_47334368_63362.185.png" ], "turn_id": 9, "scan_id": "Validation_47334368", "origin_question": "What will be the first-time appearance order of the following categories: bathtub, sink, toiletry?", "option": [ "toiletry, sink, bathtub", "sink, toiletry, bathtub", "sink, bathtub, toiletry", "toiletry, bathtub, sink" ], "answer": "sink, toiletry, bathtub", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bathtub, sink, toiletry? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['toiletry, sink, bathtub', 'sink, toiletry, bathtub', 'sink, bathtub, toiletry', 'toiletry, bathtub, sink'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_47334368/Validation_47334368_47334368_63364.184.png", "Validation_47334368/Validation_47334368_47334368_63366.183.png", "Validation_47334368/Validation_47334368_47334368_63368.182.png", "Validation_47334368/Validation_47334368_47334368_63370.181.png", "Validation_47334368/Validation_47334368_47334368_63372.197.png" ], "turn_id": 10, "scan_id": "Validation_47334368", "origin_question": "Which two objects are on the same side of you now? \"the sleek white front-loading washing machine\", \"the colorful array of cylindrical and oval toiletries\", \"the bathtub\". ", "option": [ "the sleek white front-loading washing machine, the colorful array of cylindrical and oval toiletries", "the sleek white front-loading washing machine, the bathtub", "the colorful array of cylindrical and oval toiletries, the bathtub" ], "answer": "the colorful array of cylindrical and oval toiletries, the bathtub", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the sleek white front-loading washing machine\", \"the colorful array of cylindrical and oval toiletries\", \"the bathtub\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the sleek white front-loading washing machine, the colorful array of cylindrical and oval toiletries', 'the sleek white front-loading washing machine, the bathtub', 'the colorful array of cylindrical and oval toiletries, the bathtub'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_48018381/Validation_48018381_48018381_389935.845.png", "Validation_48018381/Validation_48018381_48018381_389937.828.png", "Validation_48018381/Validation_48018381_48018381_389939.844.png", "Validation_48018381/Validation_48018381_48018381_389941.843.png" ], "turn_id": 1, "scan_id": "Validation_48018381", "origin_question": "Remember, have you seen any cabinet(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any cabinet(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_48018381/Validation_48018381_48018381_389943.842.png", "Validation_48018381/Validation_48018381_48018381_389945.841.png", "Validation_48018381/Validation_48018381_48018381_389947.840.png", "Validation_48018381/Validation_48018381_48018381_389949.840.png" ], "turn_id": 2, "scan_id": "Validation_48018381", "origin_question": "In which turn were you closest to \"the dome-shaped glass light with warm glow\"? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the dome-shaped glass light with warm glow\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_48018381/Validation_48018381_48018381_389951.839.png", "Validation_48018381/Validation_48018381_48018381_389953.838.png", "Validation_48018381/Validation_48018381_48018381_389955.837.png", "Validation_48018381/Validation_48018381_48018381_389957.836.png" ], "turn_id": 3, "scan_id": "Validation_48018381", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "136.82", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_48018381/Validation_48018381_48018381_389959.835.png", "Validation_48018381/Validation_48018381_48018381_389961.835.png", "Validation_48018381/Validation_48018381_48018381_389963.834.png", "Validation_48018381/Validation_48018381_48018381_389965.833.png" ], "turn_id": 4, "scan_id": "Validation_48018381", "origin_question": "At the end of which turn were both of \"the dome-shaped glass light with warm glow\" and \"the sleek, rectangular monitor with the matte black frame\" on your right side?", "option": [], "answer": "3", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the dome-shaped glass light with warm glow\" and \"the sleek, rectangular monitor with the matte black frame\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_48018381/Validation_48018381_48018381_389967.832.png", "Validation_48018381/Validation_48018381_48018381_389969.831.png", "Validation_48018381/Validation_48018381_48018381_389971.830.png", "Validation_48018381/Validation_48018381_48018381_389973.830.png" ], "turn_id": 5, "scan_id": "Validation_48018381", "origin_question": "In which turn were you closest to \"the rectangular white-framed window with clear panes\"? ", "option": [], "answer": "5", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the rectangular white-framed window with clear panes\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_48018381/Validation_48018381_48018381_389975.829.png", "Validation_48018381/Validation_48018381_48018381_389977.828.png", "Validation_48018381/Validation_48018381_48018381_389979.844.png", "Validation_48018381/Validation_48018381_48018381_389981.843.png" ], "turn_id": 6, "scan_id": "Validation_48018381", "origin_question": "What will be the first-time appearance order of the following categories: cabinet, flowerpot, pillow?", "option": [ "pillow, flowerpot, cabinet", "cabinet, flowerpot, pillow", "cabinet, pillow, flowerpot", "pillow, cabinet, flowerpot" ], "answer": "cabinet, pillow, flowerpot", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: cabinet, flowerpot, pillow? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['pillow, flowerpot, cabinet', 'cabinet, flowerpot, pillow', 'cabinet, pillow, flowerpot', 'pillow, cabinet, flowerpot'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_48018381/Validation_48018381_48018381_389983.842.png", "Validation_48018381/Validation_48018381_48018381_389985.841.png", "Validation_48018381/Validation_48018381_48018381_389987.841.png", "Validation_48018381/Validation_48018381_48018381_389989.840.png" ], "turn_id": 7, "scan_id": "Validation_48018381", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the rectangular white-framed window with clear panes\" from your position?", "option": [], "answer": "76.05", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the rectangular white-framed window with clear panes\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42445429/Validation_42445429_42445429_50317.089.png", "Validation_42445429/Validation_42445429_42445429_50319.088.png", "Validation_42445429/Validation_42445429_42445429_50321.088.png", "Validation_42445429/Validation_42445429_42445429_50323.087.png", "Validation_42445429/Validation_42445429_42445429_50325.086.png" ], "turn_id": 1, "scan_id": "Validation_42445429", "origin_question": "Remember, have you seen any pan(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any pan(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_42445429/Validation_42445429_42445429_50327.085.png", "Validation_42445429/Validation_42445429_42445429_50329.084.png", "Validation_42445429/Validation_42445429_42445429_50331.084.png", "Validation_42445429/Validation_42445429_42445429_50333.083.png", "Validation_42445429/Validation_42445429_42445429_50335.082.png" ], "turn_id": 2, "scan_id": "Validation_42445429", "origin_question": "Please recall \"the glossy black square table in pristine condition\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.52", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the glossy black square table in pristine condition\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_42445429/Validation_42445429_42445429_50337.081.png", "Validation_42445429/Validation_42445429_42445429_50339.081.png", "Validation_42445429/Validation_42445429_42445429_50341.080.png", "Validation_42445429/Validation_42445429_42445429_50343.096.png", "Validation_42445429/Validation_42445429_42445429_50345.095.png" ], "turn_id": 3, "scan_id": "Validation_42445429", "origin_question": "In which turn were you farthest from \"the glossy black square table in pristine condition\"? ", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the glossy black square table in pristine condition\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_42445429/Validation_42445429_42445429_50347.094.png", "Validation_42445429/Validation_42445429_42445429_50349.093.png", "Validation_42445429/Validation_42445429_42445429_50351.092.png", "Validation_42445429/Validation_42445429_42445429_50353.092.png", "Validation_42445429/Validation_42445429_42445429_50355.091.png" ], "turn_id": 4, "scan_id": "Validation_42445429", "origin_question": "In which turn were you closest to \"the couch\"? ", "option": [], "answer": "2", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the couch\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_42445429/Validation_42445429_42445429_50357.090.png", "Validation_42445429/Validation_42445429_42445429_50359.089.png", "Validation_42445429/Validation_42445429_42445429_50361.088.png", "Validation_42445429/Validation_42445429_42445429_50363.088.png", "Validation_42445429/Validation_42445429_42445429_50365.087.png" ], "turn_id": 5, "scan_id": "Validation_42445429", "origin_question": "At the end of which turn were both of \"the large, black monitor with the wide screen\" and \"an inverted dark green plant\" on your right side?", "option": [], "answer": "1", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the large, black monitor with the wide screen\" and \"an inverted dark green plant\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_42445429/Validation_42445429_42445429_50367.086.png", "Validation_42445429/Validation_42445429_42445429_50369.085.png", "Validation_42445429/Validation_42445429_42445429_50371.085.png", "Validation_42445429/Validation_42445429_42445429_50373.084.png", "Validation_42445429/Validation_42445429_42445429_50375.083.png" ], "turn_id": 6, "scan_id": "Validation_42445429", "origin_question": "Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_42445429/Validation_42445429_42445429_50377.082.png", "Validation_42445429/Validation_42445429_42445429_50379.081.png", "Validation_42445429/Validation_42445429_42445429_50381.081.png", "Validation_42445429/Validation_42445429_42445429_50383.096.png", "Validation_42445429/Validation_42445429_42445429_50385.096.png" ], "turn_id": 7, "scan_id": "Validation_42445429", "origin_question": "Which direction is the glossy black square table in pristine condition in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the glossy black square table in pristine condition in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_42445429/Validation_42445429_42445429_50387.095.png", "Validation_42445429/Validation_42445429_42445429_50389.094.png", "Validation_42445429/Validation_42445429_42445429_50391.093.png", "Validation_42445429/Validation_42445429_42445429_50393.092.png", "Validation_42445429/Validation_42445429_42445429_50395.092.png" ], "turn_id": 8, "scan_id": "Validation_42445429", "origin_question": "Assuming the direction you are facing at the end of the turn 6 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Right", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 6 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47895353/Validation_47895353_47895353_98648.846.png", "Validation_47895353/Validation_47895353_47895353_98650.829.png", "Validation_47895353/Validation_47895353_47895353_98652.828.png", "Validation_47895353/Validation_47895353_47895353_98654.827.png", "Validation_47895353/Validation_47895353_47895353_98656.826.png" ], "turn_id": 1, "scan_id": "Validation_47895353", "origin_question": "", "option": [], "answer": "", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_47895353/Validation_47895353_47895353_98658.825.png", "Validation_47895353/Validation_47895353_47895353_98660.841.png", "Validation_47895353/Validation_47895353_47895353_98662.840.png", "Validation_47895353/Validation_47895353_47895353_98664.840.png", "Validation_47895353/Validation_47895353_47895353_98666.839.png" ], "turn_id": 2, "scan_id": "Validation_47895353", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the dark wooden chair with the curved backrest\", Y: \"the modern oval table with the white top and green accent\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the dark wooden chair with the curved backrest\", Y: \"the modern oval table with the white top and green accent\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_47895353/Validation_47895353_47895353_98668.838.png", "Validation_47895353/Validation_47895353_47895353_98670.837.png", "Validation_47895353/Validation_47895353_47895353_98672.836.png", "Validation_47895353/Validation_47895353_47895353_98674.836.png", "Validation_47895353/Validation_47895353_47895353_98676.835.png" ], "turn_id": 3, "scan_id": "Validation_47895353", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the round stainless steel pan with the flat base\" from your position?", "option": [], "answer": "67.73", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the round stainless steel pan with the flat base\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47895353/Validation_47895353_47895353_98678.834.png", "Validation_47895353/Validation_47895353_47895353_98680.833.png", "Validation_47895353/Validation_47895353_47895353_98682.832.png", "Validation_47895353/Validation_47895353_47895353_98684.832.png", "Validation_47895353/Validation_47895353_47895353_98686.831.png" ], "turn_id": 4, "scan_id": "Validation_47895353", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the smooth, round ceramic bowl in the muted color\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the smooth, round ceramic bowl in the muted color\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "Validation_47895353/Validation_47895353_47895353_98688.830.png", "Validation_47895353/Validation_47895353_47895353_98690.829.png", "Validation_47895353/Validation_47895353_47895353_98692.828.png", "Validation_47895353/Validation_47895353_47895353_98694.827.png", "Validation_47895353/Validation_47895353_47895353_98696.827.png" ], "turn_id": 5, "scan_id": "Validation_47895353", "origin_question": "Among these three objects(\"the sleek dark-toned vertical shelf made of wood-like material\",\"the dark wooden chair with the curved backrest\",\"the sleek black stove with the flat surface\" ), which one is the closest to you?", "option": [ "the sleek dark-toned vertical shelf made of wood-like material", "the dark wooden chair with the curved backrest", "the sleek black stove with the flat surface" ], "answer": "the dark wooden chair with the curved backrest", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the sleek dark-toned vertical shelf made of wood-like material\",\"the dark wooden chair with the curved backrest\",\"the sleek black stove with the flat surface\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the sleek dark-toned vertical shelf made of wood-like material', 'the dark wooden chair with the curved backrest', 'the sleek black stove with the flat surface'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_47895353/Validation_47895353_47895353_98698.826.png", "Validation_47895353/Validation_47895353_47895353_98700.842.png", "Validation_47895353/Validation_47895353_47895353_98702.841.png", "Validation_47895353/Validation_47895353_47895353_98704.840.png", "Validation_47895353/Validation_47895353_47895353_98706.839.png" ], "turn_id": 6, "scan_id": "Validation_47895353", "origin_question": "Which two objects are on the same side of you now? \"the sleek dark-toned vertical shelf made of wood-like material\", \"the sleek black stove with the flat surface\", \"the glossy black rectangular oven with stainless steel accents\". ", "option": [ "the sleek dark-toned vertical shelf made of wood-like material, the sleek black stove with the flat surface", "the sleek dark-toned vertical shelf made of wood-like material, the glossy black rectangular oven with stainless steel accents", "the sleek black stove with the flat surface, the glossy black rectangular oven with stainless steel accents" ], "answer": "the sleek dark-toned vertical shelf made of wood-like material, the sleek black stove with the flat surface", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the sleek dark-toned vertical shelf made of wood-like material\", \"the sleek black stove with the flat surface\", \"the glossy black rectangular oven with stainless steel accents\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the sleek dark-toned vertical shelf made of wood-like material, the sleek black stove with the flat surface', 'the sleek dark-toned vertical shelf made of wood-like material, the glossy black rectangular oven with stainless steel accents', 'the sleek black stove with the flat surface, the glossy black rectangular oven with stainless steel accents'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "Validation_47895353/Validation_47895353_47895353_98708.838.png", "Validation_47895353/Validation_47895353_47895353_98710.838.png", "Validation_47895353/Validation_47895353_47895353_98712.837.png", "Validation_47895353/Validation_47895353_47895353_98714.836.png", "Validation_47895353/Validation_47895353_47895353_98716.835.png" ], "turn_id": 7, "scan_id": "Validation_47895353", "origin_question": "Among these three objects(\"the sleek dark-toned vertical shelf made of wood-like material\",\"the round stainless steel pan with the flat base\",\"the sleek black stove with the flat surface\" ), which one is the closest to you?", "option": [ "the sleek dark-toned vertical shelf made of wood-like material", "the round stainless steel pan with the flat base", "the sleek black stove with the flat surface" ], "answer": "the sleek black stove with the flat surface", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the sleek dark-toned vertical shelf made of wood-like material\",\"the round stainless steel pan with the flat base\",\"the sleek black stove with the flat surface\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the sleek dark-toned vertical shelf made of wood-like material', 'the round stainless steel pan with the flat base', 'the sleek black stove with the flat surface'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47895353/Validation_47895353_47895353_98718.834.png", "Validation_47895353/Validation_47895353_47895353_98720.833.png", "Validation_47895353/Validation_47895353_47895353_98722.833.png", "Validation_47895353/Validation_47895353_47895353_98724.832.png", "Validation_47895353/Validation_47895353_47895353_98726.831.png" ], "turn_id": 8, "scan_id": "Validation_47895353", "origin_question": "Compared to your position at the end of the 5 turn, are you now closer or farther away from \"the round stainless steel pan with the flat base\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 5 turn, are you now closer or farther away from \"the round stainless steel pan with the flat base\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_47895353/Validation_47895353_47895353_98728.830.png", "Validation_47895353/Validation_47895353_47895353_98730.829.png", "Validation_47895353/Validation_47895353_47895353_98732.829.png", "Validation_47895353/Validation_47895353_47895353_98734.828.png", "Validation_47895353/Validation_47895353_47895353_98736.827.png" ], "turn_id": 9, "scan_id": "Validation_47895353", "origin_question": "At the end of which turn were both of \"the glossy black rectangular oven with stainless steel accents\" and \"the smooth, round ceramic bowl in the muted color\" on your left side?", "option": [], "answer": "3", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the glossy black rectangular oven with stainless steel accents\" and \"the smooth, round ceramic bowl in the muted color\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_47895353/Validation_47895353_47895353_98738.826.png", "Validation_47895353/Validation_47895353_47895353_98740.825.png", "Validation_47895353/Validation_47895353_47895353_98742.841.png", "Validation_47895353/Validation_47895353_47895353_98744.840.png", "Validation_47895353/Validation_47895353_47895353_98746.840.png" ], "turn_id": 10, "scan_id": "Validation_47895353", "origin_question": "Is the round stainless steel pan with the flat base to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the round stainless steel pan with the flat base to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47333435/Validation_47333435_47333435_52364.117.png", "Validation_47333435/Validation_47333435_47333435_52366.100.png", "Validation_47333435/Validation_47333435_47333435_52368.099.png", "Validation_47333435/Validation_47333435_47333435_52370.115.png" ], "turn_id": 1, "scan_id": "Validation_47333435", "origin_question": "Remember, have you seen any monitor(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any monitor(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47333435/Validation_47333435_47333435_52372.114.png", "Validation_47333435/Validation_47333435_47333435_52374.113.png", "Validation_47333435/Validation_47333435_47333435_52376.112.png", "Validation_47333435/Validation_47333435_47333435_52378.111.png" ], "turn_id": 2, "scan_id": "Validation_47333435", "origin_question": "In which turn were you closest to \"the sleek white rectangular picture frame for three photos\"? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the sleek white rectangular picture frame for three photos\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_47333435/Validation_47333435_47333435_52380.111.png", "Validation_47333435/Validation_47333435_47333435_52382.110.png", "Validation_47333435/Validation_47333435_47333435_52384.109.png", "Validation_47333435/Validation_47333435_47333435_52386.108.png" ], "turn_id": 3, "scan_id": "Validation_47333435", "origin_question": "What will be the first-time appearance order of the following categories: flowerpot, monitor, picture?", "option": [ "monitor, picture, flowerpot", "flowerpot, monitor, picture", "flowerpot, picture, monitor", "monitor, flowerpot, picture" ], "answer": "monitor, picture, flowerpot", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: flowerpot, monitor, picture? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['monitor, picture, flowerpot', 'flowerpot, monitor, picture', 'flowerpot, picture, monitor', 'monitor, flowerpot, picture'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_47333435/Validation_47333435_47333435_52388.107.png", "Validation_47333435/Validation_47333435_47333435_52390.107.png", "Validation_47333435/Validation_47333435_47333435_52392.106.png", "Validation_47333435/Validation_47333435_47333435_52394.105.png" ], "turn_id": 4, "scan_id": "Validation_47333435", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47333435/Validation_47333435_47333435_52396.104.png", "Validation_47333435/Validation_47333435_47333435_52398.103.png", "Validation_47333435/Validation_47333435_47333435_52400.102.png", "Validation_47333435/Validation_47333435_47333435_52402.102.png" ], "turn_id": 5, "scan_id": "Validation_47333435", "origin_question": "", "option": [], "answer": "", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_47333435/Validation_47333435_47333435_52404.101.png", "Validation_47333435/Validation_47333435_47333435_52406.100.png", "Validation_47333435/Validation_47333435_47333435_52408.099.png", "Validation_47333435/Validation_47333435_47333435_52410.115.png" ], "turn_id": 6, "scan_id": "Validation_47333435", "origin_question": "Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "132.73", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_47333435/Validation_47333435_47333435_52412.114.png", "Validation_47333435/Validation_47333435_47333435_52414.113.png", "Validation_47333435/Validation_47333435_47333435_52416.112.png", "Validation_47333435/Validation_47333435_47333435_52418.112.png" ], "turn_id": 7, "scan_id": "Validation_47333435", "origin_question": "Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Left", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_47333435/Validation_47333435_47333435_52420.111.png", "Validation_47333435/Validation_47333435_47333435_52422.110.png", "Validation_47333435/Validation_47333435_47333435_52424.109.png", "Validation_47333435/Validation_47333435_47333435_52426.108.png" ], "turn_id": 8, "scan_id": "Validation_47333435", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the flowerpot\" from your position?", "option": [], "answer": "163.42", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the flowerpot\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_47333435/Validation_47333435_47333435_52428.107.png", "Validation_47333435/Validation_47333435_47333435_52430.107.png", "Validation_47333435/Validation_47333435_47333435_52432.106.png", "Validation_47333435/Validation_47333435_47333435_52434.105.png" ], "turn_id": 9, "scan_id": "Validation_47333435", "origin_question": "When did you first discover \"the door\"? (index of the turn)", "option": [], "answer": "6", "user_message": "For the 9 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the door\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47334361/Validation_47334361_47334361_61716.845.png", "Validation_47334361/Validation_47334361_47334361_61718.828.png", "Validation_47334361/Validation_47334361_47334361_61720.827.png", "Validation_47334361/Validation_47334361_47334361_61722.826.png", "Validation_47334361/Validation_47334361_47334361_61724.825.png" ], "turn_id": 1, "scan_id": "Validation_47334361", "origin_question": "Remember, have you seen any curtain(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any curtain(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47334361/Validation_47334361_47334361_61726.824.png", "Validation_47334361/Validation_47334361_47334361_61728.824.png", "Validation_47334361/Validation_47334361_47334361_61730.823.png", "Validation_47334361/Validation_47334361_47334361_61732.822.png", "Validation_47334361/Validation_47334361_47334361_61734.821.png" ], "turn_id": 2, "scan_id": "Validation_47334361", "origin_question": "In which turn were you farthest from \"the sleek, rectangular door with the neutral color\"? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the sleek, rectangular door with the neutral color\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_47334361/Validation_47334361_47334361_61736.820.png", "Validation_47334361/Validation_47334361_47334361_61738.819.png", "Validation_47334361/Validation_47334361_47334361_61740.819.png", "Validation_47334361/Validation_47334361_47334361_61742.818.png", "Validation_47334361/Validation_47334361_47334361_61744.817.png" ], "turn_id": 3, "scan_id": "Validation_47334361", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_47334361/Validation_47334361_47334361_61746.816.png", "Validation_47334361/Validation_47334361_47334361_61748.815.png", "Validation_47334361/Validation_47334361_47334361_61750.831.png", "Validation_47334361/Validation_47334361_47334361_61752.830.png", "Validation_47334361/Validation_47334361_47334361_61754.830.png" ], "turn_id": 4, "scan_id": "Validation_47334361", "origin_question": "In which turn did you see both \"the sleek, rectangular door with the neutral color\" and \"the large rectangular window with the neutral-colored frame\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the sleek, rectangular door with the neutral color\" and \"the large rectangular window with the neutral-colored frame\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47334361/Validation_47334361_47334361_61756.829.png", "Validation_47334361/Validation_47334361_47334361_61758.828.png", "Validation_47334361/Validation_47334361_47334361_61760.827.png", "Validation_47334361/Validation_47334361_47334361_61762.827.png", "Validation_47334361/Validation_47334361_47334361_61764.826.png" ], "turn_id": 5, "scan_id": "Validation_47334361", "origin_question": "In which turn were you closest to \"the large, sleek black monitor with the rectangular shape\"? ", "option": [], "answer": "5", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the large, sleek black monitor with the rectangular shape\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_47334361/Validation_47334361_47334361_61766.825.png", "Validation_47334361/Validation_47334361_47334361_61768.824.png", "Validation_47334361/Validation_47334361_47334361_61770.823.png", "Validation_47334361/Validation_47334361_47334361_61772.823.png", "Validation_47334361/Validation_47334361_47334361_61774.822.png" ], "turn_id": 6, "scan_id": "Validation_47334361", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the light blue plastic bottle with the tapered shape\" from your position?", "option": [], "answer": "178.4", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the light blue plastic bottle with the tapered shape\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47334361/Validation_47334361_47334361_61776.821.png", "Validation_47334361/Validation_47334361_47334361_61778.820.png", "Validation_47334361/Validation_47334361_47334361_61780.819.png", "Validation_47334361/Validation_47334361_47334361_61782.818.png", "Validation_47334361/Validation_47334361_47334361_61784.818.png" ], "turn_id": 7, "scan_id": "Validation_47334361", "origin_question": "", "option": [], "answer": "", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_47334361/Validation_47334361_47334361_61786.817.png", "Validation_47334361/Validation_47334361_47334361_61788.816.png", "Validation_47334361/Validation_47334361_47334361_61790.815.png", "Validation_47334361/Validation_47334361_47334361_61792.831.png", "Validation_47334361/Validation_47334361_47334361_61794.830.png" ], "turn_id": 8, "scan_id": "Validation_47334361", "origin_question": "Is the light blue plastic bottle with the tapered shape to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the light blue plastic bottle with the tapered shape to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_47334361/Validation_47334361_47334361_61796.829.png", "Validation_47334361/Validation_47334361_47334361_61798.829.png", "Validation_47334361/Validation_47334361_47334361_61800.828.png", "Validation_47334361/Validation_47334361_47334361_61802.827.png", "Validation_47334361/Validation_47334361_47334361_61804.826.png" ], "turn_id": 9, "scan_id": "Validation_47334361", "origin_question": "How far is your current position from where you were at the end of the turn 5? (in meters)", "option": [], "answer": "1.86", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 5? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_47334361/Validation_47334361_47334361_61806.825.png", "Validation_47334361/Validation_47334361_47334361_61808.825.png", "Validation_47334361/Validation_47334361_47334361_61810.824.png", "Validation_47334361/Validation_47334361_47334361_61812.823.png", "Validation_47334361/Validation_47334361_47334361_61814.822.png" ], "turn_id": 10, "scan_id": "Validation_47334361", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the light blue plastic bottle with the tapered shape\" from your position?", "option": [], "answer": "166.09", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the light blue plastic bottle with the tapered shape\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42446163/Validation_42446163_42446163_79566.524.png", "Validation_42446163/Validation_42446163_42446163_79568.523.png", "Validation_42446163/Validation_42446163_42446163_79570.522.png", "Validation_42446163/Validation_42446163_42446163_79572.521.png" ], "turn_id": 1, "scan_id": "Validation_42446163", "origin_question": "Remember, have you seen any bed(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any bed(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_42446163/Validation_42446163_42446163_79574.520.png", "Validation_42446163/Validation_42446163_42446163_79576.520.png", "Validation_42446163/Validation_42446163_42446163_79578.519.png", "Validation_42446163/Validation_42446163_42446163_79580.518.png" ], "turn_id": 2, "scan_id": "Validation_42446163", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the sleek, curved charcoal gray metal door\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the sleek, curved charcoal gray metal door\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_42446163/Validation_42446163_42446163_79582.517.png", "Validation_42446163/Validation_42446163_42446163_79584.533.png", "Validation_42446163/Validation_42446163_42446163_79586.532.png", "Validation_42446163/Validation_42446163_42446163_79588.532.png" ], "turn_id": 3, "scan_id": "Validation_42446163", "origin_question": "", "option": [], "answer": "", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_42446163/Validation_42446163_42446163_79590.531.png", "Validation_42446163/Validation_42446163_42446163_79592.530.png", "Validation_42446163/Validation_42446163_42446163_79594.529.png", "Validation_42446163/Validation_42446163_42446163_79596.529.png" ], "turn_id": 4, "scan_id": "Validation_42446163", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the sleek gray bed with the sturdy frame\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the sleek gray bed with the sturdy frame\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_42446163/Validation_42446163_42446163_79598.528.png", "Validation_42446163/Validation_42446163_42446163_79600.527.png", "Validation_42446163/Validation_42446163_42446163_79602.526.png", "Validation_42446163/Validation_42446163_42446163_79604.525.png" ], "turn_id": 5, "scan_id": "Validation_42446163", "origin_question": "", "option": [], "answer": "", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_42446163/Validation_42446163_42446163_79606.524.png", "Validation_42446163/Validation_42446163_42446163_79608.524.png", "Validation_42446163/Validation_42446163_42446163_79610.523.png", "Validation_42446163/Validation_42446163_42446163_79612.522.png" ], "turn_id": 6, "scan_id": "Validation_42446163", "origin_question": "Which direction is the sleek cylindrical gray or white bin in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the sleek cylindrical gray or white bin in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42446163/Validation_42446163_42446163_79614.521.png", "Validation_42446163/Validation_42446163_42446163_79616.520.png", "Validation_42446163/Validation_42446163_42446163_79618.520.png", "Validation_42446163/Validation_42446163_42446163_79620.519.png" ], "turn_id": 7, "scan_id": "Validation_42446163", "origin_question": "Remember, have you seen any bin(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any bin(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_42446163/Validation_42446163_42446163_79622.518.png", "Validation_42446163/Validation_42446163_42446163_79624.517.png", "Validation_42446163/Validation_42446163_42446163_79626.533.png", "Validation_42446163/Validation_42446163_42446163_79628.532.png" ], "turn_id": 8, "scan_id": "Validation_42446163", "origin_question": "What will be the first-time appearance order of the following categories: bed, pillow, table?", "option": [ "pillow, table, bed", "table, pillow, bed", "table, bed, pillow", "bed, table, pillow" ], "answer": "bed, table, pillow", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bed, pillow, table? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['pillow, table, bed', 'table, pillow, bed', 'table, bed, pillow', 'bed, table, pillow'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42446540/Validation_42446540_42446540_201452.352.png", "Validation_42446540/Validation_42446540_42446540_201454.351.png", "Validation_42446540/Validation_42446540_42446540_201456.351.png", "Validation_42446540/Validation_42446540_42446540_201458.350.png", "Validation_42446540/Validation_42446540_42446540_201460.349.png" ], "turn_id": 1, "scan_id": "Validation_42446540", "origin_question": "Remember, have you seen any helmets(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any helmets(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "Validation_42446540/Validation_42446540_42446540_201462.348.png", "Validation_42446540/Validation_42446540_42446540_201464.347.png", "Validation_42446540/Validation_42446540_42446540_201466.347.png", "Validation_42446540/Validation_42446540_42446540_201468.346.png", "Validation_42446540/Validation_42446540_42446540_201470.345.png" ], "turn_id": 2, "scan_id": "Validation_42446540", "origin_question": "Remember, how many pillow(s) have you seen so far? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many pillow(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_42446540/Validation_42446540_42446540_201472.344.png", "Validation_42446540/Validation_42446540_42446540_201474.343.png", "Validation_42446540/Validation_42446540_42446540_201476.342.png", "Validation_42446540/Validation_42446540_42446540_201478.342.png", "Validation_42446540/Validation_42446540_42446540_201480.341.png" ], "turn_id": 3, "scan_id": "Validation_42446540", "origin_question": "What will be the first-time appearance order of the following categories: couch, monitor, oven?", "option": [ "monitor, couch, oven", "monitor, oven, couch", "couch, monitor, oven", "oven, couch, monitor" ], "answer": "couch, monitor, oven", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: couch, monitor, oven? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['monitor, couch, oven', 'monitor, oven, couch', 'couch, monitor, oven', 'oven, couch, monitor'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_42446540/Validation_42446540_42446540_201482.340.png", "Validation_42446540/Validation_42446540_42446540_201484.339.png", "Validation_42446540/Validation_42446540_42446540_201486.338.png", "Validation_42446540/Validation_42446540_42446540_201488.337.png", "Validation_42446540/Validation_42446540_42446540_201490.337.png" ], "turn_id": 4, "scan_id": "Validation_42446540", "origin_question": "What will be the first-time appearance order of the following categories: couch, plant, stove?", "option": [ "plant, couch, stove", "plant, stove, couch", "stove, couch, plant", "couch, plant, stove" ], "answer": "couch, plant, stove", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: couch, plant, stove? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['plant, couch, stove', 'plant, stove, couch', 'stove, couch, plant', 'couch, plant, stove'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_42446540/Validation_42446540_42446540_201492.336.png", "Validation_42446540/Validation_42446540_42446540_201494.335.png", "Validation_42446540/Validation_42446540_42446540_201496.351.png", "Validation_42446540/Validation_42446540_42446540_201498.350.png", "Validation_42446540/Validation_42446540_42446540_201500.349.png" ], "turn_id": 5, "scan_id": "Validation_42446540", "origin_question": "Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance forward or backward from that position?", "option": [ "Forward", "Backward" ], "answer": "Backward", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance forward or backward from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Forward', 'Backward'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_42446540/Validation_42446540_42446540_201502.348.png", "Validation_42446540/Validation_42446540_42446540_201504.348.png", "Validation_42446540/Validation_42446540_42446540_201506.347.png", "Validation_42446540/Validation_42446540_42446540_201508.346.png", "Validation_42446540/Validation_42446540_42446540_201510.345.png" ], "turn_id": 6, "scan_id": "Validation_42446540", "origin_question": "By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the sleek black modern monitor with the slim profile\", Y: \"the modern rectangular stainless steel oven\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to Y and farther away from X", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the sleek black modern monitor with the slim profile\", Y: \"the modern rectangular stainless steel oven\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_42446540/Validation_42446540_42446540_201512.344.png", "Validation_42446540/Validation_42446540_42446540_201514.343.png", "Validation_42446540/Validation_42446540_42446540_201516.343.png", "Validation_42446540/Validation_42446540_42446540_201518.342.png", "Validation_42446540/Validation_42446540_42446540_201520.341.png" ], "turn_id": 7, "scan_id": "Validation_42446540", "origin_question": "Please recall \"the sleek stainless steel refrigerator with adjustable shelving\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.08", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the sleek stainless steel refrigerator with adjustable shelving\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_42446540/Validation_42446540_42446540_201522.340.png", "Validation_42446540/Validation_42446540_42446540_201524.339.png", "Validation_42446540/Validation_42446540_42446540_201526.338.png", "Validation_42446540/Validation_42446540_42446540_201528.338.png", "Validation_42446540/Validation_42446540_42446540_201530.337.png" ], "turn_id": 8, "scan_id": "Validation_42446540", "origin_question": "Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "Validation_47333452/Validation_47333452_47333452_54952.123.png", "Validation_47333452/Validation_47333452_47333452_54954.122.png", "Validation_47333452/Validation_47333452_47333452_54956.121.png", "Validation_47333452/Validation_47333452_47333452_54958.120.png" ], "turn_id": 1, "scan_id": "Validation_47333452", "origin_question": "Remember, how many fruit(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many fruit(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47333452/Validation_47333452_47333452_54960.120.png", "Validation_47333452/Validation_47333452_47333452_54962.119.png", "Validation_47333452/Validation_47333452_47333452_54964.218.png", "Validation_47333452/Validation_47333452_47333452_54966.217.png" ], "turn_id": 2, "scan_id": "Validation_47333452", "origin_question": "In which turn were you farthest from \"the white stainless steel refrigerator with heart-shaped stickers\"? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the white stainless steel refrigerator with heart-shaped stickers\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_47333452/Validation_47333452_47333452_54968.216.png", "Validation_47333452/Validation_47333452_47333452_54970.215.png", "Validation_47333452/Validation_47333452_47333452_54972.215.png", "Validation_47333452/Validation_47333452_47333452_54974.214.png" ], "turn_id": 3, "scan_id": "Validation_47333452", "origin_question": "Please recall \"the small, pink translucent case\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.86", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the small, pink translucent case\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_47333452/Validation_47333452_47333452_54976.213.png", "Validation_47333452/Validation_47333452_47333452_54978.212.png", "Validation_47333452/Validation_47333452_47333452_54980.211.png", "Validation_47333452/Validation_47333452_47333452_54982.211.png" ], "turn_id": 4, "scan_id": "Validation_47333452", "origin_question": "Is the small, pink translucent case to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the small, pink translucent case to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_47333452/Validation_47333452_47333452_54984.210.png", "Validation_47333452/Validation_47333452_47333452_54986.209.png", "Validation_47333452/Validation_47333452_47333452_54988.208.png", "Validation_47333452/Validation_47333452_47333452_54990.224.png" ], "turn_id": 5, "scan_id": "Validation_47333452", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the spherical light made of lightweight material\", Y: \"the white stainless steel refrigerator with heart-shaped stickers\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the spherical light made of lightweight material\", Y: \"the white stainless steel refrigerator with heart-shaped stickers\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_47333452/Validation_47333452_47333452_54992.223.png", "Validation_47333452/Validation_47333452_47333452_54994.222.png", "Validation_47333452/Validation_47333452_47333452_54996.221.png", "Validation_47333452/Validation_47333452_47333452_54998.221.png" ], "turn_id": 6, "scan_id": "Validation_47333452", "origin_question": "When did you first discover \"the small, pink translucent case\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the small, pink translucent case\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_41159558/Validation_41159558_41159558_32991.471.png", "Validation_41159558/Validation_41159558_41159558_32993.453.png", "Validation_41159558/Validation_41159558_41159558_32995.452.png", "Validation_41159558/Validation_41159558_41159558_32997.452.png", "Validation_41159558/Validation_41159558_41159558_32999.451.png" ], "turn_id": 1, "scan_id": "Validation_41159558", "origin_question": "Remember, have you seen any paper cutter(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any paper cutter(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "Validation_41159558/Validation_41159558_41159558_33001.450.png", "Validation_41159558/Validation_41159558_41159558_33003.449.png", "Validation_41159558/Validation_41159558_41159558_33005.448.png", "Validation_41159558/Validation_41159558_41159558_33007.448.png", "Validation_41159558/Validation_41159558_41159558_33009.447.png" ], "turn_id": 2, "scan_id": "Validation_41159558", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the rustic wooden bed with the natural finish\"; \"the vibrant geometric-patterned carpet with colorful triangles\"; \"the sleek rectangular synthetic backpack in neutral tone\" ", "option": [ "the rustic wooden bed with the natural finish", "the vibrant geometric-patterned carpet with colorful triangles", "the sleek rectangular synthetic backpack in neutral tone" ], "answer": "the sleek rectangular synthetic backpack in neutral tone", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the rustic wooden bed with the natural finish\"; \"the vibrant geometric-patterned carpet with colorful triangles\"; \"the sleek rectangular synthetic backpack in neutral tone\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rustic wooden bed with the natural finish', 'the vibrant geometric-patterned carpet with colorful triangles', 'the sleek rectangular synthetic backpack in neutral tone'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_41159558/Validation_41159558_41159558_33011.446.png", "Validation_41159558/Validation_41159558_41159558_33013.445.png", "Validation_41159558/Validation_41159558_41159558_33015.444.png", "Validation_41159558/Validation_41159558_41159558_33017.444.png", "Validation_41159558/Validation_41159558_41159558_33019.443.png" ], "turn_id": 3, "scan_id": "Validation_41159558", "origin_question": "Which two objects are on the same side of you now? \"the rustic wooden bed with the natural finish\", \"the rustic wood door with intricate grain patterns\", \"the white steel panel radiator with vertical grooves\". ", "option": [ "the rustic wooden bed with the natural finish, the rustic wood door with intricate grain patterns", "the rustic wooden bed with the natural finish, the white steel panel radiator with vertical grooves", "the rustic wood door with intricate grain patterns, the white steel panel radiator with vertical grooves" ], "answer": "the rustic wood door with intricate grain patterns, the white steel panel radiator with vertical grooves", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the rustic wooden bed with the natural finish\", \"the rustic wood door with intricate grain patterns\", \"the white steel panel radiator with vertical grooves\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rustic wooden bed with the natural finish, the rustic wood door with intricate grain patterns', 'the rustic wooden bed with the natural finish, the white steel panel radiator with vertical grooves', 'the rustic wood door with intricate grain patterns, the white steel panel radiator with vertical grooves'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "Validation_41159558/Validation_41159558_41159558_33021.459.png", "Validation_41159558/Validation_41159558_41159558_33023.458.png", "Validation_41159558/Validation_41159558_41159558_33025.457.png", "Validation_41159558/Validation_41159558_41159558_33027.456.png", "Validation_41159558/Validation_41159558_41159558_33029.455.png" ], "turn_id": 4, "scan_id": "Validation_41159558", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the dark athletic shoe with the rubber sole\"; \"the muted purple-gray cap with the curved visor\"; \"the vibrant geometric-patterned carpet with colorful triangles\" ", "option": [ "the dark athletic shoe with the rubber sole", "the muted purple-gray cap with the curved visor", "the vibrant geometric-patterned carpet with colorful triangles" ], "answer": "the dark athletic shoe with the rubber sole", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the dark athletic shoe with the rubber sole\"; \"the muted purple-gray cap with the curved visor\"; \"the vibrant geometric-patterned carpet with colorful triangles\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the dark athletic shoe with the rubber sole', 'the muted purple-gray cap with the curved visor', 'the vibrant geometric-patterned carpet with colorful triangles'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_41159558/Validation_41159558_41159558_33031.455.png", "Validation_41159558/Validation_41159558_41159558_33033.454.png", "Validation_41159558/Validation_41159558_41159558_33035.453.png", "Validation_41159558/Validation_41159558_41159558_33037.452.png", "Validation_41159558/Validation_41159558_41159558_33039.452.png" ], "turn_id": 5, "scan_id": "Validation_41159558", "origin_question": "Please recall \"the sleek rectangular synthetic backpack in neutral tone\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.55", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the sleek rectangular synthetic backpack in neutral tone\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_41159558/Validation_41159558_41159558_33041.451.png", "Validation_41159558/Validation_41159558_41159558_33043.450.png", "Validation_41159558/Validation_41159558_41159558_33045.449.png", "Validation_41159558/Validation_41159558_41159558_33047.448.png", "Validation_41159558/Validation_41159558_41159558_33049.448.png" ], "turn_id": 6, "scan_id": "Validation_41159558", "origin_question": "At the end of which turn were both of \"the natural wood grain table with the light finish\" and \"the sleek white oval bathtub\" on your left side?", "option": [], "answer": "5", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the natural wood grain table with the light finish\" and \"the sleek white oval bathtub\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_41159558/Validation_41159558_41159558_33051.447.png", "Validation_41159558/Validation_41159558_41159558_33053.446.png", "Validation_41159558/Validation_41159558_41159558_33055.445.png", "Validation_41159558/Validation_41159558_41159558_33057.444.png", "Validation_41159558/Validation_41159558_41159558_33059.444.png" ], "turn_id": 7, "scan_id": "Validation_41159558", "origin_question": "What will be the first-time appearance order of the following categories: bathtub, curtain, shelf?", "option": [ "bathtub, shelf, curtain", "curtain, bathtub, shelf", "curtain, shelf, bathtub", "bathtub, curtain, shelf" ], "answer": "curtain, shelf, bathtub", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bathtub, curtain, shelf? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['bathtub, shelf, curtain', 'curtain, bathtub, shelf', 'curtain, shelf, bathtub', 'bathtub, curtain, shelf'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_41159558/Validation_41159558_41159558_33061.443.png", "Validation_41159558/Validation_41159558_41159558_33063.459.png", "Validation_41159558/Validation_41159558_41159558_33065.458.png", "Validation_41159558/Validation_41159558_41159558_33067.457.png", "Validation_41159558/Validation_41159558_41159558_33069.456.png" ], "turn_id": 8, "scan_id": "Validation_41159558", "origin_question": "Please recall \"the sleek white oval bathtub\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "4.15", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the sleek white oval bathtub\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_41159558/Validation_41159558_41159558_33071.455.png", "Validation_41159558/Validation_41159558_41159558_33073.455.png", "Validation_41159558/Validation_41159558_41159558_33075.454.png", "Validation_41159558/Validation_41159558_41159558_33077.453.png", "Validation_41159558/Validation_41159558_41159558_33079.452.png" ], "turn_id": 9, "scan_id": "Validation_41159558", "origin_question": "Which two objects are on the same side of you now? \"the vibrant geometric-patterned carpet with colorful triangles\", \"the rustic wood door with intricate grain patterns\", \"the muted purple-gray cap with the curved visor\". ", "option": [ "the vibrant geometric-patterned carpet with colorful triangles, the rustic wood door with intricate grain patterns", "the vibrant geometric-patterned carpet with colorful triangles, the muted purple-gray cap with the curved visor", "the rustic wood door with intricate grain patterns, the muted purple-gray cap with the curved visor" ], "answer": "the vibrant geometric-patterned carpet with colorful triangles, the rustic wood door with intricate grain patterns", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the vibrant geometric-patterned carpet with colorful triangles\", \"the rustic wood door with intricate grain patterns\", \"the muted purple-gray cap with the curved visor\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the vibrant geometric-patterned carpet with colorful triangles, the rustic wood door with intricate grain patterns', 'the vibrant geometric-patterned carpet with colorful triangles, the muted purple-gray cap with the curved visor', 'the rustic wood door with intricate grain patterns, the muted purple-gray cap with the curved visor'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_41159558/Validation_41159558_41159558_33081.452.png", "Validation_41159558/Validation_41159558_41159558_33083.451.png", "Validation_41159558/Validation_41159558_41159558_33085.450.png", "Validation_41159558/Validation_41159558_41159558_33087.449.png", "Validation_41159558/Validation_41159558_41159558_33089.448.png" ], "turn_id": 10, "scan_id": "Validation_41159558", "origin_question": "Using your orientation at the end of turn 6 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 6 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42898527/Validation_42898527_42898527_750594.480.png", "Validation_42898527/Validation_42898527_42898527_750596.462.png", "Validation_42898527/Validation_42898527_42898527_750598.478.png", "Validation_42898527/Validation_42898527_42898527_750600.477.png", "Validation_42898527/Validation_42898527_42898527_750602.476.png" ], "turn_id": 1, "scan_id": "Validation_42898527", "origin_question": "Remember, have you seen any soap(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any soap(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42898527/Validation_42898527_42898527_750604.476.png", "Validation_42898527/Validation_42898527_42898527_750606.475.png", "Validation_42898527/Validation_42898527_42898527_750608.474.png", "Validation_42898527/Validation_42898527_42898527_750610.473.png", "Validation_42898527/Validation_42898527_42898527_750612.472.png" ], "turn_id": 2, "scan_id": "Validation_42898527", "origin_question": "Remember, have you seen any soap dish(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any soap dish(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_42898527/Validation_42898527_42898527_750614.471.png", "Validation_42898527/Validation_42898527_42898527_750616.471.png", "Validation_42898527/Validation_42898527_42898527_750618.470.png", "Validation_42898527/Validation_42898527_42898527_750620.469.png", "Validation_42898527/Validation_42898527_42898527_750622.468.png" ], "turn_id": 3, "scan_id": "Validation_42898527", "origin_question": "Please recall \"the rectangular wooden door with glass panes\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.59", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the rectangular wooden door with glass panes\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_42898527/Validation_42898527_42898527_750624.467.png", "Validation_42898527/Validation_42898527_42898527_750626.466.png", "Validation_42898527/Validation_42898527_42898527_750628.466.png", "Validation_42898527/Validation_42898527_42898527_750630.465.png", "Validation_42898527/Validation_42898527_42898527_750632.464.png" ], "turn_id": 4, "scan_id": "Validation_42898527", "origin_question": "Is the dark rectangular stainless steel bin with dual compartments to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the dark rectangular stainless steel bin with dual compartments to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_42898527/Validation_42898527_42898527_750634.463.png", "Validation_42898527/Validation_42898527_42898527_750636.462.png", "Validation_42898527/Validation_42898527_42898527_750638.462.png", "Validation_42898527/Validation_42898527_42898527_750640.461.png", "Validation_42898527/Validation_42898527_42898527_750642.477.png" ], "turn_id": 5, "scan_id": "Validation_42898527", "origin_question": "Which direction is the compact modern stainless steel and glass oven in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the compact modern stainless steel and glass oven in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_42898527/Validation_42898527_42898527_750644.476.png", "Validation_42898527/Validation_42898527_42898527_750646.475.png", "Validation_42898527/Validation_42898527_42898527_750648.474.png", "Validation_42898527/Validation_42898527_42898527_750650.473.png", "Validation_42898527/Validation_42898527_42898527_750652.473.png" ], "turn_id": 6, "scan_id": "Validation_42898527", "origin_question": "In which turn did you see both \"the rectangular wooden door with glass panes\" and \"the dark rectangular stainless steel bin with dual compartments\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the rectangular wooden door with glass panes\" and \"the dark rectangular stainless steel bin with dual compartments\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_42898527/Validation_42898527_42898527_750654.472.png", "Validation_42898527/Validation_42898527_42898527_750656.471.png", "Validation_42898527/Validation_42898527_42898527_750658.470.png", "Validation_42898527/Validation_42898527_42898527_750660.469.png", "Validation_42898527/Validation_42898527_42898527_750662.468.png" ], "turn_id": 7, "scan_id": "Validation_42898527", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular wooden door with glass panes\", Y: \"the medium-sized angular range hood\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to Y and farther away from X", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular wooden door with glass panes\", Y: \"the medium-sized angular range hood\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_42898527/Validation_42898527_42898527_750664.468.png", "Validation_42898527/Validation_42898527_42898527_750666.467.png", "Validation_42898527/Validation_42898527_42898527_750668.466.png", "Validation_42898527/Validation_42898527_42898527_750670.465.png", "Validation_42898527/Validation_42898527_42898527_750672.464.png" ], "turn_id": 8, "scan_id": "Validation_42898527", "origin_question": "Which direction is the compact modern stainless steel and glass oven in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the compact modern stainless steel and glass oven in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_42898527/Validation_42898527_42898527_750674.463.png", "Validation_42898527/Validation_42898527_42898527_750676.463.png", "Validation_42898527/Validation_42898527_42898527_750678.462.png", "Validation_42898527/Validation_42898527_42898527_750680.461.png", "Validation_42898527/Validation_42898527_42898527_750682.477.png" ], "turn_id": 9, "scan_id": "Validation_42898527", "origin_question": "Is the compact modern stainless steel and glass oven to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the compact modern stainless steel and glass oven to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_42898527/Validation_42898527_42898527_750684.476.png", "Validation_42898527/Validation_42898527_42898527_750686.475.png", "Validation_42898527/Validation_42898527_42898527_750688.474.png", "Validation_42898527/Validation_42898527_42898527_750690.474.png", "Validation_42898527/Validation_42898527_42898527_750692.473.png" ], "turn_id": 10, "scan_id": "Validation_42898527", "origin_question": "Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "102.17", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42445991/Validation_42445991_42445991_75256.441.png", "Validation_42445991/Validation_42445991_42445991_75258.424.png", "Validation_42445991/Validation_42445991_42445991_75260.423.png", "Validation_42445991/Validation_42445991_42445991_75262.422.png", "Validation_42445991/Validation_42445991_42445991_75264.421.png" ], "turn_id": 1, "scan_id": "Validation_42445991", "origin_question": "Remember, have you seen any curtain(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any curtain(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_42445991/Validation_42445991_42445991_75266.437.png", "Validation_42445991/Validation_42445991_42445991_75268.436.png", "Validation_42445991/Validation_42445991_42445991_75270.435.png", "Validation_42445991/Validation_42445991_42445991_75272.435.png", "Validation_42445991/Validation_42445991_42445991_75274.434.png" ], "turn_id": 2, "scan_id": "Validation_42445991", "origin_question": "At the end of which turn were both of \"an elegant rectangular window with white wooden frame\" and \"the compact black metal wood-burning fireplace\" on your right side?", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"an elegant rectangular window with white wooden frame\" and \"the compact black metal wood-burning fireplace\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_42445991/Validation_42445991_42445991_75276.433.png", "Validation_42445991/Validation_42445991_42445991_75278.432.png", "Validation_42445991/Validation_42445991_42445991_75280.431.png", "Validation_42445991/Validation_42445991_42445991_75282.430.png", "Validation_42445991/Validation_42445991_42445991_75284.429.png" ], "turn_id": 3, "scan_id": "Validation_42445991", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_42445991/Validation_42445991_42445991_75286.429.png", "Validation_42445991/Validation_42445991_42445991_75288.428.png", "Validation_42445991/Validation_42445991_42445991_75290.427.png", "Validation_42445991/Validation_42445991_42445991_75292.426.png", "Validation_42445991/Validation_42445991_42445991_75294.425.png" ], "turn_id": 4, "scan_id": "Validation_42445991", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the large rectangular table with the blue and white cloth\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the large rectangular table with the blue and white cloth\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_42445991/Validation_42445991_42445991_75296.425.png", "Validation_42445991/Validation_42445991_42445991_75298.424.png", "Validation_42445991/Validation_42445991_42445991_75300.423.png", "Validation_42445991/Validation_42445991_42445991_75302.422.png", "Validation_42445991/Validation_42445991_42445991_75304.421.png" ], "turn_id": 5, "scan_id": "Validation_42445991", "origin_question": "In which turn were you closest to \"the large dark wood cabinet with the polished finish\"? ", "option": [], "answer": "1", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the large dark wood cabinet with the polished finish\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_42445991/Validation_42445991_42445991_75306.437.png", "Validation_42445991/Validation_42445991_42445991_75308.436.png", "Validation_42445991/Validation_42445991_42445991_75310.435.png", "Validation_42445991/Validation_42445991_42445991_75312.435.png", "Validation_42445991/Validation_42445991_42445991_75314.434.png" ], "turn_id": 6, "scan_id": "Validation_42445991", "origin_question": "Is the compact black metal wood-burning fireplace to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the compact black metal wood-burning fireplace to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_42445991/Validation_42445991_42445991_75316.433.png", "Validation_42445991/Validation_42445991_42445991_75318.432.png", "Validation_42445991/Validation_42445991_42445991_75320.431.png", "Validation_42445991/Validation_42445991_42445991_75322.430.png", "Validation_42445991/Validation_42445991_42445991_75324.430.png" ], "turn_id": 7, "scan_id": "Validation_42445991", "origin_question": "Compared to your position at the end of the 3 turn, are you now closer or farther away from \"an elegant rectangular window with white wooden frame\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 3 turn, are you now closer or farther away from \"an elegant rectangular window with white wooden frame\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_42445991/Validation_42445991_42445991_75326.429.png", "Validation_42445991/Validation_42445991_42445991_75328.428.png", "Validation_42445991/Validation_42445991_42445991_75330.427.png", "Validation_42445991/Validation_42445991_42445991_75332.426.png", "Validation_42445991/Validation_42445991_42445991_75334.425.png" ], "turn_id": 8, "scan_id": "Validation_42445991", "origin_question": "How far is your current position from where you were at the end of the turn 6? (in meters)", "option": [], "answer": "1.17", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 6? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_42445991/Validation_42445991_42445991_75336.425.png", "Validation_42445991/Validation_42445991_42445991_75338.424.png", "Validation_42445991/Validation_42445991_42445991_75340.423.png", "Validation_42445991/Validation_42445991_42445991_75342.422.png", "Validation_42445991/Validation_42445991_42445991_75344.421.png" ], "turn_id": 9, "scan_id": "Validation_42445991", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the large dark wood cabinet with the polished finish\" from your position?", "option": [], "answer": "35.33", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the large dark wood cabinet with the polished finish\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47895549/Validation_47895549_47895549_455872.624.png", "Validation_47895549/Validation_47895549_47895549_455874.107.png", "Validation_47895549/Validation_47895549_47895549_455875.606.png", "Validation_47895549/Validation_47895549_47895549_455877.105.png" ], "turn_id": 1, "scan_id": "Validation_47895549", "origin_question": "Remember, have you seen any chair(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any chair(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_47895549/Validation_47895549_47895549_455878.605.png", "Validation_47895549/Validation_47895549_47895549_455880.104.png", "Validation_47895549/Validation_47895549_47895549_455881.604.png", "Validation_47895549/Validation_47895549_47895549_455883.103.png" ], "turn_id": 2, "scan_id": "Validation_47895549", "origin_question": "When did you first discover \"the modern stainless steel refrigerator with ample storage\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the modern stainless steel refrigerator with ample storage\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_47895549/Validation_47895549_47895549_455884.602.png", "Validation_47895549/Validation_47895549_47895549_455886.102.png", "Validation_47895549/Validation_47895549_47895549_455887.601.png", "Validation_47895549/Validation_47895549_47895549_455889.101.png" ], "turn_id": 3, "scan_id": "Validation_47895549", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "59.29", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_47895549/Validation_47895549_47895549_455890.600.png", "Validation_47895549/Validation_47895549_47895549_455892.099.png", "Validation_47895549/Validation_47895549_47895549_455893.599.png", "Validation_47895549/Validation_47895549_47895549_455895.098.png" ], "turn_id": 4, "scan_id": "Validation_47895549", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_47895549/Validation_47895549_47895549_455896.597.png", "Validation_47895549/Validation_47895549_47895549_455898.097.png", "Validation_47895549/Validation_47895549_47895549_455899.596.png", "Validation_47895549/Validation_47895549_47895549_455901.112.png" ], "turn_id": 5, "scan_id": "Validation_47895549", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "129.71", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_47895549/Validation_47895549_47895549_455902.612.png", "Validation_47895549/Validation_47895549_47895549_455904.111.png", "Validation_47895549/Validation_47895549_47895549_455905.611.png", "Validation_47895549/Validation_47895549_47895549_455907.110.png" ], "turn_id": 6, "scan_id": "Validation_47895549", "origin_question": "Using your orientation at the end of turn 5 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 5 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_47895549/Validation_47895549_47895549_455908.609.png", "Validation_47895549/Validation_47895549_47895549_455910.109.png", "Validation_47895549/Validation_47895549_47895549_455911.608.png", "Validation_47895549/Validation_47895549_47895549_455913.107.png" ], "turn_id": 7, "scan_id": "Validation_47895549", "origin_question": "By the end of the 3 turn, were you getting closer to or farther away from these two objects: X: \"the modern stainless steel refrigerator with ample storage\", Y: \"the vibrant cylindrical plastic bottle with colorful accents\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to Y and farther away from X", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 3 turn, were you getting closer to or farther away from these two objects: X: \"the modern stainless steel refrigerator with ample storage\", Y: \"the vibrant cylindrical plastic bottle with colorful accents\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_47895549/Validation_47895549_47895549_455914.607.png", "Validation_47895549/Validation_47895549_47895549_455916.106.png", "Validation_47895549/Validation_47895549_47895549_455917.606.png", "Validation_47895549/Validation_47895549_47895549_455919.105.png" ], "turn_id": 8, "scan_id": "Validation_47895549", "origin_question": "What will be the first-time appearance order of the following categories: oven, range hood, refrigerator?", "option": [ "oven, refrigerator, range hood", "refrigerator, range hood, oven", "refrigerator, oven, range hood", "oven, range hood, refrigerator" ], "answer": "refrigerator, range hood, oven", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: oven, range hood, refrigerator? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['oven, refrigerator, range hood', 'refrigerator, range hood, oven', 'refrigerator, oven, range hood', 'oven, range hood, refrigerator'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42897552/Validation_42897552_42897552_277361.813.png", "Validation_42897552/Validation_42897552_42897552_277363.296.png", "Validation_42897552/Validation_42897552_42897552_277364.795.png", "Validation_42897552/Validation_42897552_42897552_277366.295.png" ], "turn_id": 1, "scan_id": "Validation_42897552", "origin_question": "Remember, have you seen any hair dryer(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any hair dryer(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 3 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_42897552/Validation_42897552_42897552_277367.794.png", "Validation_42897552/Validation_42897552_42897552_277369.294.png", "Validation_42897552/Validation_42897552_42897552_277370.793.png", "Validation_42897552/Validation_42897552_42897552_277372.309.png" ], "turn_id": 2, "scan_id": "Validation_42897552", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 3 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_42897552/Validation_42897552_42897552_277373.808.png", "Validation_42897552/Validation_42897552_42897552_277375.308.png", "Validation_42897552/Validation_42897552_42897552_277376.807.png", "Validation_42897552/Validation_42897552_42897552_277378.307.png" ], "turn_id": 3, "scan_id": "Validation_42897552", "origin_question": "In which turn did you see both \"the sleek rectangular wood door with the metal handle\" and \"the large rectangular window with light-colored blinds\" simultaneously? ", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the sleek rectangular wood door with the metal handle\" and \"the large rectangular window with light-colored blinds\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 3 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42446056/Validation_42446056_42446056_349091.393.png", "Validation_42446056/Validation_42446056_42446056_349093.375.png", "Validation_42446056/Validation_42446056_42446056_349095.374.png", "Validation_42446056/Validation_42446056_42446056_349097.374.png", "Validation_42446056/Validation_42446056_42446056_349099.373.png" ], "turn_id": 1, "scan_id": "Validation_42446056", "origin_question": "Remember, have you seen any table(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any table(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_42446056/Validation_42446056_42446056_349101.389.png", "Validation_42446056/Validation_42446056_42446056_349103.388.png", "Validation_42446056/Validation_42446056_42446056_349105.387.png", "Validation_42446056/Validation_42446056_42446056_349107.386.png", "Validation_42446056/Validation_42446056_42446056_349109.385.png" ], "turn_id": 2, "scan_id": "Validation_42446056", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the tall, glossy hardcover book\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the tall, glossy hardcover book\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "Validation_42446056/Validation_42446056_42446056_349111.384.png", "Validation_42446056/Validation_42446056_42446056_349113.384.png", "Validation_42446056/Validation_42446056_42446056_349115.383.png", "Validation_42446056/Validation_42446056_42446056_349117.382.png", "Validation_42446056/Validation_42446056_42446056_349119.381.png" ], "turn_id": 3, "scan_id": "Validation_42446056", "origin_question": "Remember, how many table(s) have you seen so far? ", "option": [], "answer": "2", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many table(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_42446056/Validation_42446056_42446056_349121.380.png", "Validation_42446056/Validation_42446056_42446056_349123.380.png", "Validation_42446056/Validation_42446056_42446056_349125.379.png", "Validation_42446056/Validation_42446056_42446056_349127.378.png", "Validation_42446056/Validation_42446056_42446056_349129.377.png" ], "turn_id": 4, "scan_id": "Validation_42446056", "origin_question": "In which turn did you see both \"the modern deep gray rectangular plastic bin\" and \"the tall, glossy hardcover book\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the modern deep gray rectangular plastic bin\" and \"the tall, glossy hardcover book\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_42446056/Validation_42446056_42446056_349131.376.png", "Validation_42446056/Validation_42446056_42446056_349133.376.png", "Validation_42446056/Validation_42446056_42446056_349135.375.png", "Validation_42446056/Validation_42446056_42446056_349137.374.png", "Validation_42446056/Validation_42446056_42446056_349139.373.png" ], "turn_id": 5, "scan_id": "Validation_42446056", "origin_question": "Please recall \"the tall, glossy hardcover book\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.57", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the tall, glossy hardcover book\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_42446056/Validation_42446056_42446056_349141.372.png", "Validation_42446056/Validation_42446056_42446056_349143.388.png", "Validation_42446056/Validation_42446056_42446056_349145.387.png", "Validation_42446056/Validation_42446056_42446056_349147.386.png", "Validation_42446056/Validation_42446056_42446056_349149.385.png" ], "turn_id": 6, "scan_id": "Validation_42446056", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the bed with the vibrant red bedspread and purple canopy\" from your position?", "option": [], "answer": "171.64", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the bed with the vibrant red bedspread and purple canopy\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_42446056/Validation_42446056_42446056_349151.385.png", "Validation_42446056/Validation_42446056_42446056_349153.384.png", "Validation_42446056/Validation_42446056_42446056_349155.383.png", "Validation_42446056/Validation_42446056_42446056_349157.382.png", "Validation_42446056/Validation_42446056_42446056_349159.381.png" ], "turn_id": 7, "scan_id": "Validation_42446056", "origin_question": "", "option": [], "answer": "", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_42446056/Validation_42446056_42446056_349161.381.png", "Validation_42446056/Validation_42446056_42446056_349163.380.png", "Validation_42446056/Validation_42446056_42446056_349165.379.png", "Validation_42446056/Validation_42446056_42446056_349167.378.png", "Validation_42446056/Validation_42446056_42446056_349169.377.png" ], "turn_id": 8, "scan_id": "Validation_42446056", "origin_question": "At the end of which turn were both of \"the modern deep gray rectangular plastic bin\" and \"the tall, glossy hardcover book\" on your right side?", "option": [], "answer": "4", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the modern deep gray rectangular plastic bin\" and \"the tall, glossy hardcover book\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_42446056/Validation_42446056_42446056_349171.377.png", "Validation_42446056/Validation_42446056_42446056_349173.376.png", "Validation_42446056/Validation_42446056_42446056_349175.375.png", "Validation_42446056/Validation_42446056_42446056_349177.374.png", "Validation_42446056/Validation_42446056_42446056_349179.373.png" ], "turn_id": 9, "scan_id": "Validation_42446056", "origin_question": "By the end of the 5 turn, were you getting closer to or farther away from these two objects: X: \"the modern deep gray rectangular plastic bin\", Y: \"the bed with the vibrant red bedspread and purple canopy\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to Y and farther away from X", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 5 turn, were you getting closer to or farther away from these two objects: X: \"the modern deep gray rectangular plastic bin\", Y: \"the bed with the vibrant red bedspread and purple canopy\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_42446056/Validation_42446056_42446056_349181.372.png", "Validation_42446056/Validation_42446056_42446056_349183.388.png", "Validation_42446056/Validation_42446056_42446056_349185.387.png", "Validation_42446056/Validation_42446056_42446056_349187.387.png", "Validation_42446056/Validation_42446056_42446056_349189.386.png" ], "turn_id": 10, "scan_id": "Validation_42446056", "origin_question": "", "option": [], "answer": "", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_41069025/Validation_41069025_41069025_589.877.png", "Validation_41069025/Validation_41069025_41069025_591.860.png", "Validation_41069025/Validation_41069025_41069025_593.859.png", "Validation_41069025/Validation_41069025_41069025_595.858.png", "Validation_41069025/Validation_41069025_41069025_597.857.png" ], "turn_id": 1, "scan_id": "Validation_41069025", "origin_question": "Remember, have you seen any microwave(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any microwave(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_41069025/Validation_41069025_41069025_599.856.png", "Validation_41069025/Validation_41069025_41069025_601.856.png", "Validation_41069025/Validation_41069025_41069025_603.855.png", "Validation_41069025/Validation_41069025_41069025_605.854.png", "Validation_41069025/Validation_41069025_41069025_607.870.png" ], "turn_id": 2, "scan_id": "Validation_41069025", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_41069025/Validation_41069025_41069025_609.869.png", "Validation_41069025/Validation_41069025_41069025_611.868.png", "Validation_41069025/Validation_41069025_41069025_613.867.png", "Validation_41069025/Validation_41069025_41069025_615.867.png", "Validation_41069025/Validation_41069025_41069025_617.866.png" ], "turn_id": 3, "scan_id": "Validation_41069025", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_41069025/Validation_41069025_41069025_619.865.png", "Validation_41069025/Validation_41069025_41069025_621.864.png", "Validation_41069025/Validation_41069025_41069025_623.863.png", "Validation_41069025/Validation_41069025_41069025_625.862.png", "Validation_41069025/Validation_41069025_41069025_627.862.png" ], "turn_id": 4, "scan_id": "Validation_41069025", "origin_question": "At the end of which turn were both of \"the compact silver microwave with the black glass door\" and \"the sleek metallic silver stainless steel bin\" on your left side?", "option": [], "answer": "2", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the compact silver microwave with the black glass door\" and \"the sleek metallic silver stainless steel bin\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_41069025/Validation_41069025_41069025_629.861.png", "Validation_41069025/Validation_41069025_41069025_631.860.png", "Validation_41069025/Validation_41069025_41069025_633.859.png", "Validation_41069025/Validation_41069025_41069025_635.858.png", "Validation_41069025/Validation_41069025_41069025_637.857.png" ], "turn_id": 5, "scan_id": "Validation_41069025", "origin_question": "Please recall \"the sleek, cylindrical silver teapot\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.88", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the sleek, cylindrical silver teapot\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_41069025/Validation_41069025_41069025_639.857.png", "Validation_41069025/Validation_41069025_41069025_641.856.png", "Validation_41069025/Validation_41069025_41069025_643.855.png", "Validation_41069025/Validation_41069025_41069025_645.854.png", "Validation_41069025/Validation_41069025_41069025_647.870.png" ], "turn_id": 6, "scan_id": "Validation_41069025", "origin_question": "Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the compact silver microwave with the black glass door\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the compact silver microwave with the black glass door\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_41069025/Validation_41069025_41069025_649.869.png", "Validation_41069025/Validation_41069025_41069025_651.868.png", "Validation_41069025/Validation_41069025_41069025_653.868.png", "Validation_41069025/Validation_41069025_41069025_655.867.png", "Validation_41069025/Validation_41069025_41069025_657.866.png" ], "turn_id": 7, "scan_id": "Validation_41069025", "origin_question": "At the end of which turn were both of \"the sleek, cylindrical silver teapot\" and \"the large, sleek black monitor with the reflective glass screen\" on your left side?", "option": [], "answer": "7", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the sleek, cylindrical silver teapot\" and \"the large, sleek black monitor with the reflective glass screen\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_41069025/Validation_41069025_41069025_659.865.png", "Validation_41069025/Validation_41069025_41069025_661.864.png", "Validation_41069025/Validation_41069025_41069025_663.864.png", "Validation_41069025/Validation_41069025_41069025_665.863.png", "Validation_41069025/Validation_41069025_41069025_667.862.png" ], "turn_id": 8, "scan_id": "Validation_41069025", "origin_question": "At the end of which turn were both of \"the sleek rectangular stainless steel oven\" and \"the large, sleek black monitor with the reflective glass screen\" on your right side?", "option": [], "answer": "2", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the sleek rectangular stainless steel oven\" and \"the large, sleek black monitor with the reflective glass screen\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_41069025/Validation_41069025_41069025_669.861.png", "Validation_41069025/Validation_41069025_41069025_671.860.png", "Validation_41069025/Validation_41069025_41069025_673.859.png", "Validation_41069025/Validation_41069025_41069025_675.859.png", "Validation_41069025/Validation_41069025_41069025_677.858.png" ], "turn_id": 9, "scan_id": "Validation_41069025", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the sleek, stainless steel stove with the rectangular shape\" from your position?", "option": [], "answer": "174.03", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the sleek, stainless steel stove with the rectangular shape\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_41069025/Validation_41069025_41069025_679.857.png", "Validation_41069025/Validation_41069025_41069025_681.856.png", "Validation_41069025/Validation_41069025_41069025_683.855.png", "Validation_41069025/Validation_41069025_41069025_685.854.png", "Validation_41069025/Validation_41069025_41069025_687.854.png" ], "turn_id": 10, "scan_id": "Validation_41069025", "origin_question": "Which direction is the sleek rectangular stainless steel oven in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the sleek rectangular stainless steel oven in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47895534/Validation_47895534_47895534_455559.265.png", "Validation_47895534/Validation_47895534_47895534_455560.748.png", "Validation_47895534/Validation_47895534_47895534_455562.247.png", "Validation_47895534/Validation_47895534_47895534_455563.747.png" ], "turn_id": 1, "scan_id": "Validation_47895534", "origin_question": "Remember, have you seen any mailbox(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any mailbox(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47895534/Validation_47895534_47895534_455565.246.png", "Validation_47895534/Validation_47895534_47895534_455566.746.png", "Validation_47895534/Validation_47895534_47895534_455568.245.png", "Validation_47895534/Validation_47895534_47895534_455569.744.png" ], "turn_id": 2, "scan_id": "Validation_47895534", "origin_question": "In which turn were you closest to \"the medium brown rectangular table with wood grain\"? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the medium brown rectangular table with wood grain\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_47895534/Validation_47895534_47895534_455571.244.png", "Validation_47895534/Validation_47895534_47895534_455572.743.png", "Validation_47895534/Validation_47895534_47895534_455574.242.png", "Validation_47895534/Validation_47895534_47895534_455575.742.png" ], "turn_id": 3, "scan_id": "Validation_47895534", "origin_question": "Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Right", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_47895534/Validation_47895534_47895534_455577.241.png", "Validation_47895534/Validation_47895534_47895534_455578.741.png", "Validation_47895534/Validation_47895534_47895534_455580.240.png", "Validation_47895534/Validation_47895534_47895534_455581.740.png" ], "turn_id": 4, "scan_id": "Validation_47895534", "origin_question": "At the end of which turn were both of \"the medium brown rectangular table with wood grain\" and \"the tall picture with elegant glass doors\" on your left side?", "option": [], "answer": "1", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the medium brown rectangular table with wood grain\" and \"the tall picture with elegant glass doors\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_47895534/Validation_47895534_47895534_455583.239.png", "Validation_47895534/Validation_47895534_47895534_455584.738.png", "Validation_47895534/Validation_47895534_47895534_455586.238.png", "Validation_47895534/Validation_47895534_47895534_455587.737.png" ], "turn_id": 5, "scan_id": "Validation_47895534", "origin_question": "What will be the first-time appearance order of the following categories: cabinet, shelf, table?", "option": [ "shelf, cabinet, table", "cabinet, table, shelf", "shelf, table, cabinet", "table, cabinet, shelf" ], "answer": "table, cabinet, shelf", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: cabinet, shelf, table? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['shelf, cabinet, table', 'cabinet, table, shelf', 'shelf, table, cabinet', 'table, cabinet, shelf'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_47895534/Validation_47895534_47895534_455589.236.png", "Validation_47895534/Validation_47895534_47895534_455590.736.png", "Validation_47895534/Validation_47895534_47895534_455592.235.png", "Validation_47895534/Validation_47895534_47895534_455593.735.png" ], "turn_id": 6, "scan_id": "Validation_47895534", "origin_question": "Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Right", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_47895534/Validation_47895534_47895534_455595.234.png", "Validation_47895534/Validation_47895534_47895534_455596.733.png", "Validation_47895534/Validation_47895534_47895534_455598.233.png", "Validation_47895534/Validation_47895534_47895534_455599.749.png" ], "turn_id": 7, "scan_id": "Validation_47895534", "origin_question": "In which turn did you see both \"the natural-finished wooden door with the solid texture\" and \"the rectangular window with the white frame\" simultaneously? ", "option": [], "answer": "6", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the natural-finished wooden door with the solid texture\" and \"the rectangular window with the white frame\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47430005/Validation_47430005_47430005_17513.379.png", "Validation_47430005/Validation_47430005_47430005_17514.861.png", "Validation_47430005/Validation_47430005_47430005_17516.361.png", "Validation_47430005/Validation_47430005_47430005_17517.860.png" ], "turn_id": 1, "scan_id": "Validation_47430005", "origin_question": "Remember, have you seen any washing machine(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any washing machine(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 3 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_47430005/Validation_47430005_47430005_17519.360.png", "Validation_47430005/Validation_47430005_47430005_17520.876.png", "Validation_47430005/Validation_47430005_47430005_17522.375.png", "Validation_47430005/Validation_47430005_47430005_17523.874.png" ], "turn_id": 2, "scan_id": "Validation_47430005", "origin_question": "In which turn did you see both \"the washing machine\" and \"the medium-sized soft pink synthetic leather bag\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the washing machine\" and \"the medium-sized soft pink synthetic leather bag\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 3 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_47430005/Validation_47430005_47430005_17525.374.png", "Validation_47430005/Validation_47430005_47430005_17526.873.png", "Validation_47430005/Validation_47430005_47430005_17528.373.png", "Validation_47430005/Validation_47430005_47430005_17529.872.png" ], "turn_id": 3, "scan_id": "Validation_47430005", "origin_question": "At the end of which turn were both of \"the body loofah\" and \"the modern white circular light made of frosted glass\" on your right side?", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the body loofah\" and \"the modern white circular light made of frosted glass\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 3 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_44358455/Validation_44358455_44358455_43394.312.png", "Validation_44358455/Validation_44358455_44358455_43396.295.png", "Validation_44358455/Validation_44358455_44358455_43398.294.png", "Validation_44358455/Validation_44358455_44358455_43400.293.png", "Validation_44358455/Validation_44358455_44358455_43402.292.png" ], "turn_id": 1, "scan_id": "Validation_44358455", "origin_question": "Remember, have you seen any cabinet(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any cabinet(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_44358455/Validation_44358455_44358455_43404.292.png", "Validation_44358455/Validation_44358455_44358455_43406.291.png", "Validation_44358455/Validation_44358455_44358455_43408.290.png", "Validation_44358455/Validation_44358455_44358455_43410.289.png", "Validation_44358455/Validation_44358455_44358455_43412.288.png" ], "turn_id": 2, "scan_id": "Validation_44358455", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_44358455/Validation_44358455_44358455_43414.287.png", "Validation_44358455/Validation_44358455_44358455_43416.287.png", "Validation_44358455/Validation_44358455_44358455_43418.286.png", "Validation_44358455/Validation_44358455_44358455_43420.285.png", "Validation_44358455/Validation_44358455_44358455_43422.301.png" ], "turn_id": 3, "scan_id": "Validation_44358455", "origin_question": "Remember, have you seen any detergent(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any detergent(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_44358455/Validation_44358455_44358455_43424.283.png", "Validation_44358455/Validation_44358455_44358455_43426.299.png", "Validation_44358455/Validation_44358455_44358455_43428.298.png", "Validation_44358455/Validation_44358455_44358455_43430.298.png", "Validation_44358455/Validation_44358455_44358455_43432.297.png" ], "turn_id": 4, "scan_id": "Validation_44358455", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "123.21", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_44358455/Validation_44358455_44358455_43434.296.png", "Validation_44358455/Validation_44358455_44358455_43436.295.png", "Validation_44358455/Validation_44358455_44358455_43438.294.png", "Validation_44358455/Validation_44358455_44358455_43440.293.png", "Validation_44358455/Validation_44358455_44358455_43442.293.png" ], "turn_id": 5, "scan_id": "Validation_44358455", "origin_question": "", "option": [], "answer": "", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_44358455/Validation_44358455_44358455_43444.292.png", "Validation_44358455/Validation_44358455_44358455_43446.291.png", "Validation_44358455/Validation_44358455_44358455_43448.290.png", "Validation_44358455/Validation_44358455_44358455_43450.289.png", "Validation_44358455/Validation_44358455_44358455_43452.288.png" ], "turn_id": 6, "scan_id": "Validation_44358455", "origin_question": "In which turn were you farthest from \"the rack\"? ", "option": [], "answer": "1", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the rack\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_44358455/Validation_44358455_44358455_43454.287.png", "Validation_44358455/Validation_44358455_44358455_43456.287.png", "Validation_44358455/Validation_44358455_44358455_43458.286.png", "Validation_44358455/Validation_44358455_44358455_43460.285.png", "Validation_44358455/Validation_44358455_44358455_43462.284.png" ], "turn_id": 7, "scan_id": "Validation_44358455", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the rack\", Y: \"the mix of dark and light-colored clothes\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the rack\", Y: \"the mix of dark and light-colored clothes\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_44358455/Validation_44358455_44358455_43464.283.png", "Validation_44358455/Validation_44358455_44358455_43466.299.png", "Validation_44358455/Validation_44358455_44358455_43468.298.png", "Validation_44358455/Validation_44358455_44358455_43470.298.png", "Validation_44358455/Validation_44358455_44358455_43472.297.png" ], "turn_id": 8, "scan_id": "Validation_44358455", "origin_question": "What will be the first-time appearance order of the following categories: cabinet, case, washing machine?", "option": [ "washing machine, cabinet, case", "washing machine, case, cabinet", "case, washing machine, cabinet", "cabinet, washing machine, case" ], "answer": "cabinet, washing machine, case", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: cabinet, case, washing machine? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['washing machine, cabinet, case', 'washing machine, case, cabinet', 'case, washing machine, cabinet', 'cabinet, washing machine, case'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42898867/Validation_42898867_42898867_197709.545.png", "Validation_42898867/Validation_42898867_42898867_197711.527.png", "Validation_42898867/Validation_42898867_42898867_197713.527.png", "Validation_42898867/Validation_42898867_42898867_197715.526.png" ], "turn_id": 1, "scan_id": "Validation_42898867", "origin_question": "Remember, have you seen any pillow(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any pillow(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_42898867/Validation_42898867_42898867_197717.525.png", "Validation_42898867/Validation_42898867_42898867_197719.524.png", "Validation_42898867/Validation_42898867_42898867_197721.523.png", "Validation_42898867/Validation_42898867_42898867_197723.539.png" ], "turn_id": 2, "scan_id": "Validation_42898867", "origin_question": "When did you first discover \"the creamy white cylindrical light\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the creamy white cylindrical light\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_42898867/Validation_42898867_42898867_197725.538.png", "Validation_42898867/Validation_42898867_42898867_197727.538.png", "Validation_42898867/Validation_42898867_42898867_197729.537.png", "Validation_42898867/Validation_42898867_42898867_197731.536.png" ], "turn_id": 3, "scan_id": "Validation_42898867", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "78.9", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_42898867/Validation_42898867_42898867_197733.535.png", "Validation_42898867/Validation_42898867_42898867_197735.535.png", "Validation_42898867/Validation_42898867_42898867_197737.534.png", "Validation_42898867/Validation_42898867_42898867_197739.533.png" ], "turn_id": 4, "scan_id": "Validation_42898867", "origin_question": "Which direction is the sleek black rectangular monitor with the plastic design in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the sleek black rectangular monitor with the plastic design in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_42898867/Validation_42898867_42898867_197741.532.png", "Validation_42898867/Validation_42898867_42898867_197743.531.png", "Validation_42898867/Validation_42898867_42898867_197745.530.png", "Validation_42898867/Validation_42898867_42898867_197747.530.png" ], "turn_id": 5, "scan_id": "Validation_42898867", "origin_question": "Which direction is the creamy white cylindrical light in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the creamy white cylindrical light in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_42898867/Validation_42898867_42898867_197749.529.png", "Validation_42898867/Validation_42898867_42898867_197751.528.png", "Validation_42898867/Validation_42898867_42898867_197753.527.png", "Validation_42898867/Validation_42898867_42898867_197755.527.png" ], "turn_id": 6, "scan_id": "Validation_42898867", "origin_question": "Which direction is the creamy white cylindrical light in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the creamy white cylindrical light in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_42898867/Validation_42898867_42898867_197757.526.png", "Validation_42898867/Validation_42898867_42898867_197759.525.png", "Validation_42898867/Validation_42898867_42898867_197761.524.png", "Validation_42898867/Validation_42898867_42898867_197763.523.png" ], "turn_id": 7, "scan_id": "Validation_42898867", "origin_question": "", "option": [], "answer": "", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47331336/Validation_47331336_47331336_183439.834.png", "Validation_47331336/Validation_47331336_47331336_183441.334.png", "Validation_47331336/Validation_47331336_47331336_183442.833.png", "Validation_47331336/Validation_47331336_47331336_183444.332.png" ], "turn_id": 1, "scan_id": "Validation_47331336", "origin_question": "Remember, have you seen any towel(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any towel(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 3 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47331336/Validation_47331336_47331336_183445.832.png", "Validation_47331336/Validation_47331336_47331336_183447.331.png", "Validation_47331336/Validation_47331336_47331336_183448.831.png", "Validation_47331336/Validation_47331336_47331336_183450.330.png" ], "turn_id": 2, "scan_id": "Validation_47331336", "origin_question": "Remember, have you seen any player(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any player(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 3 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47331336/Validation_47331336_47331336_183451.829.png", "Validation_47331336/Validation_47331336_47331336_183453.329.png", "Validation_47331336/Validation_47331336_47331336_183454.828.png", "Validation_47331336/Validation_47331336_47331336_183456.327.png" ], "turn_id": 3, "scan_id": "Validation_47331336", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the medium-sized light grey rectangular towel\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the medium-sized light grey rectangular towel\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 3 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47333924/Validation_47333924_47333924_83659.928.png", "Validation_47333924/Validation_47333924_47333924_83661.911.png", "Validation_47333924/Validation_47333924_47333924_83663.910.png", "Validation_47333924/Validation_47333924_47333924_83665.909.png" ], "turn_id": 1, "scan_id": "Validation_47333924", "origin_question": "Remember, have you seen any copier(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any copier(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47333924/Validation_47333924_47333924_83667.909.png", "Validation_47333924/Validation_47333924_47333924_83669.908.png", "Validation_47333924/Validation_47333924_47333924_83671.907.png", "Validation_47333924/Validation_47333924_47333924_83673.906.png" ], "turn_id": 2, "scan_id": "Validation_47333924", "origin_question": "In which turn were you closest to \"the sleek cylindrical light with frosted glass\"? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the sleek cylindrical light with frosted glass\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_47333924/Validation_47333924_47333924_83675.905.png", "Validation_47333924/Validation_47333924_47333924_83677.904.png", "Validation_47333924/Validation_47333924_47333924_83679.904.png", "Validation_47333924/Validation_47333924_47333924_83681.903.png" ], "turn_id": 3, "scan_id": "Validation_47333924", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the contemporary rectangular white ceramic sink\" from your position?", "option": [], "answer": "141.92", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the contemporary rectangular white ceramic sink\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_47333924/Validation_47333924_47333924_83683.902.png", "Validation_47333924/Validation_47333924_47333924_83685.901.png", "Validation_47333924/Validation_47333924_47333924_83687.900.png", "Validation_47333924/Validation_47333924_47333924_83689.899.png" ], "turn_id": 4, "scan_id": "Validation_47333924", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the contemporary rectangular white ceramic sink\", Y: \"the sleek cylindrical light with frosted glass\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the contemporary rectangular white ceramic sink\", Y: \"the sleek cylindrical light with frosted glass\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_47333924/Validation_47333924_47333924_83691.899.png", "Validation_47333924/Validation_47333924_47333924_83693.915.png", "Validation_47333924/Validation_47333924_47333924_83695.914.png", "Validation_47333924/Validation_47333924_47333924_83697.913.png" ], "turn_id": 5, "scan_id": "Validation_47333924", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the sleek cylindrical light with frosted glass\" from your position?", "option": [], "answer": "97.36", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the sleek cylindrical light with frosted glass\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47333924/Validation_47333924_47333924_83699.912.png", "Validation_47333924/Validation_47333924_47333924_83701.911.png", "Validation_47333924/Validation_47333924_47333924_83703.910.png", "Validation_47333924/Validation_47333924_47333924_83705.910.png" ], "turn_id": 6, "scan_id": "Validation_47333924", "origin_question": "Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the rustic oval-shaped basket\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the rustic oval-shaped basket\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47333924/Validation_47333924_47333924_83707.909.png", "Validation_47333924/Validation_47333924_47333924_83709.908.png", "Validation_47333924/Validation_47333924_47333924_83711.907.png", "Validation_47333924/Validation_47333924_47333924_83713.906.png" ], "turn_id": 7, "scan_id": "Validation_47333924", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the compact rectangular white washing machine\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the compact rectangular white washing machine\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47333924/Validation_47333924_47333924_83715.906.png", "Validation_47333924/Validation_47333924_47333924_83717.905.png", "Validation_47333924/Validation_47333924_47333924_83719.904.png", "Validation_47333924/Validation_47333924_47333924_83721.903.png" ], "turn_id": 8, "scan_id": "Validation_47333924", "origin_question": "Which direction is the sleek cylindrical light with frosted glass in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the sleek cylindrical light with frosted glass in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47895365/Validation_47895365_47895365_100099.358.png", "Validation_47895365/Validation_47895365_47895365_100101.340.png", "Validation_47895365/Validation_47895365_47895365_100103.339.png", "Validation_47895365/Validation_47895365_47895365_100105.338.png" ], "turn_id": 1, "scan_id": "Validation_47895365", "origin_question": "Remember, have you seen any scissors(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any scissors(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_47895365/Validation_47895365_47895365_100107.338.png", "Validation_47895365/Validation_47895365_47895365_100109.337.png", "Validation_47895365/Validation_47895365_47895365_100111.336.png", "Validation_47895365/Validation_47895365_47895365_100113.335.png" ], "turn_id": 2, "scan_id": "Validation_47895365", "origin_question": "When did you first discover \"an elegant oval white ceramic sink\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"an elegant oval white ceramic sink\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47895365/Validation_47895365_47895365_100115.334.png", "Validation_47895365/Validation_47895365_47895365_100117.334.png", "Validation_47895365/Validation_47895365_47895365_100119.333.png", "Validation_47895365/Validation_47895365_47895365_100121.332.png" ], "turn_id": 3, "scan_id": "Validation_47895365", "origin_question": "In which turn were you closest to \"an elegant oval white ceramic sink\"? ", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"an elegant oval white ceramic sink\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_47895365/Validation_47895365_47895365_100123.331.png", "Validation_47895365/Validation_47895365_47895365_100125.330.png", "Validation_47895365/Validation_47895365_47895365_100127.330.png", "Validation_47895365/Validation_47895365_47895365_100129.345.png" ], "turn_id": 4, "scan_id": "Validation_47895365", "origin_question": "What will be the first-time appearance order of the following categories: cabinet, sink, toilet?", "option": [ "toilet, sink, cabinet", "sink, toilet, cabinet", "cabinet, toilet, sink", "cabinet, sink, toilet" ], "answer": "cabinet, sink, toilet", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: cabinet, sink, toilet? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['toilet, sink, cabinet', 'sink, toilet, cabinet', 'cabinet, toilet, sink', 'cabinet, sink, toilet'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47895365/Validation_47895365_47895365_100131.345.png", "Validation_47895365/Validation_47895365_47895365_100133.344.png", "Validation_47895365/Validation_47895365_47895365_100135.343.png", "Validation_47895365/Validation_47895365_47895365_100137.342.png" ], "turn_id": 5, "scan_id": "Validation_47895365", "origin_question": "", "option": [], "answer": "", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47895365/Validation_47895365_47895365_100139.341.png", "Validation_47895365/Validation_47895365_47895365_100141.341.png", "Validation_47895365/Validation_47895365_47895365_100143.340.png", "Validation_47895365/Validation_47895365_47895365_100145.339.png" ], "turn_id": 6, "scan_id": "Validation_47895365", "origin_question": "Compared to your position at the end of the 4 turn, are you now closer or farther away from \"an elegant oval white ceramic sink\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 4 turn, are you now closer or farther away from \"an elegant oval white ceramic sink\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47895365/Validation_47895365_47895365_100147.338.png", "Validation_47895365/Validation_47895365_47895365_100149.337.png", "Validation_47895365/Validation_47895365_47895365_100151.337.png", "Validation_47895365/Validation_47895365_47895365_100153.336.png" ], "turn_id": 7, "scan_id": "Validation_47895365", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the contemporary cubic washing machine in muted silver\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the contemporary cubic washing machine in muted silver\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_47895365/Validation_47895365_47895365_100155.335.png", "Validation_47895365/Validation_47895365_47895365_100157.334.png", "Validation_47895365/Validation_47895365_47895365_100159.333.png", "Validation_47895365/Validation_47895365_47895365_100161.333.png" ], "turn_id": 8, "scan_id": "Validation_47895365", "origin_question": "Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "118.69", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47333937/Validation_47333937_47333937_85490.399.png", "Validation_47333937/Validation_47333937_47333937_85492.365.png", "Validation_47333937/Validation_47333937_47333937_85494.364.png", "Validation_47333937/Validation_47333937_47333937_85496.363.png" ], "turn_id": 1, "scan_id": "Validation_47333937", "origin_question": "Remember, have you seen any flowerpot(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any flowerpot(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_47333937/Validation_47333937_47333937_85498.362.png", "Validation_47333937/Validation_47333937_47333937_85500.378.png", "Validation_47333937/Validation_47333937_47333937_85502.377.png", "Validation_47333937/Validation_47333937_47333937_85504.376.png" ], "turn_id": 2, "scan_id": "Validation_47333937", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the medium-sized rectangular white radiator with slats\", Y: \"the light\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the medium-sized rectangular white radiator with slats\", Y: \"the light\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "Validation_47333937/Validation_47333937_47333937_85506.375.png", "Validation_47333937/Validation_47333937_47333937_85508.375.png", "Validation_47333937/Validation_47333937_47333937_85510.374.png", "Validation_47333937/Validation_47333937_47333937_85512.373.png" ], "turn_id": 3, "scan_id": "Validation_47333937", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the medium-sized rectangular white radiator with slats\"; \"the light\"; \"the large black flat screen monitor with the thin frame\" ", "option": [ "the medium-sized rectangular white radiator with slats", "the light", "the large black flat screen monitor with the thin frame" ], "answer": "the large black flat screen monitor with the thin frame", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the medium-sized rectangular white radiator with slats\"; \"the light\"; \"the large black flat screen monitor with the thin frame\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the medium-sized rectangular white radiator with slats', 'the light', 'the large black flat screen monitor with the thin frame'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_47333937/Validation_47333937_47333937_85514.372.png", "Validation_47333937/Validation_47333937_47333937_85516.371.png", "Validation_47333937/Validation_47333937_47333937_85518.371.png", "Validation_47333937/Validation_47333937_47333937_85520.370.png" ], "turn_id": 4, "scan_id": "Validation_47333937", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the medium-sized rectangular white radiator with slats\" from your position?", "option": [], "answer": "136.67", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the medium-sized rectangular white radiator with slats\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_47333937/Validation_47333937_47333937_85522.369.png", "Validation_47333937/Validation_47333937_47333937_85524.368.png", "Validation_47333937/Validation_47333937_47333937_85526.367.png", "Validation_47333937/Validation_47333937_47333937_85528.367.png" ], "turn_id": 5, "scan_id": "Validation_47333937", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "53.12", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_47333937/Validation_47333937_47333937_85530.366.png", "Validation_47333937/Validation_47333937_47333937_85532.365.png", "Validation_47333937/Validation_47333937_47333937_85534.364.png", "Validation_47333937/Validation_47333937_47333937_85536.363.png" ], "turn_id": 6, "scan_id": "Validation_47333937", "origin_question": "What will be the first-time appearance order of the following categories: monitor, pillow, radiator?", "option": [ "pillow, monitor, radiator", "radiator, pillow, monitor", "pillow, radiator, monitor", "monitor, radiator, pillow" ], "answer": "radiator, pillow, monitor", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: monitor, pillow, radiator? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['pillow, monitor, radiator', 'radiator, pillow, monitor', 'pillow, radiator, monitor', 'monitor, radiator, pillow'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_47333937/Validation_47333937_47333937_85538.363.png", "Validation_47333937/Validation_47333937_47333937_85540.362.png", "Validation_47333937/Validation_47333937_47333937_85542.378.png", "Validation_47333937/Validation_47333937_47333937_85544.377.png" ], "turn_id": 7, "scan_id": "Validation_47333937", "origin_question": "How far is your current position from where you were at the end of the turn 5? (in meters)", "option": [], "answer": "1.32", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 5? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47895779/Validation_47895779_47895779_386972.680.png", "Validation_47895779/Validation_47895779_47895779_386974.162.png", "Validation_47895779/Validation_47895779_47895779_386975.662.png", "Validation_47895779/Validation_47895779_47895779_386977.161.png" ], "turn_id": 1, "scan_id": "Validation_47895779", "origin_question": "Remember, have you seen any roof(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any roof(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_47895779/Validation_47895779_47895779_386978.661.png", "Validation_47895779/Validation_47895779_47895779_386980.160.png", "Validation_47895779/Validation_47895779_47895779_386981.659.png", "Validation_47895779/Validation_47895779_47895779_386983.159.png" ], "turn_id": 2, "scan_id": "Validation_47895779", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the sleek square ceiling-mounted light fixture\" from your position?", "option": [], "answer": "159.5", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the sleek square ceiling-mounted light fixture\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_47895779/Validation_47895779_47895779_386984.658.png", "Validation_47895779/Validation_47895779_47895779_386986.158.png", "Validation_47895779/Validation_47895779_47895779_386987.657.png", "Validation_47895779/Validation_47895779_47895779_386989.156.png" ], "turn_id": 3, "scan_id": "Validation_47895779", "origin_question": "What will be the first-time appearance order of the following categories: cabinet, mirror, shelf?", "option": [ "mirror, shelf, cabinet", "shelf, cabinet, mirror", "cabinet, shelf, mirror", "cabinet, mirror, shelf" ], "answer": "cabinet, shelf, mirror", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: cabinet, mirror, shelf? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['mirror, shelf, cabinet', 'shelf, cabinet, mirror', 'cabinet, shelf, mirror', 'cabinet, mirror, shelf'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_47895779/Validation_47895779_47895779_386990.656.png", "Validation_47895779/Validation_47895779_47895779_386992.155.png", "Validation_47895779/Validation_47895779_47895779_386993.655.png", "Validation_47895779/Validation_47895779_47895779_386995.154.png" ], "turn_id": 4, "scan_id": "Validation_47895779", "origin_question": "In which turn did you see both \"an elegant oval white ceramic sink\" and \"the rectangular mirror with the metallic frame\" simultaneously? ", "option": [], "answer": "3", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"an elegant oval white ceramic sink\" and \"the rectangular mirror with the metallic frame\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_44358582/Validation_44358582_44358582_55768.746.png", "Validation_44358582/Validation_44358582_44358582_55770.729.png", "Validation_44358582/Validation_44358582_44358582_55772.728.png", "Validation_44358582/Validation_44358582_44358582_55774.727.png" ], "turn_id": 1, "scan_id": "Validation_44358582", "origin_question": "Remember, have you seen any camera(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any camera(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_44358582/Validation_44358582_44358582_55776.727.png", "Validation_44358582/Validation_44358582_44358582_55778.726.png", "Validation_44358582/Validation_44358582_44358582_55780.725.png", "Validation_44358582/Validation_44358582_44358582_55782.724.png" ], "turn_id": 2, "scan_id": "Validation_44358582", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the compact wooden shelf with the light natural finish\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the compact wooden shelf with the light natural finish\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_44358582/Validation_44358582_44358582_55784.723.png", "Validation_44358582/Validation_44358582_44358582_55786.723.png", "Validation_44358582/Validation_44358582_44358582_55788.722.png", "Validation_44358582/Validation_44358582_44358582_55790.721.png" ], "turn_id": 3, "scan_id": "Validation_44358582", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the compact wooden shelf with the light natural finish\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the compact wooden shelf with the light natural finish\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_44358582/Validation_44358582_44358582_55792.720.png", "Validation_44358582/Validation_44358582_44358582_55794.719.png", "Validation_44358582/Validation_44358582_44358582_55796.735.png", "Validation_44358582/Validation_44358582_44358582_55798.735.png" ], "turn_id": 4, "scan_id": "Validation_44358582", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "2.63", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_44358582/Validation_44358582_44358582_55800.734.png", "Validation_44358582/Validation_44358582_44358582_55802.733.png", "Validation_44358582/Validation_44358582_44358582_55804.732.png", "Validation_44358582/Validation_44358582_44358582_55806.731.png" ], "turn_id": 5, "scan_id": "Validation_44358582", "origin_question": "How far is your current position from where you were at the end of the turn 3? (in meters)", "option": [], "answer": "1.25", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 3? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_44358582/Validation_44358582_44358582_55808.731.png", "Validation_44358582/Validation_44358582_44358582_55810.730.png", "Validation_44358582/Validation_44358582_44358582_55812.729.png", "Validation_44358582/Validation_44358582_44358582_55814.728.png" ], "turn_id": 6, "scan_id": "Validation_44358582", "origin_question": "When did you first discover \"the soft, light-colored duvet on the wooden bed\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the soft, light-colored duvet on the wooden bed\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_44358582/Validation_44358582_44358582_55816.727.png", "Validation_44358582/Validation_44358582_44358582_55818.727.png", "Validation_44358582/Validation_44358582_44358582_55820.726.png", "Validation_44358582/Validation_44358582_44358582_55822.725.png" ], "turn_id": 7, "scan_id": "Validation_44358582", "origin_question": "", "option": [], "answer": "", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_44358582/Validation_44358582_44358582_55824.724.png", "Validation_44358582/Validation_44358582_44358582_55826.724.png", "Validation_44358582/Validation_44358582_44358582_55828.723.png", "Validation_44358582/Validation_44358582_44358582_55830.722.png" ], "turn_id": 8, "scan_id": "Validation_44358582", "origin_question": "Using your orientation at the end of turn 7 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "53.48", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 7 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47895732/Validation_47895732_47895732_384525.051.png", "Validation_47895732/Validation_47895732_47895732_384527.033.png", "Validation_47895732/Validation_47895732_47895732_384529.032.png", "Validation_47895732/Validation_47895732_47895732_384531.031.png" ], "turn_id": 1, "scan_id": "Validation_47895732", "origin_question": "", "option": [], "answer": "", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_47895732/Validation_47895732_47895732_384533.031.png", "Validation_47895732/Validation_47895732_47895732_384535.030.png", "Validation_47895732/Validation_47895732_47895732_384537.029.png", "Validation_47895732/Validation_47895732_47895732_384539.028.png" ], "turn_id": 2, "scan_id": "Validation_47895732", "origin_question": "When did you first discover \"the rustic wooden rectangular table with metal legs\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the rustic wooden rectangular table with metal legs\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47895732/Validation_47895732_47895732_384541.027.png", "Validation_47895732/Validation_47895732_47895732_384543.027.png", "Validation_47895732/Validation_47895732_47895732_384545.026.png", "Validation_47895732/Validation_47895732_47895732_384547.025.png" ], "turn_id": 3, "scan_id": "Validation_47895732", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the soft pastel fleece blanket, compact and cozy\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the soft pastel fleece blanket, compact and cozy\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_47895732/Validation_47895732_47895732_384549.024.png", "Validation_47895732/Validation_47895732_47895732_384551.023.png", "Validation_47895732/Validation_47895732_47895732_384553.023.png", "Validation_47895732/Validation_47895732_47895732_384555.022.png" ], "turn_id": 4, "scan_id": "Validation_47895732", "origin_question": "When did you first discover \"the large rectangular, plush cushion in deep green or black\"? (index of the turn)", "option": [], "answer": "4", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the large rectangular, plush cushion in deep green or black\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_47895732/Validation_47895732_47895732_384557.021.png", "Validation_47895732/Validation_47895732_47895732_384559.020.png", "Validation_47895732/Validation_47895732_47895732_384561.019.png", "Validation_47895732/Validation_47895732_47895732_384563.018.png" ], "turn_id": 5, "scan_id": "Validation_47895732", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "91.6", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_47895732/Validation_47895732_47895732_384565.034.png", "Validation_47895732/Validation_47895732_47895732_384567.017.png", "Validation_47895732/Validation_47895732_47895732_384569.032.png", "Validation_47895732/Validation_47895732_47895732_384571.032.png" ], "turn_id": 6, "scan_id": "Validation_47895732", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "158.81", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_47895732/Validation_47895732_47895732_384573.031.png", "Validation_47895732/Validation_47895732_47895732_384575.030.png", "Validation_47895732/Validation_47895732_47895732_384577.029.png", "Validation_47895732/Validation_47895732_47895732_384579.028.png" ], "turn_id": 7, "scan_id": "Validation_47895732", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the large rectangular, plush cushion in deep green or black\" from your position?", "option": [], "answer": "95.45", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the large rectangular, plush cushion in deep green or black\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_47895732/Validation_47895732_47895732_384581.028.png", "Validation_47895732/Validation_47895732_47895732_384583.027.png", "Validation_47895732/Validation_47895732_47895732_384585.026.png", "Validation_47895732/Validation_47895732_47895732_384587.025.png" ], "turn_id": 8, "scan_id": "Validation_47895732", "origin_question": "In which turn did you see both \"the rustic wooden rectangular table with metal legs\" and \"the soft pastel fleece blanket, compact and cozy\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the rustic wooden rectangular table with metal legs\" and \"the soft pastel fleece blanket, compact and cozy\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47430033/Validation_47430033_47430033_19408.406.png", "Validation_47430033/Validation_47430033_47430033_19409.872.png", "Validation_47430033/Validation_47430033_47430033_19411.371.png", "Validation_47430033/Validation_47430033_47430033_19412.887.png" ], "turn_id": 1, "scan_id": "Validation_47430033", "origin_question": "Remember, have you seen any monitor(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any monitor(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47430033/Validation_47430033_47430033_19414.387.png", "Validation_47430033/Validation_47430033_47430033_19415.886.png", "Validation_47430033/Validation_47430033_47430033_19417.386.png", "Validation_47430033/Validation_47430033_47430033_19418.885.png" ], "turn_id": 2, "scan_id": "Validation_47430033", "origin_question": "Which direction is the modern, tall white cabinet with the sleek design in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the modern, tall white cabinet with the sleek design in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_47430033/Validation_47430033_47430033_19420.384.png", "Validation_47430033/Validation_47430033_47430033_19421.884.png", "Validation_47430033/Validation_47430033_47430033_19423.383.png", "Validation_47430033/Validation_47430033_47430033_19424.882.png" ], "turn_id": 3, "scan_id": "Validation_47430033", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "80.76", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47430033/Validation_47430033_47430033_19426.382.png", "Validation_47430033/Validation_47430033_47430033_19427.881.png", "Validation_47430033/Validation_47430033_47430033_19429.381.png", "Validation_47430033/Validation_47430033_47430033_19430.880.png" ], "turn_id": 4, "scan_id": "Validation_47430033", "origin_question": "Which direction is the medium-sized light brown plush toy in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the medium-sized light brown plush toy in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_47430033/Validation_47430033_47430033_19432.379.png", "Validation_47430033/Validation_47430033_47430033_19433.879.png", "Validation_47430033/Validation_47430033_47430033_19435.378.png", "Validation_47430033/Validation_47430033_47430033_19436.878.png" ], "turn_id": 5, "scan_id": "Validation_47430033", "origin_question": "How far is your current position from where you were at the end of the turn 1? (in meters)", "option": [], "answer": "1.55", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 1? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_47430033/Validation_47430033_47430033_19438.377.png", "Validation_47430033/Validation_47430033_47430033_19439.876.png", "Validation_47430033/Validation_47430033_47430033_19441.376.png", "Validation_47430033/Validation_47430033_47430033_19442.875.png" ], "turn_id": 6, "scan_id": "Validation_47430033", "origin_question": "In which turn did you see both \"the warm light wood bed with sturdy construction\" and \"the clean, oblong, light-colored cotton pillow\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the warm light wood bed with sturdy construction\" and \"the clean, oblong, light-colored cotton pillow\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_45260900/Validation_45260900_45260900_804.492.png", "Validation_45260900/Validation_45260900_45260900_805.992.png", "Validation_45260900/Validation_45260900_45260900_807.491.png", "Validation_45260900/Validation_45260900_45260900_808.991.png" ], "turn_id": 1, "scan_id": "Validation_45260900", "origin_question": "Remember, have you seen any telephone(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any telephone(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_45260900/Validation_45260900_45260900_810.490.png", "Validation_45260900/Validation_45260900_45260900_811.990.png", "Validation_45260900/Validation_45260900_45260900_813.489.png", "Validation_45260900/Validation_45260900_45260900_814.988.png" ], "turn_id": 2, "scan_id": "Validation_45260900", "origin_question": "In which turn were you closest to \"the glossy white oval freestanding bathtub\"? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the glossy white oval freestanding bathtub\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_45260900/Validation_45260900_45260900_816.488.png", "Validation_45260900/Validation_45260900_45260900_817.987.png", "Validation_45260900/Validation_45260900_45260900_819.487.png", "Validation_45260900/Validation_45260900_45260900_820.986.png" ], "turn_id": 3, "scan_id": "Validation_45260900", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular white ceramic sink with chrome legs\", Y: \"the sleek, stainless steel bin\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "farther away from both X and Y", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular white ceramic sink with chrome legs\", Y: \"the sleek, stainless steel bin\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_45260900/Validation_45260900_45260900_822.485.png", "Validation_45260900/Validation_45260900_45260900_823.985.png", "Validation_45260900/Validation_45260900_45260900_825.484.png", "Validation_45260900/Validation_45260900_45260900_826.984.png" ], "turn_id": 4, "scan_id": "Validation_45260900", "origin_question": "Please recall \"the rectangular white ceramic sink with chrome legs\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.1", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the rectangular white ceramic sink with chrome legs\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_45260900/Validation_45260900_45260900_828.483.png", "Validation_45260900/Validation_45260900_45260900_829.982.png", "Validation_45260900/Validation_45260900_45260900_831.482.png", "Validation_45260900/Validation_45260900_45260900_832.981.png" ], "turn_id": 5, "scan_id": "Validation_45260900", "origin_question": "Remember, have you seen any sink(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any sink(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_45260900/Validation_45260900_45260900_834.481.png", "Validation_45260900/Validation_45260900_45260900_835.980.png", "Validation_45260900/Validation_45260900_45260900_837.480.png", "Validation_45260900/Validation_45260900_45260900_838.979.png" ], "turn_id": 6, "scan_id": "Validation_45260900", "origin_question": "Which direction is the modern white ceramic elongated toilet in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the modern white ceramic elongated toilet in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47333441/Validation_47333441_47333441_53851.740.png", "Validation_47333441/Validation_47333441_47333441_53853.722.png", "Validation_47333441/Validation_47333441_47333441_53855.721.png", "Validation_47333441/Validation_47333441_47333441_53857.721.png" ], "turn_id": 1, "scan_id": "Validation_47333441", "origin_question": "Remember, have you seen any glove(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any glove(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_47333441/Validation_47333441_47333441_53859.720.png", "Validation_47333441/Validation_47333441_47333441_53861.719.png", "Validation_47333441/Validation_47333441_47333441_53863.718.png", "Validation_47333441/Validation_47333441_47333441_53865.717.png" ], "turn_id": 2, "scan_id": "Validation_47333441", "origin_question": "In which turn did you see both \"the pink rounded plastic basin for toddlers\" and \"the rectangular plastic basket with the white patterned design\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the pink rounded plastic basin for toddlers\" and \"the rectangular plastic basket with the white patterned design\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47333441/Validation_47333441_47333441_53867.716.png", "Validation_47333441/Validation_47333441_47333441_53869.716.png", "Validation_47333441/Validation_47333441_47333441_53871.731.png", "Validation_47333441/Validation_47333441_47333441_53873.731.png" ], "turn_id": 3, "scan_id": "Validation_47333441", "origin_question": "In which turn were you closest to \"the classic light with the round opaque bulb\"? ", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the classic light with the round opaque bulb\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47333441/Validation_47333441_47333441_53875.730.png", "Validation_47333441/Validation_47333441_47333441_53877.729.png", "Validation_47333441/Validation_47333441_47333441_53879.728.png", "Validation_47333441/Validation_47333441_47333441_53881.727.png" ], "turn_id": 4, "scan_id": "Validation_47333441", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the classic light with the round opaque bulb\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the classic light with the round opaque bulb\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47333441/Validation_47333441_47333441_53883.726.png", "Validation_47333441/Validation_47333441_47333441_53885.726.png", "Validation_47333441/Validation_47333441_47333441_53887.725.png", "Validation_47333441/Validation_47333441_47333441_53889.724.png" ], "turn_id": 5, "scan_id": "Validation_47333441", "origin_question": "Which direction is the sleek rectangular white ceramic sink in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the sleek rectangular white ceramic sink in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47333441/Validation_47333441_47333441_53891.723.png", "Validation_47333441/Validation_47333441_47333441_53893.722.png", "Validation_47333441/Validation_47333441_47333441_53895.722.png", "Validation_47333441/Validation_47333441_47333441_53897.721.png" ], "turn_id": 6, "scan_id": "Validation_47333441", "origin_question": "Which direction is the pink rounded plastic basin for toddlers in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the pink rounded plastic basin for toddlers in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "Validation_47333441/Validation_47333441_47333441_53899.720.png", "Validation_47333441/Validation_47333441_47333441_53901.719.png", "Validation_47333441/Validation_47333441_47333441_53903.718.png", "Validation_47333441/Validation_47333441_47333441_53905.717.png" ], "turn_id": 7, "scan_id": "Validation_47333441", "origin_question": "Among these three objects(\"the sleek rectangular white ceramic sink\",\"the large rectangular cream-colored flowerpot\",\"the rectangular plastic basket with the white patterned design\" ), which one is the farthest from you?", "option": [ "the sleek rectangular white ceramic sink", "the large rectangular cream-colored flowerpot", "the rectangular plastic basket with the white patterned design" ], "answer": "the large rectangular cream-colored flowerpot", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects(\"the sleek rectangular white ceramic sink\",\"the large rectangular cream-colored flowerpot\",\"the rectangular plastic basket with the white patterned design\" ), which one is the farthest from you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the sleek rectangular white ceramic sink', 'the large rectangular cream-colored flowerpot', 'the rectangular plastic basket with the white patterned design'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_41069042/Validation_41069042_41069042_3044.239.png", "Validation_41069042/Validation_41069042_41069042_3046.221.png", "Validation_41069042/Validation_41069042_41069042_3048.237.png", "Validation_41069042/Validation_41069042_41069042_3050.236.png", "Validation_41069042/Validation_41069042_41069042_3052.235.png" ], "turn_id": 1, "scan_id": "Validation_41069042", "origin_question": "Remember, have you seen any copier(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any copier(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_41069042/Validation_41069042_41069042_3054.234.png", "Validation_41069042/Validation_41069042_41069042_3056.234.png", "Validation_41069042/Validation_41069042_41069042_3058.233.png", "Validation_41069042/Validation_41069042_41069042_3060.232.png", "Validation_41069042/Validation_41069042_41069042_3062.231.png" ], "turn_id": 2, "scan_id": "Validation_41069042", "origin_question": "When did you first discover \"the compact gray table with the laminate finish\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the compact gray table with the laminate finish\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_41069042/Validation_41069042_41069042_3064.230.png", "Validation_41069042/Validation_41069042_41069042_3066.229.png", "Validation_41069042/Validation_41069042_41069042_3068.229.png", "Validation_41069042/Validation_41069042_41069042_3070.228.png", "Validation_41069042/Validation_41069042_41069042_3072.227.png" ], "turn_id": 3, "scan_id": "Validation_41069042", "origin_question": "Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "82.59", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_41069042/Validation_41069042_41069042_3074.226.png", "Validation_41069042/Validation_41069042_41069042_3076.225.png", "Validation_41069042/Validation_41069042_41069042_3078.224.png", "Validation_41069042/Validation_41069042_41069042_3080.224.png", "Validation_41069042/Validation_41069042_41069042_3082.223.png" ], "turn_id": 4, "scan_id": "Validation_41069042", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "33.6", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_41069042/Validation_41069042_41069042_3084.222.png", "Validation_41069042/Validation_41069042_41069042_3086.221.png", "Validation_41069042/Validation_41069042_41069042_3088.220.png", "Validation_41069042/Validation_41069042_41069042_3090.236.png", "Validation_41069042/Validation_41069042_41069042_3092.235.png" ], "turn_id": 5, "scan_id": "Validation_41069042", "origin_question": "Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "110.35", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_41069042/Validation_41069042_41069042_3094.235.png", "Validation_41069042/Validation_41069042_41069042_3096.234.png", "Validation_41069042/Validation_41069042_41069042_3098.233.png", "Validation_41069042/Validation_41069042_41069042_3100.232.png", "Validation_41069042/Validation_41069042_41069042_3102.231.png" ], "turn_id": 6, "scan_id": "Validation_41069042", "origin_question": "In which turn did you see both \"the soft rectangular bed in the light color\" and \"the window\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the soft rectangular bed in the light color\" and \"the window\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_41069042/Validation_41069042_41069042_3104.230.png", "Validation_41069042/Validation_41069042_41069042_3106.229.png", "Validation_41069042/Validation_41069042_41069042_3108.229.png", "Validation_41069042/Validation_41069042_41069042_3110.228.png", "Validation_41069042/Validation_41069042_41069042_3112.227.png" ], "turn_id": 7, "scan_id": "Validation_41069042", "origin_question": "Using your orientation at the end of turn 6 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "64.16", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 6 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_41069042/Validation_41069042_41069042_3114.226.png", "Validation_41069042/Validation_41069042_41069042_3116.225.png", "Validation_41069042/Validation_41069042_41069042_3118.225.png", "Validation_41069042/Validation_41069042_41069042_3120.224.png", "Validation_41069042/Validation_41069042_41069042_3122.223.png" ], "turn_id": 8, "scan_id": "Validation_41069042", "origin_question": "How far is your current position from where you were at the end of the turn 1? (in meters)", "option": [], "answer": "2.13", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 1? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_41069042/Validation_41069042_41069042_3124.222.png", "Validation_41069042/Validation_41069042_41069042_3126.221.png", "Validation_41069042/Validation_41069042_41069042_3128.220.png", "Validation_41069042/Validation_41069042_41069042_3130.220.png", "Validation_41069042/Validation_41069042_41069042_3132.235.png" ], "turn_id": 9, "scan_id": "Validation_41069042", "origin_question": "By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular picture frame with the blue border\", Y: \"the sleek white compact sink with defined edges\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to Y and farther away from X", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular picture frame with the blue border\", Y: \"the sleek white compact sink with defined edges\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_41125718/Validation_41125718_41125718_4716.347.png", "Validation_41125718/Validation_41125718_41125718_4718.329.png", "Validation_41125718/Validation_41125718_41125718_4720.328.png", "Validation_41125718/Validation_41125718_41125718_4722.328.png", "Validation_41125718/Validation_41125718_41125718_4724.327.png" ], "turn_id": 1, "scan_id": "Validation_41125718", "origin_question": "Remember, have you seen any picture(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any picture(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_41125718/Validation_41125718_41125718_4726.326.png", "Validation_41125718/Validation_41125718_41125718_4728.325.png", "Validation_41125718/Validation_41125718_41125718_4730.324.png", "Validation_41125718/Validation_41125718_41125718_4732.324.png", "Validation_41125718/Validation_41125718_41125718_4734.323.png" ], "turn_id": 2, "scan_id": "Validation_41125718", "origin_question": "When did you first discover \"the modern bed with soft pastel pink bedding\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the modern bed with soft pastel pink bedding\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_41125718/Validation_41125718_41125718_4736.322.png", "Validation_41125718/Validation_41125718_41125718_4738.321.png", "Validation_41125718/Validation_41125718_41125718_4740.320.png", "Validation_41125718/Validation_41125718_41125718_4742.319.png", "Validation_41125718/Validation_41125718_41125718_4744.319.png" ], "turn_id": 3, "scan_id": "Validation_41125718", "origin_question": "Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "37.41", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_41125718/Validation_41125718_41125718_4746.318.png", "Validation_41125718/Validation_41125718_41125718_4748.317.png", "Validation_41125718/Validation_41125718_41125718_4750.316.png", "Validation_41125718/Validation_41125718_41125718_4752.332.png", "Validation_41125718/Validation_41125718_41125718_4754.331.png" ], "turn_id": 4, "scan_id": "Validation_41125718", "origin_question": "Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance forward or backward from that position?", "option": [ "Forward", "Backward" ], "answer": "Backward", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance forward or backward from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Forward', 'Backward'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_41125718/Validation_41125718_41125718_4756.330.png", "Validation_41125718/Validation_41125718_41125718_4758.330.png", "Validation_41125718/Validation_41125718_41125718_4760.329.png", "Validation_41125718/Validation_41125718_41125718_4762.328.png", "Validation_41125718/Validation_41125718_41125718_4764.327.png" ], "turn_id": 5, "scan_id": "Validation_41125718", "origin_question": "Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Left", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_41125718/Validation_41125718_41125718_4766.326.png", "Validation_41125718/Validation_41125718_41125718_4768.325.png", "Validation_41125718/Validation_41125718_41125718_4770.325.png", "Validation_41125718/Validation_41125718_41125718_4772.324.png", "Validation_41125718/Validation_41125718_41125718_4774.323.png" ], "turn_id": 6, "scan_id": "Validation_41125718", "origin_question": "When did you first discover \"the large rectangular window with the light-colored frame\"? (index of the turn)", "option": [], "answer": "6", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the large rectangular window with the light-colored frame\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_41125718/Validation_41125718_41125718_4776.322.png", "Validation_41125718/Validation_41125718_41125718_4778.321.png", "Validation_41125718/Validation_41125718_41125718_4780.321.png", "Validation_41125718/Validation_41125718_41125718_4782.320.png", "Validation_41125718/Validation_41125718_41125718_4784.319.png" ], "turn_id": 7, "scan_id": "Validation_41125718", "origin_question": "What will be the first-time appearance order of the following categories: bed, picture, towel?", "option": [ "picture, bed, towel", "towel, bed, picture", "bed, picture, towel", "picture, towel, bed" ], "answer": "picture, bed, towel", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bed, picture, towel? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['picture, bed, towel', 'towel, bed, picture', 'bed, picture, towel', 'picture, towel, bed'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_41125718/Validation_41125718_41125718_4786.318.png", "Validation_41125718/Validation_41125718_41125718_4788.317.png", "Validation_41125718/Validation_41125718_41125718_4790.316.png", "Validation_41125718/Validation_41125718_41125718_4792.316.png", "Validation_41125718/Validation_41125718_41125718_4794.315.png" ], "turn_id": 8, "scan_id": "Validation_41125718", "origin_question": "Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "146.26", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_41125718/Validation_41125718_41125718_4796.331.png", "Validation_41125718/Validation_41125718_41125718_4798.330.png", "Validation_41125718/Validation_41125718_41125718_4800.329.png", "Validation_41125718/Validation_41125718_41125718_4802.328.png", "Validation_41125718/Validation_41125718_41125718_4804.328.png" ], "turn_id": 9, "scan_id": "Validation_41125718", "origin_question": "Please recall \"the large rectangular window with the light-colored frame\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.51", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the large rectangular window with the light-colored frame\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_41125718/Validation_41125718_41125718_4806.327.png", "Validation_41125718/Validation_41125718_41125718_4808.326.png", "Validation_41125718/Validation_41125718_41125718_4810.325.png", "Validation_41125718/Validation_41125718_41125718_4812.324.png", "Validation_41125718/Validation_41125718_41125718_4814.323.png" ], "turn_id": 10, "scan_id": "Validation_41125718", "origin_question": "Which direction is an abstract picture with blue and off-white colors in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is an abstract picture with blue and off-white colors in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "Validation_47204552/Validation_47204552_47204552_3353.094.png", "Validation_47204552/Validation_47204552_47204552_3355.093.png", "Validation_47204552/Validation_47204552_47204552_3357.092.png", "Validation_47204552/Validation_47204552_47204552_3359.091.png", "Validation_47204552/Validation_47204552_47204552_3361.091.png" ], "turn_id": 1, "scan_id": "Validation_47204552", "origin_question": "Remember, how many light(s) have you seen so far? ", "option": [], "answer": "3", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many light(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "Validation_47204552/Validation_47204552_47204552_3363.090.png", "Validation_47204552/Validation_47204552_47204552_3365.089.png", "Validation_47204552/Validation_47204552_47204552_3367.088.png", "Validation_47204552/Validation_47204552_47204552_3369.087.png", "Validation_47204552/Validation_47204552_47204552_3371.103.png" ], "turn_id": 2, "scan_id": "Validation_47204552", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the soft beige leather couch with cushions and pillows\"; \"the large, sleek black glossy monitor\"; \"the tall, polished wooden door with minimal design\" ", "option": [ "the soft beige leather couch with cushions and pillows", "the large, sleek black glossy monitor", "the tall, polished wooden door with minimal design" ], "answer": "the large, sleek black glossy monitor", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the soft beige leather couch with cushions and pillows\"; \"the large, sleek black glossy monitor\"; \"the tall, polished wooden door with minimal design\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the soft beige leather couch with cushions and pillows', 'the large, sleek black glossy monitor', 'the tall, polished wooden door with minimal design'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_47204552/Validation_47204552_47204552_3373.102.png", "Validation_47204552/Validation_47204552_47204552_3375.101.png", "Validation_47204552/Validation_47204552_47204552_3377.101.png", "Validation_47204552/Validation_47204552_47204552_3379.100.png", "Validation_47204552/Validation_47204552_47204552_3381.099.png" ], "turn_id": 3, "scan_id": "Validation_47204552", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "1.7", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_47204552/Validation_47204552_47204552_3383.098.png", "Validation_47204552/Validation_47204552_47204552_3385.097.png", "Validation_47204552/Validation_47204552_47204552_3387.097.png", "Validation_47204552/Validation_47204552_47204552_3389.096.png", "Validation_47204552/Validation_47204552_47204552_3391.095.png" ], "turn_id": 4, "scan_id": "Validation_47204552", "origin_question": "Is the large, sleek black glossy monitor to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the large, sleek black glossy monitor to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "Validation_47204552/Validation_47204552_47204552_3393.094.png", "Validation_47204552/Validation_47204552_47204552_3395.093.png", "Validation_47204552/Validation_47204552_47204552_3397.092.png", "Validation_47204552/Validation_47204552_47204552_3399.091.png", "Validation_47204552/Validation_47204552_47204552_3401.091.png" ], "turn_id": 5, "scan_id": "Validation_47204552", "origin_question": "Among these three objects(\"the soft beige leather couch with cushions and pillows\",\"the tall, polished wooden door with minimal design\",\"the large, sleek black glossy monitor\" ), which one is the closest to you?", "option": [ "the soft beige leather couch with cushions and pillows", "the tall, polished wooden door with minimal design", "the large, sleek black glossy monitor" ], "answer": "the soft beige leather couch with cushions and pillows", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the soft beige leather couch with cushions and pillows\",\"the tall, polished wooden door with minimal design\",\"the large, sleek black glossy monitor\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the soft beige leather couch with cushions and pillows', 'the tall, polished wooden door with minimal design', 'the large, sleek black glossy monitor'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47204552/Validation_47204552_47204552_3403.090.png", "Validation_47204552/Validation_47204552_47204552_3405.089.png", "Validation_47204552/Validation_47204552_47204552_3407.088.png", "Validation_47204552/Validation_47204552_47204552_3409.087.png", "Validation_47204552/Validation_47204552_47204552_3411.086.png" ], "turn_id": 6, "scan_id": "Validation_47204552", "origin_question": "Which direction is the tall, polished wooden door with minimal design in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the tall, polished wooden door with minimal design in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47204552/Validation_47204552_47204552_3413.086.png", "Validation_47204552/Validation_47204552_47204552_3415.102.png", "Validation_47204552/Validation_47204552_47204552_3417.101.png", "Validation_47204552/Validation_47204552_47204552_3419.100.png", "Validation_47204552/Validation_47204552_47204552_3421.099.png" ], "turn_id": 7, "scan_id": "Validation_47204552", "origin_question": "", "option": [], "answer": "", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_47204552/Validation_47204552_47204552_3423.098.png", "Validation_47204552/Validation_47204552_47204552_3425.097.png", "Validation_47204552/Validation_47204552_47204552_3427.097.png", "Validation_47204552/Validation_47204552_47204552_3429.096.png", "Validation_47204552/Validation_47204552_47204552_3431.095.png" ], "turn_id": 8, "scan_id": "Validation_47204552", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_47204552/Validation_47204552_47204552_3433.094.png", "Validation_47204552/Validation_47204552_47204552_3435.093.png", "Validation_47204552/Validation_47204552_47204552_3437.092.png", "Validation_47204552/Validation_47204552_47204552_3439.091.png", "Validation_47204552/Validation_47204552_47204552_3441.091.png" ], "turn_id": 9, "scan_id": "Validation_47204552", "origin_question": "How far is your current position from where you were at the end of the turn 3? (in meters)", "option": [], "answer": "1.15", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 3? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "Validation_47204552/Validation_47204552_47204552_3443.090.png", "Validation_47204552/Validation_47204552_47204552_3445.089.png", "Validation_47204552/Validation_47204552_47204552_3447.088.png", "Validation_47204552/Validation_47204552_47204552_3449.087.png", "Validation_47204552/Validation_47204552_47204552_3451.087.png" ], "turn_id": 10, "scan_id": "Validation_47204552", "origin_question": "Among these three objects(\"the soft beige leather couch with cushions and pillows\",\"the tall, polished wooden door with minimal design\",\"the modern black stainless steel oven\" ), which one is the farthest from you?", "option": [ "the soft beige leather couch with cushions and pillows", "the tall, polished wooden door with minimal design", "the modern black stainless steel oven" ], "answer": "the modern black stainless steel oven", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the soft beige leather couch with cushions and pillows\",\"the tall, polished wooden door with minimal design\",\"the modern black stainless steel oven\" ), which one is the farthest from you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the soft beige leather couch with cushions and pillows', 'the tall, polished wooden door with minimal design', 'the modern black stainless steel oven'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42446527/Validation_42446527_42446527_198626.868.png", "Validation_42446527/Validation_42446527_42446527_198628.850.png", "Validation_42446527/Validation_42446527_42446527_198630.850.png", "Validation_42446527/Validation_42446527_42446527_198632.849.png", "Validation_42446527/Validation_42446527_42446527_198634.848.png" ], "turn_id": 1, "scan_id": "Validation_42446527", "origin_question": "Remember, have you seen any rope(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any rope(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_42446527/Validation_42446527_42446527_198636.847.png", "Validation_42446527/Validation_42446527_42446527_198638.846.png", "Validation_42446527/Validation_42446527_42446527_198640.845.png", "Validation_42446527/Validation_42446527_42446527_198642.845.png", "Validation_42446527/Validation_42446527_42446527_198644.844.png" ], "turn_id": 2, "scan_id": "Validation_42446527", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the tall, narrow wooden shelf with open compartments\", Y: \"the sleek light-colored door\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to X and farther away from Y", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the tall, narrow wooden shelf with open compartments\", Y: \"the sleek light-colored door\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_42446527/Validation_42446527_42446527_198646.843.png", "Validation_42446527/Validation_42446527_42446527_198648.842.png", "Validation_42446527/Validation_42446527_42446527_198650.841.png", "Validation_42446527/Validation_42446527_42446527_198652.841.png", "Validation_42446527/Validation_42446527_42446527_198654.840.png" ], "turn_id": 3, "scan_id": "Validation_42446527", "origin_question": "In which turn were you farthest from \"the rectangular contemporary bed in neutral colors\"? ", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the rectangular contemporary bed in neutral colors\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_42446527/Validation_42446527_42446527_198656.839.png", "Validation_42446527/Validation_42446527_42446527_198658.838.png", "Validation_42446527/Validation_42446527_42446527_198660.837.png", "Validation_42446527/Validation_42446527_42446527_198662.837.png", "Validation_42446527/Validation_42446527_42446527_198664.852.png" ], "turn_id": 4, "scan_id": "Validation_42446527", "origin_question": "In which turn were you farthest from \"the tall, narrow wooden shelf with open compartments\"? ", "option": [], "answer": "1", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the tall, narrow wooden shelf with open compartments\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_42446527/Validation_42446527_42446527_198666.851.png", "Validation_42446527/Validation_42446527_42446527_198668.851.png", "Validation_42446527/Validation_42446527_42446527_198670.850.png", "Validation_42446527/Validation_42446527_42446527_198672.849.png", "Validation_42446527/Validation_42446527_42446527_198674.848.png" ], "turn_id": 5, "scan_id": "Validation_42446527", "origin_question": "In which turn were you closest to \"the tall, circular light gray chair\"? ", "option": [], "answer": "5", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the tall, circular light gray chair\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "Validation_42446527/Validation_42446527_42446527_198676.847.png", "Validation_42446527/Validation_42446527_42446527_198678.847.png", "Validation_42446527/Validation_42446527_42446527_198680.846.png", "Validation_42446527/Validation_42446527_42446527_198682.845.png", "Validation_42446527/Validation_42446527_42446527_198684.844.png" ], "turn_id": 6, "scan_id": "Validation_42446527", "origin_question": "Among these three objects(\"the tall, narrow wooden shelf with open compartments\",\"the rectangular smooth pillow in the light color\",\"the tall, circular light gray chair\" ), which one is the farthest from you?", "option": [ "the tall, narrow wooden shelf with open compartments", "the rectangular smooth pillow in the light color", "the tall, circular light gray chair" ], "answer": "the tall, circular light gray chair", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the tall, narrow wooden shelf with open compartments\",\"the rectangular smooth pillow in the light color\",\"the tall, circular light gray chair\" ), which one is the farthest from you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the tall, narrow wooden shelf with open compartments', 'the rectangular smooth pillow in the light color', 'the tall, circular light gray chair'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "Validation_42446527/Validation_42446527_42446527_198686.843.png", "Validation_42446527/Validation_42446527_42446527_198688.842.png", "Validation_42446527/Validation_42446527_42446527_198690.842.png", "Validation_42446527/Validation_42446527_42446527_198692.841.png", "Validation_42446527/Validation_42446527_42446527_198694.840.png" ], "turn_id": 7, "scan_id": "Validation_42446527", "origin_question": "Among these three objects(\"the rectangular smooth pillow in the light color\",\"the sleek light-colored door\",\"the rectangular window with the modern white frame\" ), which one is the farthest from you?", "option": [ "the rectangular smooth pillow in the light color", "the sleek light-colored door", "the rectangular window with the modern white frame" ], "answer": "the sleek light-colored door", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the rectangular smooth pillow in the light color\",\"the sleek light-colored door\",\"the rectangular window with the modern white frame\" ), which one is the farthest from you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular smooth pillow in the light color', 'the sleek light-colored door', 'the rectangular window with the modern white frame'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "Validation_42446527/Validation_42446527_42446527_198696.839.png", "Validation_42446527/Validation_42446527_42446527_198698.838.png", "Validation_42446527/Validation_42446527_42446527_198700.837.png", "Validation_42446527/Validation_42446527_42446527_198702.837.png", "Validation_42446527/Validation_42446527_42446527_198704.853.png" ], "turn_id": 8, "scan_id": "Validation_42446527", "origin_question": "Among these three objects(\"the table\",\"the sleek light-colored door\",\"the tall, circular light gray chair\" ), which one is the farthest from you?", "option": [ "the table", "the sleek light-colored door", "the tall, circular light gray chair" ], "answer": "the table", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the table\",\"the sleek light-colored door\",\"the tall, circular light gray chair\" ), which one is the farthest from you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the table', 'the sleek light-colored door', 'the tall, circular light gray chair'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47429913/Validation_47429913_47429913_135970.108.png", "Validation_47429913/Validation_47429913_47429913_135971.591.png", "Validation_47429913/Validation_47429913_47429913_135973.090.png", "Validation_47429913/Validation_47429913_47429913_135974.590.png" ], "turn_id": 1, "scan_id": "Validation_47429913", "origin_question": "Remember, have you seen any ladder(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any ladder(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_47429913/Validation_47429913_47429913_135976.089.png", "Validation_47429913/Validation_47429913_47429913_135977.588.png", "Validation_47429913/Validation_47429913_47429913_135979.088.png", "Validation_47429913/Validation_47429913_47429913_135980.587.png" ], "turn_id": 2, "scan_id": "Validation_47429913", "origin_question": "When did you first discover \"the carpet\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the carpet\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "Validation_47429913/Validation_47429913_47429913_135982.087.png", "Validation_47429913/Validation_47429913_47429913_135983.586.png", "Validation_47429913/Validation_47429913_47429913_135985.085.png", "Validation_47429913/Validation_47429913_47429913_135986.585.png" ], "turn_id": 3, "scan_id": "Validation_47429913", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the high-gloss white oval-shaped ceramic toilet\"; \"the flowerpot\"; \"the carpet\" ", "option": [ "the high-gloss white oval-shaped ceramic toilet", "the flowerpot", "the carpet" ], "answer": "the flowerpot", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the high-gloss white oval-shaped ceramic toilet\"; \"the flowerpot\"; \"the carpet\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the high-gloss white oval-shaped ceramic toilet', 'the flowerpot', 'the carpet'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_47429913/Validation_47429913_47429913_135988.084.png", "Validation_47429913/Validation_47429913_47429913_135989.584.png", "Validation_47429913/Validation_47429913_47429913_135991.083.png", "Validation_47429913/Validation_47429913_47429913_135992.582.png" ], "turn_id": 4, "scan_id": "Validation_47429913", "origin_question": "When did you first discover \"the flowerpot\"? (index of the turn)", "option": [], "answer": "3", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the flowerpot\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_42899471/Validation_42899471_42899471_195658.555.png", "Validation_42899471/Validation_42899471_42899471_195660.537.png", "Validation_42899471/Validation_42899471_42899471_195662.536.png", "Validation_42899471/Validation_42899471_42899471_195664.535.png", "Validation_42899471/Validation_42899471_42899471_195666.535.png" ], "turn_id": 1, "scan_id": "Validation_42899471", "origin_question": "", "option": [], "answer": "", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_42899471/Validation_42899471_42899471_195668.534.png", "Validation_42899471/Validation_42899471_42899471_195670.533.png", "Validation_42899471/Validation_42899471_42899471_195672.532.png", "Validation_42899471/Validation_42899471_42899471_195674.531.png", "Validation_42899471/Validation_42899471_42899471_195676.530.png" ], "turn_id": 2, "scan_id": "Validation_42899471", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_42899471/Validation_42899471_42899471_195678.530.png", "Validation_42899471/Validation_42899471_42899471_195680.529.png", "Validation_42899471/Validation_42899471_42899471_195682.528.png", "Validation_42899471/Validation_42899471_42899471_195684.527.png", "Validation_42899471/Validation_42899471_42899471_195686.526.png" ], "turn_id": 3, "scan_id": "Validation_42899471", "origin_question": "Please recall \"the sleek, modern cabinet with the neutral color palette\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.67", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the sleek, modern cabinet with the neutral color palette\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_42899471/Validation_42899471_42899471_195688.525.png", "Validation_42899471/Validation_42899471_42899471_195690.525.png", "Validation_42899471/Validation_42899471_42899471_195692.524.png", "Validation_42899471/Validation_42899471_42899471_195694.523.png", "Validation_42899471/Validation_42899471_42899471_195696.522.png" ], "turn_id": 4, "scan_id": "Validation_42899471", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the medium-sized, glossy black rectangular monitor\" from your position?", "option": [], "answer": "69.34", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the medium-sized, glossy black rectangular monitor\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_42899471/Validation_42899471_42899471_195698.521.png", "Validation_42899471/Validation_42899471_42899471_195700.521.png", "Validation_42899471/Validation_42899471_42899471_195702.536.png", "Validation_42899471/Validation_42899471_42899471_195704.535.png", "Validation_42899471/Validation_42899471_42899471_195706.535.png" ], "turn_id": 5, "scan_id": "Validation_42899471", "origin_question": "Is the sleek, compact rectangular fireplace with the dark interior to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the sleek, compact rectangular fireplace with the dark interior to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_42899471/Validation_42899471_42899471_195708.534.png", "Validation_42899471/Validation_42899471_42899471_195710.533.png", "Validation_42899471/Validation_42899471_42899471_195712.532.png", "Validation_42899471/Validation_42899471_42899471_195714.531.png", "Validation_42899471/Validation_42899471_42899471_195716.531.png" ], "turn_id": 6, "scan_id": "Validation_42899471", "origin_question": "Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_42899471/Validation_42899471_42899471_195718.530.png", "Validation_42899471/Validation_42899471_42899471_195720.529.png", "Validation_42899471/Validation_42899471_42899471_195722.528.png", "Validation_42899471/Validation_42899471_42899471_195724.527.png", "Validation_42899471/Validation_42899471_42899471_195726.526.png" ], "turn_id": 7, "scan_id": "Validation_42899471", "origin_question": "In which turn did you see both \"the medium-sized, glossy black rectangular monitor\" and \"the rectangular white ceramic sink\" simultaneously? ", "option": [], "answer": "6", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the medium-sized, glossy black rectangular monitor\" and \"the rectangular white ceramic sink\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_42899471/Validation_42899471_42899471_195728.526.png", "Validation_42899471/Validation_42899471_42899471_195730.525.png", "Validation_42899471/Validation_42899471_42899471_195732.524.png", "Validation_42899471/Validation_42899471_42899471_195734.523.png", "Validation_42899471/Validation_42899471_42899471_195736.522.png" ], "turn_id": 8, "scan_id": "Validation_42899471", "origin_question": "Assuming the direction you are facing at the end of the turn 4 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Right", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 4 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_42899471/Validation_42899471_42899471_195738.522.png", "Validation_42899471/Validation_42899471_42899471_195740.521.png", "Validation_42899471/Validation_42899471_42899471_195742.536.png", "Validation_42899471/Validation_42899471_42899471_195744.536.png", "Validation_42899471/Validation_42899471_42899471_195746.535.png" ], "turn_id": 9, "scan_id": "Validation_42899471", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "147.89", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47429925/Validation_47429925_47429925_136543.972.png", "Validation_47429925/Validation_47429925_47429925_136545.455.png", "Validation_47429925/Validation_47429925_47429925_136546.954.png", "Validation_47429925/Validation_47429925_47429925_136548.454.png" ], "turn_id": 1, "scan_id": "Validation_47429925", "origin_question": "Remember, have you seen any cabinet(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any cabinet(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47429925/Validation_47429925_47429925_136549.953.png", "Validation_47429925/Validation_47429925_47429925_136551.453.png", "Validation_47429925/Validation_47429925_47429925_136552.952.png", "Validation_47429925/Validation_47429925_47429925_136554.451.png" ], "turn_id": 2, "scan_id": "Validation_47429925", "origin_question": "In which turn were you closest to \"the cabinet\"? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the cabinet\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_47429925/Validation_47429925_47429925_136555.951.png", "Validation_47429925/Validation_47429925_47429925_136557.450.png", "Validation_47429925/Validation_47429925_47429925_136558.949.png", "Validation_47429925/Validation_47429925_47429925_136560.449.png" ], "turn_id": 3, "scan_id": "Validation_47429925", "origin_question": "When did you first discover \"the clothes\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the clothes\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_47429925/Validation_47429925_47429925_136561.948.png", "Validation_47429925/Validation_47429925_47429925_136563.448.png", "Validation_47429925/Validation_47429925_47429925_136564.947.png", "Validation_47429925/Validation_47429925_47429925_136566.446.png" ], "turn_id": 4, "scan_id": "Validation_47429925", "origin_question": "Is the clothes to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the clothes to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_47429925/Validation_47429925_47429925_136567.946.png", "Validation_47429925/Validation_47429925_47429925_136569.445.png", "Validation_47429925/Validation_47429925_47429925_136570.945.png", "Validation_47429925/Validation_47429925_47429925_136572.444.png" ], "turn_id": 5, "scan_id": "Validation_47429925", "origin_question": "Is the clothes to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the clothes to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47334109/Validation_47334109_47334109_93380.700.png", "Validation_47334109/Validation_47334109_47334109_93382.699.png", "Validation_47334109/Validation_47334109_47334109_93384.698.png", "Validation_47334109/Validation_47334109_47334109_93386.697.png" ], "turn_id": 1, "scan_id": "Validation_47334109", "origin_question": "Remember, have you seen any monitor(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any monitor(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47334109/Validation_47334109_47334109_93388.697.png", "Validation_47334109/Validation_47334109_47334109_93390.696.png", "Validation_47334109/Validation_47334109_47334109_93392.695.png", "Validation_47334109/Validation_47334109_47334109_93394.694.png" ], "turn_id": 2, "scan_id": "Validation_47334109", "origin_question": "In which turn were you closest to \"the modern rectangular white acrylic bathtub\"? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the modern rectangular white acrylic bathtub\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47334109/Validation_47334109_47334109_93396.693.png", "Validation_47334109/Validation_47334109_47334109_93398.693.png", "Validation_47334109/Validation_47334109_47334109_93400.692.png", "Validation_47334109/Validation_47334109_47334109_93402.691.png" ], "turn_id": 3, "scan_id": "Validation_47334109", "origin_question": "", "option": [], "answer": "", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47334109/Validation_47334109_47334109_93404.690.png", "Validation_47334109/Validation_47334109_47334109_93406.689.png", "Validation_47334109/Validation_47334109_47334109_93408.689.png", "Validation_47334109/Validation_47334109_47334109_93410.688.png" ], "turn_id": 4, "scan_id": "Validation_47334109", "origin_question": "In which turn were you closest to \"the sink\"? ", "option": [], "answer": "4", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the sink\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_47334109/Validation_47334109_47334109_93412.687.png", "Validation_47334109/Validation_47334109_47334109_93414.703.png", "Validation_47334109/Validation_47334109_47334109_93416.702.png", "Validation_47334109/Validation_47334109_47334109_93418.701.png" ], "turn_id": 5, "scan_id": "Validation_47334109", "origin_question": "What will be the first-time appearance order of the following categories: bathtub, bin, couch?", "option": [ "bathtub, couch, bin", "bin, couch, bathtub", "couch, bathtub, bin", "bin, bathtub, couch" ], "answer": "bathtub, couch, bin", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bathtub, bin, couch? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['bathtub, couch, bin', 'bin, couch, bathtub', 'couch, bathtub, bin', 'bin, bathtub, couch'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47334109/Validation_47334109_47334109_93420.700.png", "Validation_47334109/Validation_47334109_47334109_93422.700.png", "Validation_47334109/Validation_47334109_47334109_93424.699.png", "Validation_47334109/Validation_47334109_47334109_93426.698.png" ], "turn_id": 6, "scan_id": "Validation_47334109", "origin_question": "Compared to your position at the end of the 5 turn, are you now closer or farther away from \"the bin\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 5 turn, are you now closer or farther away from \"the bin\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_47334109/Validation_47334109_47334109_93428.697.png", "Validation_47334109/Validation_47334109_47334109_93430.696.png", "Validation_47334109/Validation_47334109_47334109_93432.696.png", "Validation_47334109/Validation_47334109_47334109_93434.695.png" ], "turn_id": 7, "scan_id": "Validation_47334109", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the sink\" from your position?", "option": [], "answer": "164.6", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the sink\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_47334109/Validation_47334109_47334109_93436.694.png", "Validation_47334109/Validation_47334109_47334109_93438.693.png", "Validation_47334109/Validation_47334109_47334109_93440.692.png", "Validation_47334109/Validation_47334109_47334109_93442.691.png" ], "turn_id": 8, "scan_id": "Validation_47334109", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the sink\", Y: \"the soft rectangular velvet pillow in light grey\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to X and farther away from Y", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the sink\", Y: \"the soft rectangular velvet pillow in light grey\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_47334109/Validation_47334109_47334109_93444.691.png", "Validation_47334109/Validation_47334109_47334109_93446.690.png", "Validation_47334109/Validation_47334109_47334109_93448.689.png", "Validation_47334109/Validation_47334109_47334109_93450.688.png" ], "turn_id": 9, "scan_id": "Validation_47334109", "origin_question": "Please recall \"the front-loading white washing machine with the sleek shape\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.59", "user_message": "For the 9 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the front-loading white washing machine with the sleek shape\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47430419/Validation_47430419_47430419_696840.568.png", "Validation_47430419/Validation_47430419_47430419_696842.550.png", "Validation_47430419/Validation_47430419_47430419_696844.550.png", "Validation_47430419/Validation_47430419_47430419_696846.549.png" ], "turn_id": 1, "scan_id": "Validation_47430419", "origin_question": "Remember, have you seen any screen(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any screen(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_47430419/Validation_47430419_47430419_696848.548.png", "Validation_47430419/Validation_47430419_47430419_696850.547.png", "Validation_47430419/Validation_47430419_47430419_696852.546.png", "Validation_47430419/Validation_47430419_47430419_696854.546.png" ], "turn_id": 2, "scan_id": "Validation_47430419", "origin_question": "In which turn did you see both \"the modern, wall-mounted, rounded white toilet\" and \"the sleek white metal front-loading washing machine\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the modern, wall-mounted, rounded white toilet\" and \"the sleek white metal front-loading washing machine\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_47430419/Validation_47430419_47430419_696856.545.png", "Validation_47430419/Validation_47430419_47430419_696858.544.png", "Validation_47430419/Validation_47430419_47430419_696860.543.png", "Validation_47430419/Validation_47430419_47430419_696862.542.png" ], "turn_id": 3, "scan_id": "Validation_47430419", "origin_question": "In which turn did you see both \"the bathtub\" and \"the sleek white metal front-loading washing machine\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the bathtub\" and \"the sleek white metal front-loading washing machine\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_47430419/Validation_47430419_47430419_696864.542.png", "Validation_47430419/Validation_47430419_47430419_696866.557.png", "Validation_47430419/Validation_47430419_47430419_696868.556.png", "Validation_47430419/Validation_47430419_47430419_696870.556.png" ], "turn_id": 4, "scan_id": "Validation_47430419", "origin_question": "Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "113.24", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_47430419/Validation_47430419_47430419_696872.555.png", "Validation_47430419/Validation_47430419_47430419_696874.554.png", "Validation_47430419/Validation_47430419_47430419_696876.553.png", "Validation_47430419/Validation_47430419_47430419_696878.552.png" ], "turn_id": 5, "scan_id": "Validation_47430419", "origin_question": "Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "Validation_47430419/Validation_47430419_47430419_696880.552.png", "Validation_47430419/Validation_47430419_47430419_696882.551.png", "Validation_47430419/Validation_47430419_47430419_696884.550.png", "Validation_47430419/Validation_47430419_47430419_696886.549.png" ], "turn_id": 6, "scan_id": "Validation_47430419", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the bathtub\"; \"the natural brown woven rectangular basket\"; \"the deep purple, soft cotton towel\" ", "option": [ "the bathtub", "the natural brown woven rectangular basket", "the deep purple, soft cotton towel" ], "answer": "the deep purple, soft cotton towel", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the bathtub\"; \"the natural brown woven rectangular basket\"; \"the deep purple, soft cotton towel\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the bathtub', 'the natural brown woven rectangular basket', 'the deep purple, soft cotton towel'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47430419/Validation_47430419_47430419_696888.548.png", "Validation_47430419/Validation_47430419_47430419_696890.548.png", "Validation_47430419/Validation_47430419_47430419_696892.547.png", "Validation_47430419/Validation_47430419_47430419_696894.546.png" ], "turn_id": 7, "scan_id": "Validation_47430419", "origin_question": "", "option": [], "answer": "", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_47430419/Validation_47430419_47430419_696896.545.png", "Validation_47430419/Validation_47430419_47430419_696898.544.png", "Validation_47430419/Validation_47430419_47430419_696900.544.png", "Validation_47430419/Validation_47430419_47430419_696902.543.png" ], "turn_id": 8, "scan_id": "Validation_47430419", "origin_question": "Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_47430419/Validation_47430419_47430419_696904.542.png", "Validation_47430419/Validation_47430419_47430419_696906.541.png", "Validation_47430419/Validation_47430419_47430419_696908.540.png", "Validation_47430419/Validation_47430419_47430419_696910.556.png" ], "turn_id": 9, "scan_id": "Validation_47430419", "origin_question": "Which two objects are on the same side of you now? \"the modern, wall-mounted, rounded white toilet\", \"the bathtub\", \"the sleek white metal front-loading washing machine\". ", "option": [ "the modern, wall-mounted, rounded white toilet, the bathtub", "the modern, wall-mounted, rounded white toilet, the sleek white metal front-loading washing machine", "the bathtub, the sleek white metal front-loading washing machine" ], "answer": "the modern, wall-mounted, rounded white toilet, the bathtub", "user_message": "For the 9 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the modern, wall-mounted, rounded white toilet\", \"the bathtub\", \"the sleek white metal front-loading washing machine\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the modern, wall-mounted, rounded white toilet, the bathtub', 'the modern, wall-mounted, rounded white toilet, the sleek white metal front-loading washing machine', 'the bathtub, the sleek white metal front-loading washing machine'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47331662/Validation_47331662_47331662_23421.100.png", "Validation_47331662/Validation_47331662_47331662_23423.082.png", "Validation_47331662/Validation_47331662_47331662_23425.082.png", "Validation_47331662/Validation_47331662_47331662_23427.081.png", "Validation_47331662/Validation_47331662_47331662_23429.080.png" ], "turn_id": 1, "scan_id": "Validation_47331662", "origin_question": "Remember, have you seen any cart(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any cart(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "Validation_47331662/Validation_47331662_47331662_23431.079.png", "Validation_47331662/Validation_47331662_47331662_23433.078.png", "Validation_47331662/Validation_47331662_47331662_23435.094.png", "Validation_47331662/Validation_47331662_47331662_23437.093.png", "Validation_47331662/Validation_47331662_47331662_23439.093.png" ], "turn_id": 2, "scan_id": "Validation_47331662", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the sleek black and gray angular backpack\"; \"the sleek cream door with refined panel design\"; \"the versatile light wooden table with the minimalist design\" ", "option": [ "the sleek black and gray angular backpack", "the sleek cream door with refined panel design", "the versatile light wooden table with the minimalist design" ], "answer": "the versatile light wooden table with the minimalist design", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the sleek black and gray angular backpack\"; \"the sleek cream door with refined panel design\"; \"the versatile light wooden table with the minimalist design\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the sleek black and gray angular backpack', 'the sleek cream door with refined panel design', 'the versatile light wooden table with the minimalist design'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47331662/Validation_47331662_47331662_23441.092.png", "Validation_47331662/Validation_47331662_47331662_23443.091.png", "Validation_47331662/Validation_47331662_47331662_23445.090.png", "Validation_47331662/Validation_47331662_47331662_23447.089.png", "Validation_47331662/Validation_47331662_47331662_23449.089.png" ], "turn_id": 3, "scan_id": "Validation_47331662", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the sleek cream door with refined panel design\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the sleek cream door with refined panel design\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_47331662/Validation_47331662_47331662_23451.088.png", "Validation_47331662/Validation_47331662_47331662_23453.087.png", "Validation_47331662/Validation_47331662_47331662_23455.086.png", "Validation_47331662/Validation_47331662_47331662_23457.085.png", "Validation_47331662/Validation_47331662_47331662_23459.085.png" ], "turn_id": 4, "scan_id": "Validation_47331662", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the sleek black and gray angular backpack\" from your position?", "option": [], "answer": "147.34", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the sleek black and gray angular backpack\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_47331662/Validation_47331662_47331662_23461.084.png", "Validation_47331662/Validation_47331662_47331662_23463.083.png", "Validation_47331662/Validation_47331662_47331662_23465.082.png", "Validation_47331662/Validation_47331662_47331662_23467.081.png", "Validation_47331662/Validation_47331662_47331662_23469.081.png" ], "turn_id": 5, "scan_id": "Validation_47331662", "origin_question": "Please recall \"the sleek cream door with refined panel design\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.36", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the sleek cream door with refined panel design\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_47331662/Validation_47331662_47331662_23471.080.png", "Validation_47331662/Validation_47331662_47331662_23473.079.png", "Validation_47331662/Validation_47331662_47331662_23475.078.png", "Validation_47331662/Validation_47331662_47331662_23477.094.png", "Validation_47331662/Validation_47331662_47331662_23479.093.png" ], "turn_id": 6, "scan_id": "Validation_47331662", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"an oval clock with the dark metallic or wooden frame\" from your position?", "option": [], "answer": "167.83", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"an oval clock with the dark metallic or wooden frame\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_47331662/Validation_47331662_47331662_23481.092.png", "Validation_47331662/Validation_47331662_47331662_23483.091.png", "Validation_47331662/Validation_47331662_47331662_23485.091.png", "Validation_47331662/Validation_47331662_47331662_23487.090.png", "Validation_47331662/Validation_47331662_47331662_23489.089.png" ], "turn_id": 7, "scan_id": "Validation_47331662", "origin_question": "Is the classic vertical metal radiator in neutral color to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the classic vertical metal radiator in neutral color to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47331662/Validation_47331662_47331662_23491.088.png", "Validation_47331662/Validation_47331662_47331662_23493.087.png", "Validation_47331662/Validation_47331662_47331662_23495.087.png", "Validation_47331662/Validation_47331662_47331662_23497.086.png", "Validation_47331662/Validation_47331662_47331662_23499.085.png" ], "turn_id": 8, "scan_id": "Validation_47331662", "origin_question": "Compared to your position at the end of the 6 turn, are you now closer or farther away from \"the versatile light wooden table with the minimalist design\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 6 turn, are you now closer or farther away from \"the versatile light wooden table with the minimalist design\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_48018346/Validation_48018346_48018346_389194.156.png", "Validation_48018346/Validation_48018346_48018346_389196.155.png", "Validation_48018346/Validation_48018346_48018346_389198.154.png", "Validation_48018346/Validation_48018346_48018346_389200.153.png" ], "turn_id": 1, "scan_id": "Validation_48018346", "origin_question": "Remember, have you seen any shampoo(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any shampoo(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_48018346/Validation_48018346_48018346_389202.152.png", "Validation_48018346/Validation_48018346_48018346_389204.151.png", "Validation_48018346/Validation_48018346_48018346_389206.151.png", "Validation_48018346/Validation_48018346_48018346_389208.150.png" ], "turn_id": 2, "scan_id": "Validation_48018346", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_48018346/Validation_48018346_48018346_389210.149.png", "Validation_48018346/Validation_48018346_48018346_389212.148.png", "Validation_48018346/Validation_48018346_48018346_389214.147.png", "Validation_48018346/Validation_48018346_48018346_389216.146.png" ], "turn_id": 3, "scan_id": "Validation_48018346", "origin_question": "In which turn did you see both \"the small bed with the vibrant plaid pattern\" and \"the sheer white curtain that diffuses light\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the small bed with the vibrant plaid pattern\" and \"the sheer white curtain that diffuses light\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_48018346/Validation_48018346_48018346_389218.146.png", "Validation_48018346/Validation_48018346_48018346_389220.145.png", "Validation_48018346/Validation_48018346_48018346_389222.144.png", "Validation_48018346/Validation_48018346_48018346_389224.143.png" ], "turn_id": 4, "scan_id": "Validation_48018346", "origin_question": "Is the sheer white curtain that diffuses light to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the sheer white curtain that diffuses light to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "Validation_48018346/Validation_48018346_48018346_389226.142.png", "Validation_48018346/Validation_48018346_48018346_389228.142.png", "Validation_48018346/Validation_48018346_48018346_389230.141.png", "Validation_48018346/Validation_48018346_48018346_389232.140.png" ], "turn_id": 5, "scan_id": "Validation_48018346", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the sleek, minimalist chair with the black metal frame\"; \"the sleek rectangular light-colored frame\"; \"the sleek rectangular table in the subtle tone\" ", "option": [ "the sleek, minimalist chair with the black metal frame", "the sleek rectangular light-colored frame", "the sleek rectangular table in the subtle tone" ], "answer": "the sleek rectangular light-colored frame", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the sleek, minimalist chair with the black metal frame\"; \"the sleek rectangular light-colored frame\"; \"the sleek rectangular table in the subtle tone\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the sleek, minimalist chair with the black metal frame', 'the sleek rectangular light-colored frame', 'the sleek rectangular table in the subtle tone'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_48018346/Validation_48018346_48018346_389234.156.png", "Validation_48018346/Validation_48018346_48018346_389236.155.png", "Validation_48018346/Validation_48018346_48018346_389238.154.png", "Validation_48018346/Validation_48018346_48018346_389240.153.png" ], "turn_id": 6, "scan_id": "Validation_48018346", "origin_question": "Remember, have you seen any chair(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any chair(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_48018346/Validation_48018346_48018346_389242.152.png", "Validation_48018346/Validation_48018346_48018346_389244.152.png", "Validation_48018346/Validation_48018346_48018346_389246.151.png", "Validation_48018346/Validation_48018346_48018346_389248.150.png" ], "turn_id": 7, "scan_id": "Validation_48018346", "origin_question": "When did you first discover \"the sleek rectangular table in the subtle tone\"? (index of the turn)", "option": [], "answer": "4", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the sleek rectangular table in the subtle tone\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_48018346/Validation_48018346_48018346_389250.149.png", "Validation_48018346/Validation_48018346_48018346_389252.148.png", "Validation_48018346/Validation_48018346_48018346_389254.147.png", "Validation_48018346/Validation_48018346_48018346_389256.147.png" ], "turn_id": 8, "scan_id": "Validation_48018346", "origin_question": "Is the small bed with the vibrant plaid pattern to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the small bed with the vibrant plaid pattern to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42898497/Validation_42898497_42898497_749617.897.png", "Validation_42898497/Validation_42898497_42898497_749619.880.png", "Validation_42898497/Validation_42898497_42898497_749621.879.png", "Validation_42898497/Validation_42898497_42898497_749623.878.png", "Validation_42898497/Validation_42898497_42898497_749625.878.png" ], "turn_id": 1, "scan_id": "Validation_42898497", "origin_question": "Remember, have you seen any hanger(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any hanger(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_42898497/Validation_42898497_42898497_749627.877.png", "Validation_42898497/Validation_42898497_42898497_749629.876.png", "Validation_42898497/Validation_42898497_42898497_749631.875.png", "Validation_42898497/Validation_42898497_42898497_749633.874.png", "Validation_42898497/Validation_42898497_42898497_749635.890.png" ], "turn_id": 2, "scan_id": "Validation_42898497", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_42898497/Validation_42898497_42898497_749637.889.png", "Validation_42898497/Validation_42898497_42898497_749639.889.png", "Validation_42898497/Validation_42898497_42898497_749641.888.png", "Validation_42898497/Validation_42898497_42898497_749643.887.png", "Validation_42898497/Validation_42898497_42898497_749645.886.png" ], "turn_id": 3, "scan_id": "Validation_42898497", "origin_question": "Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "99.82", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_42898497/Validation_42898497_42898497_749647.885.png", "Validation_42898497/Validation_42898497_42898497_749649.884.png", "Validation_42898497/Validation_42898497_42898497_749651.883.png", "Validation_42898497/Validation_42898497_42898497_749653.883.png", "Validation_42898497/Validation_42898497_42898497_749655.882.png" ], "turn_id": 4, "scan_id": "Validation_42898497", "origin_question": "Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "71.66", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_42898497/Validation_42898497_42898497_749657.881.png", "Validation_42898497/Validation_42898497_42898497_749659.880.png", "Validation_42898497/Validation_42898497_42898497_749661.879.png", "Validation_42898497/Validation_42898497_42898497_749663.879.png", "Validation_42898497/Validation_42898497_42898497_749665.878.png" ], "turn_id": 5, "scan_id": "Validation_42898497", "origin_question": "Using your orientation at the end of turn 4 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "90.87", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 4 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_42898497/Validation_42898497_42898497_749667.877.png", "Validation_42898497/Validation_42898497_42898497_749669.876.png", "Validation_42898497/Validation_42898497_42898497_749671.875.png", "Validation_42898497/Validation_42898497_42898497_749673.875.png", "Validation_42898497/Validation_42898497_42898497_749675.890.png" ], "turn_id": 6, "scan_id": "Validation_42898497", "origin_question": "Compared to your position at the end of the 5 turn, are you now closer or farther away from \"the sleek, polished metal lamp with the drum-shaped shade\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 5 turn, are you now closer or farther away from \"the sleek, polished metal lamp with the drum-shaped shade\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_42898497/Validation_42898497_42898497_749677.890.png", "Validation_42898497/Validation_42898497_42898497_749679.889.png", "Validation_42898497/Validation_42898497_42898497_749681.888.png", "Validation_42898497/Validation_42898497_42898497_749683.887.png", "Validation_42898497/Validation_42898497_42898497_749685.886.png" ], "turn_id": 7, "scan_id": "Validation_42898497", "origin_question": "How far is your current position from where you were at the end of the turn 4? (in meters)", "option": [], "answer": "1.64", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 4? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_42898497/Validation_42898497_42898497_749687.885.png", "Validation_42898497/Validation_42898497_42898497_749689.885.png", "Validation_42898497/Validation_42898497_42898497_749691.884.png", "Validation_42898497/Validation_42898497_42898497_749693.883.png", "Validation_42898497/Validation_42898497_42898497_749695.882.png" ], "turn_id": 8, "scan_id": "Validation_42898497", "origin_question": "Which direction is an artificial plant with rich green foliage in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is an artificial plant with rich green foliage in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_42898497/Validation_42898497_42898497_749697.881.png", "Validation_42898497/Validation_42898497_42898497_749699.880.png", "Validation_42898497/Validation_42898497_42898497_749701.880.png", "Validation_42898497/Validation_42898497_42898497_749703.879.png", "Validation_42898497/Validation_42898497_42898497_749705.878.png" ], "turn_id": 9, "scan_id": "Validation_42898497", "origin_question": "At the end of which turn were both of \"an artificial plant with rich green foliage\" and \"the dark moose-patterned rectangular towel\" on your left side?", "option": [], "answer": "4", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"an artificial plant with rich green foliage\" and \"the dark moose-patterned rectangular towel\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_42898497/Validation_42898497_42898497_749707.877.png", "Validation_42898497/Validation_42898497_42898497_749709.876.png", "Validation_42898497/Validation_42898497_42898497_749711.875.png", "Validation_42898497/Validation_42898497_42898497_749713.875.png", "Validation_42898497/Validation_42898497_42898497_749715.890.png" ], "turn_id": 10, "scan_id": "Validation_42898497", "origin_question": "Is an artificial plant with rich green foliage to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Is an artificial plant with rich green foliage to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47331654/Validation_47331654_47331654_22736.848.png", "Validation_47331654/Validation_47331654_47331654_22738.831.png", "Validation_47331654/Validation_47331654_47331654_22740.830.png", "Validation_47331654/Validation_47331654_47331654_22742.829.png", "Validation_47331654/Validation_47331654_47331654_22744.828.png" ], "turn_id": 1, "scan_id": "Validation_47331654", "origin_question": "Remember, have you seen any dumbbell(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any dumbbell(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_47331654/Validation_47331654_47331654_22746.827.png", "Validation_47331654/Validation_47331654_47331654_22748.827.png", "Validation_47331654/Validation_47331654_47331654_22750.826.png", "Validation_47331654/Validation_47331654_47331654_22752.825.png", "Validation_47331654/Validation_47331654_47331654_22754.824.png" ], "turn_id": 2, "scan_id": "Validation_47331654", "origin_question": "In which turn did you see both \"the sleek front-loading washing machine with the white exterior\" and \"the cylindrical white kettle with the plastic body\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the sleek front-loading washing machine with the white exterior\" and \"the cylindrical white kettle with the plastic body\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_47331654/Validation_47331654_47331654_22756.823.png", "Validation_47331654/Validation_47331654_47331654_22758.839.png", "Validation_47331654/Validation_47331654_47331654_22760.838.png", "Validation_47331654/Validation_47331654_47331654_22762.838.png", "Validation_47331654/Validation_47331654_47331654_22764.837.png" ], "turn_id": 3, "scan_id": "Validation_47331654", "origin_question": "Please recall \"the white rectangular oven with glass window and knobs\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.49", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the white rectangular oven with glass window and knobs\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "Validation_47331654/Validation_47331654_47331654_22766.836.png", "Validation_47331654/Validation_47331654_47331654_22768.835.png", "Validation_47331654/Validation_47331654_47331654_22770.834.png", "Validation_47331654/Validation_47331654_47331654_22772.834.png", "Validation_47331654/Validation_47331654_47331654_22774.833.png" ], "turn_id": 4, "scan_id": "Validation_47331654", "origin_question": "Among these three objects(\"the solid modern door with the white frame and glass panels\",\"the white rectangular oven with glass window and knobs\",\"the cylindrical white kettle with the plastic body\" ), which one is the closest to you?", "option": [ "the solid modern door with the white frame and glass panels", "the white rectangular oven with glass window and knobs", "the cylindrical white kettle with the plastic body" ], "answer": "the white rectangular oven with glass window and knobs", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the solid modern door with the white frame and glass panels\",\"the white rectangular oven with glass window and knobs\",\"the cylindrical white kettle with the plastic body\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the solid modern door with the white frame and glass panels', 'the white rectangular oven with glass window and knobs', 'the cylindrical white kettle with the plastic body'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_47331654/Validation_47331654_47331654_22776.832.png", "Validation_47331654/Validation_47331654_47331654_22778.831.png", "Validation_47331654/Validation_47331654_47331654_22780.830.png", "Validation_47331654/Validation_47331654_47331654_22782.829.png", "Validation_47331654/Validation_47331654_47331654_22784.829.png" ], "turn_id": 5, "scan_id": "Validation_47331654", "origin_question": "At the end of which turn were both of \"the sleek front-loading washing machine with the white exterior\" and \"the white rectangular oven with glass window and knobs\" on your left side?", "option": [], "answer": "1", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the sleek front-loading washing machine with the white exterior\" and \"the white rectangular oven with glass window and knobs\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "Validation_47331654/Validation_47331654_47331654_22786.828.png", "Validation_47331654/Validation_47331654_47331654_22788.827.png", "Validation_47331654/Validation_47331654_47331654_22790.826.png", "Validation_47331654/Validation_47331654_47331654_22792.825.png", "Validation_47331654/Validation_47331654_47331654_22794.825.png" ], "turn_id": 6, "scan_id": "Validation_47331654", "origin_question": "Among these three objects(\"the white chair with the rounded backrest and patterned cushion\",\"the window\",\"the solid modern door with the white frame and glass panels\" ), which one is the farthest from you?", "option": [ "the white chair with the rounded backrest and patterned cushion", "the window", "the solid modern door with the white frame and glass panels" ], "answer": "the solid modern door with the white frame and glass panels", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the white chair with the rounded backrest and patterned cushion\",\"the window\",\"the solid modern door with the white frame and glass panels\" ), which one is the farthest from you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the white chair with the rounded backrest and patterned cushion', 'the window', 'the solid modern door with the white frame and glass panels'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47331654/Validation_47331654_47331654_22796.824.png", "Validation_47331654/Validation_47331654_47331654_22798.840.png", "Validation_47331654/Validation_47331654_47331654_22800.839.png", "Validation_47331654/Validation_47331654_47331654_22802.838.png", "Validation_47331654/Validation_47331654_47331654_22804.837.png" ], "turn_id": 7, "scan_id": "Validation_47331654", "origin_question": "Remember, have you seen any duct(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any duct(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_47331654/Validation_47331654_47331654_22806.836.png", "Validation_47331654/Validation_47331654_47331654_22808.836.png", "Validation_47331654/Validation_47331654_47331654_22810.835.png", "Validation_47331654/Validation_47331654_47331654_22812.834.png", "Validation_47331654/Validation_47331654_47331654_22814.833.png" ], "turn_id": 8, "scan_id": "Validation_47331654", "origin_question": "Is the solid modern door with the white frame and glass panels to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the solid modern door with the white frame and glass panels to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_47331654/Validation_47331654_47331654_22816.832.png", "Validation_47331654/Validation_47331654_47331654_22818.832.png", "Validation_47331654/Validation_47331654_47331654_22820.831.png", "Validation_47331654/Validation_47331654_47331654_22822.830.png", "Validation_47331654/Validation_47331654_47331654_22824.829.png" ], "turn_id": 9, "scan_id": "Validation_47331654", "origin_question": "Please recall \"the sleek, rectangular dark and metallic stove\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.12", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the sleek, rectangular dark and metallic stove\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47331654/Validation_47331654_47331654_22826.828.png", "Validation_47331654/Validation_47331654_47331654_22828.827.png", "Validation_47331654/Validation_47331654_47331654_22830.827.png", "Validation_47331654/Validation_47331654_47331654_22832.826.png", "Validation_47331654/Validation_47331654_47331654_22834.825.png" ], "turn_id": 10, "scan_id": "Validation_47331654", "origin_question": "Which direction is the sleek front-loading washing machine with the white exterior in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the sleek front-loading washing machine with the white exterior in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42899619/Validation_42899619_42899619_228888.890.png", "Validation_42899619/Validation_42899619_42899619_228890.372.png", "Validation_42899619/Validation_42899619_42899619_228891.872.png", "Validation_42899619/Validation_42899619_42899619_228893.371.png" ], "turn_id": 1, "scan_id": "Validation_42899619", "origin_question": "Remember, have you seen any water heater(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any water heater(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "Validation_42899619/Validation_42899619_42899619_228894.871.png", "Validation_42899619/Validation_42899619_42899619_228896.370.png", "Validation_42899619/Validation_42899619_42899619_228897.869.png", "Validation_42899619/Validation_42899619_42899619_228899.369.png" ], "turn_id": 2, "scan_id": "Validation_42899619", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the rectangular white wooden door with modern hardware\"; \"the bathtub\"; \"the compact round white porcelain toilet\" ", "option": [ "the rectangular white wooden door with modern hardware", "the bathtub", "the compact round white porcelain toilet" ], "answer": "the bathtub", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the rectangular white wooden door with modern hardware\"; \"the bathtub\"; \"the compact round white porcelain toilet\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular white wooden door with modern hardware', 'the bathtub', 'the compact round white porcelain toilet'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_42899619/Validation_42899619_42899619_228900.868.png", "Validation_42899619/Validation_42899619_42899619_228902.368.png", "Validation_42899619/Validation_42899619_42899619_228903.867.png", "Validation_42899619/Validation_42899619_42899619_228905.366.png" ], "turn_id": 3, "scan_id": "Validation_42899619", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_42899619/Validation_42899619_42899619_228906.866.png", "Validation_42899619/Validation_42899619_42899619_228908.365.png", "Validation_42899619/Validation_42899619_42899619_228909.865.png", "Validation_42899619/Validation_42899619_42899619_228911.364.png" ], "turn_id": 4, "scan_id": "Validation_42899619", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the standard white soft multi-ply toilet paper\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the standard white soft multi-ply toilet paper\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_42899619/Validation_42899619_42899619_228912.863.png", "Validation_42899619/Validation_42899619_42899619_228914.363.png", "Validation_42899619/Validation_42899619_42899619_228915.862.png", "Validation_42899619/Validation_42899619_42899619_228917.362.png" ], "turn_id": 5, "scan_id": "Validation_42899619", "origin_question": "Which direction is the compact round white porcelain toilet in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the compact round white porcelain toilet in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42897541/Validation_42897541_42897541_275377.551.png", "Validation_42897541/Validation_42897541_42897541_275379.534.png", "Validation_42897541/Validation_42897541_42897541_275381.533.png", "Validation_42897541/Validation_42897541_42897541_275383.532.png" ], "turn_id": 1, "scan_id": "Validation_42897541", "origin_question": "Remember, have you seen any garage door(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any garage door(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_42897541/Validation_42897541_42897541_275385.531.png", "Validation_42897541/Validation_42897541_42897541_275387.530.png", "Validation_42897541/Validation_42897541_42897541_275389.530.png", "Validation_42897541/Validation_42897541_42897541_275391.529.png" ], "turn_id": 2, "scan_id": "Validation_42897541", "origin_question": "At the end of which turn were both of \"the square beige upholstered stool\" and \"the soft light beige couch\" on your right side?", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the square beige upholstered stool\" and \"the soft light beige couch\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_42897541/Validation_42897541_42897541_275393.528.png", "Validation_42897541/Validation_42897541_42897541_275395.527.png", "Validation_42897541/Validation_42897541_42897541_275397.526.png", "Validation_42897541/Validation_42897541_42897541_275399.525.png" ], "turn_id": 3, "scan_id": "Validation_42897541", "origin_question": "By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the soft light beige couch\", Y: \"the square, light lavender pillow\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "farther away from both X and Y", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the soft light beige couch\", Y: \"the square, light lavender pillow\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_42897541/Validation_42897541_42897541_275401.541.png", "Validation_42897541/Validation_42897541_42897541_275403.541.png", "Validation_42897541/Validation_42897541_42897541_275405.540.png", "Validation_42897541/Validation_42897541_42897541_275407.539.png" ], "turn_id": 4, "scan_id": "Validation_42897541", "origin_question": "Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_42897541/Validation_42897541_42897541_275409.538.png", "Validation_42897541/Validation_42897541_42897541_275411.537.png", "Validation_42897541/Validation_42897541_42897541_275413.536.png", "Validation_42897541/Validation_42897541_42897541_275415.536.png" ], "turn_id": 5, "scan_id": "Validation_42897541", "origin_question": "Please recall \"the sleek white rectangular steel radiator\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.27", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the sleek white rectangular steel radiator\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_42897541/Validation_42897541_42897541_275417.535.png", "Validation_42897541/Validation_42897541_42897541_275419.534.png", "Validation_42897541/Validation_42897541_42897541_275421.533.png", "Validation_42897541/Validation_42897541_42897541_275423.532.png" ], "turn_id": 6, "scan_id": "Validation_42897541", "origin_question": "By the end of the 4 turn, were you getting closer to or farther away from these two objects: X: \"the sleek white rectangular steel radiator\", Y: \"the glossy black flat-screen monitor with the metallic frame\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to Y and farther away from X", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 4 turn, were you getting closer to or farther away from these two objects: X: \"the sleek white rectangular steel radiator\", Y: \"the glossy black flat-screen monitor with the metallic frame\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_42897541/Validation_42897541_42897541_275425.532.png", "Validation_42897541/Validation_42897541_42897541_275427.531.png", "Validation_42897541/Validation_42897541_42897541_275429.530.png", "Validation_42897541/Validation_42897541_42897541_275431.529.png" ], "turn_id": 7, "scan_id": "Validation_42897541", "origin_question": "In which turn did you see both \"the glossy black flat-screen monitor with the metallic frame\" and \"an antiqued bronze lamp with the light fabric shade\" simultaneously? ", "option": [], "answer": "6", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the glossy black flat-screen monitor with the metallic frame\" and \"an antiqued bronze lamp with the light fabric shade\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_42897541/Validation_42897541_42897541_275433.528.png", "Validation_42897541/Validation_42897541_42897541_275435.528.png", "Validation_42897541/Validation_42897541_42897541_275437.527.png", "Validation_42897541/Validation_42897541_42897541_275439.526.png" ], "turn_id": 8, "scan_id": "Validation_42897541", "origin_question": "Which direction is the refrigerator in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the refrigerator in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47331071/Validation_47331071_47331071_130994.382.png", "Validation_47331071/Validation_47331071_47331071_130996.365.png", "Validation_47331071/Validation_47331071_47331071_130998.364.png", "Validation_47331071/Validation_47331071_47331071_131000.363.png" ], "turn_id": 1, "scan_id": "Validation_47331071", "origin_question": "Remember, have you seen any desk(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any desk(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_47331071/Validation_47331071_47331071_131002.362.png", "Validation_47331071/Validation_47331071_47331071_131004.361.png", "Validation_47331071/Validation_47331071_47331071_131006.361.png", "Validation_47331071/Validation_47331071_47331071_131008.360.png" ], "turn_id": 2, "scan_id": "Validation_47331071", "origin_question": "At the end of which turn were both of \"the sink\" and \"the classic round toilet with the wood-like seat\" on your right side?", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the sink\" and \"the classic round toilet with the wood-like seat\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_47331071/Validation_47331071_47331071_131010.359.png", "Validation_47331071/Validation_47331071_47331071_131012.358.png", "Validation_47331071/Validation_47331071_47331071_131014.357.png", "Validation_47331071/Validation_47331071_47331071_131016.356.png" ], "turn_id": 3, "scan_id": "Validation_47331071", "origin_question": "In which turn did you see both \"the sink\" and \"the classic round toilet with the wood-like seat\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the sink\" and \"the classic round toilet with the wood-like seat\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_47331071/Validation_47331071_47331071_131018.356.png", "Validation_47331071/Validation_47331071_47331071_131020.355.png", "Validation_47331071/Validation_47331071_47331071_131022.354.png", "Validation_47331071/Validation_47331071_47331071_131024.353.png" ], "turn_id": 4, "scan_id": "Validation_47331071", "origin_question": "Which two objects are on the same side of you now? \"the sleek, compact light fixture with the glass bulb\", \"the glossy blue rectangular detergent bottle\", \"the towel\". ", "option": [ "the sleek, compact light fixture with the glass bulb, the glossy blue rectangular detergent bottle", "the sleek, compact light fixture with the glass bulb, the towel", "the glossy blue rectangular detergent bottle, the towel" ], "answer": "the sleek, compact light fixture with the glass bulb, the glossy blue rectangular detergent bottle", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the sleek, compact light fixture with the glass bulb\", \"the glossy blue rectangular detergent bottle\", \"the towel\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the sleek, compact light fixture with the glass bulb, the glossy blue rectangular detergent bottle', 'the sleek, compact light fixture with the glass bulb, the towel', 'the glossy blue rectangular detergent bottle, the towel'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_47331071/Validation_47331071_47331071_131026.369.png", "Validation_47331071/Validation_47331071_47331071_131028.368.png", "Validation_47331071/Validation_47331071_47331071_131030.367.png", "Validation_47331071/Validation_47331071_47331071_131032.367.png" ], "turn_id": 5, "scan_id": "Validation_47331071", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the classic round toilet with the wood-like seat\", Y: \"the glossy blue rectangular detergent bottle\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the classic round toilet with the wood-like seat\", Y: \"the glossy blue rectangular detergent bottle\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "Validation_47331071/Validation_47331071_47331071_131034.366.png", "Validation_47331071/Validation_47331071_47331071_131036.365.png", "Validation_47331071/Validation_47331071_47331071_131038.364.png", "Validation_47331071/Validation_47331071_47331071_131040.364.png" ], "turn_id": 6, "scan_id": "Validation_47331071", "origin_question": "Among these three objects(\"the sink\",\"the classic round toilet with the wood-like seat\",\"the glossy blue rectangular detergent bottle\" ), which one is the farthest from you?", "option": [ "the sink", "the classic round toilet with the wood-like seat", "the glossy blue rectangular detergent bottle" ], "answer": "the glossy blue rectangular detergent bottle", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects(\"the sink\",\"the classic round toilet with the wood-like seat\",\"the glossy blue rectangular detergent bottle\" ), which one is the farthest from you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the sink', 'the classic round toilet with the wood-like seat', 'the glossy blue rectangular detergent bottle'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47331071/Validation_47331071_47331071_131042.363.png", "Validation_47331071/Validation_47331071_47331071_131044.362.png", "Validation_47331071/Validation_47331071_47331071_131046.361.png", "Validation_47331071/Validation_47331071_47331071_131048.360.png" ], "turn_id": 7, "scan_id": "Validation_47331071", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the classic round toilet with the wood-like seat\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the classic round toilet with the wood-like seat\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47331071/Validation_47331071_47331071_131050.360.png", "Validation_47331071/Validation_47331071_47331071_131052.359.png", "Validation_47331071/Validation_47331071_47331071_131054.358.png", "Validation_47331071/Validation_47331071_47331071_131056.357.png" ], "turn_id": 8, "scan_id": "Validation_47331071", "origin_question": "In which turn were you closest to \"the classic round toilet with the wood-like seat\"? ", "option": [], "answer": "4", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the classic round toilet with the wood-like seat\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_44358538/Validation_44358538_44358538_53552.695.png", "Validation_44358538/Validation_44358538_44358538_53554.178.png", "Validation_44358538/Validation_44358538_44358538_53555.677.png", "Validation_44358538/Validation_44358538_44358538_53557.177.png" ], "turn_id": 1, "scan_id": "Validation_44358538", "origin_question": "Remember, have you seen any cleanser(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any cleanser(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_44358538/Validation_44358538_44358538_53558.676.png", "Validation_44358538/Validation_44358538_44358538_53560.175.png", "Validation_44358538/Validation_44358538_44358538_53561.675.png", "Validation_44358538/Validation_44358538_44358538_53563.174.png" ], "turn_id": 2, "scan_id": "Validation_44358538", "origin_question": "In which turn were you closest to \"the spacious brown leather rectangular bag\"? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the spacious brown leather rectangular bag\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_44358538/Validation_44358538_44358538_53564.673.png", "Validation_44358538/Validation_44358538_44358538_53566.173.png", "Validation_44358538/Validation_44358538_44358538_53567.672.png", "Validation_44358538/Validation_44358538_44358538_53569.172.png" ], "turn_id": 3, "scan_id": "Validation_44358538", "origin_question": "At the end of which turn were both of \"the spacious brown leather rectangular bag\" and \"the small, white cylindrical bin\" on your left side?", "option": [], "answer": "2", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the spacious brown leather rectangular bag\" and \"the small, white cylindrical bin\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_44358538/Validation_44358538_44358538_53570.671.png", "Validation_44358538/Validation_44358538_44358538_53572.170.png", "Validation_44358538/Validation_44358538_44358538_53573.670.png", "Validation_44358538/Validation_44358538_44358538_53575.169.png" ], "turn_id": 4, "scan_id": "Validation_44358538", "origin_question": "Remember, have you seen any bathtub(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any bathtub(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_48458489/Validation_48458489_48458489_111538.311.png", "Validation_48458489/Validation_48458489_48458489_111539.811.png", "Validation_48458489/Validation_48458489_48458489_111541.310.png", "Validation_48458489/Validation_48458489_48458489_111542.810.png" ], "turn_id": 1, "scan_id": "Validation_48458489", "origin_question": "Remember, have you seen any trophy(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any trophy(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_48458489/Validation_48458489_48458489_111544.309.png", "Validation_48458489/Validation_48458489_48458489_111545.809.png", "Validation_48458489/Validation_48458489_48458489_111547.308.png", "Validation_48458489/Validation_48458489_48458489_111548.807.png" ], "turn_id": 2, "scan_id": "Validation_48458489", "origin_question": "In which turn were you farthest from \"the sleek circular recessed ceiling light with white trim\"? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the sleek circular recessed ceiling light with white trim\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_48458489/Validation_48458489_48458489_111550.307.png", "Validation_48458489/Validation_48458489_48458489_111551.806.png", "Validation_48458489/Validation_48458489_48458489_111553.306.png", "Validation_48458489/Validation_48458489_48458489_111554.805.png" ], "turn_id": 3, "scan_id": "Validation_48458489", "origin_question": "In which turn were you closest to \"the sleek circular recessed ceiling light with white trim\"? ", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the sleek circular recessed ceiling light with white trim\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_48458489/Validation_48458489_48458489_111556.304.png", "Validation_48458489/Validation_48458489_48458489_111557.804.png", "Validation_48458489/Validation_48458489_48458489_111559.303.png", "Validation_48458489/Validation_48458489_48458489_111560.803.png" ], "turn_id": 4, "scan_id": "Validation_48458489", "origin_question": "How far is your current position from where you were at the end of the turn 1? (in meters)", "option": [], "answer": "1.68", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 1? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42445998/Validation_42445998_42445998_75359.532.png", "Validation_42445998/Validation_42445998_42445998_75361.531.png", "Validation_42445998/Validation_42445998_42445998_75363.530.png", "Validation_42445998/Validation_42445998_42445998_75365.529.png", "Validation_42445998/Validation_42445998_42445998_75367.528.png" ], "turn_id": 1, "scan_id": "Validation_42445998", "origin_question": "Remember, have you seen any magazine(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any magazine(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_42445998/Validation_42445998_42445998_75369.527.png", "Validation_42445998/Validation_42445998_42445998_75371.527.png", "Validation_42445998/Validation_42445998_42445998_75373.526.png", "Validation_42445998/Validation_42445998_42445998_75375.525.png", "Validation_42445998/Validation_42445998_42445998_75377.524.png" ], "turn_id": 2, "scan_id": "Validation_42445998", "origin_question": "At the end of which turn were both of \"the large rectangular wooden table with the patterned cloth\" and \"the compact black fireplace\" on your right side?", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the large rectangular wooden table with the patterned cloth\" and \"the compact black fireplace\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_42445998/Validation_42445998_42445998_75379.523.png", "Validation_42445998/Validation_42445998_42445998_75381.523.png", "Validation_42445998/Validation_42445998_42445998_75383.522.png", "Validation_42445998/Validation_42445998_42445998_75385.521.png", "Validation_42445998/Validation_42445998_42445998_75387.520.png" ], "turn_id": 3, "scan_id": "Validation_42445998", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the spacious rectangular wicker basket with sturdy handles\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the spacious rectangular wicker basket with sturdy handles\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_42445998/Validation_42445998_42445998_75389.536.png", "Validation_42445998/Validation_42445998_42445998_75391.535.png", "Validation_42445998/Validation_42445998_42445998_75393.534.png", "Validation_42445998/Validation_42445998_42445998_75395.533.png", "Validation_42445998/Validation_42445998_42445998_75397.533.png" ], "turn_id": 4, "scan_id": "Validation_42445998", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the spacious rectangular wicker basket with sturdy handles\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the spacious rectangular wicker basket with sturdy handles\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_42445998/Validation_42445998_42445998_75399.532.png", "Validation_42445998/Validation_42445998_42445998_75401.531.png", "Validation_42445998/Validation_42445998_42445998_75403.530.png", "Validation_42445998/Validation_42445998_42445998_75405.529.png", "Validation_42445998/Validation_42445998_42445998_75407.528.png" ], "turn_id": 5, "scan_id": "Validation_42445998", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the medium-sized cardboard box with the black logo\" from your position?", "option": [], "answer": "153.12", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the medium-sized cardboard box with the black logo\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_42445998/Validation_42445998_42445998_75409.527.png", "Validation_42445998/Validation_42445998_42445998_75411.527.png", "Validation_42445998/Validation_42445998_42445998_75413.526.png", "Validation_42445998/Validation_42445998_42445998_75415.525.png", "Validation_42445998/Validation_42445998_42445998_75417.524.png" ], "turn_id": 6, "scan_id": "Validation_42445998", "origin_question": "In which turn were you farthest from \"the large rectangular wooden table with the patterned cloth\"? ", "option": [], "answer": "1", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the large rectangular wooden table with the patterned cloth\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_42445998/Validation_42445998_42445998_75419.523.png", "Validation_42445998/Validation_42445998_42445998_75421.523.png", "Validation_42445998/Validation_42445998_42445998_75423.522.png", "Validation_42445998/Validation_42445998_42445998_75425.521.png", "Validation_42445998/Validation_42445998_42445998_75427.520.png" ], "turn_id": 7, "scan_id": "Validation_42445998", "origin_question": "Please recall \"the spacious rectangular wicker basket with sturdy handles\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.15", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the spacious rectangular wicker basket with sturdy handles\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_42445998/Validation_42445998_42445998_75429.536.png", "Validation_42445998/Validation_42445998_42445998_75431.535.png", "Validation_42445998/Validation_42445998_42445998_75433.534.png", "Validation_42445998/Validation_42445998_42445998_75435.533.png", "Validation_42445998/Validation_42445998_42445998_75437.533.png" ], "turn_id": 8, "scan_id": "Validation_42445998", "origin_question": "Which two objects are on the same side of you now? \"the large rectangular wooden table with the patterned cloth\", \"the spacious rectangular wicker basket with sturdy handles\", \"the compact black fireplace\". ", "option": [ "the large rectangular wooden table with the patterned cloth, the spacious rectangular wicker basket with sturdy handles", "the large rectangular wooden table with the patterned cloth, the compact black fireplace", "the spacious rectangular wicker basket with sturdy handles, the compact black fireplace" ], "answer": "the large rectangular wooden table with the patterned cloth, the spacious rectangular wicker basket with sturdy handles", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the large rectangular wooden table with the patterned cloth\", \"the spacious rectangular wicker basket with sturdy handles\", \"the compact black fireplace\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the large rectangular wooden table with the patterned cloth, the spacious rectangular wicker basket with sturdy handles', 'the large rectangular wooden table with the patterned cloth, the compact black fireplace', 'the spacious rectangular wicker basket with sturdy handles, the compact black fireplace'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_42445998/Validation_42445998_42445998_75439.532.png", "Validation_42445998/Validation_42445998_42445998_75441.531.png", "Validation_42445998/Validation_42445998_42445998_75443.530.png", "Validation_42445998/Validation_42445998_42445998_75445.529.png", "Validation_42445998/Validation_42445998_42445998_75447.528.png" ], "turn_id": 9, "scan_id": "Validation_42445998", "origin_question": "What will be the first-time appearance order of the following categories: curtain, fireplace, window?", "option": [ "curtain, fireplace, window", "window, curtain, fireplace", "fireplace, window, curtain", "fireplace, curtain, window" ], "answer": "fireplace, curtain, window", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: curtain, fireplace, window? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['curtain, fireplace, window', 'window, curtain, fireplace', 'fireplace, window, curtain', 'fireplace, curtain, window'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_42445998/Validation_42445998_42445998_75449.527.png", "Validation_42445998/Validation_42445998_42445998_75451.527.png", "Validation_42445998/Validation_42445998_42445998_75453.526.png", "Validation_42445998/Validation_42445998_42445998_75455.525.png", "Validation_42445998/Validation_42445998_42445998_75457.524.png" ], "turn_id": 10, "scan_id": "Validation_42445998", "origin_question": "Assuming the direction you are facing at the end of the turn 3 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Left", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 3 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47332899/Validation_47332899_47332899_9120.200.png", "Validation_47332899/Validation_47332899_47332899_9122.199.png", "Validation_47332899/Validation_47332899_47332899_9124.198.png", "Validation_47332899/Validation_47332899_47332899_9126.197.png" ], "turn_id": 1, "scan_id": "Validation_47332899", "origin_question": "Remember, have you seen any dress(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any dress(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47332899/Validation_47332899_47332899_9128.196.png", "Validation_47332899/Validation_47332899_47332899_9130.196.png", "Validation_47332899/Validation_47332899_47332899_9132.195.png", "Validation_47332899/Validation_47332899_47332899_9134.194.png" ], "turn_id": 2, "scan_id": "Validation_47332899", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47332899/Validation_47332899_47332899_9136.193.png", "Validation_47332899/Validation_47332899_47332899_9138.192.png", "Validation_47332899/Validation_47332899_47332899_9140.192.png", "Validation_47332899/Validation_47332899_47332899_9142.191.png" ], "turn_id": 3, "scan_id": "Validation_47332899", "origin_question": "In which turn were you farthest from \"the modern lofted bed with the light cream frame\"? ", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the modern lofted bed with the light cream frame\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47332899/Validation_47332899_47332899_9144.190.png", "Validation_47332899/Validation_47332899_47332899_9146.189.png", "Validation_47332899/Validation_47332899_47332899_9148.188.png", "Validation_47332899/Validation_47332899_47332899_9150.188.png" ], "turn_id": 4, "scan_id": "Validation_47332899", "origin_question": "In which turn were you farthest from \"the vibrant yellow corrugated cardboard box\"? ", "option": [], "answer": "4", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the vibrant yellow corrugated cardboard box\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47332899/Validation_47332899_47332899_9152.187.png", "Validation_47332899/Validation_47332899_47332899_9154.186.png", "Validation_47332899/Validation_47332899_47332899_9156.185.png", "Validation_47332899/Validation_47332899_47332899_9158.184.png" ], "turn_id": 5, "scan_id": "Validation_47332899", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the light wooden rectangular table with open shelving\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the light wooden rectangular table with open shelving\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47332899/Validation_47332899_47332899_9160.183.png", "Validation_47332899/Validation_47332899_47332899_9162.199.png", "Validation_47332899/Validation_47332899_47332899_9164.199.png", "Validation_47332899/Validation_47332899_47332899_9166.198.png" ], "turn_id": 6, "scan_id": "Validation_47332899", "origin_question": "", "option": [], "answer": "", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47332899/Validation_47332899_47332899_9168.197.png", "Validation_47332899/Validation_47332899_47332899_9170.196.png", "Validation_47332899/Validation_47332899_47332899_9172.195.png", "Validation_47332899/Validation_47332899_47332899_9174.194.png" ], "turn_id": 7, "scan_id": "Validation_47332899", "origin_question": "Which direction is the sleek black rectangular monitor made of durable plastic in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the sleek black rectangular monitor made of durable plastic in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_47332899/Validation_47332899_47332899_9176.194.png", "Validation_47332899/Validation_47332899_47332899_9178.193.png", "Validation_47332899/Validation_47332899_47332899_9180.192.png", "Validation_47332899/Validation_47332899_47332899_9182.191.png" ], "turn_id": 8, "scan_id": "Validation_47332899", "origin_question": "When did you first discover \"the compact white metal radiator with vertical fins\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the compact white metal radiator with vertical fins\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47332916/Validation_47332916_47332916_13391.587.png", "Validation_47332916/Validation_47332916_47332916_13393.570.png", "Validation_47332916/Validation_47332916_47332916_13395.569.png", "Validation_47332916/Validation_47332916_47332916_13397.568.png" ], "turn_id": 1, "scan_id": "Validation_47332916", "origin_question": "Remember, have you seen any copier(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any copier(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_47332916/Validation_47332916_47332916_13399.567.png", "Validation_47332916/Validation_47332916_47332916_13401.567.png", "Validation_47332916/Validation_47332916_47332916_13403.566.png", "Validation_47332916/Validation_47332916_47332916_13405.565.png" ], "turn_id": 2, "scan_id": "Validation_47332916", "origin_question": "When did you first discover \"the modern bed with the sleek black or white metal frame\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the modern bed with the sleek black or white metal frame\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_47332916/Validation_47332916_47332916_13407.564.png", "Validation_47332916/Validation_47332916_47332916_13409.563.png", "Validation_47332916/Validation_47332916_47332916_13411.562.png", "Validation_47332916/Validation_47332916_47332916_13413.578.png" ], "turn_id": 3, "scan_id": "Validation_47332916", "origin_question": "In which turn did you see both \"the radiator\" and \"the large rectangular window with the neutral frame\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the radiator\" and \"the large rectangular window with the neutral frame\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47332916/Validation_47332916_47332916_13415.577.png", "Validation_47332916/Validation_47332916_47332916_13417.577.png", "Validation_47332916/Validation_47332916_47332916_13419.576.png", "Validation_47332916/Validation_47332916_47332916_13421.575.png" ], "turn_id": 4, "scan_id": "Validation_47332916", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the large rectangular window with the neutral frame\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the large rectangular window with the neutral frame\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47332916/Validation_47332916_47332916_13423.574.png", "Validation_47332916/Validation_47332916_47332916_13425.574.png", "Validation_47332916/Validation_47332916_47332916_13427.573.png", "Validation_47332916/Validation_47332916_47332916_13429.572.png" ], "turn_id": 5, "scan_id": "Validation_47332916", "origin_question": "In which turn were you closest to \"the modern bed with the sleek black or white metal frame\"? ", "option": [], "answer": "3", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the modern bed with the sleek black or white metal frame\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_47332916/Validation_47332916_47332916_13431.571.png", "Validation_47332916/Validation_47332916_47332916_13433.570.png", "Validation_47332916/Validation_47332916_47332916_13435.569.png", "Validation_47332916/Validation_47332916_47332916_13437.569.png" ], "turn_id": 6, "scan_id": "Validation_47332916", "origin_question": "Is the large rectangular window with the neutral frame to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the large rectangular window with the neutral frame to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_47332916/Validation_47332916_47332916_13439.568.png", "Validation_47332916/Validation_47332916_47332916_13441.567.png", "Validation_47332916/Validation_47332916_47332916_13443.566.png", "Validation_47332916/Validation_47332916_47332916_13445.565.png" ], "turn_id": 7, "scan_id": "Validation_47332916", "origin_question": "When did you first discover \"the radiator\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the radiator\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_47332916/Validation_47332916_47332916_13447.565.png", "Validation_47332916/Validation_47332916_47332916_13449.564.png", "Validation_47332916/Validation_47332916_47332916_13451.563.png", "Validation_47332916/Validation_47332916_47332916_13453.579.png" ], "turn_id": 8, "scan_id": "Validation_47332916", "origin_question": "Please recall \"the large rectangular window with the neutral frame\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.52", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the large rectangular window with the neutral frame\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42899700/Validation_42899700_42899700_793154.467.png", "Validation_42899700/Validation_42899700_42899700_793155.949.png", "Validation_42899700/Validation_42899700_42899700_793157.449.png", "Validation_42899700/Validation_42899700_42899700_793158.948.png" ], "turn_id": 1, "scan_id": "Validation_42899700", "origin_question": "Remember, have you seen any bathtub(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any bathtub(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_42899700/Validation_42899700_42899700_793160.448.png", "Validation_42899700/Validation_42899700_42899700_793161.947.png", "Validation_42899700/Validation_42899700_42899700_793163.446.png", "Validation_42899700/Validation_42899700_42899700_793164.962.png" ], "turn_id": 2, "scan_id": "Validation_42899700", "origin_question": "In which turn were you closest to \"the sleek rectangular white acrylic bathtub\"? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the sleek rectangular white acrylic bathtub\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_42899700/Validation_42899700_42899700_793166.462.png", "Validation_42899700/Validation_42899700_42899700_793167.961.png", "Validation_42899700/Validation_42899700_42899700_793169.460.png", "Validation_42899700/Validation_42899700_42899700_793170.960.png" ], "turn_id": 3, "scan_id": "Validation_42899700", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_42899700/Validation_42899700_42899700_793172.459.png", "Validation_42899700/Validation_42899700_42899700_793173.958.png", "Validation_42899700/Validation_42899700_42899700_793175.458.png", "Validation_42899700/Validation_42899700_42899700_793176.957.png" ], "turn_id": 4, "scan_id": "Validation_42899700", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "1.2", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_42899700/Validation_42899700_42899700_793178.457.png", "Validation_42899700/Validation_42899700_42899700_793179.956.png", "Validation_42899700/Validation_42899700_42899700_793181.455.png", "Validation_42899700/Validation_42899700_42899700_793182.955.png" ], "turn_id": 5, "scan_id": "Validation_42899700", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "92.51", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_42899700/Validation_42899700_42899700_793184.454.png", "Validation_42899700/Validation_42899700_42899700_793185.953.png", "Validation_42899700/Validation_42899700_42899700_793187.453.png", "Validation_42899700/Validation_42899700_42899700_793188.952.png" ], "turn_id": 6, "scan_id": "Validation_42899700", "origin_question": "Compared to your position at the end of the 5 turn, are you now closer or farther away from \"the modern rectangular glossy white ceramic sink\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 5 turn, are you now closer or farther away from \"the modern rectangular glossy white ceramic sink\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_48018379/Validation_48018379_48018379_389996.137.png", "Validation_48018379/Validation_48018379_48018379_389998.136.png", "Validation_48018379/Validation_48018379_48018379_390000.135.png", "Validation_48018379/Validation_48018379_48018379_390002.135.png" ], "turn_id": 1, "scan_id": "Validation_48018379", "origin_question": "Remember, have you seen any cabinet(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any cabinet(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_48018379/Validation_48018379_48018379_390004.134.png", "Validation_48018379/Validation_48018379_48018379_390006.133.png", "Validation_48018379/Validation_48018379_48018379_390008.132.png", "Validation_48018379/Validation_48018379_48018379_390010.131.png" ], "turn_id": 2, "scan_id": "Validation_48018379", "origin_question": "In which turn did you see both \"the tall, glossy red vase made of glass\" and \"the rectangular mirror with the light wood frame\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the tall, glossy red vase made of glass\" and \"the rectangular mirror with the light wood frame\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_48018379/Validation_48018379_48018379_390012.131.png", "Validation_48018379/Validation_48018379_48018379_390014.130.png", "Validation_48018379/Validation_48018379_48018379_390016.129.png", "Validation_48018379/Validation_48018379_48018379_390018.128.png" ], "turn_id": 3, "scan_id": "Validation_48018379", "origin_question": "Is the large, white-framed rectangular window to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the large, white-framed rectangular window to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_48018379/Validation_48018379_48018379_390020.127.png", "Validation_48018379/Validation_48018379_48018379_390022.143.png", "Validation_48018379/Validation_48018379_48018379_390024.142.png", "Validation_48018379/Validation_48018379_48018379_390026.141.png" ], "turn_id": 4, "scan_id": "Validation_48018379", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_48018379/Validation_48018379_48018379_390028.141.png", "Validation_48018379/Validation_48018379_48018379_390030.140.png", "Validation_48018379/Validation_48018379_48018379_390032.139.png", "Validation_48018379/Validation_48018379_48018379_390034.138.png" ], "turn_id": 5, "scan_id": "Validation_48018379", "origin_question": "How far is your current position from where you were at the end of the turn 3? (in meters)", "option": [], "answer": "2.16", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 3? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_48018379/Validation_48018379_48018379_390036.137.png", "Validation_48018379/Validation_48018379_48018379_390038.137.png", "Validation_48018379/Validation_48018379_48018379_390040.136.png", "Validation_48018379/Validation_48018379_48018379_390042.135.png" ], "turn_id": 6, "scan_id": "Validation_48018379", "origin_question": "Is the tall, glossy red vase made of glass to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the tall, glossy red vase made of glass to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42899725/Validation_42899725_42899725_799579.269.png", "Validation_42899725/Validation_42899725_42899725_799581.252.png", "Validation_42899725/Validation_42899725_42899725_799583.251.png", "Validation_42899725/Validation_42899725_42899725_799585.250.png" ], "turn_id": 1, "scan_id": "Validation_42899725", "origin_question": "Remember, have you seen any drawer(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any drawer(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_42899725/Validation_42899725_42899725_799587.249.png", "Validation_42899725/Validation_42899725_42899725_799589.248.png", "Validation_42899725/Validation_42899725_42899725_799591.247.png", "Validation_42899725/Validation_42899725_42899725_799593.247.png" ], "turn_id": 2, "scan_id": "Validation_42899725", "origin_question": "In which turn were you closest to \"the modern oval-shaped white ceramic toilet\"? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the modern oval-shaped white ceramic toilet\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_42899725/Validation_42899725_42899725_799595.246.png", "Validation_42899725/Validation_42899725_42899725_799597.245.png", "Validation_42899725/Validation_42899725_42899725_799599.244.png", "Validation_42899725/Validation_42899725_42899725_799601.260.png" ], "turn_id": 3, "scan_id": "Validation_42899725", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the bathtub\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the bathtub\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_42899725/Validation_42899725_42899725_799603.259.png", "Validation_42899725/Validation_42899725_42899725_799605.258.png", "Validation_42899725/Validation_42899725_42899725_799607.257.png", "Validation_42899725/Validation_42899725_42899725_799609.257.png" ], "turn_id": 4, "scan_id": "Validation_42899725", "origin_question": "Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the bathtub\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the bathtub\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_42899725/Validation_42899725_42899725_799611.256.png", "Validation_42899725/Validation_42899725_42899725_799613.255.png", "Validation_42899725/Validation_42899725_42899725_799615.254.png", "Validation_42899725/Validation_42899725_42899725_799617.253.png" ], "turn_id": 5, "scan_id": "Validation_42899725", "origin_question": "When did you first discover \"the small, silver cylindrical bin\"? (index of the turn)", "option": [], "answer": "5", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the small, silver cylindrical bin\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_42899725/Validation_42899725_42899725_799619.252.png", "Validation_42899725/Validation_42899725_42899725_799621.252.png", "Validation_42899725/Validation_42899725_42899725_799623.251.png", "Validation_42899725/Validation_42899725_42899725_799625.250.png" ], "turn_id": 6, "scan_id": "Validation_42899725", "origin_question": "Is the small, silver cylindrical bin to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the small, silver cylindrical bin to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_45260925/Validation_45260925_45260925_2394.831.png", "Validation_45260925/Validation_45260925_45260925_2396.830.png", "Validation_45260925/Validation_45260925_45260925_2398.829.png", "Validation_45260925/Validation_45260925_45260925_2400.828.png" ], "turn_id": 1, "scan_id": "Validation_45260925", "origin_question": "Remember, have you seen any fireplace(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any fireplace(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_45260925/Validation_45260925_45260925_2402.827.png", "Validation_45260925/Validation_45260925_45260925_2404.827.png", "Validation_45260925/Validation_45260925_45260925_2406.826.png", "Validation_45260925/Validation_45260925_45260925_2408.825.png" ], "turn_id": 2, "scan_id": "Validation_45260925", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the sleek rectangular black monitor with thin edges\", Y: \"the classic dark cast iron fireplace with ceramic tiles\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the sleek rectangular black monitor with thin edges\", Y: \"the classic dark cast iron fireplace with ceramic tiles\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_45260925/Validation_45260925_45260925_2410.824.png", "Validation_45260925/Validation_45260925_45260925_2412.823.png", "Validation_45260925/Validation_45260925_45260925_2414.823.png", "Validation_45260925/Validation_45260925_45260925_2416.822.png" ], "turn_id": 3, "scan_id": "Validation_45260925", "origin_question": "Which two objects are on the same side of you now? \"the wooden chair with the warm brown finish\", \"the soft white ceramic teapot with an ergonomic handle\", \"the classic dark cast iron fireplace with ceramic tiles\". ", "option": [ "the wooden chair with the warm brown finish, the soft white ceramic teapot with an ergonomic handle", "the wooden chair with the warm brown finish, the classic dark cast iron fireplace with ceramic tiles", "the soft white ceramic teapot with an ergonomic handle, the classic dark cast iron fireplace with ceramic tiles" ], "answer": "the wooden chair with the warm brown finish, the soft white ceramic teapot with an ergonomic handle", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the wooden chair with the warm brown finish\", \"the soft white ceramic teapot with an ergonomic handle\", \"the classic dark cast iron fireplace with ceramic tiles\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the wooden chair with the warm brown finish, the soft white ceramic teapot with an ergonomic handle', 'the wooden chair with the warm brown finish, the classic dark cast iron fireplace with ceramic tiles', 'the soft white ceramic teapot with an ergonomic handle, the classic dark cast iron fireplace with ceramic tiles'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_45260925/Validation_45260925_45260925_2418.821.png", "Validation_45260925/Validation_45260925_45260925_2420.820.png", "Validation_45260925/Validation_45260925_45260925_2422.820.png", "Validation_45260925/Validation_45260925_45260925_2424.819.png" ], "turn_id": 4, "scan_id": "Validation_45260925", "origin_question": "Which two objects are on the same side of you now? \"the classic dark cast iron fireplace with ceramic tiles\", \"the vibrant, lush plant with glossy green leaves\", \"the rich brown leather couch with rolled armrests\". ", "option": [ "the classic dark cast iron fireplace with ceramic tiles, the vibrant, lush plant with glossy green leaves", "the classic dark cast iron fireplace with ceramic tiles, the rich brown leather couch with rolled armrests", "the vibrant, lush plant with glossy green leaves, the rich brown leather couch with rolled armrests" ], "answer": "the classic dark cast iron fireplace with ceramic tiles, the vibrant, lush plant with glossy green leaves", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the classic dark cast iron fireplace with ceramic tiles\", \"the vibrant, lush plant with glossy green leaves\", \"the rich brown leather couch with rolled armrests\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the classic dark cast iron fireplace with ceramic tiles, the vibrant, lush plant with glossy green leaves', 'the classic dark cast iron fireplace with ceramic tiles, the rich brown leather couch with rolled armrests', 'the vibrant, lush plant with glossy green leaves, the rich brown leather couch with rolled armrests'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_45260925/Validation_45260925_45260925_2426.818.png", "Validation_45260925/Validation_45260925_45260925_2428.817.png", "Validation_45260925/Validation_45260925_45260925_2430.816.png", "Validation_45260925/Validation_45260925_45260925_2432.816.png" ], "turn_id": 5, "scan_id": "Validation_45260925", "origin_question": "Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_45260925/Validation_45260925_45260925_2434.831.png", "Validation_45260925/Validation_45260925_45260925_2436.831.png", "Validation_45260925/Validation_45260925_45260925_2438.830.png", "Validation_45260925/Validation_45260925_45260925_2440.829.png" ], "turn_id": 6, "scan_id": "Validation_45260925", "origin_question": "When did you first discover \"the wooden chair with the warm brown finish\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the wooden chair with the warm brown finish\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_45260925/Validation_45260925_45260925_2442.828.png", "Validation_45260925/Validation_45260925_45260925_2444.827.png", "Validation_45260925/Validation_45260925_45260925_2446.827.png", "Validation_45260925/Validation_45260925_45260925_2448.826.png" ], "turn_id": 7, "scan_id": "Validation_45260925", "origin_question": "Remember, have you seen any chair(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any chair(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_45260925/Validation_45260925_45260925_2450.825.png", "Validation_45260925/Validation_45260925_45260925_2452.824.png", "Validation_45260925/Validation_45260925_45260925_2454.824.png", "Validation_45260925/Validation_45260925_45260925_2456.823.png" ], "turn_id": 8, "scan_id": "Validation_45260925", "origin_question": "Which two objects are on the same side of you now? \"the wooden chair with the warm brown finish\", \"the rich brown leather couch with rolled armrests\", \"the vibrant, lush plant with glossy green leaves\". ", "option": [ "the wooden chair with the warm brown finish, the rich brown leather couch with rolled armrests", "the wooden chair with the warm brown finish, the vibrant, lush plant with glossy green leaves", "the rich brown leather couch with rolled armrests, the vibrant, lush plant with glossy green leaves" ], "answer": "the rich brown leather couch with rolled armrests, the vibrant, lush plant with glossy green leaves", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the wooden chair with the warm brown finish\", \"the rich brown leather couch with rolled armrests\", \"the vibrant, lush plant with glossy green leaves\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the wooden chair with the warm brown finish, the rich brown leather couch with rolled armrests', 'the wooden chair with the warm brown finish, the vibrant, lush plant with glossy green leaves', 'the rich brown leather couch with rolled armrests, the vibrant, lush plant with glossy green leaves'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42899736/Validation_42899736_42899736_801360.551.png", "Validation_42899736/Validation_42899736_42899736_801362.533.png", "Validation_42899736/Validation_42899736_42899736_801364.532.png", "Validation_42899736/Validation_42899736_42899736_801366.531.png", "Validation_42899736/Validation_42899736_42899736_801368.531.png" ], "turn_id": 1, "scan_id": "Validation_42899736", "origin_question": "Remember, have you seen any bin(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any bin(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_42899736/Validation_42899736_42899736_801370.530.png", "Validation_42899736/Validation_42899736_42899736_801372.529.png", "Validation_42899736/Validation_42899736_42899736_801374.545.png", "Validation_42899736/Validation_42899736_42899736_801376.544.png", "Validation_42899736/Validation_42899736_42899736_801378.543.png" ], "turn_id": 2, "scan_id": "Validation_42899736", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the white rectangular cotton towel\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the white rectangular cotton towel\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_42899736/Validation_42899736_42899736_801380.542.png", "Validation_42899736/Validation_42899736_42899736_801382.542.png", "Validation_42899736/Validation_42899736_42899736_801384.541.png", "Validation_42899736/Validation_42899736_42899736_801386.540.png", "Validation_42899736/Validation_42899736_42899736_801388.539.png" ], "turn_id": 3, "scan_id": "Validation_42899736", "origin_question": "In which turn were you farthest from \"the modern rectangular bed with the brown bedspread\"? ", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the modern rectangular bed with the brown bedspread\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_42899736/Validation_42899736_42899736_801390.538.png", "Validation_42899736/Validation_42899736_42899736_801392.537.png", "Validation_42899736/Validation_42899736_42899736_801394.537.png", "Validation_42899736/Validation_42899736_42899736_801396.536.png", "Validation_42899736/Validation_42899736_42899736_801398.535.png" ], "turn_id": 4, "scan_id": "Validation_42899736", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "1.34", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_42899736/Validation_42899736_42899736_801400.534.png", "Validation_42899736/Validation_42899736_42899736_801402.533.png", "Validation_42899736/Validation_42899736_42899736_801404.532.png", "Validation_42899736/Validation_42899736_42899736_801406.532.png", "Validation_42899736/Validation_42899736_42899736_801408.531.png" ], "turn_id": 5, "scan_id": "Validation_42899736", "origin_question": "Which direction is the modern rectangular bed with the brown bedspread in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the modern rectangular bed with the brown bedspread in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_42899736/Validation_42899736_42899736_801410.530.png", "Validation_42899736/Validation_42899736_42899736_801412.529.png", "Validation_42899736/Validation_42899736_42899736_801414.545.png", "Validation_42899736/Validation_42899736_42899736_801416.544.png", "Validation_42899736/Validation_42899736_42899736_801418.543.png" ], "turn_id": 6, "scan_id": "Validation_42899736", "origin_question": "Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_42899736/Validation_42899736_42899736_801420.542.png", "Validation_42899736/Validation_42899736_42899736_801422.542.png", "Validation_42899736/Validation_42899736_42899736_801424.541.png", "Validation_42899736/Validation_42899736_42899736_801426.540.png", "Validation_42899736/Validation_42899736_42899736_801428.539.png" ], "turn_id": 7, "scan_id": "Validation_42899736", "origin_question": "Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the classic lamp with the round white lampshade\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the classic lamp with the round white lampshade\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_42899736/Validation_42899736_42899736_801430.538.png", "Validation_42899736/Validation_42899736_42899736_801432.537.png", "Validation_42899736/Validation_42899736_42899736_801434.537.png", "Validation_42899736/Validation_42899736_42899736_801436.536.png", "Validation_42899736/Validation_42899736_42899736_801438.535.png" ], "turn_id": 8, "scan_id": "Validation_42899736", "origin_question": "How far is your current position from where you were at the end of the turn 1? (in meters)", "option": [], "answer": "4.46", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 1? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47332905/Validation_47332905_47332905_10950.910.png", "Validation_47332905/Validation_47332905_47332905_10952.893.png", "Validation_47332905/Validation_47332905_47332905_10954.892.png", "Validation_47332905/Validation_47332905_47332905_10956.891.png" ], "turn_id": 1, "scan_id": "Validation_47332905", "origin_question": "Remember, have you seen any headboard(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any headboard(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_47332905/Validation_47332905_47332905_10958.890.png", "Validation_47332905/Validation_47332905_47332905_10960.889.png", "Validation_47332905/Validation_47332905_47332905_10962.889.png", "Validation_47332905/Validation_47332905_47332905_10964.888.png" ], "turn_id": 2, "scan_id": "Validation_47332905", "origin_question": "When did you first discover \"the light\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the light\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_47332905/Validation_47332905_47332905_10966.887.png", "Validation_47332905/Validation_47332905_47332905_10968.886.png", "Validation_47332905/Validation_47332905_47332905_10970.885.png", "Validation_47332905/Validation_47332905_47332905_10972.901.png" ], "turn_id": 3, "scan_id": "Validation_47332905", "origin_question": "Is the light to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the light to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47332905/Validation_47332905_47332905_10974.900.png", "Validation_47332905/Validation_47332905_47332905_10976.900.png", "Validation_47332905/Validation_47332905_47332905_10978.899.png", "Validation_47332905/Validation_47332905_47332905_10980.898.png" ], "turn_id": 4, "scan_id": "Validation_47332905", "origin_question": "Which direction is the medium to large stainless steel refrigerator in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the medium to large stainless steel refrigerator in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_47332905/Validation_47332905_47332905_10982.897.png", "Validation_47332905/Validation_47332905_47332905_10984.896.png", "Validation_47332905/Validation_47332905_47332905_10986.896.png", "Validation_47332905/Validation_47332905_47332905_10988.895.png" ], "turn_id": 5, "scan_id": "Validation_47332905", "origin_question": "What will be the first-time appearance order of the following categories: light, sink, stove?", "option": [ "stove, sink, light", "light, sink, stove", "stove, light, sink", "sink, stove, light" ], "answer": "light, sink, stove", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: light, sink, stove? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['stove, sink, light', 'light, sink, stove', 'stove, light, sink', 'sink, stove, light'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_47332905/Validation_47332905_47332905_10990.894.png", "Validation_47332905/Validation_47332905_47332905_10992.893.png", "Validation_47332905/Validation_47332905_47332905_10994.892.png", "Validation_47332905/Validation_47332905_47332905_10996.892.png" ], "turn_id": 6, "scan_id": "Validation_47332905", "origin_question": "Which two objects are on the same side of you now? \"the window\", \"the medium to large stainless steel refrigerator\", \"the sleek stainless steel portable stove with one or two burners\". ", "option": [ "the window, the medium to large stainless steel refrigerator", "the window, the sleek stainless steel portable stove with one or two burners", "the medium to large stainless steel refrigerator, the sleek stainless steel portable stove with one or two burners" ], "answer": "the window, the medium to large stainless steel refrigerator", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the window\", \"the medium to large stainless steel refrigerator\", \"the sleek stainless steel portable stove with one or two burners\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the window, the medium to large stainless steel refrigerator', 'the window, the sleek stainless steel portable stove with one or two burners', 'the medium to large stainless steel refrigerator, the sleek stainless steel portable stove with one or two burners'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_47332905/Validation_47332905_47332905_10998.891.png", "Validation_47332905/Validation_47332905_47332905_11000.890.png", "Validation_47332905/Validation_47332905_47332905_11002.889.png", "Validation_47332905/Validation_47332905_47332905_11004.888.png" ], "turn_id": 7, "scan_id": "Validation_47332905", "origin_question": "Assuming the direction you are facing at the end of the turn 5 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Right", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 5 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_47332905/Validation_47332905_47332905_11006.888.png", "Validation_47332905/Validation_47332905_47332905_11008.887.png", "Validation_47332905/Validation_47332905_47332905_11010.886.png", "Validation_47332905/Validation_47332905_47332905_11012.885.png" ], "turn_id": 8, "scan_id": "Validation_47332905", "origin_question": "At the end of which turn were both of \"the light\" and \"the medium-sized light wood rectangular table\" on your right side?", "option": [], "answer": "8", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the light\" and \"the medium-sized light wood rectangular table\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42897564/Validation_42897564_42897564_278973.315.png", "Validation_42897564/Validation_42897564_42897564_278975.314.png", "Validation_42897564/Validation_42897564_42897564_278977.314.png", "Validation_42897564/Validation_42897564_42897564_278979.313.png" ], "turn_id": 1, "scan_id": "Validation_42897564", "origin_question": "Remember, have you seen any magazine(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any magazine(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_42897564/Validation_42897564_42897564_278981.312.png", "Validation_42897564/Validation_42897564_42897564_278983.311.png", "Validation_42897564/Validation_42897564_42897564_278985.310.png", "Validation_42897564/Validation_42897564_42897564_278987.309.png" ], "turn_id": 2, "scan_id": "Validation_42897564", "origin_question": "In which turn were you closest to \"the compact, dark-finished table with multiple shelves\"? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the compact, dark-finished table with multiple shelves\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_42897564/Validation_42897564_42897564_278989.309.png", "Validation_42897564/Validation_42897564_42897564_278991.308.png", "Validation_42897564/Validation_42897564_42897564_278993.307.png", "Validation_42897564/Validation_42897564_42897564_278995.306.png" ], "turn_id": 3, "scan_id": "Validation_42897564", "origin_question": "In which turn did you see both \"the medium rectangular white steel radiator\" and \"the white rectangular wooden door\" simultaneously? ", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the medium rectangular white steel radiator\" and \"the white rectangular wooden door\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_42897564/Validation_42897564_42897564_278997.305.png", "Validation_42897564/Validation_42897564_42897564_278999.305.png", "Validation_42897564/Validation_42897564_42897564_279001.304.png", "Validation_42897564/Validation_42897564_42897564_279003.303.png" ], "turn_id": 4, "scan_id": "Validation_42897564", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "Validation_42897564/Validation_42897564_42897564_279005.302.png", "Validation_42897564/Validation_42897564_42897564_279007.301.png", "Validation_42897564/Validation_42897564_42897564_279009.300.png", "Validation_42897564/Validation_42897564_42897564_279011.316.png" ], "turn_id": 5, "scan_id": "Validation_42897564", "origin_question": "Among these three objects(\"the medium rectangular white steel radiator\",\"the large rectangular wooden picture frame\",\"the sleek, tall white cabinet with the glossy finish\" ), which one is the closest to you?", "option": [ "the medium rectangular white steel radiator", "the large rectangular wooden picture frame", "the sleek, tall white cabinet with the glossy finish" ], "answer": "the sleek, tall white cabinet with the glossy finish", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects(\"the medium rectangular white steel radiator\",\"the large rectangular wooden picture frame\",\"the sleek, tall white cabinet with the glossy finish\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the medium rectangular white steel radiator', 'the large rectangular wooden picture frame', 'the sleek, tall white cabinet with the glossy finish'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_42897564/Validation_42897564_42897564_279013.315.png", "Validation_42897564/Validation_42897564_42897564_279015.315.png", "Validation_42897564/Validation_42897564_42897564_279017.314.png", "Validation_42897564/Validation_42897564_42897564_279019.313.png" ], "turn_id": 6, "scan_id": "Validation_42897564", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_42897564/Validation_42897564_42897564_279021.312.png", "Validation_42897564/Validation_42897564_42897564_279023.311.png", "Validation_42897564/Validation_42897564_42897564_279025.310.png", "Validation_42897564/Validation_42897564_42897564_279027.310.png" ], "turn_id": 7, "scan_id": "Validation_42897564", "origin_question": "Which direction is the sleek rectangular mirror with the silver frame in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the sleek rectangular mirror with the silver frame in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_44358446/Validation_44358446_44358446_43008.918.png", "Validation_44358446/Validation_44358446_44358446_43010.901.png", "Validation_44358446/Validation_44358446_44358446_43012.900.png", "Validation_44358446/Validation_44358446_44358446_43014.899.png", "Validation_44358446/Validation_44358446_44358446_43016.898.png" ], "turn_id": 1, "scan_id": "Validation_44358446", "origin_question": "Remember, have you seen any clothes(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any clothes(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_44358446/Validation_44358446_44358446_43018.897.png", "Validation_44358446/Validation_44358446_44358446_43020.896.png", "Validation_44358446/Validation_44358446_44358446_43022.895.png", "Validation_44358446/Validation_44358446_44358446_43024.895.png", "Validation_44358446/Validation_44358446_44358446_43026.894.png" ], "turn_id": 2, "scan_id": "Validation_44358446", "origin_question": "Please recall \"the medium-sized, light brown woven basket\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.42", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the medium-sized, light brown woven basket\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_44358446/Validation_44358446_44358446_43028.893.png", "Validation_44358446/Validation_44358446_44358446_43030.892.png", "Validation_44358446/Validation_44358446_44358446_43032.891.png", "Validation_44358446/Validation_44358446_44358446_43034.891.png", "Validation_44358446/Validation_44358446_44358446_43036.890.png" ], "turn_id": 3, "scan_id": "Validation_44358446", "origin_question": "Which two objects are on the same side of you now? \"the table\", \"the tall white wooden cabinet with the creamy finish\", \"the collection of dark-toned clothes\". ", "option": [ "the table, the tall white wooden cabinet with the creamy finish", "the table, the collection of dark-toned clothes", "the tall white wooden cabinet with the creamy finish, the collection of dark-toned clothes" ], "answer": "the tall white wooden cabinet with the creamy finish, the collection of dark-toned clothes", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the table\", \"the tall white wooden cabinet with the creamy finish\", \"the collection of dark-toned clothes\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the table, the tall white wooden cabinet with the creamy finish', 'the table, the collection of dark-toned clothes', 'the tall white wooden cabinet with the creamy finish, the collection of dark-toned clothes'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_44358446/Validation_44358446_44358446_43038.889.png", "Validation_44358446/Validation_44358446_44358446_43040.888.png", "Validation_44358446/Validation_44358446_44358446_43042.887.png", "Validation_44358446/Validation_44358446_44358446_43044.903.png", "Validation_44358446/Validation_44358446_44358446_43046.902.png" ], "turn_id": 4, "scan_id": "Validation_44358446", "origin_question": "Which two objects are on the same side of you now? \"the sleek dark chair with armrests\", \"the table\", \"the tall white wooden cabinet with the creamy finish\". ", "option": [ "the sleek dark chair with armrests, the table", "the sleek dark chair with armrests, the tall white wooden cabinet with the creamy finish", "the table, the tall white wooden cabinet with the creamy finish" ], "answer": "the sleek dark chair with armrests, the table", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the sleek dark chair with armrests\", \"the table\", \"the tall white wooden cabinet with the creamy finish\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the sleek dark chair with armrests, the table', 'the sleek dark chair with armrests, the tall white wooden cabinet with the creamy finish', 'the table, the tall white wooden cabinet with the creamy finish'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_44358446/Validation_44358446_44358446_43048.901.png", "Validation_44358446/Validation_44358446_44358446_43050.901.png", "Validation_44358446/Validation_44358446_44358446_43052.900.png", "Validation_44358446/Validation_44358446_44358446_43054.899.png", "Validation_44358446/Validation_44358446_44358446_43056.898.png" ], "turn_id": 5, "scan_id": "Validation_44358446", "origin_question": "Please recall \"the sleek dark chair with armrests\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.31", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the sleek dark chair with armrests\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_44358446/Validation_44358446_44358446_43058.897.png", "Validation_44358446/Validation_44358446_44358446_43060.896.png", "Validation_44358446/Validation_44358446_44358446_43062.896.png", "Validation_44358446/Validation_44358446_44358446_43064.895.png", "Validation_44358446/Validation_44358446_44358446_43066.894.png" ], "turn_id": 6, "scan_id": "Validation_44358446", "origin_question": "At the end of which turn were both of \"the floral-patterned couch with no armrests\" and \"the medium-sized, light brown woven basket\" on your left side?", "option": [], "answer": "5", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the floral-patterned couch with no armrests\" and \"the medium-sized, light brown woven basket\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_44358446/Validation_44358446_44358446_43068.893.png", "Validation_44358446/Validation_44358446_44358446_43070.892.png", "Validation_44358446/Validation_44358446_44358446_43072.891.png", "Validation_44358446/Validation_44358446_44358446_43074.891.png", "Validation_44358446/Validation_44358446_44358446_43076.890.png" ], "turn_id": 7, "scan_id": "Validation_44358446", "origin_question": "Is the sleek dark chair with armrests to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the sleek dark chair with armrests to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_44358446/Validation_44358446_44358446_43078.889.png", "Validation_44358446/Validation_44358446_44358446_43080.888.png", "Validation_44358446/Validation_44358446_44358446_43082.887.png", "Validation_44358446/Validation_44358446_44358446_43084.903.png", "Validation_44358446/Validation_44358446_44358446_43086.902.png" ], "turn_id": 8, "scan_id": "Validation_44358446", "origin_question": "Using your orientation at the end of turn 7 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 7 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_44358446/Validation_44358446_44358446_43088.901.png", "Validation_44358446/Validation_44358446_44358446_43090.901.png", "Validation_44358446/Validation_44358446_44358446_43092.900.png", "Validation_44358446/Validation_44358446_44358446_43094.899.png", "Validation_44358446/Validation_44358446_44358446_43096.898.png" ], "turn_id": 9, "scan_id": "Validation_44358446", "origin_question": "Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Right", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_44358446/Validation_44358446_44358446_43098.897.png", "Validation_44358446/Validation_44358446_44358446_43100.897.png", "Validation_44358446/Validation_44358446_44358446_43102.896.png", "Validation_44358446/Validation_44358446_44358446_43104.895.png", "Validation_44358446/Validation_44358446_44358446_43106.894.png" ], "turn_id": 10, "scan_id": "Validation_44358446", "origin_question": "When did you first discover \"the floral-patterned couch with no armrests\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the floral-patterned couch with no armrests\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_48018962/Validation_48018962_48018962_556470.557.png", "Validation_48018962/Validation_48018962_48018962_556472.056.png", "Validation_48018962/Validation_48018962_48018962_556473.556.png", "Validation_48018962/Validation_48018962_48018962_556475.055.png" ], "turn_id": 1, "scan_id": "Validation_48018962", "origin_question": "", "option": [], "answer": "", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_48018962/Validation_48018962_48018962_556476.554.png", "Validation_48018962/Validation_48018962_48018962_556478.054.png", "Validation_48018962/Validation_48018962_48018962_556479.553.png", "Validation_48018962/Validation_48018962_48018962_556481.053.png" ], "turn_id": 2, "scan_id": "Validation_48018962", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the classic rounded white toilet made of porcelain\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the classic rounded white toilet made of porcelain\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_48018962/Validation_48018962_48018962_556482.552.png", "Validation_48018962/Validation_48018962_48018962_556484.051.png", "Validation_48018962/Validation_48018962_48018962_556485.551.png", "Validation_48018962/Validation_48018962_48018962_556487.050.png" ], "turn_id": 3, "scan_id": "Validation_48018962", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the sleek white front-loading washing machine\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the sleek white front-loading washing machine\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_48018962/Validation_48018962_48018962_556488.550.png", "Validation_48018962/Validation_48018962_48018962_556490.049.png", "Validation_48018962/Validation_48018962_48018962_556491.548.png", "Validation_48018962/Validation_48018962_48018962_556493.048.png" ], "turn_id": 4, "scan_id": "Validation_48018962", "origin_question": "", "option": [], "answer": "", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "Validation_41159525/Validation_41159525_41159525_38024.935.png", "Validation_41159525/Validation_41159525_41159525_38026.918.png", "Validation_41159525/Validation_41159525_41159525_38028.917.png", "Validation_41159525/Validation_41159525_41159525_38030.916.png", "Validation_41159525/Validation_41159525_41159525_38032.915.png" ], "turn_id": 1, "scan_id": "Validation_41159525", "origin_question": "Remember, how many oven(s) have you seen so far? ", "option": [], "answer": "2", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many oven(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_41159525/Validation_41159525_41159525_38034.914.png", "Validation_41159525/Validation_41159525_41159525_38036.914.png", "Validation_41159525/Validation_41159525_41159525_38038.913.png", "Validation_41159525/Validation_41159525_41159525_38040.912.png", "Validation_41159525/Validation_41159525_41159525_38042.911.png" ], "turn_id": 2, "scan_id": "Validation_41159525", "origin_question": "When did you first discover \"the tall stainless steel refrigerator with the black finish\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the tall stainless steel refrigerator with the black finish\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_41159525/Validation_41159525_41159525_38044.910.png", "Validation_41159525/Validation_41159525_41159525_38046.909.png", "Validation_41159525/Validation_41159525_41159525_38048.908.png", "Validation_41159525/Validation_41159525_41159525_38050.908.png", "Validation_41159525/Validation_41159525_41159525_38052.907.png" ], "turn_id": 3, "scan_id": "Validation_41159525", "origin_question": "How far is your current position from where you were at the end of the turn 1? (in meters)", "option": [], "answer": "2.53", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 1? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "Validation_41159525/Validation_41159525_41159525_38054.906.png", "Validation_41159525/Validation_41159525_41159525_38056.905.png", "Validation_41159525/Validation_41159525_41159525_38058.904.png", "Validation_41159525/Validation_41159525_41159525_38060.904.png", "Validation_41159525/Validation_41159525_41159525_38062.903.png" ], "turn_id": 4, "scan_id": "Validation_41159525", "origin_question": "Among these three objects(\"the pale blue industrial-style chair with the tall backrest\",\"the sleek, compact white or gray radiator\",\"the book\" ), which one is the closest to you?", "option": [ "the pale blue industrial-style chair with the tall backrest", "the sleek, compact white or gray radiator", "the book" ], "answer": "the book", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the pale blue industrial-style chair with the tall backrest\",\"the sleek, compact white or gray radiator\",\"the book\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the pale blue industrial-style chair with the tall backrest', 'the sleek, compact white or gray radiator', 'the book'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "Validation_41159525/Validation_41159525_41159525_38064.902.png", "Validation_41159525/Validation_41159525_41159525_38066.918.png", "Validation_41159525/Validation_41159525_41159525_38068.917.png", "Validation_41159525/Validation_41159525_41159525_38070.916.png", "Validation_41159525/Validation_41159525_41159525_38072.915.png" ], "turn_id": 5, "scan_id": "Validation_41159525", "origin_question": "Among these three objects(\"the pale blue industrial-style chair with the tall backrest\",\"the sleek metallic box with the dark-colored finish\",\"the shelf\" ), which one is the closest to you?", "option": [ "the pale blue industrial-style chair with the tall backrest", "the sleek metallic box with the dark-colored finish", "the shelf" ], "answer": "the shelf", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the pale blue industrial-style chair with the tall backrest\",\"the sleek metallic box with the dark-colored finish\",\"the shelf\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the pale blue industrial-style chair with the tall backrest', 'the sleek metallic box with the dark-colored finish', 'the shelf'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_41159525/Validation_41159525_41159525_38074.914.png", "Validation_41159525/Validation_41159525_41159525_38076.914.png", "Validation_41159525/Validation_41159525_41159525_38078.913.png", "Validation_41159525/Validation_41159525_41159525_38080.912.png", "Validation_41159525/Validation_41159525_41159525_38082.911.png" ], "turn_id": 6, "scan_id": "Validation_41159525", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_41159525/Validation_41159525_41159525_38084.910.png", "Validation_41159525/Validation_41159525_41159525_38086.909.png", "Validation_41159525/Validation_41159525_41159525_38088.908.png", "Validation_41159525/Validation_41159525_41159525_38090.908.png", "Validation_41159525/Validation_41159525_41159525_38092.907.png" ], "turn_id": 7, "scan_id": "Validation_41159525", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the dark brown wooden kitchen table with the simple design\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the dark brown wooden kitchen table with the simple design\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_41159525/Validation_41159525_41159525_38094.906.png", "Validation_41159525/Validation_41159525_41159525_38096.905.png", "Validation_41159525/Validation_41159525_41159525_38098.904.png", "Validation_41159525/Validation_41159525_41159525_38100.904.png", "Validation_41159525/Validation_41159525_41159525_38102.903.png" ], "turn_id": 8, "scan_id": "Validation_41159525", "origin_question": "Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "75.03", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_41159525/Validation_41159525_41159525_38104.902.png", "Validation_41159525/Validation_41159525_41159525_38106.901.png", "Validation_41159525/Validation_41159525_41159525_38108.900.png", "Validation_41159525/Validation_41159525_41159525_38110.916.png", "Validation_41159525/Validation_41159525_41159525_38112.915.png" ], "turn_id": 9, "scan_id": "Validation_41159525", "origin_question": "Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "97.77", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_41159525/Validation_41159525_41159525_38114.914.png", "Validation_41159525/Validation_41159525_41159525_38116.914.png", "Validation_41159525/Validation_41159525_41159525_38118.913.png", "Validation_41159525/Validation_41159525_41159525_38120.912.png", "Validation_41159525/Validation_41159525_41159525_38122.911.png" ], "turn_id": 10, "scan_id": "Validation_41159525", "origin_question": "Which direction is the deep green rectangular woven cushion in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the deep green rectangular woven cushion in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42897526/Validation_42897526_42897526_273846.408.png", "Validation_42897526/Validation_42897526_42897526_273847.891.png", "Validation_42897526/Validation_42897526_42897526_273849.407.png", "Validation_42897526/Validation_42897526_42897526_273850.906.png" ], "turn_id": 1, "scan_id": "Validation_42897526", "origin_question": "Remember, have you seen any coffee maker(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any coffee maker(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_42897526/Validation_42897526_42897526_273852.406.png", "Validation_42897526/Validation_42897526_42897526_273853.905.png", "Validation_42897526/Validation_42897526_42897526_273855.404.png", "Validation_42897526/Validation_42897526_42897526_273856.904.png" ], "turn_id": 2, "scan_id": "Validation_42897526", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_42897526/Validation_42897526_42897526_273858.403.png", "Validation_42897526/Validation_42897526_42897526_273859.903.png", "Validation_42897526/Validation_42897526_42897526_273861.402.png", "Validation_42897526/Validation_42897526_42897526_273862.901.png" ], "turn_id": 3, "scan_id": "Validation_42897526", "origin_question": "Which two objects are on the same side of you now? \"the refrigerator\", \"the sleek black glass-ceramic oven with multiple cooking zones\", \"the sleek stainless-steel sink with the single basin\". ", "option": [ "the refrigerator, the sleek black glass-ceramic oven with multiple cooking zones", "the refrigerator, the sleek stainless-steel sink with the single basin", "the sleek black glass-ceramic oven with multiple cooking zones, the sleek stainless-steel sink with the single basin" ], "answer": "the refrigerator, the sleek black glass-ceramic oven with multiple cooking zones", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the refrigerator\", \"the sleek black glass-ceramic oven with multiple cooking zones\", \"the sleek stainless-steel sink with the single basin\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the refrigerator, the sleek black glass-ceramic oven with multiple cooking zones', 'the refrigerator, the sleek stainless-steel sink with the single basin', 'the sleek black glass-ceramic oven with multiple cooking zones, the sleek stainless-steel sink with the single basin'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_42897526/Validation_42897526_42897526_273864.401.png", "Validation_42897526/Validation_42897526_42897526_273865.900.png", "Validation_42897526/Validation_42897526_42897526_273867.399.png", "Validation_42897526/Validation_42897526_42897526_273868.899.png" ], "turn_id": 4, "scan_id": "Validation_42897526", "origin_question": "Please recall \"the rectangular white window with the classic frame\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.03", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the rectangular white window with the classic frame\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_42897526/Validation_42897526_42897526_273870.398.png", "Validation_42897526/Validation_42897526_42897526_273871.898.png", "Validation_42897526/Validation_42897526_42897526_273873.397.png", "Validation_42897526/Validation_42897526_42897526_273874.896.png" ], "turn_id": 5, "scan_id": "Validation_42897526", "origin_question": "In which turn did you see both \"the refrigerator\" and \"the cylindrical stainless steel kettle with the metallic finish\" simultaneously? ", "option": [], "answer": "5", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the refrigerator\" and \"the cylindrical stainless steel kettle with the metallic finish\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_42897526/Validation_42897526_42897526_273876.396.png", "Validation_42897526/Validation_42897526_42897526_273877.895.png", "Validation_42897526/Validation_42897526_42897526_273879.394.png", "Validation_42897526/Validation_42897526_42897526_273880.894.png" ], "turn_id": 6, "scan_id": "Validation_42897526", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the refrigerator\", Y: \"the sleek stainless-steel sink with the single basin\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the refrigerator\", Y: \"the sleek stainless-steel sink with the single basin\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_42897526/Validation_42897526_42897526_273882.393.png", "Validation_42897526/Validation_42897526_42897526_273883.893.png", "Validation_42897526/Validation_42897526_42897526_273885.392.png", "Validation_42897526/Validation_42897526_42897526_273886.891.png" ], "turn_id": 7, "scan_id": "Validation_42897526", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the moderately-sized mirror with the warm brown frame\" from your position?", "option": [], "answer": "176.05", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the moderately-sized mirror with the warm brown frame\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47430047/Validation_47430047_47430047_21091.855.png", "Validation_47430047/Validation_47430047_47430047_21093.338.png", "Validation_47430047/Validation_47430047_47430047_21094.837.png", "Validation_47430047/Validation_47430047_47430047_21096.337.png" ], "turn_id": 1, "scan_id": "Validation_47430047", "origin_question": "Remember, have you seen any hamper(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any hamper(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_47430047/Validation_47430047_47430047_21097.836.png", "Validation_47430047/Validation_47430047_47430047_21099.335.png", "Validation_47430047/Validation_47430047_47430047_21100.835.png", "Validation_47430047/Validation_47430047_47430047_21102.334.png" ], "turn_id": 2, "scan_id": "Validation_47430047", "origin_question": "When did you first discover \"the sleek rectangular window with the white frame\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the sleek rectangular window with the white frame\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_47430047/Validation_47430047_47430047_21103.834.png", "Validation_47430047/Validation_47430047_47430047_21105.333.png", "Validation_47430047/Validation_47430047_47430047_21106.832.png", "Validation_47430047/Validation_47430047_47430047_21108.332.png" ], "turn_id": 3, "scan_id": "Validation_47430047", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "1.1", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_47430047/Validation_47430047_47430047_21109.831.png", "Validation_47430047/Validation_47430047_47430047_21111.331.png", "Validation_47430047/Validation_47430047_47430047_21112.830.png", "Validation_47430047/Validation_47430047_47430047_21114.329.png" ], "turn_id": 4, "scan_id": "Validation_47430047", "origin_question": "Please recall \"the medium-sized oval white ceramic sink\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.75", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the medium-sized oval white ceramic sink\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_47430047/Validation_47430047_47430047_21115.829.png", "Validation_47430047/Validation_47430047_47430047_21117.328.png", "Validation_47430047/Validation_47430047_47430047_21118.827.png", "Validation_47430047/Validation_47430047_47430047_21120.327.png" ], "turn_id": 5, "scan_id": "Validation_47430047", "origin_question": "Is the small cylindrical white plastic bin to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the small cylindrical white plastic bin to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_41254412/Validation_41254412_41254412_95737.701.png", "Validation_41254412/Validation_41254412_41254412_95739.683.png", "Validation_41254412/Validation_41254412_41254412_95741.683.png", "Validation_41254412/Validation_41254412_41254412_95743.682.png" ], "turn_id": 1, "scan_id": "Validation_41254412", "origin_question": "Remember, have you seen any book(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any book(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_41254412/Validation_41254412_41254412_95745.681.png", "Validation_41254412/Validation_41254412_41254412_95747.680.png", "Validation_41254412/Validation_41254412_41254412_95749.679.png", "Validation_41254412/Validation_41254412_41254412_95751.695.png" ], "turn_id": 2, "scan_id": "Validation_41254412", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the bed\", Y: \"the white, soft cotton pillow\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the bed\", Y: \"the white, soft cotton pillow\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_41254412/Validation_41254412_41254412_95753.694.png", "Validation_41254412/Validation_41254412_41254412_95755.694.png", "Validation_41254412/Validation_41254412_41254412_95757.693.png", "Validation_41254412/Validation_41254412_41254412_95759.692.png" ], "turn_id": 3, "scan_id": "Validation_41254412", "origin_question": "In which turn were you farthest from \"the white, soft cotton pillow\"? ", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the white, soft cotton pillow\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_41254412/Validation_41254412_41254412_95761.691.png", "Validation_41254412/Validation_41254412_41254412_95763.690.png", "Validation_41254412/Validation_41254412_41254412_95765.690.png", "Validation_41254412/Validation_41254412_41254412_95767.689.png" ], "turn_id": 4, "scan_id": "Validation_41254412", "origin_question": "Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the bed\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the bed\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_41254412/Validation_41254412_41254412_95769.688.png", "Validation_41254412/Validation_41254412_41254412_95771.687.png", "Validation_41254412/Validation_41254412_41254412_95773.687.png", "Validation_41254412/Validation_41254412_41254412_95775.686.png" ], "turn_id": 5, "scan_id": "Validation_41254412", "origin_question": "By the end of the 3 turn, were you getting closer to or farther away from these two objects: X: \"the bed\", Y: \"the sleek silver-gray metal table with the vertical design\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "farther away from both X and Y", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 3 turn, were you getting closer to or farther away from these two objects: X: \"the bed\", Y: \"the sleek silver-gray metal table with the vertical design\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_41254412/Validation_41254412_41254412_95777.685.png", "Validation_41254412/Validation_41254412_41254412_95779.684.png", "Validation_41254412/Validation_41254412_41254412_95781.683.png", "Validation_41254412/Validation_41254412_41254412_95783.683.png" ], "turn_id": 6, "scan_id": "Validation_41254412", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the sleek silver-gray metal table with the vertical design\" from your position?", "option": [], "answer": "96.94", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the sleek silver-gray metal table with the vertical design\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42897672/Validation_42897672_42897672_479336.485.png", "Validation_42897672/Validation_42897672_42897672_479338.484.png", "Validation_42897672/Validation_42897672_42897672_479340.483.png", "Validation_42897672/Validation_42897672_42897672_479342.482.png", "Validation_42897672/Validation_42897672_42897672_479344.482.png" ], "turn_id": 1, "scan_id": "Validation_42897672", "origin_question": "Remember, have you seen any bed(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any bed(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_42897672/Validation_42897672_42897672_479346.481.png", "Validation_42897672/Validation_42897672_42897672_479348.480.png", "Validation_42897672/Validation_42897672_42897672_479350.479.png", "Validation_42897672/Validation_42897672_42897672_479352.478.png", "Validation_42897672/Validation_42897672_42897672_479354.477.png" ], "turn_id": 2, "scan_id": "Validation_42897672", "origin_question": "At the end of which turn were both of \"the rectangular window framed in reddish-brown\" and \"the sleek white steel radiator\" on your left side?", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the rectangular window framed in reddish-brown\" and \"the sleek white steel radiator\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42897672/Validation_42897672_42897672_479356.477.png", "Validation_42897672/Validation_42897672_42897672_479358.476.png", "Validation_42897672/Validation_42897672_42897672_479360.475.png", "Validation_42897672/Validation_42897672_42897672_479362.474.png", "Validation_42897672/Validation_42897672_42897672_479364.490.png" ], "turn_id": 3, "scan_id": "Validation_42897672", "origin_question": "Remember, have you seen any basket(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any basket(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_42897672/Validation_42897672_42897672_479366.489.png", "Validation_42897672/Validation_42897672_42897672_479368.488.png", "Validation_42897672/Validation_42897672_42897672_479370.487.png", "Validation_42897672/Validation_42897672_42897672_479372.487.png", "Validation_42897672/Validation_42897672_42897672_479374.486.png" ], "turn_id": 4, "scan_id": "Validation_42897672", "origin_question": "Is the rectangular window framed in reddish-brown to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the rectangular window framed in reddish-brown to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_42897672/Validation_42897672_42897672_479376.485.png", "Validation_42897672/Validation_42897672_42897672_479378.484.png", "Validation_42897672/Validation_42897672_42897672_479380.483.png", "Validation_42897672/Validation_42897672_42897672_479382.483.png", "Validation_42897672/Validation_42897672_42897672_479384.482.png" ], "turn_id": 5, "scan_id": "Validation_42897672", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the rectangular window framed in reddish-brown\" from your position?", "option": [], "answer": "148.66", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the rectangular window framed in reddish-brown\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_42897672/Validation_42897672_42897672_479386.481.png", "Validation_42897672/Validation_42897672_42897672_479388.480.png", "Validation_42897672/Validation_42897672_42897672_479390.479.png", "Validation_42897672/Validation_42897672_42897672_479392.479.png", "Validation_42897672/Validation_42897672_42897672_479394.478.png" ], "turn_id": 6, "scan_id": "Validation_42897672", "origin_question": "", "option": [], "answer": "", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_42897672/Validation_42897672_42897672_479396.477.png", "Validation_42897672/Validation_42897672_42897672_479398.476.png", "Validation_42897672/Validation_42897672_42897672_479400.475.png", "Validation_42897672/Validation_42897672_42897672_479402.474.png", "Validation_42897672/Validation_42897672_42897672_479404.474.png" ], "turn_id": 7, "scan_id": "Validation_42897672", "origin_question": "Is the vibrant patterned stool with geometric upholstery to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the vibrant patterned stool with geometric upholstery to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_42897672/Validation_42897672_42897672_479406.473.png", "Validation_42897672/Validation_42897672_42897672_479408.488.png", "Validation_42897672/Validation_42897672_42897672_479410.488.png", "Validation_42897672/Validation_42897672_42897672_479412.487.png", "Validation_42897672/Validation_42897672_42897672_479414.486.png" ], "turn_id": 8, "scan_id": "Validation_42897672", "origin_question": "How far is your current position from where you were at the end of the turn 7? (in meters)", "option": [], "answer": "1.18", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 7? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_42897672/Validation_42897672_42897672_479416.485.png", "Validation_42897672/Validation_42897672_42897672_479418.484.png", "Validation_42897672/Validation_42897672_42897672_479420.483.png", "Validation_42897672/Validation_42897672_42897672_479422.483.png", "Validation_42897672/Validation_42897672_42897672_479424.482.png" ], "turn_id": 9, "scan_id": "Validation_42897672", "origin_question": "When did you first discover \"the rectangular window framed in reddish-brown\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the rectangular window framed in reddish-brown\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_42897672/Validation_42897672_42897672_479426.481.png", "Validation_42897672/Validation_42897672_42897672_479428.480.png", "Validation_42897672/Validation_42897672_42897672_479430.479.png", "Validation_42897672/Validation_42897672_42897672_479432.479.png", "Validation_42897672/Validation_42897672_42897672_479434.478.png" ], "turn_id": 10, "scan_id": "Validation_42897672", "origin_question": "When did you first discover \"the vibrant patterned stool with geometric upholstery\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the vibrant patterned stool with geometric upholstery\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42446104/Validation_42446104_42446104_349793.388.png", "Validation_42446104/Validation_42446104_42446104_349795.371.png", "Validation_42446104/Validation_42446104_42446104_349797.370.png", "Validation_42446104/Validation_42446104_42446104_349799.369.png", "Validation_42446104/Validation_42446104_42446104_349801.368.png" ], "turn_id": 1, "scan_id": "Validation_42446104", "origin_question": "Remember, have you seen any chandelier(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any chandelier(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_42446104/Validation_42446104_42446104_349803.367.png", "Validation_42446104/Validation_42446104_42446104_349805.367.png", "Validation_42446104/Validation_42446104_42446104_349807.366.png", "Validation_42446104/Validation_42446104_42446104_349809.365.png", "Validation_42446104/Validation_42446104_42446104_349811.364.png" ], "turn_id": 2, "scan_id": "Validation_42446104", "origin_question": "In which turn did you see both \"the modern cylindrical fabric chandelier in neutral colors\" and \"the vibrant green and red cylindrical lamp\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the modern cylindrical fabric chandelier in neutral colors\" and \"the vibrant green and red cylindrical lamp\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_42446104/Validation_42446104_42446104_349813.363.png", "Validation_42446104/Validation_42446104_42446104_349815.363.png", "Validation_42446104/Validation_42446104_42446104_349817.362.png", "Validation_42446104/Validation_42446104_42446104_349819.378.png", "Validation_42446104/Validation_42446104_42446104_349821.377.png" ], "turn_id": 3, "scan_id": "Validation_42446104", "origin_question": "Which direction is the modern cylindrical fabric chandelier in neutral colors in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the modern cylindrical fabric chandelier in neutral colors in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_42446104/Validation_42446104_42446104_349823.376.png", "Validation_42446104/Validation_42446104_42446104_349825.375.png", "Validation_42446104/Validation_42446104_42446104_349827.374.png", "Validation_42446104/Validation_42446104_42446104_349829.373.png", "Validation_42446104/Validation_42446104_42446104_349831.373.png" ], "turn_id": 4, "scan_id": "Validation_42446104", "origin_question": "Is the vibrant green and red cylindrical lamp to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the vibrant green and red cylindrical lamp to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_42446104/Validation_42446104_42446104_349833.372.png", "Validation_42446104/Validation_42446104_42446104_349835.371.png", "Validation_42446104/Validation_42446104_42446104_349837.370.png", "Validation_42446104/Validation_42446104_42446104_349839.369.png", "Validation_42446104/Validation_42446104_42446104_349841.369.png" ], "turn_id": 5, "scan_id": "Validation_42446104", "origin_question": "Which direction is the sleek white rectangular door panel in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the sleek white rectangular door panel in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_42446104/Validation_42446104_42446104_349843.368.png", "Validation_42446104/Validation_42446104_42446104_349845.367.png", "Validation_42446104/Validation_42446104_42446104_349847.366.png", "Validation_42446104/Validation_42446104_42446104_349849.365.png", "Validation_42446104/Validation_42446104_42446104_349851.365.png" ], "turn_id": 6, "scan_id": "Validation_42446104", "origin_question": "At the end of which turn were both of \"the bed\" and \"the sleek white rectangular door panel\" on your right side?", "option": [], "answer": "5", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the bed\" and \"the sleek white rectangular door panel\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_42446104/Validation_42446104_42446104_349853.364.png", "Validation_42446104/Validation_42446104_42446104_349855.363.png", "Validation_42446104/Validation_42446104_42446104_349857.362.png", "Validation_42446104/Validation_42446104_42446104_349859.361.png", "Validation_42446104/Validation_42446104_42446104_349861.377.png" ], "turn_id": 7, "scan_id": "Validation_42446104", "origin_question": "In which turn were you closest to \"the bed\"? ", "option": [], "answer": "4", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the bed\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_42446104/Validation_42446104_42446104_349863.376.png", "Validation_42446104/Validation_42446104_42446104_349865.376.png", "Validation_42446104/Validation_42446104_42446104_349867.375.png", "Validation_42446104/Validation_42446104_42446104_349869.374.png", "Validation_42446104/Validation_42446104_42446104_349871.373.png" ], "turn_id": 8, "scan_id": "Validation_42446104", "origin_question": "Please recall \"the vibrant green and red cylindrical lamp\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.59", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the vibrant green and red cylindrical lamp\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_42446104/Validation_42446104_42446104_349873.372.png", "Validation_42446104/Validation_42446104_42446104_349875.371.png", "Validation_42446104/Validation_42446104_42446104_349877.371.png", "Validation_42446104/Validation_42446104_42446104_349879.370.png", "Validation_42446104/Validation_42446104_42446104_349881.369.png" ], "turn_id": 9, "scan_id": "Validation_42446104", "origin_question": "Using your orientation at the end of turn 8 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 8 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_42446104/Validation_42446104_42446104_349883.368.png", "Validation_42446104/Validation_42446104_42446104_349885.367.png", "Validation_42446104/Validation_42446104_42446104_349887.367.png", "Validation_42446104/Validation_42446104_42446104_349889.366.png", "Validation_42446104/Validation_42446104_42446104_349891.365.png" ], "turn_id": 10, "scan_id": "Validation_42446104", "origin_question": "In which turn did you see both \"the sleek white rectangular door panel\" and \"the modern cylindrical fabric chandelier in neutral colors\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the sleek white rectangular door panel\" and \"the modern cylindrical fabric chandelier in neutral colors\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "Validation_47332890/Validation_47332890_47332890_7597.386.png", "Validation_47332890/Validation_47332890_47332890_7599.369.png", "Validation_47332890/Validation_47332890_47332890_7601.368.png", "Validation_47332890/Validation_47332890_47332890_7603.367.png", "Validation_47332890/Validation_47332890_47332890_7605.367.png" ], "turn_id": 1, "scan_id": "Validation_47332890", "origin_question": "Remember, how many water cooler(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many water cooler(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_47332890/Validation_47332890_47332890_7607.366.png", "Validation_47332890/Validation_47332890_47332890_7609.365.png", "Validation_47332890/Validation_47332890_47332890_7611.364.png", "Validation_47332890/Validation_47332890_47332890_7613.363.png", "Validation_47332890/Validation_47332890_47332890_7615.363.png" ], "turn_id": 2, "scan_id": "Validation_47332890", "origin_question": "Is the vibrant orange circular plastic basin to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the vibrant orange circular plastic basin to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_47332890/Validation_47332890_47332890_7617.362.png", "Validation_47332890/Validation_47332890_47332890_7619.361.png", "Validation_47332890/Validation_47332890_47332890_7621.360.png", "Validation_47332890/Validation_47332890_47332890_7623.359.png", "Validation_47332890/Validation_47332890_47332890_7625.358.png" ], "turn_id": 3, "scan_id": "Validation_47332890", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the light wood rectangular cabinet with closed doors\", Y: \"the soft pink, durable synthetic backpack\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the light wood rectangular cabinet with closed doors\", Y: \"the soft pink, durable synthetic backpack\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_47332890/Validation_47332890_47332890_7627.358.png", "Validation_47332890/Validation_47332890_47332890_7629.357.png", "Validation_47332890/Validation_47332890_47332890_7631.356.png", "Validation_47332890/Validation_47332890_47332890_7633.372.png", "Validation_47332890/Validation_47332890_47332890_7635.371.png" ], "turn_id": 4, "scan_id": "Validation_47332890", "origin_question": "Which two objects are on the same side of you now? \"the light wood rectangular cabinet with closed doors\", \"the soft pink, durable synthetic backpack\", \"the vibrant orange circular plastic basin\". ", "option": [ "the light wood rectangular cabinet with closed doors, the soft pink, durable synthetic backpack", "the light wood rectangular cabinet with closed doors, the vibrant orange circular plastic basin", "the soft pink, durable synthetic backpack, the vibrant orange circular plastic basin" ], "answer": "the light wood rectangular cabinet with closed doors, the soft pink, durable synthetic backpack", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the light wood rectangular cabinet with closed doors\", \"the soft pink, durable synthetic backpack\", \"the vibrant orange circular plastic basin\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the light wood rectangular cabinet with closed doors, the soft pink, durable synthetic backpack', 'the light wood rectangular cabinet with closed doors, the vibrant orange circular plastic basin', 'the soft pink, durable synthetic backpack, the vibrant orange circular plastic basin'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_47332890/Validation_47332890_47332890_7637.370.png", "Validation_47332890/Validation_47332890_47332890_7639.369.png", "Validation_47332890/Validation_47332890_47332890_7641.369.png", "Validation_47332890/Validation_47332890_47332890_7643.368.png", "Validation_47332890/Validation_47332890_47332890_7645.367.png" ], "turn_id": 5, "scan_id": "Validation_47332890", "origin_question": "Which two objects are on the same side of you now? \"the soft pink, durable synthetic backpack\", \"the medium-sized rectangular light brown table with drawers\", \"the vibrant orange circular plastic basin\". ", "option": [ "the soft pink, durable synthetic backpack, the medium-sized rectangular light brown table with drawers", "the soft pink, durable synthetic backpack, the vibrant orange circular plastic basin", "the medium-sized rectangular light brown table with drawers, the vibrant orange circular plastic basin" ], "answer": "the medium-sized rectangular light brown table with drawers, the vibrant orange circular plastic basin", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the soft pink, durable synthetic backpack\", \"the medium-sized rectangular light brown table with drawers\", \"the vibrant orange circular plastic basin\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the soft pink, durable synthetic backpack, the medium-sized rectangular light brown table with drawers', 'the soft pink, durable synthetic backpack, the vibrant orange circular plastic basin', 'the medium-sized rectangular light brown table with drawers, the vibrant orange circular plastic basin'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47332890/Validation_47332890_47332890_7647.366.png", "Validation_47332890/Validation_47332890_47332890_7649.365.png", "Validation_47332890/Validation_47332890_47332890_7651.364.png", "Validation_47332890/Validation_47332890_47332890_7653.364.png", "Validation_47332890/Validation_47332890_47332890_7655.363.png" ], "turn_id": 6, "scan_id": "Validation_47332890", "origin_question": "In which turn were you closest to \"the blanket\"? ", "option": [], "answer": "3", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the blanket\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_47332890/Validation_47332890_47332890_7657.362.png", "Validation_47332890/Validation_47332890_47332890_7659.361.png", "Validation_47332890/Validation_47332890_47332890_7661.360.png", "Validation_47332890/Validation_47332890_47332890_7663.360.png", "Validation_47332890/Validation_47332890_47332890_7665.359.png" ], "turn_id": 7, "scan_id": "Validation_47332890", "origin_question": "What will be the first-time appearance order of the following categories: bed, blanket, clothes?", "option": [ "clothes, bed, blanket", "blanket, bed, clothes", "bed, blanket, clothes", "clothes, blanket, bed" ], "answer": "bed, blanket, clothes", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bed, blanket, clothes? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['clothes, bed, blanket', 'blanket, bed, clothes', 'bed, blanket, clothes', 'clothes, blanket, bed'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_47332890/Validation_47332890_47332890_7667.358.png", "Validation_47332890/Validation_47332890_47332890_7669.357.png", "Validation_47332890/Validation_47332890_47332890_7671.356.png", "Validation_47332890/Validation_47332890_47332890_7673.356.png", "Validation_47332890/Validation_47332890_47332890_7675.371.png" ], "turn_id": 8, "scan_id": "Validation_47332890", "origin_question": "Which two objects are on the same side of you now? \"the vibrant orange circular plastic basin\", \"the sleek silver monitor with the matte finish\", \"the framed picture with muted colors and the rectangular shape\". ", "option": [ "the vibrant orange circular plastic basin, the sleek silver monitor with the matte finish", "the vibrant orange circular plastic basin, the framed picture with muted colors and the rectangular shape", "the sleek silver monitor with the matte finish, the framed picture with muted colors and the rectangular shape" ], "answer": "the vibrant orange circular plastic basin, the sleek silver monitor with the matte finish", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the vibrant orange circular plastic basin\", \"the sleek silver monitor with the matte finish\", \"the framed picture with muted colors and the rectangular shape\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the vibrant orange circular plastic basin, the sleek silver monitor with the matte finish', 'the vibrant orange circular plastic basin, the framed picture with muted colors and the rectangular shape', 'the sleek silver monitor with the matte finish, the framed picture with muted colors and the rectangular shape'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47332890/Validation_47332890_47332890_7677.371.png", "Validation_47332890/Validation_47332890_47332890_7679.370.png", "Validation_47332890/Validation_47332890_47332890_7681.369.png", "Validation_47332890/Validation_47332890_47332890_7683.368.png", "Validation_47332890/Validation_47332890_47332890_7685.367.png" ], "turn_id": 9, "scan_id": "Validation_47332890", "origin_question": "Which direction is the blanket in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the blanket in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47332890/Validation_47332890_47332890_7687.367.png", "Validation_47332890/Validation_47332890_47332890_7689.366.png", "Validation_47332890/Validation_47332890_47332890_7691.365.png", "Validation_47332890/Validation_47332890_47332890_7693.364.png", "Validation_47332890/Validation_47332890_47332890_7695.363.png" ], "turn_id": 10, "scan_id": "Validation_47332890", "origin_question": "Compared to your position at the end of the 6 turn, are you now closer or farther away from \"the blanket\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 6 turn, are you now closer or farther away from \"the blanket\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47334254/Validation_47334254_47334254_111908.912.png", "Validation_47334254/Validation_47334254_47334254_111910.894.png", "Validation_47334254/Validation_47334254_47334254_111912.894.png", "Validation_47334254/Validation_47334254_47334254_111914.893.png", "Validation_47334254/Validation_47334254_47334254_111916.892.png" ], "turn_id": 1, "scan_id": "Validation_47334254", "origin_question": "Remember, have you seen any chair(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any chair(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_47334254/Validation_47334254_47334254_111918.891.png", "Validation_47334254/Validation_47334254_47334254_111920.890.png", "Validation_47334254/Validation_47334254_47334254_111922.890.png", "Validation_47334254/Validation_47334254_47334254_111924.889.png", "Validation_47334254/Validation_47334254_47334254_111926.888.png" ], "turn_id": 2, "scan_id": "Validation_47334254", "origin_question": "When did you first discover \"the sleek modern oven with the glass front panel\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the sleek modern oven with the glass front panel\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_47334254/Validation_47334254_47334254_111928.887.png", "Validation_47334254/Validation_47334254_47334254_111930.886.png", "Validation_47334254/Validation_47334254_47334254_111932.886.png", "Validation_47334254/Validation_47334254_47334254_111934.885.png", "Validation_47334254/Validation_47334254_47334254_111936.884.png" ], "turn_id": 3, "scan_id": "Validation_47334254", "origin_question": "At the end of which turn were both of \"the range hood\" and \"the sleek rectangular stainless steel stove with black burners\" on your right side?", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the range hood\" and \"the sleek rectangular stainless steel stove with black burners\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_47334254/Validation_47334254_47334254_111938.883.png", "Validation_47334254/Validation_47334254_47334254_111940.899.png", "Validation_47334254/Validation_47334254_47334254_111942.898.png", "Validation_47334254/Validation_47334254_47334254_111944.898.png", "Validation_47334254/Validation_47334254_47334254_111946.897.png" ], "turn_id": 4, "scan_id": "Validation_47334254", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "107.07", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47334254/Validation_47334254_47334254_111948.896.png", "Validation_47334254/Validation_47334254_47334254_111950.895.png", "Validation_47334254/Validation_47334254_47334254_111952.894.png", "Validation_47334254/Validation_47334254_47334254_111954.894.png", "Validation_47334254/Validation_47334254_47334254_111956.893.png" ], "turn_id": 5, "scan_id": "Validation_47334254", "origin_question": "Which direction is the range hood in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the range hood in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_47334254/Validation_47334254_47334254_111958.892.png", "Validation_47334254/Validation_47334254_47334254_111960.891.png", "Validation_47334254/Validation_47334254_47334254_111962.890.png", "Validation_47334254/Validation_47334254_47334254_111964.889.png", "Validation_47334254/Validation_47334254_47334254_111966.889.png" ], "turn_id": 6, "scan_id": "Validation_47334254", "origin_question": "Is the sleek rectangular stainless steel stove with black burners to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the sleek rectangular stainless steel stove with black burners to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47334254/Validation_47334254_47334254_111968.888.png", "Validation_47334254/Validation_47334254_47334254_111970.887.png", "Validation_47334254/Validation_47334254_47334254_111972.886.png", "Validation_47334254/Validation_47334254_47334254_111974.885.png", "Validation_47334254/Validation_47334254_47334254_111976.885.png" ], "turn_id": 7, "scan_id": "Validation_47334254", "origin_question": "Which direction is the medium-sized oval stainless steel sink in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the medium-sized oval stainless steel sink in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_47334254/Validation_47334254_47334254_111978.884.png", "Validation_47334254/Validation_47334254_47334254_111980.883.png", "Validation_47334254/Validation_47334254_47334254_111982.899.png", "Validation_47334254/Validation_47334254_47334254_111984.898.png", "Validation_47334254/Validation_47334254_47334254_111986.897.png" ], "turn_id": 8, "scan_id": "Validation_47334254", "origin_question": "At the end of which turn were both of \"the sleek rectangular stainless steel stove with black burners\" and \"the shelf\" on your right side?", "option": [], "answer": "5", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the sleek rectangular stainless steel stove with black burners\" and \"the shelf\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42446038/Validation_42446038_42446038_347859.978.png", "Validation_42446038/Validation_42446038_42446038_347861.977.png", "Validation_42446038/Validation_42446038_42446038_347863.976.png", "Validation_42446038/Validation_42446038_42446038_347865.975.png", "Validation_42446038/Validation_42446038_42446038_347867.975.png" ], "turn_id": 1, "scan_id": "Validation_42446038", "origin_question": "Remember, have you seen any camera(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any camera(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_42446038/Validation_42446038_42446038_347869.974.png", "Validation_42446038/Validation_42446038_42446038_347871.973.png", "Validation_42446038/Validation_42446038_42446038_347873.972.png", "Validation_42446038/Validation_42446038_42446038_347875.971.png", "Validation_42446038/Validation_42446038_42446038_347877.970.png" ], "turn_id": 2, "scan_id": "Validation_42446038", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the stainless steel built-in gas stove with four burners\", Y: \"the washing machine\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the stainless steel built-in gas stove with four burners\", Y: \"the washing machine\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_42446038/Validation_42446038_42446038_347879.970.png", "Validation_42446038/Validation_42446038_42446038_347881.969.png", "Validation_42446038/Validation_42446038_42446038_347883.985.png", "Validation_42446038/Validation_42446038_42446038_347885.984.png", "Validation_42446038/Validation_42446038_42446038_347887.983.png" ], "turn_id": 3, "scan_id": "Validation_42446038", "origin_question": "Please recall \"the medium-sized stainless steel cooking pan with the glass lid\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.97", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the medium-sized stainless steel cooking pan with the glass lid\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_42446038/Validation_42446038_42446038_347889.982.png", "Validation_42446038/Validation_42446038_42446038_347891.981.png", "Validation_42446038/Validation_42446038_42446038_347893.981.png", "Validation_42446038/Validation_42446038_42446038_347895.980.png", "Validation_42446038/Validation_42446038_42446038_347897.979.png" ], "turn_id": 4, "scan_id": "Validation_42446038", "origin_question": "In which turn did you see both \"the washing machine\" and \"the medium-sized stainless steel cooking pan with the glass lid\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the washing machine\" and \"the medium-sized stainless steel cooking pan with the glass lid\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_42446038/Validation_42446038_42446038_347899.978.png", "Validation_42446038/Validation_42446038_42446038_347901.977.png", "Validation_42446038/Validation_42446038_42446038_347903.976.png", "Validation_42446038/Validation_42446038_42446038_347905.976.png", "Validation_42446038/Validation_42446038_42446038_347907.975.png" ], "turn_id": 5, "scan_id": "Validation_42446038", "origin_question": "In which turn did you see both \"the washing machine\" and \"the large rectangular window with white upvc frames\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the washing machine\" and \"the large rectangular window with white upvc frames\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_42446038/Validation_42446038_42446038_347909.974.png", "Validation_42446038/Validation_42446038_42446038_347911.973.png", "Validation_42446038/Validation_42446038_42446038_347913.972.png", "Validation_42446038/Validation_42446038_42446038_347915.971.png", "Validation_42446038/Validation_42446038_42446038_347917.971.png" ], "turn_id": 6, "scan_id": "Validation_42446038", "origin_question": "How far is your current position from where you were at the end of the turn 1? (in meters)", "option": [], "answer": "3.14", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 1? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_42446038/Validation_42446038_42446038_347919.970.png", "Validation_42446038/Validation_42446038_42446038_347921.969.png", "Validation_42446038/Validation_42446038_42446038_347923.968.png", "Validation_42446038/Validation_42446038_42446038_347925.967.png", "Validation_42446038/Validation_42446038_42446038_347927.983.png" ], "turn_id": 7, "scan_id": "Validation_42446038", "origin_question": "In which turn did you see both \"the oven\" and \"the washing machine\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the oven\" and \"the washing machine\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_42446038/Validation_42446038_42446038_347929.982.png", "Validation_42446038/Validation_42446038_42446038_347931.982.png", "Validation_42446038/Validation_42446038_42446038_347933.981.png", "Validation_42446038/Validation_42446038_42446038_347935.980.png", "Validation_42446038/Validation_42446038_42446038_347937.979.png" ], "turn_id": 8, "scan_id": "Validation_42446038", "origin_question": "Which two objects are on the same side of you now? \"the oven\", \"the refrigerator\", \"the large rectangular window with white upvc frames\". ", "option": [ "the oven, the refrigerator", "the oven, the large rectangular window with white upvc frames", "the refrigerator, the large rectangular window with white upvc frames" ], "answer": "the oven, the refrigerator", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the oven\", \"the refrigerator\", \"the large rectangular window with white upvc frames\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the oven, the refrigerator', 'the oven, the large rectangular window with white upvc frames', 'the refrigerator, the large rectangular window with white upvc frames'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "Validation_42446038/Validation_42446038_42446038_347939.978.png", "Validation_42446038/Validation_42446038_42446038_347941.977.png", "Validation_42446038/Validation_42446038_42446038_347943.977.png", "Validation_42446038/Validation_42446038_42446038_347945.976.png", "Validation_42446038/Validation_42446038_42446038_347947.975.png" ], "turn_id": 9, "scan_id": "Validation_42446038", "origin_question": "Among these three objects(\"the oven\",\"the medium-sized stainless steel cooking pan with the glass lid\",\"the large rectangular window with white upvc frames\" ), which one is the farthest from you?", "option": [ "the oven", "the medium-sized stainless steel cooking pan with the glass lid", "the large rectangular window with white upvc frames" ], "answer": "the large rectangular window with white upvc frames", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the oven\",\"the medium-sized stainless steel cooking pan with the glass lid\",\"the large rectangular window with white upvc frames\" ), which one is the farthest from you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the oven', 'the medium-sized stainless steel cooking pan with the glass lid', 'the large rectangular window with white upvc frames'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_42446038/Validation_42446038_42446038_347949.974.png", "Validation_42446038/Validation_42446038_42446038_347951.973.png", "Validation_42446038/Validation_42446038_42446038_347953.973.png", "Validation_42446038/Validation_42446038_42446038_347955.972.png", "Validation_42446038/Validation_42446038_42446038_347957.971.png" ], "turn_id": 10, "scan_id": "Validation_42446038", "origin_question": "How far is your current position from where you were at the end of the turn 7? (in meters)", "option": [], "answer": "1.7", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 7? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47334372/Validation_47334372_47334372_63087.013.png", "Validation_47334372/Validation_47334372_47334372_63088.995.png", "Validation_47334372/Validation_47334372_47334372_63090.995.png", "Validation_47334372/Validation_47334372_47334372_63092.994.png", "Validation_47334372/Validation_47334372_47334372_63094.993.png" ], "turn_id": 1, "scan_id": "Validation_47334372", "origin_question": "Remember, have you seen any clothes(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any clothes(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47334372/Validation_47334372_47334372_63096.992.png", "Validation_47334372/Validation_47334372_47334372_63098.991.png", "Validation_47334372/Validation_47334372_47334372_63100.991.png", "Validation_47334372/Validation_47334372_47334372_63102.990.png", "Validation_47334372/Validation_47334372_47334372_63104.989.png" ], "turn_id": 2, "scan_id": "Validation_47334372", "origin_question": "In which turn were you closest to \"the sleek white washing machine\"? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the sleek white washing machine\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "Validation_47334372/Validation_47334372_47334372_63106.988.png", "Validation_47334372/Validation_47334372_47334372_63108.987.png", "Validation_47334372/Validation_47334372_47334372_63110.986.png", "Validation_47334372/Validation_47334372_47334372_63112.986.png", "Validation_47334372/Validation_47334372_47334372_63114.985.png" ], "turn_id": 3, "scan_id": "Validation_47334372", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the compact white cabinet with the sleek design\"; \"the clothes\"; \"the sleek white washing machine\" ", "option": [ "the compact white cabinet with the sleek design", "the clothes", "the sleek white washing machine" ], "answer": "the compact white cabinet with the sleek design", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the compact white cabinet with the sleek design\"; \"the clothes\"; \"the sleek white washing machine\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the compact white cabinet with the sleek design', 'the clothes', 'the sleek white washing machine'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_47334372/Validation_47334372_47334372_63116.984.png", "Validation_47334372/Validation_47334372_47334372_63118.983.png", "Validation_47334372/Validation_47334372_47334372_63120.982.png", "Validation_47334372/Validation_47334372_47334372_63122.998.png", "Validation_47334372/Validation_47334372_47334372_63124.997.png" ], "turn_id": 4, "scan_id": "Validation_47334372", "origin_question": "What will be the first-time appearance order of the following categories: bathtub, clothes, washing machine?", "option": [ "clothes, bathtub, washing machine", "washing machine, bathtub, clothes", "bathtub, washing machine, clothes", "clothes, washing machine, bathtub" ], "answer": "clothes, washing machine, bathtub", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bathtub, clothes, washing machine? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['clothes, bathtub, washing machine', 'washing machine, bathtub, clothes', 'bathtub, washing machine, clothes', 'clothes, washing machine, bathtub'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_47334372/Validation_47334372_47334372_63126.997.png", "Validation_47334372/Validation_47334372_47334372_63128.996.png", "Validation_47334372/Validation_47334372_47334372_63130.995.png", "Validation_47334372/Validation_47334372_47334372_63132.994.png", "Validation_47334372/Validation_47334372_47334372_63134.993.png" ], "turn_id": 5, "scan_id": "Validation_47334372", "origin_question": "Please recall \"the clothes\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.77", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the clothes\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "Validation_47334372/Validation_47334372_47334372_63136.993.png", "Validation_47334372/Validation_47334372_47334372_63138.992.png", "Validation_47334372/Validation_47334372_47334372_63140.991.png", "Validation_47334372/Validation_47334372_47334372_63142.990.png", "Validation_47334372/Validation_47334372_47334372_63144.989.png" ], "turn_id": 6, "scan_id": "Validation_47334372", "origin_question": "Among these three objects(\"the clothes\",\"the sleek white washing machine\",\"the bright orange round basin with smooth edges\" ), which one is the farthest from you?", "option": [ "the clothes", "the sleek white washing machine", "the bright orange round basin with smooth edges" ], "answer": "the bright orange round basin with smooth edges", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the clothes\",\"the sleek white washing machine\",\"the bright orange round basin with smooth edges\" ), which one is the farthest from you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the clothes', 'the sleek white washing machine', 'the bright orange round basin with smooth edges'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "Validation_47334372/Validation_47334372_47334372_63146.989.png", "Validation_47334372/Validation_47334372_47334372_63148.988.png", "Validation_47334372/Validation_47334372_47334372_63150.987.png", "Validation_47334372/Validation_47334372_47334372_63152.986.png", "Validation_47334372/Validation_47334372_47334372_63154.985.png" ], "turn_id": 7, "scan_id": "Validation_47334372", "origin_question": "Among these three objects(\"the bathtub\",\"the sleek white ceramic sink with modern lines\",\"the bright orange round basin with smooth edges\" ), which one is the farthest from you?", "option": [ "the bathtub", "the sleek white ceramic sink with modern lines", "the bright orange round basin with smooth edges" ], "answer": "the bright orange round basin with smooth edges", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the bathtub\",\"the sleek white ceramic sink with modern lines\",\"the bright orange round basin with smooth edges\" ), which one is the farthest from you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the bathtub', 'the sleek white ceramic sink with modern lines', 'the bright orange round basin with smooth edges'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "Validation_47334372/Validation_47334372_47334372_63156.984.png", "Validation_47334372/Validation_47334372_47334372_63158.984.png", "Validation_47334372/Validation_47334372_47334372_63160.983.png", "Validation_47334372/Validation_47334372_47334372_63162.982.png", "Validation_47334372/Validation_47334372_47334372_63164.998.png" ], "turn_id": 8, "scan_id": "Validation_47334372", "origin_question": "Among these three objects(\"the clothes\",\"the compact white cabinet with the sleek design\",\"the bathtub\" ), which one is the farthest from you?", "option": [ "the clothes", "the compact white cabinet with the sleek design", "the bathtub" ], "answer": "the bathtub", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the clothes\",\"the compact white cabinet with the sleek design\",\"the bathtub\" ), which one is the farthest from you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the clothes', 'the compact white cabinet with the sleek design', 'the bathtub'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42897550/Validation_42897550_42897550_276535.190.png", "Validation_42897550/Validation_42897550_42897550_276536.673.png", "Validation_42897550/Validation_42897550_42897550_276538.172.png", "Validation_42897550/Validation_42897550_42897550_276539.671.png" ], "turn_id": 1, "scan_id": "Validation_42897550", "origin_question": "Remember, have you seen any bathtub(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any bathtub(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_42897550/Validation_42897550_42897550_276541.171.png", "Validation_42897550/Validation_42897550_42897550_276542.670.png", "Validation_42897550/Validation_42897550_42897550_276544.170.png", "Validation_42897550/Validation_42897550_42897550_276545.669.png" ], "turn_id": 2, "scan_id": "Validation_42897550", "origin_question": "In which turn were you closest to \"the classic rectangular white bathtub made of durable materials\"? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the classic rectangular white bathtub made of durable materials\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "Validation_42897550/Validation_42897550_42897550_276547.169.png", "Validation_42897550/Validation_42897550_42897550_276548.668.png", "Validation_42897550/Validation_42897550_42897550_276550.167.png", "Validation_42897550/Validation_42897550_42897550_276551.667.png" ], "turn_id": 3, "scan_id": "Validation_42897550", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the classic rectangular white bathtub made of durable materials\"; \"an oval-shaped white ceramic toilet\"; \"the sink\" ", "option": [ "the classic rectangular white bathtub made of durable materials", "an oval-shaped white ceramic toilet", "the sink" ], "answer": "an oval-shaped white ceramic toilet", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the classic rectangular white bathtub made of durable materials\"; \"an oval-shaped white ceramic toilet\"; \"the sink\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the classic rectangular white bathtub made of durable materials', 'an oval-shaped white ceramic toilet', 'the sink'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_42897550/Validation_42897550_42897550_276553.166.png", "Validation_42897550/Validation_42897550_42897550_276554.665.png", "Validation_42897550/Validation_42897550_42897550_276556.165.png", "Validation_42897550/Validation_42897550_42897550_276557.664.png" ], "turn_id": 4, "scan_id": "Validation_42897550", "origin_question": "What will be the first-time appearance order of the following categories: bathtub, sink, teapot?", "option": [ "sink, teapot, bathtub", "bathtub, teapot, sink", "teapot, bathtub, sink", "sink, bathtub, teapot" ], "answer": "bathtub, teapot, sink", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bathtub, sink, teapot? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['sink, teapot, bathtub', 'bathtub, teapot, sink', 'teapot, bathtub, sink', 'sink, bathtub, teapot'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_42897550/Validation_42897550_42897550_276559.164.png", "Validation_42897550/Validation_42897550_42897550_276560.663.png", "Validation_42897550/Validation_42897550_42897550_276562.162.png", "Validation_42897550/Validation_42897550_42897550_276563.662.png" ], "turn_id": 5, "scan_id": "Validation_42897550", "origin_question": "When did you first discover \"the window\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the window\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_42897550/Validation_42897550_42897550_276565.161.png", "Validation_42897550/Validation_42897550_42897550_276566.660.png", "Validation_42897550/Validation_42897550_42897550_276568.160.png", "Validation_42897550/Validation_42897550_42897550_276569.659.png" ], "turn_id": 6, "scan_id": "Validation_42897550", "origin_question": "What will be the first-time appearance order of the following categories: bathtub, teapot, window?", "option": [ "bathtub, teapot, window", "teapot, bathtub, window", "teapot, window, bathtub", "window, teapot, bathtub" ], "answer": "bathtub, teapot, window", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bathtub, teapot, window? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['bathtub, teapot, window', 'teapot, bathtub, window', 'teapot, window, bathtub', 'window, teapot, bathtub'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_42897550/Validation_42897550_42897550_276571.159.png", "Validation_42897550/Validation_42897550_42897550_276572.658.png", "Validation_42897550/Validation_42897550_42897550_276574.157.png", "Validation_42897550/Validation_42897550_42897550_276575.673.png" ], "turn_id": 7, "scan_id": "Validation_42897550", "origin_question": "Is the sleek, vibrant pink teapot to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the sleek, vibrant pink teapot to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_42897550/Validation_42897550_42897550_276577.173.png", "Validation_42897550/Validation_42897550_42897550_276578.672.png", "Validation_42897550/Validation_42897550_42897550_276580.172.png", "Validation_42897550/Validation_42897550_42897550_276581.671.png" ], "turn_id": 8, "scan_id": "Validation_42897550", "origin_question": "At the end of which turn were both of \"the sink\" and \"the window\" on your right side?", "option": [], "answer": "3", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the sink\" and \"the window\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_45261150/Validation_45261150_45261150_4526.329.png", "Validation_45261150/Validation_45261150_45261150_4527.811.png", "Validation_45261150/Validation_45261150_45261150_4529.311.png", "Validation_45261150/Validation_45261150_45261150_4530.810.png" ], "turn_id": 1, "scan_id": "Validation_45261150", "origin_question": "Remember, have you seen any bed(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any bed(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_45261150/Validation_45261150_45261150_4532.310.png", "Validation_45261150/Validation_45261150_45261150_4533.809.png", "Validation_45261150/Validation_45261150_45261150_4535.308.png", "Validation_45261150/Validation_45261150_45261150_4536.808.png" ], "turn_id": 2, "scan_id": "Validation_45261150", "origin_question": "When did you first discover \"the soft pastel pink bed with the cotton blend bedspread\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the soft pastel pink bed with the cotton blend bedspread\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_45261150/Validation_45261150_45261150_4538.307.png", "Validation_45261150/Validation_45261150_45261150_4539.806.png", "Validation_45261150/Validation_45261150_45261150_4541.306.png", "Validation_45261150/Validation_45261150_45261150_4542.805.png" ], "turn_id": 3, "scan_id": "Validation_45261150", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the pristine white rectangular pillow with soft fabric\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the pristine white rectangular pillow with soft fabric\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_45261150/Validation_45261150_45261150_4544.305.png", "Validation_45261150/Validation_45261150_45261150_4545.804.png", "Validation_45261150/Validation_45261150_45261150_4547.303.png", "Validation_45261150/Validation_45261150_45261150_4548.803.png" ], "turn_id": 4, "scan_id": "Validation_45261150", "origin_question": "Is the soft pastel pink bed with the cotton blend bedspread to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the soft pastel pink bed with the cotton blend bedspread to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_45261150/Validation_45261150_45261150_4550.302.png", "Validation_45261150/Validation_45261150_45261150_4551.801.png", "Validation_45261150/Validation_45261150_45261150_4553.301.png", "Validation_45261150/Validation_45261150_45261150_4554.800.png" ], "turn_id": 5, "scan_id": "Validation_45261150", "origin_question": "", "option": [], "answer": "", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_45261150/Validation_45261150_45261150_4556.300.png", "Validation_45261150/Validation_45261150_45261150_4557.799.png", "Validation_45261150/Validation_45261150_45261150_4559.298.png", "Validation_45261150/Validation_45261150_45261150_4560.798.png" ], "turn_id": 6, "scan_id": "Validation_45261150", "origin_question": "Using your orientation at the end of turn 5 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 5 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_45261150/Validation_45261150_45261150_4562.314.png", "Validation_45261150/Validation_45261150_45261150_4563.813.png", "Validation_45261150/Validation_45261150_45261150_4565.312.png", "Validation_45261150/Validation_45261150_45261150_4566.812.png" ], "turn_id": 7, "scan_id": "Validation_45261150", "origin_question": "When did you first discover \"the soft pastel pink bed with the cotton blend bedspread\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the soft pastel pink bed with the cotton blend bedspread\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_45261150/Validation_45261150_45261150_4568.311.png", "Validation_45261150/Validation_45261150_45261150_4569.811.png", "Validation_45261150/Validation_45261150_45261150_4571.310.png", "Validation_45261150/Validation_45261150_45261150_4572.809.png" ], "turn_id": 8, "scan_id": "Validation_45261150", "origin_question": "", "option": [], "answer": "", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42898574/Validation_42898574_42898574_752946.194.png", "Validation_42898574/Validation_42898574_42898574_752948.176.png", "Validation_42898574/Validation_42898574_42898574_752950.176.png", "Validation_42898574/Validation_42898574_42898574_752952.175.png", "Validation_42898574/Validation_42898574_42898574_752954.174.png" ], "turn_id": 1, "scan_id": "Validation_42898574", "origin_question": "Remember, have you seen any pack(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any pack(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_42898574/Validation_42898574_42898574_752956.173.png", "Validation_42898574/Validation_42898574_42898574_752958.172.png", "Validation_42898574/Validation_42898574_42898574_752960.172.png", "Validation_42898574/Validation_42898574_42898574_752962.171.png", "Validation_42898574/Validation_42898574_42898574_752964.170.png" ], "turn_id": 2, "scan_id": "Validation_42898574", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_42898574/Validation_42898574_42898574_752966.186.png", "Validation_42898574/Validation_42898574_42898574_752968.185.png", "Validation_42898574/Validation_42898574_42898574_752970.184.png", "Validation_42898574/Validation_42898574_42898574_752972.183.png", "Validation_42898574/Validation_42898574_42898574_752974.182.png" ], "turn_id": 3, "scan_id": "Validation_42898574", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "34.09", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_42898574/Validation_42898574_42898574_752976.182.png", "Validation_42898574/Validation_42898574_42898574_752978.181.png", "Validation_42898574/Validation_42898574_42898574_752980.180.png", "Validation_42898574/Validation_42898574_42898574_752982.179.png", "Validation_42898574/Validation_42898574_42898574_752984.178.png" ], "turn_id": 4, "scan_id": "Validation_42898574", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "2.19", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_42898574/Validation_42898574_42898574_752986.178.png", "Validation_42898574/Validation_42898574_42898574_752988.177.png", "Validation_42898574/Validation_42898574_42898574_752990.176.png", "Validation_42898574/Validation_42898574_42898574_752992.175.png", "Validation_42898574/Validation_42898574_42898574_752994.174.png" ], "turn_id": 5, "scan_id": "Validation_42898574", "origin_question": "Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "149.21", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_42898574/Validation_42898574_42898574_752996.174.png", "Validation_42898574/Validation_42898574_42898574_752998.173.png", "Validation_42898574/Validation_42898574_42898574_753000.172.png", "Validation_42898574/Validation_42898574_42898574_753002.171.png", "Validation_42898574/Validation_42898574_42898574_753004.170.png" ], "turn_id": 6, "scan_id": "Validation_42898574", "origin_question": "Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Left", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_42898574/Validation_42898574_42898574_753006.186.png", "Validation_42898574/Validation_42898574_42898574_753008.185.png", "Validation_42898574/Validation_42898574_42898574_753010.184.png", "Validation_42898574/Validation_42898574_42898574_753012.184.png", "Validation_42898574/Validation_42898574_42898574_753014.183.png" ], "turn_id": 7, "scan_id": "Validation_42898574", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "145.74", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_42898574/Validation_42898574_42898574_753016.182.png", "Validation_42898574/Validation_42898574_42898574_753018.181.png", "Validation_42898574/Validation_42898574_42898574_753020.180.png", "Validation_42898574/Validation_42898574_42898574_753022.179.png", "Validation_42898574/Validation_42898574_42898574_753024.179.png" ], "turn_id": 8, "scan_id": "Validation_42898574", "origin_question": "How far is your current position from where you were at the end of the turn 1? (in meters)", "option": [], "answer": "4.82", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 1? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_42898574/Validation_42898574_42898574_753026.178.png", "Validation_42898574/Validation_42898574_42898574_753028.177.png", "Validation_42898574/Validation_42898574_42898574_753030.176.png", "Validation_42898574/Validation_42898574_42898574_753032.175.png", "Validation_42898574/Validation_42898574_42898574_753034.174.png" ], "turn_id": 9, "scan_id": "Validation_42898574", "origin_question": "Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "99.76", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_42898574/Validation_42898574_42898574_753036.174.png", "Validation_42898574/Validation_42898574_42898574_753038.173.png", "Validation_42898574/Validation_42898574_42898574_753040.172.png", "Validation_42898574/Validation_42898574_42898574_753042.171.png", "Validation_42898574/Validation_42898574_42898574_753044.170.png" ], "turn_id": 10, "scan_id": "Validation_42898574", "origin_question": "Which direction is the rectangular light brown wooden table with drawers in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the rectangular light brown wooden table with drawers in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_48458650/Validation_48458650_48458650_7059.622.png", "Validation_48458650/Validation_48458650_48458650_7061.604.png", "Validation_48458650/Validation_48458650_48458650_7063.604.png", "Validation_48458650/Validation_48458650_48458650_7065.603.png" ], "turn_id": 1, "scan_id": "Validation_48458650", "origin_question": "Remember, have you seen any bin(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any bin(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_48458650/Validation_48458650_48458650_7067.602.png", "Validation_48458650/Validation_48458650_48458650_7069.618.png", "Validation_48458650/Validation_48458650_48458650_7071.617.png", "Validation_48458650/Validation_48458650_48458650_7073.616.png" ], "turn_id": 2, "scan_id": "Validation_48458650", "origin_question": "At the end of which turn were both of \"the cylindrical metal waste paper bin with crisscross patterns\" and \"the compact cabinet with the light wooden top and gray sides\" on your right side?", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the cylindrical metal waste paper bin with crisscross patterns\" and \"the compact cabinet with the light wooden top and gray sides\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_48458650/Validation_48458650_48458650_7075.615.png", "Validation_48458650/Validation_48458650_48458650_7077.614.png", "Validation_48458650/Validation_48458650_48458650_7079.614.png", "Validation_48458650/Validation_48458650_48458650_7081.613.png" ], "turn_id": 3, "scan_id": "Validation_48458650", "origin_question": "At the end of which turn were both of \"the rectangular window in neutral color with sturdy frame\" and \"the bed\" on your right side?", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the rectangular window in neutral color with sturdy frame\" and \"the bed\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_48458650/Validation_48458650_48458650_7083.612.png", "Validation_48458650/Validation_48458650_48458650_7085.611.png", "Validation_48458650/Validation_48458650_48458650_7087.610.png", "Validation_48458650/Validation_48458650_48458650_7089.609.png" ], "turn_id": 4, "scan_id": "Validation_48458650", "origin_question": "Which direction is the large rectangular mirror with the light-colored frame in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the large rectangular mirror with the light-colored frame in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_48458650/Validation_48458650_48458650_7091.609.png", "Validation_48458650/Validation_48458650_48458650_7093.608.png", "Validation_48458650/Validation_48458650_48458650_7095.607.png", "Validation_48458650/Validation_48458650_48458650_7097.606.png" ], "turn_id": 5, "scan_id": "Validation_48458650", "origin_question": "Is the vibrant collage of small rectangular pictures to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the vibrant collage of small rectangular pictures to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_48458650/Validation_48458650_48458650_7099.605.png", "Validation_48458650/Validation_48458650_48458650_7101.605.png", "Validation_48458650/Validation_48458650_48458650_7103.604.png", "Validation_48458650/Validation_48458650_48458650_7105.603.png" ], "turn_id": 6, "scan_id": "Validation_48458650", "origin_question": "Which direction is the small white rectangular table made of wood and laminate in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the small white rectangular table made of wood and laminate in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "Validation_41159540/Validation_41159540_41159540_39872.352.png", "Validation_41159540/Validation_41159540_41159540_39874.335.png", "Validation_41159540/Validation_41159540_41159540_39876.334.png", "Validation_41159540/Validation_41159540_41159540_39878.333.png", "Validation_41159540/Validation_41159540_41159540_39880.332.png" ], "turn_id": 1, "scan_id": "Validation_41159540", "origin_question": "Remember, how many chair(s) have you seen so far? ", "option": [], "answer": "3", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many chair(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_41159540/Validation_41159540_41159540_39882.331.png", "Validation_41159540/Validation_41159540_41159540_39884.330.png", "Validation_41159540/Validation_41159540_41159540_39886.330.png", "Validation_41159540/Validation_41159540_41159540_39888.329.png", "Validation_41159540/Validation_41159540_41159540_39890.328.png" ], "turn_id": 2, "scan_id": "Validation_41159540", "origin_question": "Is the sleek light wood door with the modern handle to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the sleek light wood door with the modern handle to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_41159540/Validation_41159540_41159540_39892.327.png", "Validation_41159540/Validation_41159540_41159540_39894.326.png", "Validation_41159540/Validation_41159540_41159540_39896.325.png", "Validation_41159540/Validation_41159540_41159540_39898.325.png", "Validation_41159540/Validation_41159540_41159540_39900.324.png" ], "turn_id": 3, "scan_id": "Validation_41159540", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the sleek rectangular off-white cabinet with round knobs\" from your position?", "option": [], "answer": "141.74", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the sleek rectangular off-white cabinet with round knobs\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_41159540/Validation_41159540_41159540_39902.323.png", "Validation_41159540/Validation_41159540_41159540_39904.322.png", "Validation_41159540/Validation_41159540_41159540_39906.321.png", "Validation_41159540/Validation_41159540_41159540_39908.320.png", "Validation_41159540/Validation_41159540_41159540_39910.320.png" ], "turn_id": 4, "scan_id": "Validation_41159540", "origin_question": "How far is your current position from where you were at the end of the turn 3? (in meters)", "option": [], "answer": "1.66", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 3? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_41159540/Validation_41159540_41159540_39912.335.png", "Validation_41159540/Validation_41159540_41159540_39914.335.png", "Validation_41159540/Validation_41159540_41159540_39916.334.png", "Validation_41159540/Validation_41159540_41159540_39918.333.png", "Validation_41159540/Validation_41159540_41159540_39920.332.png" ], "turn_id": 5, "scan_id": "Validation_41159540", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_41159540/Validation_41159540_41159540_39922.331.png", "Validation_41159540/Validation_41159540_41159540_39924.330.png", "Validation_41159540/Validation_41159540_41159540_39926.330.png", "Validation_41159540/Validation_41159540_41159540_39928.329.png", "Validation_41159540/Validation_41159540_41159540_39930.328.png" ], "turn_id": 6, "scan_id": "Validation_41159540", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the sleek light wood door with the modern handle\" from your position?", "option": [], "answer": "108.77", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the sleek light wood door with the modern handle\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_41159540/Validation_41159540_41159540_39932.327.png", "Validation_41159540/Validation_41159540_41159540_39934.326.png", "Validation_41159540/Validation_41159540_41159540_39936.325.png", "Validation_41159540/Validation_41159540_41159540_39938.325.png", "Validation_41159540/Validation_41159540_41159540_39940.324.png" ], "turn_id": 7, "scan_id": "Validation_41159540", "origin_question": "Remember, have you seen any soap(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any soap(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_41159540/Validation_41159540_41159540_39942.323.png", "Validation_41159540/Validation_41159540_41159540_39944.322.png", "Validation_41159540/Validation_41159540_41159540_39946.321.png", "Validation_41159540/Validation_41159540_41159540_39948.321.png", "Validation_41159540/Validation_41159540_41159540_39950.320.png" ], "turn_id": 8, "scan_id": "Validation_41159540", "origin_question": "Assuming the direction you are facing at the end of the turn 4 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Right", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 4 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "Validation_41159540/Validation_41159540_41159540_39952.336.png", "Validation_41159540/Validation_41159540_41159540_39954.335.png", "Validation_41159540/Validation_41159540_41159540_39956.334.png", "Validation_41159540/Validation_41159540_41159540_39958.333.png", "Validation_41159540/Validation_41159540_41159540_39960.332.png" ], "turn_id": 9, "scan_id": "Validation_41159540", "origin_question": "Among these three objects(\"the large black sleek rectangular monitor\",\"the sturdy blue rectangular plastic folder\",\"the spherical, soft pink globe with brass accents\" ), which one is the closest to you?", "option": [ "the large black sleek rectangular monitor", "the sturdy blue rectangular plastic folder", "the spherical, soft pink globe with brass accents" ], "answer": "the large black sleek rectangular monitor", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the large black sleek rectangular monitor\",\"the sturdy blue rectangular plastic folder\",\"the spherical, soft pink globe with brass accents\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the large black sleek rectangular monitor', 'the sturdy blue rectangular plastic folder', 'the spherical, soft pink globe with brass accents'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_41159540/Validation_41159540_41159540_39962.331.png", "Validation_41159540/Validation_41159540_41159540_39964.330.png", "Validation_41159540/Validation_41159540_41159540_39966.330.png", "Validation_41159540/Validation_41159540_41159540_39968.329.png", "Validation_41159540/Validation_41159540_41159540_39970.328.png" ], "turn_id": 10, "scan_id": "Validation_41159540", "origin_question": "Using your orientation at the end of turn 5 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "129.37", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 5 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47895739/Validation_47895739_47895739_384893.864.png", "Validation_47895739/Validation_47895739_47895739_384895.346.png", "Validation_47895739/Validation_47895739_47895739_384896.846.png", "Validation_47895739/Validation_47895739_47895739_384898.345.png" ], "turn_id": 1, "scan_id": "Validation_47895739", "origin_question": "Remember, have you seen any cabinet(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any cabinet(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47895739/Validation_47895739_47895739_384899.845.png", "Validation_47895739/Validation_47895739_47895739_384901.344.png", "Validation_47895739/Validation_47895739_47895739_384902.843.png", "Validation_47895739/Validation_47895739_47895739_384904.343.png" ], "turn_id": 2, "scan_id": "Validation_47895739", "origin_question": "In which turn were you closest to \"the glossy white rectangular sink made of porcelain\"? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the glossy white rectangular sink made of porcelain\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47895739/Validation_47895739_47895739_384905.842.png", "Validation_47895739/Validation_47895739_47895739_384907.342.png", "Validation_47895739/Validation_47895739_47895739_384908.841.png", "Validation_47895739/Validation_47895739_47895739_384910.340.png" ], "turn_id": 3, "scan_id": "Validation_47895739", "origin_question": "", "option": [], "answer": "", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_47895739/Validation_47895739_47895739_384911.840.png", "Validation_47895739/Validation_47895739_47895739_384913.339.png", "Validation_47895739/Validation_47895739_47895739_384914.839.png", "Validation_47895739/Validation_47895739_47895739_384916.338.png" ], "turn_id": 4, "scan_id": "Validation_47895739", "origin_question": "When did you first discover \"the glossy white rectangular sink made of porcelain\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the glossy white rectangular sink made of porcelain\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47895739/Validation_47895739_47895739_384917.837.png", "Validation_47895739/Validation_47895739_47895739_384919.353.png", "Validation_47895739/Validation_47895739_47895739_384920.853.png", "Validation_47895739/Validation_47895739_47895739_384922.352.png" ], "turn_id": 5, "scan_id": "Validation_47895739", "origin_question": "Remember, have you seen any bathtub(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any bathtub(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47115473/Validation_47115473_47115473_14072.784.png", "Validation_47115473/Validation_47115473_47115473_14074.784.png", "Validation_47115473/Validation_47115473_47115473_14076.783.png", "Validation_47115473/Validation_47115473_47115473_14078.782.png", "Validation_47115473/Validation_47115473_47115473_14080.781.png" ], "turn_id": 1, "scan_id": "Validation_47115473", "origin_question": "Remember, have you seen any door(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any door(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_47115473/Validation_47115473_47115473_14082.780.png", "Validation_47115473/Validation_47115473_47115473_14084.780.png", "Validation_47115473/Validation_47115473_47115473_14086.779.png", "Validation_47115473/Validation_47115473_47115473_14088.778.png", "Validation_47115473/Validation_47115473_47115473_14090.777.png" ], "turn_id": 2, "scan_id": "Validation_47115473", "origin_question": "Please recall \"the sleek oval glossy white bathtub\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.69", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the sleek oval glossy white bathtub\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_47115473/Validation_47115473_47115473_14092.776.png", "Validation_47115473/Validation_47115473_47115473_14094.792.png", "Validation_47115473/Validation_47115473_47115473_14096.791.png", "Validation_47115473/Validation_47115473_47115473_14098.790.png", "Validation_47115473/Validation_47115473_47115473_14100.790.png" ], "turn_id": 3, "scan_id": "Validation_47115473", "origin_question": "At the end of which turn were both of \"the classic white ceramic toilet with the glossy finish\" and \"the rectangular mirror with the sleek silver frame\" on your right side?", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the classic white ceramic toilet with the glossy finish\" and \"the rectangular mirror with the sleek silver frame\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_47115473/Validation_47115473_47115473_14102.789.png", "Validation_47115473/Validation_47115473_47115473_14104.788.png", "Validation_47115473/Validation_47115473_47115473_14106.787.png", "Validation_47115473/Validation_47115473_47115473_14108.786.png", "Validation_47115473/Validation_47115473_47115473_14110.785.png" ], "turn_id": 4, "scan_id": "Validation_47115473", "origin_question": "Please recall \"the sleek oval glossy white bathtub\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.03", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the sleek oval glossy white bathtub\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_47115473/Validation_47115473_47115473_14112.785.png", "Validation_47115473/Validation_47115473_47115473_14114.784.png", "Validation_47115473/Validation_47115473_47115473_14116.783.png", "Validation_47115473/Validation_47115473_47115473_14118.782.png", "Validation_47115473/Validation_47115473_47115473_14120.781.png" ], "turn_id": 5, "scan_id": "Validation_47115473", "origin_question": "Is the rustic light brown wooden stool with the polished surface to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the rustic light brown wooden stool with the polished surface to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47115473/Validation_47115473_47115473_14122.781.png", "Validation_47115473/Validation_47115473_47115473_14124.780.png", "Validation_47115473/Validation_47115473_47115473_14126.779.png", "Validation_47115473/Validation_47115473_47115473_14128.778.png", "Validation_47115473/Validation_47115473_47115473_14130.777.png" ], "turn_id": 6, "scan_id": "Validation_47115473", "origin_question": "Which direction is the rustic light brown wooden stool with the polished surface in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the rustic light brown wooden stool with the polished surface in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "Validation_47115473/Validation_47115473_47115473_14132.776.png", "Validation_47115473/Validation_47115473_47115473_14134.792.png", "Validation_47115473/Validation_47115473_47115473_14136.791.png", "Validation_47115473/Validation_47115473_47115473_14138.790.png", "Validation_47115473/Validation_47115473_47115473_14140.790.png" ], "turn_id": 7, "scan_id": "Validation_47115473", "origin_question": "Among these three objects(\"the sleek oval glossy white bathtub\",\"the glossy white oval ceramic sink on the pedestal\",\"the rectangular mirror with the sleek silver frame\" ), which one is the closest to you?", "option": [ "the sleek oval glossy white bathtub", "the glossy white oval ceramic sink on the pedestal", "the rectangular mirror with the sleek silver frame" ], "answer": "the rectangular mirror with the sleek silver frame", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the sleek oval glossy white bathtub\",\"the glossy white oval ceramic sink on the pedestal\",\"the rectangular mirror with the sleek silver frame\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the sleek oval glossy white bathtub', 'the glossy white oval ceramic sink on the pedestal', 'the rectangular mirror with the sleek silver frame'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47115473/Validation_47115473_47115473_14142.789.png", "Validation_47115473/Validation_47115473_47115473_14144.788.png", "Validation_47115473/Validation_47115473_47115473_14146.787.png", "Validation_47115473/Validation_47115473_47115473_14148.786.png", "Validation_47115473/Validation_47115473_47115473_14150.785.png" ], "turn_id": 8, "scan_id": "Validation_47115473", "origin_question": "Which direction is the glossy white oval ceramic sink on the pedestal in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the glossy white oval ceramic sink on the pedestal in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47115473/Validation_47115473_47115473_14152.785.png", "Validation_47115473/Validation_47115473_47115473_14154.784.png", "Validation_47115473/Validation_47115473_47115473_14156.783.png", "Validation_47115473/Validation_47115473_47115473_14158.782.png", "Validation_47115473/Validation_47115473_47115473_14160.781.png" ], "turn_id": 9, "scan_id": "Validation_47115473", "origin_question": "Which direction is the rectangular mirror with the sleek silver frame in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the rectangular mirror with the sleek silver frame in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_47115473/Validation_47115473_47115473_14162.781.png", "Validation_47115473/Validation_47115473_47115473_14164.780.png", "Validation_47115473/Validation_47115473_47115473_14166.779.png", "Validation_47115473/Validation_47115473_47115473_14168.778.png", "Validation_47115473/Validation_47115473_47115473_14170.777.png" ], "turn_id": 10, "scan_id": "Validation_47115473", "origin_question": "Using your orientation at the end of turn 8 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "40.8", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 8 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_41125709/Validation_41125709_41125709_3816.462.png", "Validation_41125709/Validation_41125709_41125709_3818.445.png", "Validation_41125709/Validation_41125709_41125709_3820.444.png", "Validation_41125709/Validation_41125709_41125709_3822.443.png", "Validation_41125709/Validation_41125709_41125709_3824.442.png" ], "turn_id": 1, "scan_id": "Validation_41125709", "origin_question": "Remember, have you seen any eraser(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any eraser(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_41125709/Validation_41125709_41125709_3826.441.png", "Validation_41125709/Validation_41125709_41125709_3828.441.png", "Validation_41125709/Validation_41125709_41125709_3830.440.png", "Validation_41125709/Validation_41125709_41125709_3832.439.png", "Validation_41125709/Validation_41125709_41125709_3834.438.png" ], "turn_id": 2, "scan_id": "Validation_41125709", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the silver rectangular microwave with the glass door\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the silver rectangular microwave with the glass door\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_41125709/Validation_41125709_41125709_3836.437.png", "Validation_41125709/Validation_41125709_41125709_3838.436.png", "Validation_41125709/Validation_41125709_41125709_3840.436.png", "Validation_41125709/Validation_41125709_41125709_3842.452.png", "Validation_41125709/Validation_41125709_41125709_3844.451.png" ], "turn_id": 3, "scan_id": "Validation_41125709", "origin_question": "Which direction is the vintage grayscale picture with red and white text in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the vintage grayscale picture with red and white text in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_41125709/Validation_41125709_41125709_3846.450.png", "Validation_41125709/Validation_41125709_41125709_3848.449.png", "Validation_41125709/Validation_41125709_41125709_3850.448.png", "Validation_41125709/Validation_41125709_41125709_3852.447.png", "Validation_41125709/Validation_41125709_41125709_3854.447.png" ], "turn_id": 4, "scan_id": "Validation_41125709", "origin_question": "In which turn were you farthest from \"the sleek, black monitor with thin bezels\"? ", "option": [], "answer": "1", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the sleek, black monitor with thin bezels\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_41125709/Validation_41125709_41125709_3856.446.png", "Validation_41125709/Validation_41125709_41125709_3858.445.png", "Validation_41125709/Validation_41125709_41125709_3860.444.png", "Validation_41125709/Validation_41125709_41125709_3862.443.png", "Validation_41125709/Validation_41125709_41125709_3864.443.png" ], "turn_id": 5, "scan_id": "Validation_41125709", "origin_question": "Using your orientation at the end of turn 4 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "114.96", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 4 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_41125709/Validation_41125709_41125709_3866.442.png", "Validation_41125709/Validation_41125709_41125709_3868.441.png", "Validation_41125709/Validation_41125709_41125709_3870.440.png", "Validation_41125709/Validation_41125709_41125709_3872.439.png", "Validation_41125709/Validation_41125709_41125709_3874.439.png" ], "turn_id": 6, "scan_id": "Validation_41125709", "origin_question": "Which two objects are on the same side of you now? \"the silver rectangular microwave with the glass door\", \"the sleek white metal radiator with horizontal fins\", \"the sleek rectangular stainless-steel sink\". ", "option": [ "the silver rectangular microwave with the glass door, the sleek white metal radiator with horizontal fins", "the silver rectangular microwave with the glass door, the sleek rectangular stainless-steel sink", "the sleek white metal radiator with horizontal fins, the sleek rectangular stainless-steel sink" ], "answer": "the silver rectangular microwave with the glass door, the sleek white metal radiator with horizontal fins", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the silver rectangular microwave with the glass door\", \"the sleek white metal radiator with horizontal fins\", \"the sleek rectangular stainless-steel sink\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the silver rectangular microwave with the glass door, the sleek white metal radiator with horizontal fins', 'the silver rectangular microwave with the glass door, the sleek rectangular stainless-steel sink', 'the sleek white metal radiator with horizontal fins, the sleek rectangular stainless-steel sink'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_41125709/Validation_41125709_41125709_3876.438.png", "Validation_41125709/Validation_41125709_41125709_3878.437.png", "Validation_41125709/Validation_41125709_41125709_3880.436.png", "Validation_41125709/Validation_41125709_41125709_3882.435.png", "Validation_41125709/Validation_41125709_41125709_3884.451.png" ], "turn_id": 7, "scan_id": "Validation_41125709", "origin_question": "Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_41125709/Validation_41125709_41125709_3886.450.png", "Validation_41125709/Validation_41125709_41125709_3888.450.png", "Validation_41125709/Validation_41125709_41125709_3890.449.png", "Validation_41125709/Validation_41125709_41125709_3892.448.png", "Validation_41125709/Validation_41125709_41125709_3894.447.png" ], "turn_id": 8, "scan_id": "Validation_41125709", "origin_question": "What will be the first-time appearance order of the following categories: chair, monitor, sink?", "option": [ "monitor, chair, sink", "chair, monitor, sink", "sink, monitor, chair", "monitor, sink, chair" ], "answer": "sink, monitor, chair", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: chair, monitor, sink? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['monitor, chair, sink', 'chair, monitor, sink', 'sink, monitor, chair', 'monitor, sink, chair'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_41125709/Validation_41125709_41125709_3896.446.png", "Validation_41125709/Validation_41125709_41125709_3898.445.png", "Validation_41125709/Validation_41125709_41125709_3900.445.png", "Validation_41125709/Validation_41125709_41125709_3902.444.png", "Validation_41125709/Validation_41125709_41125709_3904.443.png" ], "turn_id": 9, "scan_id": "Validation_41125709", "origin_question": "Please recall \"the modern stainless steel range hood with the metallic gray finish\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "3.27", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the modern stainless steel range hood with the metallic gray finish\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_41125709/Validation_41125709_41125709_3906.442.png", "Validation_41125709/Validation_41125709_41125709_3908.441.png", "Validation_41125709/Validation_41125709_41125709_3910.441.png", "Validation_41125709/Validation_41125709_41125709_3912.440.png", "Validation_41125709/Validation_41125709_41125709_3914.439.png" ], "turn_id": 10, "scan_id": "Validation_41125709", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the modern stainless steel range hood with the metallic gray finish\" from your position?", "option": [], "answer": "134.77", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the modern stainless steel range hood with the metallic gray finish\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47895367/Validation_47895367_47895367_100028.253.png", "Validation_47895367/Validation_47895367_47895367_100030.235.png", "Validation_47895367/Validation_47895367_47895367_100032.235.png", "Validation_47895367/Validation_47895367_47895367_100034.234.png" ], "turn_id": 1, "scan_id": "Validation_47895367", "origin_question": "Remember, have you seen any washing machine(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any washing machine(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_47895367/Validation_47895367_47895367_100036.233.png", "Validation_47895367/Validation_47895367_47895367_100038.232.png", "Validation_47895367/Validation_47895367_47895367_100040.231.png", "Validation_47895367/Validation_47895367_47895367_100042.231.png" ], "turn_id": 2, "scan_id": "Validation_47895367", "origin_question": "In which turn did you see both \"the sleek white front-loading washing machine\" and \"the sleek, dark-colored cotton towel\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the sleek white front-loading washing machine\" and \"the sleek, dark-colored cotton towel\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47895367/Validation_47895367_47895367_100044.230.png", "Validation_47895367/Validation_47895367_47895367_100046.246.png", "Validation_47895367/Validation_47895367_47895367_100048.245.png", "Validation_47895367/Validation_47895367_47895367_100050.244.png" ], "turn_id": 3, "scan_id": "Validation_47895367", "origin_question": "In which turn were you closest to \"the sleek, dark-colored cotton towel\"? ", "option": [], "answer": "2", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the sleek, dark-colored cotton towel\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47895367/Validation_47895367_47895367_100052.243.png", "Validation_47895367/Validation_47895367_47895367_100054.242.png", "Validation_47895367/Validation_47895367_47895367_100056.242.png", "Validation_47895367/Validation_47895367_47895367_100058.241.png" ], "turn_id": 4, "scan_id": "Validation_47895367", "origin_question": "Remember, have you seen any bathtub(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any bathtub(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_47895367/Validation_47895367_47895367_100060.240.png", "Validation_47895367/Validation_47895367_47895367_100062.239.png", "Validation_47895367/Validation_47895367_47895367_100064.239.png", "Validation_47895367/Validation_47895367_47895367_100066.238.png" ], "turn_id": 5, "scan_id": "Validation_47895367", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_47895367/Validation_47895367_47895367_100068.237.png", "Validation_47895367/Validation_47895367_47895367_100070.236.png", "Validation_47895367/Validation_47895367_47895367_100072.235.png", "Validation_47895367/Validation_47895367_47895367_100074.234.png" ], "turn_id": 6, "scan_id": "Validation_47895367", "origin_question": "By the end of the 3 turn, were you getting closer to or farther away from these two objects: X: \"the sleek white front-loading washing machine\", Y: \"the sleek, dark-colored cotton towel\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "farther away from both X and Y", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 3 turn, were you getting closer to or farther away from these two objects: X: \"the sleek white front-loading washing machine\", Y: \"the sleek, dark-colored cotton towel\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_47895367/Validation_47895367_47895367_100076.234.png", "Validation_47895367/Validation_47895367_47895367_100078.233.png", "Validation_47895367/Validation_47895367_47895367_100080.232.png", "Validation_47895367/Validation_47895367_47895367_100082.231.png" ], "turn_id": 7, "scan_id": "Validation_47895367", "origin_question": "Using your orientation at the end of turn 5 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "121.29", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 5 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47895367/Validation_47895367_47895367_100084.230.png", "Validation_47895367/Validation_47895367_47895367_100086.246.png", "Validation_47895367/Validation_47895367_47895367_100088.245.png", "Validation_47895367/Validation_47895367_47895367_100090.245.png" ], "turn_id": 8, "scan_id": "Validation_47895367", "origin_question": "Which direction is the sleek white front-loading washing machine in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the sleek white front-loading washing machine in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42444950/Validation_42444950_42444950_223798.586.png", "Validation_42444950/Validation_42444950_42444950_223800.568.png", "Validation_42444950/Validation_42444950_42444950_223802.568.png", "Validation_42444950/Validation_42444950_42444950_223804.567.png", "Validation_42444950/Validation_42444950_42444950_223806.566.png" ], "turn_id": 1, "scan_id": "Validation_42444950", "origin_question": "Remember, have you seen any fireplace(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any fireplace(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_42444950/Validation_42444950_42444950_223808.565.png", "Validation_42444950/Validation_42444950_42444950_223810.564.png", "Validation_42444950/Validation_42444950_42444950_223812.564.png", "Validation_42444950/Validation_42444950_42444950_223814.579.png", "Validation_42444950/Validation_42444950_42444950_223816.579.png" ], "turn_id": 2, "scan_id": "Validation_42444950", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_42444950/Validation_42444950_42444950_223818.578.png", "Validation_42444950/Validation_42444950_42444950_223820.577.png", "Validation_42444950/Validation_42444950_42444950_223822.576.png", "Validation_42444950/Validation_42444950_42444950_223824.575.png", "Validation_42444950/Validation_42444950_42444950_223826.575.png" ], "turn_id": 3, "scan_id": "Validation_42444950", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_42444950/Validation_42444950_42444950_223828.574.png", "Validation_42444950/Validation_42444950_42444950_223830.573.png", "Validation_42444950/Validation_42444950_42444950_223832.572.png", "Validation_42444950/Validation_42444950_42444950_223834.571.png", "Validation_42444950/Validation_42444950_42444950_223836.571.png" ], "turn_id": 4, "scan_id": "Validation_42444950", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "1.09", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_42444950/Validation_42444950_42444950_223838.570.png", "Validation_42444950/Validation_42444950_42444950_223840.569.png", "Validation_42444950/Validation_42444950_42444950_223842.568.png", "Validation_42444950/Validation_42444950_42444950_223844.567.png", "Validation_42444950/Validation_42444950_42444950_223846.566.png" ], "turn_id": 5, "scan_id": "Validation_42444950", "origin_question": "Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_42444950/Validation_42444950_42444950_223848.566.png", "Validation_42444950/Validation_42444950_42444950_223850.565.png", "Validation_42444950/Validation_42444950_42444950_223852.564.png", "Validation_42444950/Validation_42444950_42444950_223854.563.png", "Validation_42444950/Validation_42444950_42444950_223856.562.png" ], "turn_id": 6, "scan_id": "Validation_42444950", "origin_question": "In which turn were you closest to \"the compact, deep blue upholstered couch\"? ", "option": [], "answer": "3", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the compact, deep blue upholstered couch\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_42444950/Validation_42444950_42444950_223858.578.png", "Validation_42444950/Validation_42444950_42444950_223860.577.png", "Validation_42444950/Validation_42444950_42444950_223862.576.png", "Validation_42444950/Validation_42444950_42444950_223864.576.png", "Validation_42444950/Validation_42444950_42444950_223866.575.png" ], "turn_id": 7, "scan_id": "Validation_42444950", "origin_question": "In which turn were you closest to \"the white rectangular fireplace with intricate moldings\"? ", "option": [], "answer": "3", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the white rectangular fireplace with intricate moldings\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_42444950/Validation_42444950_42444950_223868.574.png", "Validation_42444950/Validation_42444950_42444950_223870.573.png", "Validation_42444950/Validation_42444950_42444950_223872.573.png", "Validation_42444950/Validation_42444950_42444950_223874.572.png", "Validation_42444950/Validation_42444950_42444950_223876.571.png" ], "turn_id": 8, "scan_id": "Validation_42444950", "origin_question": "Assuming the direction you are facing at the end of the turn 3 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Left", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 3 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "Validation_41159553/Validation_41159553_41159553_33327.775.png", "Validation_41159553/Validation_41159553_41159553_33329.758.png", "Validation_41159553/Validation_41159553_41159553_33331.757.png", "Validation_41159553/Validation_41159553_41159553_33333.756.png", "Validation_41159553/Validation_41159553_41159553_33335.756.png" ], "turn_id": 1, "scan_id": "Validation_41159553", "origin_question": "Remember, how many laptop(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many laptop(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_41159553/Validation_41159553_41159553_33337.755.png", "Validation_41159553/Validation_41159553_41159553_33339.754.png", "Validation_41159553/Validation_41159553_41159553_33341.753.png", "Validation_41159553/Validation_41159553_41159553_33343.752.png", "Validation_41159553/Validation_41159553_41159553_33345.752.png" ], "turn_id": 2, "scan_id": "Validation_41159553", "origin_question": "When did you first discover \"the vibrant geometric carpet in blue, red, and white\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the vibrant geometric carpet in blue, red, and white\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_41159553/Validation_41159553_41159553_33347.751.png", "Validation_41159553/Validation_41159553_41159553_33349.750.png", "Validation_41159553/Validation_41159553_41159553_33351.766.png", "Validation_41159553/Validation_41159553_41159553_33353.765.png", "Validation_41159553/Validation_41159553_41159553_33355.764.png" ], "turn_id": 3, "scan_id": "Validation_41159553", "origin_question": "", "option": [], "answer": "", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_41159553/Validation_41159553_41159553_33357.764.png", "Validation_41159553/Validation_41159553_41159553_33359.763.png", "Validation_41159553/Validation_41159553_41159553_33361.762.png", "Validation_41159553/Validation_41159553_41159553_33363.761.png", "Validation_41159553/Validation_41159553_41159553_33365.760.png" ], "turn_id": 4, "scan_id": "Validation_41159553", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the two-level light wood stool\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the two-level light wood stool\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_41159553/Validation_41159553_41159553_33367.760.png", "Validation_41159553/Validation_41159553_41159553_33369.759.png", "Validation_41159553/Validation_41159553_41159553_33371.758.png", "Validation_41159553/Validation_41159553_41159553_33373.757.png", "Validation_41159553/Validation_41159553_41159553_33375.756.png" ], "turn_id": 5, "scan_id": "Validation_41159553", "origin_question": "Which direction is the standard-sized beige cap with the round crown in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the standard-sized beige cap with the round crown in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_41159553/Validation_41159553_41159553_33377.756.png", "Validation_41159553/Validation_41159553_41159553_33379.755.png", "Validation_41159553/Validation_41159553_41159553_33381.754.png", "Validation_41159553/Validation_41159553_41159553_33383.753.png", "Validation_41159553/Validation_41159553_41159553_33385.752.png" ], "turn_id": 6, "scan_id": "Validation_41159553", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the light-colored wooden door with the metallic handle\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the light-colored wooden door with the metallic handle\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_41159553/Validation_41159553_41159553_33387.752.png", "Validation_41159553/Validation_41159553_41159553_33389.751.png", "Validation_41159553/Validation_41159553_41159553_33391.750.png", "Validation_41159553/Validation_41159553_41159553_33393.766.png", "Validation_41159553/Validation_41159553_41159553_33395.765.png" ], "turn_id": 7, "scan_id": "Validation_41159553", "origin_question": "In which turn were you closest to \"the vibrant geometric carpet in blue, red, and white\"? ", "option": [], "answer": "7", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the vibrant geometric carpet in blue, red, and white\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_41159553/Validation_41159553_41159553_33397.764.png", "Validation_41159553/Validation_41159553_41159553_33399.764.png", "Validation_41159553/Validation_41159553_41159553_33401.763.png", "Validation_41159553/Validation_41159553_41159553_33403.762.png", "Validation_41159553/Validation_41159553_41159553_33405.761.png" ], "turn_id": 8, "scan_id": "Validation_41159553", "origin_question": "Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Left", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_41159553/Validation_41159553_41159553_33407.760.png", "Validation_41159553/Validation_41159553_41159553_33409.760.png", "Validation_41159553/Validation_41159553_41159553_33411.759.png", "Validation_41159553/Validation_41159553_41159553_33413.758.png", "Validation_41159553/Validation_41159553_41159553_33415.757.png" ], "turn_id": 9, "scan_id": "Validation_41159553", "origin_question": "Which two objects are on the same side of you now? \"the vibrant geometric carpet in blue, red, and white\", \"the light-colored wooden door with the metallic handle\", \"the standard-sized beige cap with the round crown\". ", "option": [ "the vibrant geometric carpet in blue, red, and white, the light-colored wooden door with the metallic handle", "the vibrant geometric carpet in blue, red, and white, the standard-sized beige cap with the round crown", "the light-colored wooden door with the metallic handle, the standard-sized beige cap with the round crown" ], "answer": "the vibrant geometric carpet in blue, red, and white, the light-colored wooden door with the metallic handle", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the vibrant geometric carpet in blue, red, and white\", \"the light-colored wooden door with the metallic handle\", \"the standard-sized beige cap with the round crown\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the vibrant geometric carpet in blue, red, and white, the light-colored wooden door with the metallic handle', 'the vibrant geometric carpet in blue, red, and white, the standard-sized beige cap with the round crown', 'the light-colored wooden door with the metallic handle, the standard-sized beige cap with the round crown'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_41159553/Validation_41159553_41159553_33417.756.png", "Validation_41159553/Validation_41159553_41159553_33419.756.png", "Validation_41159553/Validation_41159553_41159553_33421.755.png", "Validation_41159553/Validation_41159553_41159553_33423.754.png", "Validation_41159553/Validation_41159553_41159553_33425.753.png" ], "turn_id": 10, "scan_id": "Validation_41159553", "origin_question": "Please recall \"the two-level light wood stool\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.32", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the two-level light wood stool\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47331645/Validation_47331645_47331645_21839.607.png", "Validation_47331645/Validation_47331645_47331645_21841.606.png", "Validation_47331645/Validation_47331645_47331645_21843.605.png", "Validation_47331645/Validation_47331645_47331645_21845.604.png", "Validation_47331645/Validation_47331645_47331645_21847.604.png" ], "turn_id": 1, "scan_id": "Validation_47331645", "origin_question": "Remember, have you seen any toiletry(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any toiletry(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47331645/Validation_47331645_47331645_21849.603.png", "Validation_47331645/Validation_47331645_47331645_21851.602.png", "Validation_47331645/Validation_47331645_47331645_21853.618.png", "Validation_47331645/Validation_47331645_47331645_21855.617.png", "Validation_47331645/Validation_47331645_47331645_21857.616.png" ], "turn_id": 2, "scan_id": "Validation_47331645", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47331645/Validation_47331645_47331645_21859.615.png", "Validation_47331645/Validation_47331645_47331645_21861.614.png", "Validation_47331645/Validation_47331645_47331645_21863.614.png", "Validation_47331645/Validation_47331645_47331645_21865.613.png", "Validation_47331645/Validation_47331645_47331645_21867.612.png" ], "turn_id": 3, "scan_id": "Validation_47331645", "origin_question": "Remember, have you seen any table(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any table(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_47331645/Validation_47331645_47331645_21869.611.png", "Validation_47331645/Validation_47331645_47331645_21871.610.png", "Validation_47331645/Validation_47331645_47331645_21873.610.png", "Validation_47331645/Validation_47331645_47331645_21875.609.png", "Validation_47331645/Validation_47331645_47331645_21877.608.png" ], "turn_id": 4, "scan_id": "Validation_47331645", "origin_question": "When did you first discover \"the light cream rectangular wooden door with the glass panel\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the light cream rectangular wooden door with the glass panel\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_47331645/Validation_47331645_47331645_21879.607.png", "Validation_47331645/Validation_47331645_47331645_21881.606.png", "Validation_47331645/Validation_47331645_47331645_21883.606.png", "Validation_47331645/Validation_47331645_47331645_21885.605.png", "Validation_47331645/Validation_47331645_47331645_21887.604.png" ], "turn_id": 5, "scan_id": "Validation_47331645", "origin_question": "Which two objects are on the same side of you now? \"the patterned couch with earthy tones and durable fabric\", \"an off-white cylindrical ceramic flowerpot\", \"the soft white sheer curtain with subtle pleats\". ", "option": [ "the patterned couch with earthy tones and durable fabric, an off-white cylindrical ceramic flowerpot", "the patterned couch with earthy tones and durable fabric, the soft white sheer curtain with subtle pleats", "an off-white cylindrical ceramic flowerpot, the soft white sheer curtain with subtle pleats" ], "answer": "the patterned couch with earthy tones and durable fabric, an off-white cylindrical ceramic flowerpot", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the patterned couch with earthy tones and durable fabric\", \"an off-white cylindrical ceramic flowerpot\", \"the soft white sheer curtain with subtle pleats\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the patterned couch with earthy tones and durable fabric, an off-white cylindrical ceramic flowerpot', 'the patterned couch with earthy tones and durable fabric, the soft white sheer curtain with subtle pleats', 'an off-white cylindrical ceramic flowerpot, the soft white sheer curtain with subtle pleats'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_47331645/Validation_47331645_47331645_21889.603.png", "Validation_47331645/Validation_47331645_47331645_21891.602.png", "Validation_47331645/Validation_47331645_47331645_21893.602.png", "Validation_47331645/Validation_47331645_47331645_21895.617.png", "Validation_47331645/Validation_47331645_47331645_21897.617.png" ], "turn_id": 6, "scan_id": "Validation_47331645", "origin_question": "Which two objects are on the same side of you now? \"the patterned couch with earthy tones and durable fabric\", \"the vintage dome-shaped lamp with the cream-colored shade\", \"the soft white sheer curtain with subtle pleats\". ", "option": [ "the patterned couch with earthy tones and durable fabric, the vintage dome-shaped lamp with the cream-colored shade", "the patterned couch with earthy tones and durable fabric, the soft white sheer curtain with subtle pleats", "the vintage dome-shaped lamp with the cream-colored shade, the soft white sheer curtain with subtle pleats" ], "answer": "the vintage dome-shaped lamp with the cream-colored shade, the soft white sheer curtain with subtle pleats", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the patterned couch with earthy tones and durable fabric\", \"the vintage dome-shaped lamp with the cream-colored shade\", \"the soft white sheer curtain with subtle pleats\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the patterned couch with earthy tones and durable fabric, the vintage dome-shaped lamp with the cream-colored shade', 'the patterned couch with earthy tones and durable fabric, the soft white sheer curtain with subtle pleats', 'the vintage dome-shaped lamp with the cream-colored shade, the soft white sheer curtain with subtle pleats'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47331645/Validation_47331645_47331645_21899.616.png", "Validation_47331645/Validation_47331645_47331645_21901.615.png", "Validation_47331645/Validation_47331645_47331645_21903.614.png", "Validation_47331645/Validation_47331645_47331645_21905.613.png", "Validation_47331645/Validation_47331645_47331645_21907.612.png" ], "turn_id": 7, "scan_id": "Validation_47331645", "origin_question": "Which direction is the soft white sheer curtain with subtle pleats in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the soft white sheer curtain with subtle pleats in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_47331645/Validation_47331645_47331645_21909.612.png", "Validation_47331645/Validation_47331645_47331645_21911.611.png", "Validation_47331645/Validation_47331645_47331645_21913.610.png", "Validation_47331645/Validation_47331645_47331645_21915.609.png", "Validation_47331645/Validation_47331645_47331645_21917.608.png" ], "turn_id": 8, "scan_id": "Validation_47331645", "origin_question": "By the end of the 5 turn, were you getting closer to or farther away from these two objects: X: \"the vintage dome-shaped lamp with the cream-colored shade\", Y: \"the soft white sheer curtain with subtle pleats\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "farther away from both X and Y", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 5 turn, were you getting closer to or farther away from these two objects: X: \"the vintage dome-shaped lamp with the cream-colored shade\", Y: \"the soft white sheer curtain with subtle pleats\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_47331645/Validation_47331645_47331645_21919.608.png", "Validation_47331645/Validation_47331645_47331645_21921.607.png", "Validation_47331645/Validation_47331645_47331645_21923.606.png", "Validation_47331645/Validation_47331645_47331645_21925.605.png", "Validation_47331645/Validation_47331645_47331645_21927.604.png" ], "turn_id": 9, "scan_id": "Validation_47331645", "origin_question": "Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47331645/Validation_47331645_47331645_21929.604.png", "Validation_47331645/Validation_47331645_47331645_21931.603.png", "Validation_47331645/Validation_47331645_47331645_21933.602.png", "Validation_47331645/Validation_47331645_47331645_21935.601.png", "Validation_47331645/Validation_47331645_47331645_21937.600.png" ], "turn_id": 10, "scan_id": "Validation_47331645", "origin_question": "Which direction is the polished medium-brown wooden shelf in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the polished medium-brown wooden shelf in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_48018361/Validation_48018361_48018361_389004.567.png", "Validation_48018361/Validation_48018361_48018361_389006.050.png", "Validation_48018361/Validation_48018361_48018361_389007.549.png", "Validation_48018361/Validation_48018361_48018361_389009.049.png" ], "turn_id": 1, "scan_id": "Validation_48018361", "origin_question": "", "option": [], "answer": "", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_48018361/Validation_48018361_48018361_389010.548.png", "Validation_48018361/Validation_48018361_48018361_389012.047.png", "Validation_48018361/Validation_48018361_48018361_389013.547.png", "Validation_48018361/Validation_48018361_48018361_389015.046.png" ], "turn_id": 2, "scan_id": "Validation_48018361", "origin_question": "When did you first discover \"the window\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the window\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_48018361/Validation_48018361_48018361_389016.546.png", "Validation_48018361/Validation_48018361_48018361_389018.045.png", "Validation_48018361/Validation_48018361_48018361_389019.544.png", "Validation_48018361/Validation_48018361_48018361_389021.044.png" ], "turn_id": 3, "scan_id": "Validation_48018361", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the sleek rectangular light wood table with drawers\" from your position?", "option": [], "answer": "69.23", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the sleek rectangular light wood table with drawers\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_48018361/Validation_48018361_48018361_389022.543.png", "Validation_48018361/Validation_48018361_48018361_389024.043.png", "Validation_48018361/Validation_48018361_48018361_389025.542.png", "Validation_48018361/Validation_48018361_48018361_389027.041.png" ], "turn_id": 4, "scan_id": "Validation_48018361", "origin_question": "When did you first discover \"an array of clothes in neutral hues\"? (index of the turn)", "option": [], "answer": "3", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"an array of clothes in neutral hues\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_48018361/Validation_48018361_48018361_389028.557.png", "Validation_48018361/Validation_48018361_48018361_389030.057.png", "Validation_48018361/Validation_48018361_48018361_389031.556.png", "Validation_48018361/Validation_48018361_48018361_389033.056.png" ], "turn_id": 5, "scan_id": "Validation_48018361", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the modern, elongated light with the glossy metallic finish\" from your position?", "option": [], "answer": "81.13", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the modern, elongated light with the glossy metallic finish\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_48018361/Validation_48018361_48018361_389034.555.png", "Validation_48018361/Validation_48018361_48018361_389036.054.png", "Validation_48018361/Validation_48018361_48018361_389037.554.png", "Validation_48018361/Validation_48018361_48018361_389039.053.png" ], "turn_id": 6, "scan_id": "Validation_48018361", "origin_question": "Which direction is the modern rectangular monitor with the sleek black finish in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the modern rectangular monitor with the sleek black finish in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_48018361/Validation_48018361_48018361_389040.552.png", "Validation_48018361/Validation_48018361_48018361_389042.052.png", "Validation_48018361/Validation_48018361_48018361_389043.551.png", "Validation_48018361/Validation_48018361_48018361_389045.051.png" ], "turn_id": 7, "scan_id": "Validation_48018361", "origin_question": "Which direction is the window in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the window in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42446536/Validation_42446536_42446536_200444.910.png", "Validation_42446536/Validation_42446536_42446536_200446.393.png", "Validation_42446536/Validation_42446536_42446536_200447.892.png", "Validation_42446536/Validation_42446536_42446536_200449.392.png" ], "turn_id": 1, "scan_id": "Validation_42446536", "origin_question": "Remember, have you seen any hanger(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any hanger(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_42446536/Validation_42446536_42446536_200450.891.png", "Validation_42446536/Validation_42446536_42446536_200452.391.png", "Validation_42446536/Validation_42446536_42446536_200453.890.png", "Validation_42446536/Validation_42446536_42446536_200455.406.png" ], "turn_id": 2, "scan_id": "Validation_42446536", "origin_question": "In which turn did you see both \"the standard-sized, lightweight door in the soft neutral hue\" and \"the soft, absorbent white cotton towel\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the standard-sized, lightweight door in the soft neutral hue\" and \"the soft, absorbent white cotton towel\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_42446536/Validation_42446536_42446536_200456.905.png", "Validation_42446536/Validation_42446536_42446536_200458.405.png", "Validation_42446536/Validation_42446536_42446536_200459.904.png", "Validation_42446536/Validation_42446536_42446536_200461.403.png" ], "turn_id": 3, "scan_id": "Validation_42446536", "origin_question": "When did you first discover \"the standard-sized, lightweight door in the soft neutral hue\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the standard-sized, lightweight door in the soft neutral hue\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_42446536/Validation_42446536_42446536_200462.903.png", "Validation_42446536/Validation_42446536_42446536_200464.402.png", "Validation_42446536/Validation_42446536_42446536_200465.902.png", "Validation_42446536/Validation_42446536_42446536_200467.401.png" ], "turn_id": 4, "scan_id": "Validation_42446536", "origin_question": "Which two objects are on the same side of you now? \"the sleek white ceramic toilet with an elongated bowl\", \"the soft, absorbent white cotton towel\", \"the small red cylindrical bin\". ", "option": [ "the sleek white ceramic toilet with an elongated bowl, the soft, absorbent white cotton towel", "the sleek white ceramic toilet with an elongated bowl, the small red cylindrical bin", "the soft, absorbent white cotton towel, the small red cylindrical bin" ], "answer": "the sleek white ceramic toilet with an elongated bowl, the soft, absorbent white cotton towel", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the sleek white ceramic toilet with an elongated bowl\", \"the soft, absorbent white cotton towel\", \"the small red cylindrical bin\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the sleek white ceramic toilet with an elongated bowl, the soft, absorbent white cotton towel', 'the sleek white ceramic toilet with an elongated bowl, the small red cylindrical bin', 'the soft, absorbent white cotton towel, the small red cylindrical bin'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47333443/Validation_47333443_47333443_53749.132.png", "Validation_47333443/Validation_47333443_47333443_53750.631.png", "Validation_47333443/Validation_47333443_47333443_53752.130.png", "Validation_47333443/Validation_47333443_47333443_53753.630.png" ], "turn_id": 1, "scan_id": "Validation_47333443", "origin_question": "", "option": [], "answer": "", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47333443/Validation_47333443_47333443_53755.129.png", "Validation_47333443/Validation_47333443_47333443_53756.628.png", "Validation_47333443/Validation_47333443_47333443_53758.128.png", "Validation_47333443/Validation_47333443_47333443_53759.627.png" ], "turn_id": 2, "scan_id": "Validation_47333443", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the standard white porcelain toilet with an artistic seat cover\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the standard white porcelain toilet with an artistic seat cover\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_47333443/Validation_47333443_47333443_53761.127.png", "Validation_47333443/Validation_47333443_47333443_53762.626.png", "Validation_47333443/Validation_47333443_47333443_53764.125.png", "Validation_47333443/Validation_47333443_47333443_53765.625.png" ], "turn_id": 3, "scan_id": "Validation_47333443", "origin_question": "Is the washing machine to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the washing machine to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47333443/Validation_47333443_47333443_53767.124.png", "Validation_47333443/Validation_47333443_47333443_53768.624.png", "Validation_47333443/Validation_47333443_47333443_53770.123.png", "Validation_47333443/Validation_47333443_47333443_53771.622.png" ], "turn_id": 4, "scan_id": "Validation_47333443", "origin_question": "", "option": [], "answer": "", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "Validation_47333443/Validation_47333443_47333443_53773.122.png", "Validation_47333443/Validation_47333443_47333443_53774.621.png", "Validation_47333443/Validation_47333443_47333443_53776.121.png", "Validation_47333443/Validation_47333443_47333443_53777.620.png" ], "turn_id": 5, "scan_id": "Validation_47333443", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the frosted glass globe light fixture with the warm glow\"; \"the rectangular case in neutral tones\"; \"the large rectangular dark-framed glass window\" ", "option": [ "the frosted glass globe light fixture with the warm glow", "the rectangular case in neutral tones", "the large rectangular dark-framed glass window" ], "answer": "the frosted glass globe light fixture with the warm glow", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the frosted glass globe light fixture with the warm glow\"; \"the rectangular case in neutral tones\"; \"the large rectangular dark-framed glass window\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the frosted glass globe light fixture with the warm glow', 'the rectangular case in neutral tones', 'the large rectangular dark-framed glass window'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "Validation_47333443/Validation_47333443_47333443_53779.119.png", "Validation_47333443/Validation_47333443_47333443_53780.619.png", "Validation_47333443/Validation_47333443_47333443_53782.118.png", "Validation_47333443/Validation_47333443_47333443_53783.618.png" ], "turn_id": 6, "scan_id": "Validation_47333443", "origin_question": "Among these three objects(\"the standard white porcelain toilet with an artistic seat cover\",\"the durable translucent plastic bottle with the cylindrical shape\",\"the large rectangular dark-framed glass window\" ), which one is the closest to you?", "option": [ "the standard white porcelain toilet with an artistic seat cover", "the durable translucent plastic bottle with the cylindrical shape", "the large rectangular dark-framed glass window" ], "answer": "the standard white porcelain toilet with an artistic seat cover", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects(\"the standard white porcelain toilet with an artistic seat cover\",\"the durable translucent plastic bottle with the cylindrical shape\",\"the large rectangular dark-framed glass window\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the standard white porcelain toilet with an artistic seat cover', 'the durable translucent plastic bottle with the cylindrical shape', 'the large rectangular dark-framed glass window'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_48018572/Validation_48018572_48018572_3797.063.png", "Validation_48018572/Validation_48018572_48018572_3799.046.png", "Validation_48018572/Validation_48018572_48018572_3801.045.png", "Validation_48018572/Validation_48018572_48018572_3803.044.png" ], "turn_id": 1, "scan_id": "Validation_48018572", "origin_question": "Remember, have you seen any monitor(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any monitor(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_48018572/Validation_48018572_48018572_3805.043.png", "Validation_48018572/Validation_48018572_48018572_3807.042.png", "Validation_48018572/Validation_48018572_48018572_3809.042.png", "Validation_48018572/Validation_48018572_48018572_3811.041.png" ], "turn_id": 2, "scan_id": "Validation_48018572", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the light wood table with sturdy legs\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the light wood table with sturdy legs\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_48018572/Validation_48018572_48018572_3813.040.png", "Validation_48018572/Validation_48018572_48018572_3815.039.png", "Validation_48018572/Validation_48018572_48018572_3817.038.png", "Validation_48018572/Validation_48018572_48018572_3819.037.png" ], "turn_id": 3, "scan_id": "Validation_48018572", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the compact, modern rectangular cabinet in light neutral color\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the compact, modern rectangular cabinet in light neutral color\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_48018572/Validation_48018572_48018572_3821.037.png", "Validation_48018572/Validation_48018572_48018572_3823.036.png", "Validation_48018572/Validation_48018572_48018572_3825.035.png", "Validation_48018572/Validation_48018572_48018572_3827.034.png" ], "turn_id": 4, "scan_id": "Validation_48018572", "origin_question": "In which turn did you see both \"the sleek, modern rectangular black monitor\" and \"the light wood table with sturdy legs\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the sleek, modern rectangular black monitor\" and \"the light wood table with sturdy legs\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_48018572/Validation_48018572_48018572_3829.034.png", "Validation_48018572/Validation_48018572_48018572_3831.033.png", "Validation_48018572/Validation_48018572_48018572_3833.032.png", "Validation_48018572/Validation_48018572_48018572_3835.031.png" ], "turn_id": 5, "scan_id": "Validation_48018572", "origin_question": "Please recall \"the vertical rectangular metal radiator in the neutral color\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.53", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the vertical rectangular metal radiator in the neutral color\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_48018572/Validation_48018572_48018572_3837.030.png", "Validation_48018572/Validation_48018572_48018572_3839.046.png", "Validation_48018572/Validation_48018572_48018572_3841.045.png", "Validation_48018572/Validation_48018572_48018572_3843.044.png" ], "turn_id": 6, "scan_id": "Validation_48018572", "origin_question": "Compared to your position at the end of the 5 turn, are you now closer or farther away from \"the sleek, modern rectangular black monitor\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 5 turn, are you now closer or farther away from \"the sleek, modern rectangular black monitor\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47895536/Validation_47895536_47895536_455437.664.png", "Validation_47895536/Validation_47895536_47895536_455439.647.png", "Validation_47895536/Validation_47895536_47895536_455441.646.png", "Validation_47895536/Validation_47895536_47895536_455443.645.png" ], "turn_id": 1, "scan_id": "Validation_47895536", "origin_question": "Remember, have you seen any shelf(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any shelf(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47895536/Validation_47895536_47895536_455445.644.png", "Validation_47895536/Validation_47895536_47895536_455447.644.png", "Validation_47895536/Validation_47895536_47895536_455449.643.png", "Validation_47895536/Validation_47895536_47895536_455451.642.png" ], "turn_id": 2, "scan_id": "Validation_47895536", "origin_question": "In which turn were you farthest from \"the minimalist rectangular wooden shelf in light and dark tones\"? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the minimalist rectangular wooden shelf in light and dark tones\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47895536/Validation_47895536_47895536_455453.641.png", "Validation_47895536/Validation_47895536_47895536_455455.640.png", "Validation_47895536/Validation_47895536_47895536_455457.640.png", "Validation_47895536/Validation_47895536_47895536_455459.639.png" ], "turn_id": 3, "scan_id": "Validation_47895536", "origin_question": "In which turn were you farthest from \"the minimalist rectangular wooden shelf in light and dark tones\"? ", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the minimalist rectangular wooden shelf in light and dark tones\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_47895536/Validation_47895536_47895536_455461.638.png", "Validation_47895536/Validation_47895536_47895536_455463.637.png", "Validation_47895536/Validation_47895536_47895536_455465.636.png", "Validation_47895536/Validation_47895536_47895536_455467.636.png" ], "turn_id": 4, "scan_id": "Validation_47895536", "origin_question": "Assuming the direction you are facing at the end of the turn 3 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Left", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 3 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_47895536/Validation_47895536_47895536_455469.635.png", "Validation_47895536/Validation_47895536_47895536_455471.634.png", "Validation_47895536/Validation_47895536_47895536_455473.633.png", "Validation_47895536/Validation_47895536_47895536_455475.649.png" ], "turn_id": 5, "scan_id": "Validation_47895536", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47895536/Validation_47895536_47895536_455477.648.png", "Validation_47895536/Validation_47895536_47895536_455479.647.png", "Validation_47895536/Validation_47895536_47895536_455481.647.png", "Validation_47895536/Validation_47895536_47895536_455483.646.png" ], "turn_id": 6, "scan_id": "Validation_47895536", "origin_question": "Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the large rectangular aluminum-framed window\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the large rectangular aluminum-framed window\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_47895536/Validation_47895536_47895536_455485.645.png", "Validation_47895536/Validation_47895536_47895536_455487.644.png", "Validation_47895536/Validation_47895536_47895536_455489.643.png", "Validation_47895536/Validation_47895536_47895536_455491.643.png" ], "turn_id": 7, "scan_id": "Validation_47895536", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_47895536/Validation_47895536_47895536_455493.642.png", "Validation_47895536/Validation_47895536_47895536_455495.641.png", "Validation_47895536/Validation_47895536_47895536_455497.640.png", "Validation_47895536/Validation_47895536_47895536_455499.639.png" ], "turn_id": 8, "scan_id": "Validation_47895536", "origin_question": "Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "131.42", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "Validation_47429987/Validation_47429987_47429987_16723.007.png", "Validation_47429987/Validation_47429987_47429987_16724.490.png", "Validation_47429987/Validation_47429987_47429987_16725.989.png", "Validation_47429987/Validation_47429987_47429987_16727.488.png" ], "turn_id": 1, "scan_id": "Validation_47429987", "origin_question": "Remember, how many chair(s) have you seen so far? ", "option": [], "answer": "2", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many chair(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_47429987/Validation_47429987_47429987_16728.988.png", "Validation_47429987/Validation_47429987_47429987_16730.487.png", "Validation_47429987/Validation_47429987_47429987_16731.987.png", "Validation_47429987/Validation_47429987_47429987_16733.486.png" ], "turn_id": 2, "scan_id": "Validation_47429987", "origin_question": "In which turn did you see both \"the rectangular table with the light wood finish\" and \"the large rectangular window with the white frame\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the rectangular table with the light wood finish\" and \"the large rectangular window with the white frame\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47429987/Validation_47429987_47429987_16734.985.png", "Validation_47429987/Validation_47429987_47429987_16736.485.png", "Validation_47429987/Validation_47429987_47429987_16737.984.png", "Validation_47429987/Validation_47429987_47429987_16739.483.png" ], "turn_id": 3, "scan_id": "Validation_47429987", "origin_question": "", "option": [], "answer": "", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "Validation_47429987/Validation_47429987_47429987_16740.983.png", "Validation_47429987/Validation_47429987_47429987_16742.482.png", "Validation_47429987/Validation_47429987_47429987_16743.982.png", "Validation_47429987/Validation_47429987_47429987_16745.481.png" ], "turn_id": 4, "scan_id": "Validation_47429987", "origin_question": "Remember, how many chair(s) have you seen so far? ", "option": [], "answer": "2", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many chair(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_47429987/Validation_47429987_47429987_16746.997.png", "Validation_47429987/Validation_47429987_47429987_16748.496.png", "Validation_47429987/Validation_47429987_47429987_16749.996.png", "Validation_47429987/Validation_47429987_47429987_16751.495.png" ], "turn_id": 5, "scan_id": "Validation_47429987", "origin_question": "Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_47429987/Validation_47429987_47429987_16752.995.png", "Validation_47429987/Validation_47429987_47429987_16754.494.png", "Validation_47429987/Validation_47429987_47429987_16755.993.png", "Validation_47429987/Validation_47429987_47429987_16757.493.png" ], "turn_id": 6, "scan_id": "Validation_47429987", "origin_question": "Is the rectangular table with the light wood finish to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the rectangular table with the light wood finish to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47895542/Validation_47895542_47895542_455830.224.png", "Validation_47895542/Validation_47895542_47895542_455831.707.png", "Validation_47895542/Validation_47895542_47895542_455833.206.png", "Validation_47895542/Validation_47895542_47895542_455834.706.png" ], "turn_id": 1, "scan_id": "Validation_47895542", "origin_question": "Remember, have you seen any sink(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any sink(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47895542/Validation_47895542_47895542_455836.205.png", "Validation_47895542/Validation_47895542_47895542_455837.705.png", "Validation_47895542/Validation_47895542_47895542_455839.204.png", "Validation_47895542/Validation_47895542_47895542_455840.704.png" ], "turn_id": 2, "scan_id": "Validation_47895542", "origin_question": "Remember, have you seen any furniture(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any furniture(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_47895542/Validation_47895542_47895542_455842.203.png", "Validation_47895542/Validation_47895542_47895542_455843.702.png", "Validation_47895542/Validation_47895542_47895542_455845.202.png", "Validation_47895542/Validation_47895542_47895542_455846.701.png" ], "turn_id": 3, "scan_id": "Validation_47895542", "origin_question": "When did you first discover \"the modern stainless steel four-burner stove\"? (index of the turn)", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the modern stainless steel four-burner stove\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "Validation_47895542/Validation_47895542_47895542_455848.200.png", "Validation_47895542/Validation_47895542_47895542_455849.700.png", "Validation_47895542/Validation_47895542_47895542_455851.199.png", "Validation_47895542/Validation_47895542_47895542_455852.699.png" ], "turn_id": 4, "scan_id": "Validation_47895542", "origin_question": "Remember, how many body loofah(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many body loofah(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_47895542/Validation_47895542_47895542_455854.198.png", "Validation_47895542/Validation_47895542_47895542_455855.697.png", "Validation_47895542/Validation_47895542_47895542_455857.197.png", "Validation_47895542/Validation_47895542_47895542_455858.696.png" ], "turn_id": 5, "scan_id": "Validation_47895542", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_47895542/Validation_47895542_47895542_455860.212.png", "Validation_47895542/Validation_47895542_47895542_455861.712.png", "Validation_47895542/Validation_47895542_47895542_455863.211.png", "Validation_47895542/Validation_47895542_47895542_455864.710.png" ], "turn_id": 6, "scan_id": "Validation_47895542", "origin_question": "When did you first discover \"the refrigerator\"? (index of the turn)", "option": [], "answer": "5", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the refrigerator\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47331069/Validation_47331069_47331069_131144.388.png", "Validation_47331069/Validation_47331069_47331069_131146.354.png", "Validation_47331069/Validation_47331069_47331069_131148.353.png", "Validation_47331069/Validation_47331069_47331069_131150.352.png", "Validation_47331069/Validation_47331069_47331069_131152.368.png" ], "turn_id": 1, "scan_id": "Validation_47331069", "origin_question": "Remember, have you seen any teapot(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any teapot(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47331069/Validation_47331069_47331069_131154.367.png", "Validation_47331069/Validation_47331069_47331069_131156.367.png", "Validation_47331069/Validation_47331069_47331069_131158.366.png", "Validation_47331069/Validation_47331069_47331069_131160.365.png", "Validation_47331069/Validation_47331069_47331069_131162.364.png" ], "turn_id": 2, "scan_id": "Validation_47331069", "origin_question": "In which turn were you closest to \"the sink\"? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the sink\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_47331069/Validation_47331069_47331069_131164.363.png", "Validation_47331069/Validation_47331069_47331069_131166.362.png", "Validation_47331069/Validation_47331069_47331069_131168.362.png", "Validation_47331069/Validation_47331069_47331069_131170.361.png", "Validation_47331069/Validation_47331069_47331069_131172.360.png" ], "turn_id": 3, "scan_id": "Validation_47331069", "origin_question": "By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the sink\", Y: \"the sleek toilet with the wooden seat cover\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the sink\", Y: \"the sleek toilet with the wooden seat cover\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47331069/Validation_47331069_47331069_131174.359.png", "Validation_47331069/Validation_47331069_47331069_131176.359.png", "Validation_47331069/Validation_47331069_47331069_131178.358.png", "Validation_47331069/Validation_47331069_47331069_131180.357.png", "Validation_47331069/Validation_47331069_47331069_131182.356.png" ], "turn_id": 4, "scan_id": "Validation_47331069", "origin_question": "Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the sleek toilet with the wooden seat cover\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the sleek toilet with the wooden seat cover\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_47331069/Validation_47331069_47331069_131184.355.png", "Validation_47331069/Validation_47331069_47331069_131186.354.png", "Validation_47331069/Validation_47331069_47331069_131188.354.png", "Validation_47331069/Validation_47331069_47331069_131190.353.png", "Validation_47331069/Validation_47331069_47331069_131192.352.png" ], "turn_id": 5, "scan_id": "Validation_47331069", "origin_question": "Is the sink to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the sink to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_47331069/Validation_47331069_47331069_131194.368.png", "Validation_47331069/Validation_47331069_47331069_131196.367.png", "Validation_47331069/Validation_47331069_47331069_131198.366.png", "Validation_47331069/Validation_47331069_47331069_131200.365.png", "Validation_47331069/Validation_47331069_47331069_131202.365.png" ], "turn_id": 6, "scan_id": "Validation_47331069", "origin_question": "In which turn did you see both \"the sink\" and \"the mop\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the sink\" and \"the mop\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_47331069/Validation_47331069_47331069_131204.364.png", "Validation_47331069/Validation_47331069_47331069_131206.363.png", "Validation_47331069/Validation_47331069_47331069_131208.362.png", "Validation_47331069/Validation_47331069_47331069_131210.361.png", "Validation_47331069/Validation_47331069_47331069_131212.361.png" ], "turn_id": 7, "scan_id": "Validation_47331069", "origin_question": "In which turn did you see both \"the sink\" and \"the mop\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the sink\" and \"the mop\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_47331069/Validation_47331069_47331069_131214.360.png", "Validation_47331069/Validation_47331069_47331069_131216.359.png", "Validation_47331069/Validation_47331069_47331069_131218.358.png", "Validation_47331069/Validation_47331069_47331069_131220.357.png", "Validation_47331069/Validation_47331069_47331069_131222.356.png" ], "turn_id": 8, "scan_id": "Validation_47331069", "origin_question": "Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "118.66", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47331069/Validation_47331069_47331069_131224.356.png", "Validation_47331069/Validation_47331069_47331069_131226.355.png", "Validation_47331069/Validation_47331069_47331069_131228.354.png", "Validation_47331069/Validation_47331069_47331069_131230.353.png", "Validation_47331069/Validation_47331069_47331069_131232.352.png" ], "turn_id": 9, "scan_id": "Validation_47331069", "origin_question": "", "option": [], "answer": "", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_47331069/Validation_47331069_47331069_131234.352.png", "Validation_47331069/Validation_47331069_47331069_131236.367.png", "Validation_47331069/Validation_47331069_47331069_131238.367.png", "Validation_47331069/Validation_47331069_47331069_131240.366.png", "Validation_47331069/Validation_47331069_47331069_131242.365.png" ], "turn_id": 10, "scan_id": "Validation_47331069", "origin_question": "Using your orientation at the end of turn 6 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "137.29", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 6 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42898811/Validation_42898811_42898811_167471.420.png", "Validation_42898811/Validation_42898811_42898811_167473.386.png", "Validation_42898811/Validation_42898811_42898811_167475.402.png", "Validation_42898811/Validation_42898811_42898811_167477.401.png", "Validation_42898811/Validation_42898811_42898811_167479.400.png" ], "turn_id": 1, "scan_id": "Validation_42898811", "origin_question": "Remember, have you seen any cabinet(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any cabinet(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_42898811/Validation_42898811_42898811_167481.399.png", "Validation_42898811/Validation_42898811_42898811_167483.398.png", "Validation_42898811/Validation_42898811_42898811_167485.397.png", "Validation_42898811/Validation_42898811_42898811_167487.397.png", "Validation_42898811/Validation_42898811_42898811_167489.396.png" ], "turn_id": 2, "scan_id": "Validation_42898811", "origin_question": "When did you first discover \"the rectangular modern frame showcasing seaside artwork\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the rectangular modern frame showcasing seaside artwork\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_42898811/Validation_42898811_42898811_167491.395.png", "Validation_42898811/Validation_42898811_42898811_167493.394.png", "Validation_42898811/Validation_42898811_42898811_167495.393.png", "Validation_42898811/Validation_42898811_42898811_167497.393.png", "Validation_42898811/Validation_42898811_42898811_167499.392.png" ], "turn_id": 3, "scan_id": "Validation_42898811", "origin_question": "What will be the first-time appearance order of the following categories: cabinet, clothes, sink?", "option": [ "sink, cabinet, clothes", "cabinet, sink, clothes", "cabinet, clothes, sink", "clothes, cabinet, sink" ], "answer": "cabinet, clothes, sink", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: cabinet, clothes, sink? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['sink, cabinet, clothes', 'cabinet, sink, clothes', 'cabinet, clothes, sink', 'clothes, cabinet, sink'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_42898811/Validation_42898811_42898811_167501.391.png", "Validation_42898811/Validation_42898811_42898811_167503.390.png", "Validation_42898811/Validation_42898811_42898811_167505.389.png", "Validation_42898811/Validation_42898811_42898811_167507.388.png", "Validation_42898811/Validation_42898811_42898811_167509.388.png" ], "turn_id": 4, "scan_id": "Validation_42898811", "origin_question": "What will be the first-time appearance order of the following categories: cabinet, picture, sink?", "option": [ "sink, cabinet, picture", "cabinet, picture, sink", "cabinet, sink, picture", "sink, picture, cabinet" ], "answer": "cabinet, picture, sink", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: cabinet, picture, sink? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['sink, cabinet, picture', 'cabinet, picture, sink', 'cabinet, sink, picture', 'sink, picture, cabinet'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_42898811/Validation_42898811_42898811_167511.387.png", "Validation_42898811/Validation_42898811_42898811_167513.386.png", "Validation_42898811/Validation_42898811_42898811_167515.402.png", "Validation_42898811/Validation_42898811_42898811_167517.401.png", "Validation_42898811/Validation_42898811_42898811_167519.400.png" ], "turn_id": 5, "scan_id": "Validation_42898811", "origin_question": "Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the vibrant red satin clothes among dark shirts\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the vibrant red satin clothes among dark shirts\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_42898811/Validation_42898811_42898811_167521.399.png", "Validation_42898811/Validation_42898811_42898811_167523.398.png", "Validation_42898811/Validation_42898811_42898811_167525.398.png", "Validation_42898811/Validation_42898811_42898811_167527.397.png", "Validation_42898811/Validation_42898811_42898811_167529.396.png" ], "turn_id": 6, "scan_id": "Validation_42898811", "origin_question": "Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_42898811/Validation_42898811_42898811_167531.395.png", "Validation_42898811/Validation_42898811_42898811_167533.394.png", "Validation_42898811/Validation_42898811_42898811_167535.394.png", "Validation_42898811/Validation_42898811_42898811_167537.393.png", "Validation_42898811/Validation_42898811_42898811_167539.392.png" ], "turn_id": 7, "scan_id": "Validation_42898811", "origin_question": "At the end of which turn were both of \"the rectangular modern frame showcasing seaside artwork\" and \"the modern rectangular stainless steel sink\" on your right side?", "option": [], "answer": "1", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the rectangular modern frame showcasing seaside artwork\" and \"the modern rectangular stainless steel sink\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_42898811/Validation_42898811_42898811_167541.391.png", "Validation_42898811/Validation_42898811_42898811_167543.390.png", "Validation_42898811/Validation_42898811_42898811_167545.389.png", "Validation_42898811/Validation_42898811_42898811_167547.389.png", "Validation_42898811/Validation_42898811_42898811_167549.388.png" ], "turn_id": 8, "scan_id": "Validation_42898811", "origin_question": "Using your orientation at the end of turn 5 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "63.53", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 5 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_42898811/Validation_42898811_42898811_167551.387.png", "Validation_42898811/Validation_42898811_42898811_167553.386.png", "Validation_42898811/Validation_42898811_42898811_167555.385.png", "Validation_42898811/Validation_42898811_42898811_167557.401.png", "Validation_42898811/Validation_42898811_42898811_167559.400.png" ], "turn_id": 9, "scan_id": "Validation_42898811", "origin_question": "In which turn were you farthest from \"the vibrant red satin clothes among dark shirts\"? ", "option": [], "answer": "1", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the vibrant red satin clothes among dark shirts\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47430007/Validation_47430007_47430007_17537.785.png", "Validation_47430007/Validation_47430007_47430007_17539.268.png", "Validation_47430007/Validation_47430007_47430007_17540.768.png", "Validation_47430007/Validation_47430007_47430007_17542.267.png" ], "turn_id": 1, "scan_id": "Validation_47430007", "origin_question": "Remember, have you seen any bathtub(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any bathtub(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_47430007/Validation_47430007_47430007_17543.766.png", "Validation_47430007/Validation_47430007_47430007_17545.266.png", "Validation_47430007/Validation_47430007_47430007_17546.765.png", "Validation_47430007/Validation_47430007_47430007_17548.264.png" ], "turn_id": 2, "scan_id": "Validation_47430007", "origin_question": "Please recall \"dark, sturdy cardboard boxes with the parallelepiped shape\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.38", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"dark, sturdy cardboard boxes with the parallelepiped shape\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_47430007/Validation_47430007_47430007_17549.764.png", "Validation_47430007/Validation_47430007_47430007_17551.263.png", "Validation_47430007/Validation_47430007_47430007_17552.763.png", "Validation_47430007/Validation_47430007_47430007_17554.262.png" ], "turn_id": 3, "scan_id": "Validation_47430007", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "84.82", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_47430007/Validation_47430007_47430007_17555.761.png", "Validation_47430007/Validation_47430007_47430007_17557.261.png", "Validation_47430007/Validation_47430007_47430007_17558.760.png", "Validation_47430007/Validation_47430007_47430007_17560.260.png" ], "turn_id": 4, "scan_id": "Validation_47430007", "origin_question": "When did you first discover \"the rectangular light neutral-colored bathtub\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the rectangular light neutral-colored bathtub\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42446093/Validation_42446093_42446093_350095.864.png", "Validation_42446093/Validation_42446093_42446093_350097.863.png", "Validation_42446093/Validation_42446093_42446093_350099.862.png", "Validation_42446093/Validation_42446093_42446093_350101.862.png", "Validation_42446093/Validation_42446093_42446093_350103.861.png" ], "turn_id": 1, "scan_id": "Validation_42446093", "origin_question": "Remember, have you seen any switch(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any switch(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_42446093/Validation_42446093_42446093_350105.860.png", "Validation_42446093/Validation_42446093_42446093_350107.859.png", "Validation_42446093/Validation_42446093_42446093_350109.875.png", "Validation_42446093/Validation_42446093_42446093_350111.874.png", "Validation_42446093/Validation_42446093_42446093_350113.873.png" ], "turn_id": 2, "scan_id": "Validation_42446093", "origin_question": "When did you first discover \"the modern cylindrical white chandelier with the minimalist design\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the modern cylindrical white chandelier with the minimalist design\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_42446093/Validation_42446093_42446093_350115.873.png", "Validation_42446093/Validation_42446093_42446093_350117.872.png", "Validation_42446093/Validation_42446093_42446093_350119.871.png", "Validation_42446093/Validation_42446093_42446093_350121.870.png", "Validation_42446093/Validation_42446093_42446093_350123.869.png" ], "turn_id": 3, "scan_id": "Validation_42446093", "origin_question": "Is the bed to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the bed to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_42446093/Validation_42446093_42446093_350125.868.png", "Validation_42446093/Validation_42446093_42446093_350127.868.png", "Validation_42446093/Validation_42446093_42446093_350129.867.png", "Validation_42446093/Validation_42446093_42446093_350131.866.png", "Validation_42446093/Validation_42446093_42446093_350133.865.png" ], "turn_id": 4, "scan_id": "Validation_42446093", "origin_question": "Is the rectangular pink case made of textile material to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the rectangular pink case made of textile material to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_42446093/Validation_42446093_42446093_350135.864.png", "Validation_42446093/Validation_42446093_42446093_350137.864.png", "Validation_42446093/Validation_42446093_42446093_350139.863.png", "Validation_42446093/Validation_42446093_42446093_350141.862.png", "Validation_42446093/Validation_42446093_42446093_350143.861.png" ], "turn_id": 5, "scan_id": "Validation_42446093", "origin_question": "Which two objects are on the same side of you now? \"the rectangular pink case made of textile material\", \"the modern cylindrical white chandelier with the minimalist design\", \"the large rectangular wooden frame with the lattice design\". ", "option": [ "the rectangular pink case made of textile material, the modern cylindrical white chandelier with the minimalist design", "the rectangular pink case made of textile material, the large rectangular wooden frame with the lattice design", "the modern cylindrical white chandelier with the minimalist design, the large rectangular wooden frame with the lattice design" ], "answer": "the rectangular pink case made of textile material, the modern cylindrical white chandelier with the minimalist design", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the rectangular pink case made of textile material\", \"the modern cylindrical white chandelier with the minimalist design\", \"the large rectangular wooden frame with the lattice design\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular pink case made of textile material, the modern cylindrical white chandelier with the minimalist design', 'the rectangular pink case made of textile material, the large rectangular wooden frame with the lattice design', 'the modern cylindrical white chandelier with the minimalist design, the large rectangular wooden frame with the lattice design'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_42446093/Validation_42446093_42446093_350145.860.png", "Validation_42446093/Validation_42446093_42446093_350147.860.png", "Validation_42446093/Validation_42446093_42446093_350149.859.png", "Validation_42446093/Validation_42446093_42446093_350151.858.png", "Validation_42446093/Validation_42446093_42446093_350153.874.png" ], "turn_id": 6, "scan_id": "Validation_42446093", "origin_question": "Which direction is the modern cylindrical white chandelier with the minimalist design in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the modern cylindrical white chandelier with the minimalist design in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "Validation_42446093/Validation_42446093_42446093_350155.873.png", "Validation_42446093/Validation_42446093_42446093_350157.872.png", "Validation_42446093/Validation_42446093_42446093_350159.871.png", "Validation_42446093/Validation_42446093_42446093_350161.870.png", "Validation_42446093/Validation_42446093_42446093_350163.870.png" ], "turn_id": 7, "scan_id": "Validation_42446093", "origin_question": "Among these three objects(\"the bed\",\"the modern cylindrical white chandelier with the minimalist design\",\"the large rectangular wooden frame with the lattice design\" ), which one is the closest to you?", "option": [ "the bed", "the modern cylindrical white chandelier with the minimalist design", "the large rectangular wooden frame with the lattice design" ], "answer": "the modern cylindrical white chandelier with the minimalist design", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the bed\",\"the modern cylindrical white chandelier with the minimalist design\",\"the large rectangular wooden frame with the lattice design\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the bed', 'the modern cylindrical white chandelier with the minimalist design', 'the large rectangular wooden frame with the lattice design'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_42446093/Validation_42446093_42446093_350165.869.png", "Validation_42446093/Validation_42446093_42446093_350167.868.png", "Validation_42446093/Validation_42446093_42446093_350169.867.png", "Validation_42446093/Validation_42446093_42446093_350171.866.png", "Validation_42446093/Validation_42446093_42446093_350173.865.png" ], "turn_id": 8, "scan_id": "Validation_42446093", "origin_question": "How far is your current position from where you were at the end of the turn 1? (in meters)", "option": [], "answer": "2.28", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 1? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_42446093/Validation_42446093_42446093_350175.865.png", "Validation_42446093/Validation_42446093_42446093_350177.864.png", "Validation_42446093/Validation_42446093_42446093_350179.863.png", "Validation_42446093/Validation_42446093_42446093_350181.862.png", "Validation_42446093/Validation_42446093_42446093_350183.861.png" ], "turn_id": 9, "scan_id": "Validation_42446093", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the modern cylindrical white chandelier with the minimalist design\" from your position?", "option": [], "answer": "109.26", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the modern cylindrical white chandelier with the minimalist design\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_42446093/Validation_42446093_42446093_350185.861.png", "Validation_42446093/Validation_42446093_42446093_350187.860.png", "Validation_42446093/Validation_42446093_42446093_350189.859.png", "Validation_42446093/Validation_42446093_42446093_350191.858.png", "Validation_42446093/Validation_42446093_42446093_350193.874.png" ], "turn_id": 10, "scan_id": "Validation_42446093", "origin_question": "In which turn were you closest to \"the bed\"? ", "option": [], "answer": "8", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the bed\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47331266/Validation_47331266_47331266_89823.527.png", "Validation_47331266/Validation_47331266_47331266_89825.493.png", "Validation_47331266/Validation_47331266_47331266_89827.492.png", "Validation_47331266/Validation_47331266_47331266_89829.491.png" ], "turn_id": 1, "scan_id": "Validation_47331266", "origin_question": "Remember, have you seen any salt(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any salt(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_47331266/Validation_47331266_47331266_89831.490.png", "Validation_47331266/Validation_47331266_47331266_89833.506.png", "Validation_47331266/Validation_47331266_47331266_89835.505.png", "Validation_47331266/Validation_47331266_47331266_89837.504.png" ], "turn_id": 2, "scan_id": "Validation_47331266", "origin_question": "At the end of which turn were both of \"the medium white cylindrical ceramic flowerpot\" and \"the rectangular window with the sleek white frame\" on your left side?", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the medium white cylindrical ceramic flowerpot\" and \"the rectangular window with the sleek white frame\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_47331266/Validation_47331266_47331266_89839.504.png", "Validation_47331266/Validation_47331266_47331266_89841.503.png", "Validation_47331266/Validation_47331266_47331266_89843.502.png", "Validation_47331266/Validation_47331266_47331266_89845.501.png" ], "turn_id": 3, "scan_id": "Validation_47331266", "origin_question": "Is the rectangular window with the sleek white frame to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the rectangular window with the sleek white frame to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_47331266/Validation_47331266_47331266_89847.500.png", "Validation_47331266/Validation_47331266_47331266_89849.500.png", "Validation_47331266/Validation_47331266_47331266_89851.499.png", "Validation_47331266/Validation_47331266_47331266_89853.498.png" ], "turn_id": 4, "scan_id": "Validation_47331266", "origin_question": "Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "88.56", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_47331266/Validation_47331266_47331266_89855.497.png", "Validation_47331266/Validation_47331266_47331266_89857.496.png", "Validation_47331266/Validation_47331266_47331266_89859.496.png", "Validation_47331266/Validation_47331266_47331266_89861.495.png" ], "turn_id": 5, "scan_id": "Validation_47331266", "origin_question": "When did you first discover \"the rectangular window with the sleek white frame\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the rectangular window with the sleek white frame\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_47331266/Validation_47331266_47331266_89863.494.png", "Validation_47331266/Validation_47331266_47331266_89865.493.png", "Validation_47331266/Validation_47331266_47331266_89867.492.png", "Validation_47331266/Validation_47331266_47331266_89869.492.png" ], "turn_id": 6, "scan_id": "Validation_47331266", "origin_question": "In which turn did you see both \"the medium-sized white ceramic teapot with the round shape\" and \"the large ceiling light with glass shades\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the medium-sized white ceramic teapot with the round shape\" and \"the large ceiling light with glass shades\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_47331266/Validation_47331266_47331266_89871.491.png", "Validation_47331266/Validation_47331266_47331266_89873.490.png", "Validation_47331266/Validation_47331266_47331266_89875.506.png", "Validation_47331266/Validation_47331266_47331266_89877.505.png" ], "turn_id": 7, "scan_id": "Validation_47331266", "origin_question": "Is the white rectangular metal radiator with vertical fins to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the white rectangular metal radiator with vertical fins to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_47331266/Validation_47331266_47331266_89879.504.png", "Validation_47331266/Validation_47331266_47331266_89881.503.png", "Validation_47331266/Validation_47331266_47331266_89883.503.png", "Validation_47331266/Validation_47331266_47331266_89885.502.png" ], "turn_id": 8, "scan_id": "Validation_47331266", "origin_question": "Which two objects are on the same side of you now? \"the medium-sized white ceramic teapot with the round shape\", \"the large ceiling light with glass shades\", \"the sleek rectangular black monitor made of plastic and glass\". ", "option": [ "the medium-sized white ceramic teapot with the round shape, the large ceiling light with glass shades", "the medium-sized white ceramic teapot with the round shape, the sleek rectangular black monitor made of plastic and glass", "the large ceiling light with glass shades, the sleek rectangular black monitor made of plastic and glass" ], "answer": "the medium-sized white ceramic teapot with the round shape, the large ceiling light with glass shades", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the medium-sized white ceramic teapot with the round shape\", \"the large ceiling light with glass shades\", \"the sleek rectangular black monitor made of plastic and glass\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the medium-sized white ceramic teapot with the round shape, the large ceiling light with glass shades', 'the medium-sized white ceramic teapot with the round shape, the sleek rectangular black monitor made of plastic and glass', 'the large ceiling light with glass shades, the sleek rectangular black monitor made of plastic and glass'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47331266/Validation_47331266_47331266_89887.501.png", "Validation_47331266/Validation_47331266_47331266_89889.500.png", "Validation_47331266/Validation_47331266_47331266_89891.499.png", "Validation_47331266/Validation_47331266_47331266_89893.499.png" ], "turn_id": 9, "scan_id": "Validation_47331266", "origin_question": "Which direction is the large ceiling light with glass shades in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 9 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the large ceiling light with glass shades in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42897688/Validation_42897688_42897688_479144.097.png", "Validation_42897688/Validation_42897688_42897688_479146.080.png", "Validation_42897688/Validation_42897688_42897688_479148.079.png", "Validation_42897688/Validation_42897688_42897688_479150.078.png", "Validation_42897688/Validation_42897688_42897688_479152.078.png" ], "turn_id": 1, "scan_id": "Validation_42897688", "origin_question": "Remember, have you seen any guitar(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any guitar(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_42897688/Validation_42897688_42897688_479154.077.png", "Validation_42897688/Validation_42897688_42897688_479156.076.png", "Validation_42897688/Validation_42897688_42897688_479158.092.png", "Validation_42897688/Validation_42897688_42897688_479160.091.png", "Validation_42897688/Validation_42897688_42897688_479162.090.png" ], "turn_id": 2, "scan_id": "Validation_42897688", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the rectangular mirror with an ornate brown wooden frame\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the rectangular mirror with an ornate brown wooden frame\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_42897688/Validation_42897688_42897688_479164.089.png", "Validation_42897688/Validation_42897688_42897688_479166.089.png", "Validation_42897688/Validation_42897688_42897688_479168.088.png", "Validation_42897688/Validation_42897688_42897688_479170.087.png", "Validation_42897688/Validation_42897688_42897688_479172.086.png" ], "turn_id": 3, "scan_id": "Validation_42897688", "origin_question": "Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Right", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "Validation_42897688/Validation_42897688_42897688_479174.085.png", "Validation_42897688/Validation_42897688_42897688_479176.084.png", "Validation_42897688/Validation_42897688_42897688_479178.084.png", "Validation_42897688/Validation_42897688_42897688_479180.083.png", "Validation_42897688/Validation_42897688_42897688_479182.082.png" ], "turn_id": 4, "scan_id": "Validation_42897688", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the medium-brown queen-sized bed with curved edges\"; \"the wooden chair with the cushioned beige seat\"; \"the white rectangular steel radiator with vertical fins\" ", "option": [ "the medium-brown queen-sized bed with curved edges", "the wooden chair with the cushioned beige seat", "the white rectangular steel radiator with vertical fins" ], "answer": "the white rectangular steel radiator with vertical fins", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the medium-brown queen-sized bed with curved edges\"; \"the wooden chair with the cushioned beige seat\"; \"the white rectangular steel radiator with vertical fins\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the medium-brown queen-sized bed with curved edges', 'the wooden chair with the cushioned beige seat', 'the white rectangular steel radiator with vertical fins'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_42897688/Validation_42897688_42897688_479184.081.png", "Validation_42897688/Validation_42897688_42897688_479186.080.png", "Validation_42897688/Validation_42897688_42897688_479188.079.png", "Validation_42897688/Validation_42897688_42897688_479190.079.png", "Validation_42897688/Validation_42897688_42897688_479192.078.png" ], "turn_id": 5, "scan_id": "Validation_42897688", "origin_question": "Using your orientation at the end of turn 4 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "98.42", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 4 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_42897688/Validation_42897688_42897688_479194.077.png", "Validation_42897688/Validation_42897688_42897688_479196.076.png", "Validation_42897688/Validation_42897688_42897688_479198.075.png", "Validation_42897688/Validation_42897688_42897688_479200.091.png", "Validation_42897688/Validation_42897688_42897688_479202.090.png" ], "turn_id": 6, "scan_id": "Validation_42897688", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the white rectangular steel radiator with vertical fins\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the white rectangular steel radiator with vertical fins\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_42897688/Validation_42897688_42897688_479204.090.png", "Validation_42897688/Validation_42897688_42897688_479206.089.png", "Validation_42897688/Validation_42897688_42897688_479208.088.png", "Validation_42897688/Validation_42897688_42897688_479210.087.png", "Validation_42897688/Validation_42897688_42897688_479212.086.png" ], "turn_id": 7, "scan_id": "Validation_42897688", "origin_question": "Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the warm medium-toned rectangular wooden door\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the warm medium-toned rectangular wooden door\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_42897688/Validation_42897688_42897688_479214.085.png", "Validation_42897688/Validation_42897688_42897688_479216.085.png", "Validation_42897688/Validation_42897688_42897688_479218.084.png", "Validation_42897688/Validation_42897688_42897688_479220.083.png", "Validation_42897688/Validation_42897688_42897688_479222.082.png" ], "turn_id": 8, "scan_id": "Validation_42897688", "origin_question": "Is the multi-pane rectangular wooden window to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the multi-pane rectangular wooden window to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_42897688/Validation_42897688_42897688_479224.081.png", "Validation_42897688/Validation_42897688_42897688_479226.080.png", "Validation_42897688/Validation_42897688_42897688_479228.080.png", "Validation_42897688/Validation_42897688_42897688_479230.079.png", "Validation_42897688/Validation_42897688_42897688_479232.078.png" ], "turn_id": 9, "scan_id": "Validation_42897688", "origin_question": "By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular mirror with an ornate brown wooden frame\", Y: \"the multi-pane rectangular wooden window\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to Y and farther away from X", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular mirror with an ornate brown wooden frame\", Y: \"the multi-pane rectangular wooden window\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_42897688/Validation_42897688_42897688_479234.077.png", "Validation_42897688/Validation_42897688_42897688_479236.076.png", "Validation_42897688/Validation_42897688_42897688_479238.075.png", "Validation_42897688/Validation_42897688_42897688_479240.091.png", "Validation_42897688/Validation_42897688_42897688_479242.090.png" ], "turn_id": 10, "scan_id": "Validation_42897688", "origin_question": "Please recall \"the rectangular mirror with an ornate brown wooden frame\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "3.82", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the rectangular mirror with an ornate brown wooden frame\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_44358499/Validation_44358499_44358499_50557.210.png", "Validation_44358499/Validation_44358499_44358499_50559.192.png", "Validation_44358499/Validation_44358499_44358499_50561.191.png", "Validation_44358499/Validation_44358499_44358499_50563.191.png", "Validation_44358499/Validation_44358499_44358499_50565.190.png" ], "turn_id": 1, "scan_id": "Validation_44358499", "origin_question": "Remember, have you seen any bathtub(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any bathtub(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_44358499/Validation_44358499_44358499_50567.206.png", "Validation_44358499/Validation_44358499_44358499_50569.205.png", "Validation_44358499/Validation_44358499_44358499_50571.204.png", "Validation_44358499/Validation_44358499_44358499_50573.203.png", "Validation_44358499/Validation_44358499_44358499_50575.202.png" ], "turn_id": 2, "scan_id": "Validation_44358499", "origin_question": "In which turn were you farthest from \"the large, metallic silver cylindrical bin\"? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the large, metallic silver cylindrical bin\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_44358499/Validation_44358499_44358499_50577.202.png", "Validation_44358499/Validation_44358499_44358499_50579.201.png", "Validation_44358499/Validation_44358499_44358499_50581.200.png", "Validation_44358499/Validation_44358499_44358499_50583.199.png", "Validation_44358499/Validation_44358499_44358499_50585.198.png" ], "turn_id": 3, "scan_id": "Validation_44358499", "origin_question": "In which turn did you see both \"the modern rectangular stainless steel sink\" and \"the polished silver machine with adjustable browning settings\" simultaneously? ", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the modern rectangular stainless steel sink\" and \"the polished silver machine with adjustable browning settings\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_44358499/Validation_44358499_44358499_50587.197.png", "Validation_44358499/Validation_44358499_44358499_50589.196.png", "Validation_44358499/Validation_44358499_44358499_50591.196.png", "Validation_44358499/Validation_44358499_44358499_50593.195.png", "Validation_44358499/Validation_44358499_44358499_50595.194.png" ], "turn_id": 4, "scan_id": "Validation_44358499", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "3.28", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_44358499/Validation_44358499_44358499_50597.193.png", "Validation_44358499/Validation_44358499_44358499_50599.192.png", "Validation_44358499/Validation_44358499_44358499_50601.192.png", "Validation_44358499/Validation_44358499_44358499_50603.191.png", "Validation_44358499/Validation_44358499_44358499_50605.190.png" ], "turn_id": 5, "scan_id": "Validation_44358499", "origin_question": "At the end of which turn were both of \"the modern cylindrical black fireplace with the glass door\" and \"the large sleek black monitor with the thin bezel\" on your left side?", "option": [], "answer": "4", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the modern cylindrical black fireplace with the glass door\" and \"the large sleek black monitor with the thin bezel\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_44358499/Validation_44358499_44358499_50607.206.png", "Validation_44358499/Validation_44358499_44358499_50609.205.png", "Validation_44358499/Validation_44358499_44358499_50611.204.png", "Validation_44358499/Validation_44358499_44358499_50613.203.png", "Validation_44358499/Validation_44358499_44358499_50615.202.png" ], "turn_id": 6, "scan_id": "Validation_44358499", "origin_question": "Assuming the direction you are facing at the end of the turn 5 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Right", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 5 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_44358499/Validation_44358499_44358499_50617.202.png", "Validation_44358499/Validation_44358499_44358499_50619.201.png", "Validation_44358499/Validation_44358499_44358499_50621.200.png", "Validation_44358499/Validation_44358499_44358499_50623.199.png", "Validation_44358499/Validation_44358499_44358499_50625.198.png" ], "turn_id": 7, "scan_id": "Validation_44358499", "origin_question": "Using your orientation at the end of turn 4 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "146.71", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 4 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_44358499/Validation_44358499_44358499_50627.197.png", "Validation_44358499/Validation_44358499_44358499_50629.197.png", "Validation_44358499/Validation_44358499_44358499_50631.196.png", "Validation_44358499/Validation_44358499_44358499_50633.195.png", "Validation_44358499/Validation_44358499_44358499_50635.194.png" ], "turn_id": 8, "scan_id": "Validation_44358499", "origin_question": "Is the modern rectangular stainless steel sink to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the modern rectangular stainless steel sink to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_44358499/Validation_44358499_44358499_50637.193.png", "Validation_44358499/Validation_44358499_44358499_50639.192.png", "Validation_44358499/Validation_44358499_44358499_50641.192.png", "Validation_44358499/Validation_44358499_44358499_50643.191.png", "Validation_44358499/Validation_44358499_44358499_50645.190.png" ], "turn_id": 9, "scan_id": "Validation_44358499", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the modern rectangular stainless steel sink\" from your position?", "option": [], "answer": "173.77", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the modern rectangular stainless steel sink\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47115469/Validation_47115469_47115469_14011.310.png", "Validation_47115469/Validation_47115469_47115469_14013.292.png", "Validation_47115469/Validation_47115469_47115469_14015.292.png", "Validation_47115469/Validation_47115469_47115469_14017.291.png" ], "turn_id": 1, "scan_id": "Validation_47115469", "origin_question": "Remember, have you seen any door(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any door(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_47115469/Validation_47115469_47115469_14019.290.png", "Validation_47115469/Validation_47115469_47115469_14021.289.png", "Validation_47115469/Validation_47115469_47115469_14023.288.png", "Validation_47115469/Validation_47115469_47115469_14025.288.png" ], "turn_id": 2, "scan_id": "Validation_47115469", "origin_question": "In which turn did you see both \"the circular glossy white bathtub made of acrylic\" and \"the sleek rectangular window with the white frame\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the circular glossy white bathtub made of acrylic\" and \"the sleek rectangular window with the white frame\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_47115469/Validation_47115469_47115469_14027.287.png", "Validation_47115469/Validation_47115469_47115469_14029.286.png", "Validation_47115469/Validation_47115469_47115469_14031.285.png", "Validation_47115469/Validation_47115469_47115469_14033.284.png" ], "turn_id": 3, "scan_id": "Validation_47115469", "origin_question": "Is the circular glossy white bathtub made of acrylic to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the circular glossy white bathtub made of acrylic to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47115469/Validation_47115469_47115469_14035.283.png", "Validation_47115469/Validation_47115469_47115469_14037.283.png", "Validation_47115469/Validation_47115469_47115469_14039.282.png", "Validation_47115469/Validation_47115469_47115469_14041.281.png" ], "turn_id": 4, "scan_id": "Validation_47115469", "origin_question": "Which direction is the slender rectangular cabinet with the soft color palette in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the slender rectangular cabinet with the soft color palette in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_47115469/Validation_47115469_47115469_14043.280.png", "Validation_47115469/Validation_47115469_47115469_14045.279.png", "Validation_47115469/Validation_47115469_47115469_14047.278.png", "Validation_47115469/Validation_47115469_47115469_14049.278.png" ], "turn_id": 5, "scan_id": "Validation_47115469", "origin_question": "Is an oval glossy white porcelain sink to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Is an oval glossy white porcelain sink to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_47115469/Validation_47115469_47115469_14051.293.png", "Validation_47115469/Validation_47115469_47115469_14053.293.png", "Validation_47115469/Validation_47115469_47115469_14055.292.png", "Validation_47115469/Validation_47115469_47115469_14057.291.png" ], "turn_id": 6, "scan_id": "Validation_47115469", "origin_question": "Using your orientation at the end of turn 5 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "132.9", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 5 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47334238/Validation_47334238_47334238_109278.104.png", "Validation_47334238/Validation_47334238_47334238_109280.070.png", "Validation_47334238/Validation_47334238_47334238_109282.069.png", "Validation_47334238/Validation_47334238_47334238_109284.085.png", "Validation_47334238/Validation_47334238_47334238_109286.084.png" ], "turn_id": 1, "scan_id": "Validation_47334238", "origin_question": "Remember, have you seen any curtain(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any curtain(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47334238/Validation_47334238_47334238_109288.083.png", "Validation_47334238/Validation_47334238_47334238_109290.082.png", "Validation_47334238/Validation_47334238_47334238_109292.082.png", "Validation_47334238/Validation_47334238_47334238_109294.081.png", "Validation_47334238/Validation_47334238_47334238_109296.080.png" ], "turn_id": 2, "scan_id": "Validation_47334238", "origin_question": "In which turn were you farthest from \"the dark brown wooden table with floral tablecloth\"? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the dark brown wooden table with floral tablecloth\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_47334238/Validation_47334238_47334238_109298.079.png", "Validation_47334238/Validation_47334238_47334238_109300.078.png", "Validation_47334238/Validation_47334238_47334238_109302.078.png", "Validation_47334238/Validation_47334238_47334238_109304.077.png", "Validation_47334238/Validation_47334238_47334238_109306.076.png" ], "turn_id": 3, "scan_id": "Validation_47334238", "origin_question": "When did you first discover \"the white rectangular door with glass panels\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the white rectangular door with glass panels\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_47334238/Validation_47334238_47334238_109308.075.png", "Validation_47334238/Validation_47334238_47334238_109310.074.png", "Validation_47334238/Validation_47334238_47334238_109312.073.png", "Validation_47334238/Validation_47334238_47334238_109314.073.png", "Validation_47334238/Validation_47334238_47334238_109316.072.png" ], "turn_id": 4, "scan_id": "Validation_47334238", "origin_question": "In which turn did you see both \"the dark brown wooden table with floral tablecloth\" and \"the white rectangular door with glass panels\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the dark brown wooden table with floral tablecloth\" and \"the white rectangular door with glass panels\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_47334238/Validation_47334238_47334238_109318.071.png", "Validation_47334238/Validation_47334238_47334238_109320.070.png", "Validation_47334238/Validation_47334238_47334238_109322.069.png", "Validation_47334238/Validation_47334238_47334238_109324.085.png", "Validation_47334238/Validation_47334238_47334238_109326.084.png" ], "turn_id": 5, "scan_id": "Validation_47334238", "origin_question": "In which turn did you see both \"the dark brown wooden table with floral tablecloth\" and \"the curtain\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the dark brown wooden table with floral tablecloth\" and \"the curtain\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47334238/Validation_47334238_47334238_109328.084.png", "Validation_47334238/Validation_47334238_47334238_109330.083.png", "Validation_47334238/Validation_47334238_47334238_109332.082.png", "Validation_47334238/Validation_47334238_47334238_109334.081.png", "Validation_47334238/Validation_47334238_47334238_109336.080.png" ], "turn_id": 6, "scan_id": "Validation_47334238", "origin_question": "In which turn were you farthest from \"the dark brown wooden table with floral tablecloth\"? ", "option": [], "answer": "1", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the dark brown wooden table with floral tablecloth\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_47334238/Validation_47334238_47334238_109338.080.png", "Validation_47334238/Validation_47334238_47334238_109340.079.png", "Validation_47334238/Validation_47334238_47334238_109342.078.png", "Validation_47334238/Validation_47334238_47334238_109344.077.png", "Validation_47334238/Validation_47334238_47334238_109346.076.png" ], "turn_id": 7, "scan_id": "Validation_47334238", "origin_question": "In which turn did you see both \"the white rectangular door with glass panels\" and \"the brown ceramic vase with floral patterns\" simultaneously? ", "option": [], "answer": "6", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the white rectangular door with glass panels\" and \"the brown ceramic vase with floral patterns\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_47334238/Validation_47334238_47334238_109348.076.png", "Validation_47334238/Validation_47334238_47334238_109350.075.png", "Validation_47334238/Validation_47334238_47334238_109352.074.png", "Validation_47334238/Validation_47334238_47334238_109354.073.png", "Validation_47334238/Validation_47334238_47334238_109356.072.png" ], "turn_id": 8, "scan_id": "Validation_47334238", "origin_question": "What will be the first-time appearance order of the following categories: curtain, door, vase?", "option": [ "vase, door, curtain", "curtain, door, vase", "door, vase, curtain", "door, curtain, vase" ], "answer": "door, curtain, vase", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: curtain, door, vase? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['vase, door, curtain', 'curtain, door, vase', 'door, vase, curtain', 'door, curtain, vase'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_47334238/Validation_47334238_47334238_109358.072.png", "Validation_47334238/Validation_47334238_47334238_109360.071.png", "Validation_47334238/Validation_47334238_47334238_109362.070.png", "Validation_47334238/Validation_47334238_47334238_109364.069.png", "Validation_47334238/Validation_47334238_47334238_109366.085.png" ], "turn_id": 9, "scan_id": "Validation_47334238", "origin_question": "When did you first discover \"the large rectangular mirror with the dark wood frame\"? (index of the turn)", "option": [], "answer": "3", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the large rectangular mirror with the dark wood frame\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_47334238/Validation_47334238_47334238_109368.084.png", "Validation_47334238/Validation_47334238_47334238_109370.083.png", "Validation_47334238/Validation_47334238_47334238_109372.083.png", "Validation_47334238/Validation_47334238_47334238_109374.082.png", "Validation_47334238/Validation_47334238_47334238_109376.081.png" ], "turn_id": 10, "scan_id": "Validation_47334238", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47895783/Validation_47895783_47895783_387003.267.png", "Validation_47895783/Validation_47895783_47895783_387004.750.png", "Validation_47895783/Validation_47895783_47895783_387006.249.png", "Validation_47895783/Validation_47895783_47895783_387007.765.png" ], "turn_id": 1, "scan_id": "Validation_47895783", "origin_question": "Remember, have you seen any bed(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any bed(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47895783/Validation_47895783_47895783_387009.265.png", "Validation_47895783/Validation_47895783_47895783_387010.764.png", "Validation_47895783/Validation_47895783_47895783_387012.264.png", "Validation_47895783/Validation_47895783_47895783_387013.763.png" ], "turn_id": 2, "scan_id": "Validation_47895783", "origin_question": "In which turn were you closest to \"the compact silver cylindrical bin with the pedal mechanism\"? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the compact silver cylindrical bin with the pedal mechanism\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47895783/Validation_47895783_47895783_387015.262.png", "Validation_47895783/Validation_47895783_47895783_387016.762.png", "Validation_47895783/Validation_47895783_47895783_387018.261.png", "Validation_47895783/Validation_47895783_47895783_387019.760.png" ], "turn_id": 3, "scan_id": "Validation_47895783", "origin_question": "In which turn were you closest to \"the tall, sleek white shelf with open compartments\"? ", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the tall, sleek white shelf with open compartments\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_47895783/Validation_47895783_47895783_387021.260.png", "Validation_47895783/Validation_47895783_47895783_387022.759.png", "Validation_47895783/Validation_47895783_47895783_387024.259.png", "Validation_47895783/Validation_47895783_47895783_387025.758.png" ], "turn_id": 4, "scan_id": "Validation_47895783", "origin_question": "In which turn did you see both \"the compact white porcelain sink with softened edges\" and \"the warm white dome-shaped light fixture\" simultaneously? ", "option": [], "answer": "3", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the compact white porcelain sink with softened edges\" and \"the warm white dome-shaped light fixture\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42444966/Validation_42444966_42444966_40329.815.png", "Validation_42444966/Validation_42444966_42444966_40331.781.png", "Validation_42444966/Validation_42444966_42444966_40333.780.png", "Validation_42444966/Validation_42444966_42444966_40335.780.png", "Validation_42444966/Validation_42444966_42444966_40337.795.png" ], "turn_id": 1, "scan_id": "Validation_42444966", "origin_question": "Remember, have you seen any dishwasher(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any dishwasher(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_42444966/Validation_42444966_42444966_40339.795.png", "Validation_42444966/Validation_42444966_42444966_40341.794.png", "Validation_42444966/Validation_42444966_42444966_40343.793.png", "Validation_42444966/Validation_42444966_42444966_40345.792.png", "Validation_42444966/Validation_42444966_42444966_40347.791.png" ], "turn_id": 2, "scan_id": "Validation_42444966", "origin_question": "When did you first discover \"the compact, glossy black stove with multiple burners\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the compact, glossy black stove with multiple burners\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_42444966/Validation_42444966_42444966_40349.790.png", "Validation_42444966/Validation_42444966_42444966_40351.790.png", "Validation_42444966/Validation_42444966_42444966_40353.789.png", "Validation_42444966/Validation_42444966_42444966_40355.788.png", "Validation_42444966/Validation_42444966_42444966_40357.787.png" ], "turn_id": 3, "scan_id": "Validation_42444966", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the refrigerator\", Y: \"the compact, glossy black stove with multiple burners\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "farther away from both X and Y", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the refrigerator\", Y: \"the compact, glossy black stove with multiple burners\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_42444966/Validation_42444966_42444966_40359.786.png", "Validation_42444966/Validation_42444966_42444966_40361.785.png", "Validation_42444966/Validation_42444966_42444966_40363.785.png", "Validation_42444966/Validation_42444966_42444966_40365.784.png", "Validation_42444966/Validation_42444966_42444966_40367.783.png" ], "turn_id": 4, "scan_id": "Validation_42444966", "origin_question": "Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Right", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_42444966/Validation_42444966_42444966_40369.782.png", "Validation_42444966/Validation_42444966_42444966_40371.781.png", "Validation_42444966/Validation_42444966_42444966_40373.780.png", "Validation_42444966/Validation_42444966_42444966_40375.780.png", "Validation_42444966/Validation_42444966_42444966_40377.795.png" ], "turn_id": 5, "scan_id": "Validation_42444966", "origin_question": "Is the medium-sized stainless steel dishwasher with the sleek design to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the medium-sized stainless steel dishwasher with the sleek design to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_42444966/Validation_42444966_42444966_40379.795.png", "Validation_42444966/Validation_42444966_42444966_40381.794.png", "Validation_42444966/Validation_42444966_42444966_40383.793.png", "Validation_42444966/Validation_42444966_42444966_40385.792.png", "Validation_42444966/Validation_42444966_42444966_40387.791.png" ], "turn_id": 6, "scan_id": "Validation_42444966", "origin_question": "Please recall \"the sleek, modern cylindrical stainless steel kettle\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.44", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the sleek, modern cylindrical stainless steel kettle\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42444966/Validation_42444966_42444966_40389.790.png", "Validation_42444966/Validation_42444966_42444966_40391.790.png", "Validation_42444966/Validation_42444966_42444966_40393.789.png", "Validation_42444966/Validation_42444966_42444966_40395.788.png", "Validation_42444966/Validation_42444966_42444966_40397.787.png" ], "turn_id": 7, "scan_id": "Validation_42444966", "origin_question": "Remember, have you seen any kitchenware(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any kitchenware(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_42444966/Validation_42444966_42444966_40399.786.png", "Validation_42444966/Validation_42444966_42444966_40401.785.png", "Validation_42444966/Validation_42444966_42444966_40403.785.png", "Validation_42444966/Validation_42444966_42444966_40405.784.png", "Validation_42444966/Validation_42444966_42444966_40407.783.png" ], "turn_id": 8, "scan_id": "Validation_42444966", "origin_question": "Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_42444966/Validation_42444966_42444966_40409.782.png", "Validation_42444966/Validation_42444966_42444966_40411.781.png", "Validation_42444966/Validation_42444966_42444966_40413.781.png", "Validation_42444966/Validation_42444966_42444966_40415.780.png", "Validation_42444966/Validation_42444966_42444966_40417.779.png" ], "turn_id": 9, "scan_id": "Validation_42444966", "origin_question": "", "option": [], "answer": "", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_42444966/Validation_42444966_42444966_40419.778.png", "Validation_42444966/Validation_42444966_42444966_40421.794.png", "Validation_42444966/Validation_42444966_42444966_40423.793.png", "Validation_42444966/Validation_42444966_42444966_40425.792.png", "Validation_42444966/Validation_42444966_42444966_40427.791.png" ], "turn_id": 10, "scan_id": "Validation_42444966", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the rectangular white glossy sink made of porcelain\" from your position?", "option": [], "answer": "125.47", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the rectangular white glossy sink made of porcelain\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47334102/Validation_47334102_47334102_92772.764.png", "Validation_47334102/Validation_47334102_47334102_92774.747.png", "Validation_47334102/Validation_47334102_47334102_92776.746.png", "Validation_47334102/Validation_47334102_47334102_92778.745.png", "Validation_47334102/Validation_47334102_47334102_92780.744.png" ], "turn_id": 1, "scan_id": "Validation_47334102", "origin_question": "Remember, have you seen any cabinet(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any cabinet(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_47334102/Validation_47334102_47334102_92782.760.png", "Validation_47334102/Validation_47334102_47334102_92784.759.png", "Validation_47334102/Validation_47334102_47334102_92786.758.png", "Validation_47334102/Validation_47334102_47334102_92788.758.png", "Validation_47334102/Validation_47334102_47334102_92790.757.png" ], "turn_id": 2, "scan_id": "Validation_47334102", "origin_question": "Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Left", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_47334102/Validation_47334102_47334102_92792.756.png", "Validation_47334102/Validation_47334102_47334102_92794.755.png", "Validation_47334102/Validation_47334102_47334102_92796.754.png", "Validation_47334102/Validation_47334102_47334102_92798.754.png", "Validation_47334102/Validation_47334102_47334102_92800.753.png" ], "turn_id": 3, "scan_id": "Validation_47334102", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the large, glossy black monitor with the sleek design\", Y: \"the rectangular, beige flowerpot\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the large, glossy black monitor with the sleek design\", Y: \"the rectangular, beige flowerpot\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47334102/Validation_47334102_47334102_92802.752.png", "Validation_47334102/Validation_47334102_47334102_92804.751.png", "Validation_47334102/Validation_47334102_47334102_92806.750.png", "Validation_47334102/Validation_47334102_47334102_92808.750.png", "Validation_47334102/Validation_47334102_47334102_92810.749.png" ], "turn_id": 4, "scan_id": "Validation_47334102", "origin_question": "Which direction is the large, glossy black monitor with the sleek design in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the large, glossy black monitor with the sleek design in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_47334102/Validation_47334102_47334102_92812.748.png", "Validation_47334102/Validation_47334102_47334102_92814.747.png", "Validation_47334102/Validation_47334102_47334102_92816.746.png", "Validation_47334102/Validation_47334102_47334102_92818.746.png", "Validation_47334102/Validation_47334102_47334102_92820.745.png" ], "turn_id": 5, "scan_id": "Validation_47334102", "origin_question": "Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_47334102/Validation_47334102_47334102_92822.761.png", "Validation_47334102/Validation_47334102_47334102_92824.760.png", "Validation_47334102/Validation_47334102_47334102_92826.759.png", "Validation_47334102/Validation_47334102_47334102_92828.758.png", "Validation_47334102/Validation_47334102_47334102_92830.757.png" ], "turn_id": 6, "scan_id": "Validation_47334102", "origin_question": "Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Left", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "Validation_47334102/Validation_47334102_47334102_92832.757.png", "Validation_47334102/Validation_47334102_47334102_92834.756.png", "Validation_47334102/Validation_47334102_47334102_92836.755.png", "Validation_47334102/Validation_47334102_47334102_92838.754.png", "Validation_47334102/Validation_47334102_47334102_92840.753.png" ], "turn_id": 7, "scan_id": "Validation_47334102", "origin_question": "Among these three objects(\"the couch\",\"the modern glass door with an aluminum frame\",\"the sleek, glossy black wooden and glass table\" ), which one is the closest to you?", "option": [ "the couch", "the modern glass door with an aluminum frame", "the sleek, glossy black wooden and glass table" ], "answer": "the couch", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the couch\",\"the modern glass door with an aluminum frame\",\"the sleek, glossy black wooden and glass table\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the couch', 'the modern glass door with an aluminum frame', 'the sleek, glossy black wooden and glass table'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_47334102/Validation_47334102_47334102_92842.753.png", "Validation_47334102/Validation_47334102_47334102_92844.752.png", "Validation_47334102/Validation_47334102_47334102_92846.751.png", "Validation_47334102/Validation_47334102_47334102_92848.750.png", "Validation_47334102/Validation_47334102_47334102_92850.749.png" ], "turn_id": 8, "scan_id": "Validation_47334102", "origin_question": "At the end of which turn were both of \"the large, glossy black monitor with the sleek design\" and \"the sleek, horizontally oriented white-framed window\" on your left side?", "option": [], "answer": "2", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the large, glossy black monitor with the sleek design\" and \"the sleek, horizontally oriented white-framed window\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_47334102/Validation_47334102_47334102_92852.749.png", "Validation_47334102/Validation_47334102_47334102_92854.748.png", "Validation_47334102/Validation_47334102_47334102_92856.747.png", "Validation_47334102/Validation_47334102_47334102_92858.746.png", "Validation_47334102/Validation_47334102_47334102_92860.745.png" ], "turn_id": 9, "scan_id": "Validation_47334102", "origin_question": "Assuming the direction you are facing at the end of the turn 5 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Left", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 5 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_47334102/Validation_47334102_47334102_92862.744.png", "Validation_47334102/Validation_47334102_47334102_92864.744.png", "Validation_47334102/Validation_47334102_47334102_92866.760.png", "Validation_47334102/Validation_47334102_47334102_92868.759.png", "Validation_47334102/Validation_47334102_47334102_92870.758.png" ], "turn_id": 10, "scan_id": "Validation_47334102", "origin_question": "By the end of the 3 turn, were you getting closer to or farther away from these two objects: X: \"the sleek, glossy black wooden and glass table\", Y: \"the sleek, horizontally oriented white-framed window\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "farther away from both X and Y", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 3 turn, were you getting closer to or farther away from these two objects: X: \"the sleek, glossy black wooden and glass table\", Y: \"the sleek, horizontally oriented white-framed window\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42446016/Validation_42446016_42446016_347976.297.png", "Validation_42446016/Validation_42446016_42446016_347978.279.png", "Validation_42446016/Validation_42446016_42446016_347980.278.png", "Validation_42446016/Validation_42446016_42446016_347982.278.png", "Validation_42446016/Validation_42446016_42446016_347984.277.png" ], "turn_id": 1, "scan_id": "Validation_42446016", "origin_question": "Remember, have you seen any person(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any person(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_42446016/Validation_42446016_42446016_347986.276.png", "Validation_42446016/Validation_42446016_42446016_347988.275.png", "Validation_42446016/Validation_42446016_42446016_347990.274.png", "Validation_42446016/Validation_42446016_42446016_347992.274.png", "Validation_42446016/Validation_42446016_42446016_347994.273.png" ], "turn_id": 2, "scan_id": "Validation_42446016", "origin_question": "At the end of which turn were both of \"the modern stainless steel oven with the sleek design\" and \"the stainless steel rectangular stove with control knobs\" on your left side?", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the modern stainless steel oven with the sleek design\" and \"the stainless steel rectangular stove with control knobs\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_42446016/Validation_42446016_42446016_347996.272.png", "Validation_42446016/Validation_42446016_42446016_347998.271.png", "Validation_42446016/Validation_42446016_42446016_348000.270.png", "Validation_42446016/Validation_42446016_42446016_348002.269.png", "Validation_42446016/Validation_42446016_42446016_348004.269.png" ], "turn_id": 3, "scan_id": "Validation_42446016", "origin_question": "Please recall \"the large rectangular window with the white frame\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "3.05", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the large rectangular window with the white frame\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_42446016/Validation_42446016_42446016_348006.268.png", "Validation_42446016/Validation_42446016_42446016_348008.267.png", "Validation_42446016/Validation_42446016_42446016_348010.283.png", "Validation_42446016/Validation_42446016_42446016_348012.282.png", "Validation_42446016/Validation_42446016_42446016_348014.281.png" ], "turn_id": 4, "scan_id": "Validation_42446016", "origin_question": "Which direction is the classic circular ceramic plate with the glossy finish in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the classic circular ceramic plate with the glossy finish in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_42446016/Validation_42446016_42446016_348016.280.png", "Validation_42446016/Validation_42446016_42446016_348018.280.png", "Validation_42446016/Validation_42446016_42446016_348020.279.png", "Validation_42446016/Validation_42446016_42446016_348022.278.png", "Validation_42446016/Validation_42446016_42446016_348024.277.png" ], "turn_id": 5, "scan_id": "Validation_42446016", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the classic circular ceramic plate with the glossy finish\" from your position?", "option": [], "answer": "43.74", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the classic circular ceramic plate with the glossy finish\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_42446016/Validation_42446016_42446016_348026.276.png", "Validation_42446016/Validation_42446016_42446016_348028.275.png", "Validation_42446016/Validation_42446016_42446016_348030.275.png", "Validation_42446016/Validation_42446016_42446016_348032.274.png", "Validation_42446016/Validation_42446016_42446016_348034.273.png" ], "turn_id": 6, "scan_id": "Validation_42446016", "origin_question": "Is the modern stainless steel oven with the sleek design to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the modern stainless steel oven with the sleek design to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_42446016/Validation_42446016_42446016_348036.272.png", "Validation_42446016/Validation_42446016_42446016_348038.271.png", "Validation_42446016/Validation_42446016_42446016_348040.270.png", "Validation_42446016/Validation_42446016_42446016_348042.270.png", "Validation_42446016/Validation_42446016_42446016_348044.269.png" ], "turn_id": 7, "scan_id": "Validation_42446016", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the vibrant orange and green ceramic teapot\" from your position?", "option": [], "answer": "88.87", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the vibrant orange and green ceramic teapot\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_42446016/Validation_42446016_42446016_348046.268.png", "Validation_42446016/Validation_42446016_42446016_348048.267.png", "Validation_42446016/Validation_42446016_42446016_348050.283.png", "Validation_42446016/Validation_42446016_42446016_348052.282.png", "Validation_42446016/Validation_42446016_42446016_348054.281.png" ], "turn_id": 8, "scan_id": "Validation_42446016", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the large rectangular window with the white frame\" from your position?", "option": [], "answer": "74.3", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the large rectangular window with the white frame\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_42446016/Validation_42446016_42446016_348056.281.png", "Validation_42446016/Validation_42446016_42446016_348058.280.png", "Validation_42446016/Validation_42446016_42446016_348060.279.png", "Validation_42446016/Validation_42446016_42446016_348062.278.png", "Validation_42446016/Validation_42446016_42446016_348064.277.png" ], "turn_id": 9, "scan_id": "Validation_42446016", "origin_question": "In which turn were you closest to \"the large rectangular window with the white frame\"? ", "option": [], "answer": "9", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the large rectangular window with the white frame\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_42446016/Validation_42446016_42446016_348066.276.png", "Validation_42446016/Validation_42446016_42446016_348068.276.png", "Validation_42446016/Validation_42446016_42446016_348070.275.png", "Validation_42446016/Validation_42446016_42446016_348072.274.png", "Validation_42446016/Validation_42446016_42446016_348074.273.png" ], "turn_id": 10, "scan_id": "Validation_42446016", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the medium-sized, round stainless steel pan\", Y: \"the large rectangular window with the white frame\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the medium-sized, round stainless steel pan\", Y: \"the large rectangular window with the white frame\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42898486/Validation_42898486_42898486_749307.391.png", "Validation_42898486/Validation_42898486_42898486_749309.391.png", "Validation_42898486/Validation_42898486_42898486_749311.390.png", "Validation_42898486/Validation_42898486_42898486_749313.389.png", "Validation_42898486/Validation_42898486_42898486_749315.388.png" ], "turn_id": 1, "scan_id": "Validation_42898486", "origin_question": "Remember, have you seen any countertop(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any countertop(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_42898486/Validation_42898486_42898486_749317.387.png", "Validation_42898486/Validation_42898486_42898486_749319.387.png", "Validation_42898486/Validation_42898486_42898486_749321.386.png", "Validation_42898486/Validation_42898486_42898486_749323.385.png", "Validation_42898486/Validation_42898486_42898486_749325.384.png" ], "turn_id": 2, "scan_id": "Validation_42898486", "origin_question": "In which turn were you closest to \"the medium-sized green artificial plant with ornaments\"? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the medium-sized green artificial plant with ornaments\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_42898486/Validation_42898486_42898486_749327.383.png", "Validation_42898486/Validation_42898486_42898486_749329.382.png", "Validation_42898486/Validation_42898486_42898486_749331.382.png", "Validation_42898486/Validation_42898486_42898486_749333.381.png", "Validation_42898486/Validation_42898486_42898486_749335.380.png" ], "turn_id": 3, "scan_id": "Validation_42898486", "origin_question": "In which turn did you see both \"the rectangular pillow with beige and gray stripes\" and \"the curtain\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the rectangular pillow with beige and gray stripes\" and \"the curtain\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_42898486/Validation_42898486_42898486_749337.379.png", "Validation_42898486/Validation_42898486_42898486_749339.378.png", "Validation_42898486/Validation_42898486_42898486_749341.377.png", "Validation_42898486/Validation_42898486_42898486_749343.393.png", "Validation_42898486/Validation_42898486_42898486_749345.393.png" ], "turn_id": 4, "scan_id": "Validation_42898486", "origin_question": "In which turn were you closest to \"an off-white wooden door with the sleek finish\"? ", "option": [], "answer": "2", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"an off-white wooden door with the sleek finish\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_42898486/Validation_42898486_42898486_749347.392.png", "Validation_42898486/Validation_42898486_42898486_749349.391.png", "Validation_42898486/Validation_42898486_42898486_749351.390.png", "Validation_42898486/Validation_42898486_42898486_749353.389.png", "Validation_42898486/Validation_42898486_42898486_749355.389.png" ], "turn_id": 5, "scan_id": "Validation_42898486", "origin_question": "Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Left", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_42898486/Validation_42898486_42898486_749357.388.png", "Validation_42898486/Validation_42898486_42898486_749359.387.png", "Validation_42898486/Validation_42898486_42898486_749361.386.png", "Validation_42898486/Validation_42898486_42898486_749363.385.png", "Validation_42898486/Validation_42898486_42898486_749365.384.png" ], "turn_id": 6, "scan_id": "Validation_42898486", "origin_question": "Which direction is the sleek vertical metal radiator in the neutral color in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the sleek vertical metal radiator in the neutral color in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_42898486/Validation_42898486_42898486_749367.384.png", "Validation_42898486/Validation_42898486_42898486_749369.383.png", "Validation_42898486/Validation_42898486_42898486_749371.382.png", "Validation_42898486/Validation_42898486_42898486_749373.381.png", "Validation_42898486/Validation_42898486_42898486_749375.380.png" ], "turn_id": 7, "scan_id": "Validation_42898486", "origin_question": "Using your orientation at the end of turn 6 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 6 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_42898486/Validation_42898486_42898486_749377.380.png", "Validation_42898486/Validation_42898486_42898486_749379.379.png", "Validation_42898486/Validation_42898486_42898486_749381.378.png", "Validation_42898486/Validation_42898486_42898486_749383.377.png", "Validation_42898486/Validation_42898486_42898486_749385.393.png" ], "turn_id": 8, "scan_id": "Validation_42898486", "origin_question": "Please recall \"the rectangular pillow with beige and gray stripes\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.27", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the rectangular pillow with beige and gray stripes\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_42898486/Validation_42898486_42898486_749387.392.png", "Validation_42898486/Validation_42898486_42898486_749389.391.png", "Validation_42898486/Validation_42898486_42898486_749391.390.png", "Validation_42898486/Validation_42898486_42898486_749393.390.png", "Validation_42898486/Validation_42898486_42898486_749395.389.png" ], "turn_id": 9, "scan_id": "Validation_42898486", "origin_question": "Which two objects are on the same side of you now? \"the medium-sized green artificial plant with ornaments\", \"the curtain\", \"the sleek vertical metal radiator in the neutral color\". ", "option": [ "the medium-sized green artificial plant with ornaments, the curtain", "the medium-sized green artificial plant with ornaments, the sleek vertical metal radiator in the neutral color", "the curtain, the sleek vertical metal radiator in the neutral color" ], "answer": "the medium-sized green artificial plant with ornaments, the curtain", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the medium-sized green artificial plant with ornaments\", \"the curtain\", \"the sleek vertical metal radiator in the neutral color\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the medium-sized green artificial plant with ornaments, the curtain', 'the medium-sized green artificial plant with ornaments, the sleek vertical metal radiator in the neutral color', 'the curtain, the sleek vertical metal radiator in the neutral color'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_42898486/Validation_42898486_42898486_749397.388.png", "Validation_42898486/Validation_42898486_42898486_749399.387.png", "Validation_42898486/Validation_42898486_42898486_749401.386.png", "Validation_42898486/Validation_42898486_42898486_749403.385.png", "Validation_42898486/Validation_42898486_42898486_749405.385.png" ], "turn_id": 10, "scan_id": "Validation_42898486", "origin_question": "Which direction is the medium-sized green artificial plant with ornaments in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the medium-sized green artificial plant with ornaments in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_45261143/Validation_45261143_45261143_4579.723.png", "Validation_45261143/Validation_45261143_45261143_4581.706.png", "Validation_45261143/Validation_45261143_45261143_4583.705.png", "Validation_45261143/Validation_45261143_45261143_4585.704.png" ], "turn_id": 1, "scan_id": "Validation_45261143", "origin_question": "Remember, have you seen any pillow(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any pillow(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_45261143/Validation_45261143_45261143_4587.703.png", "Validation_45261143/Validation_45261143_45261143_4589.702.png", "Validation_45261143/Validation_45261143_45261143_4591.702.png", "Validation_45261143/Validation_45261143_45261143_4593.701.png" ], "turn_id": 2, "scan_id": "Validation_45261143", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_45261143/Validation_45261143_45261143_4595.700.png", "Validation_45261143/Validation_45261143_45261143_4597.699.png", "Validation_45261143/Validation_45261143_45261143_4599.698.png", "Validation_45261143/Validation_45261143_45261143_4601.698.png" ], "turn_id": 3, "scan_id": "Validation_45261143", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_45261143/Validation_45261143_45261143_4603.713.png", "Validation_45261143/Validation_45261143_45261143_4605.712.png", "Validation_45261143/Validation_45261143_45261143_4607.712.png", "Validation_45261143/Validation_45261143_45261143_4609.711.png" ], "turn_id": 4, "scan_id": "Validation_45261143", "origin_question": "In which turn were you closest to \"the rectangular wooden bed with pale pink bedding\"? ", "option": [], "answer": "3", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the rectangular wooden bed with pale pink bedding\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_45261143/Validation_45261143_45261143_4611.710.png", "Validation_45261143/Validation_45261143_45261143_4613.709.png", "Validation_45261143/Validation_45261143_45261143_4615.708.png", "Validation_45261143/Validation_45261143_45261143_4617.708.png" ], "turn_id": 5, "scan_id": "Validation_45261143", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_45261143/Validation_45261143_45261143_4619.707.png", "Validation_45261143/Validation_45261143_45261143_4621.706.png", "Validation_45261143/Validation_45261143_45261143_4623.705.png", "Validation_45261143/Validation_45261143_45261143_4625.704.png" ], "turn_id": 6, "scan_id": "Validation_45261143", "origin_question": "Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_45261143/Validation_45261143_45261143_4627.703.png", "Validation_45261143/Validation_45261143_45261143_4629.703.png", "Validation_45261143/Validation_45261143_45261143_4631.702.png", "Validation_45261143/Validation_45261143_45261143_4633.701.png" ], "turn_id": 7, "scan_id": "Validation_45261143", "origin_question": "When did you first discover \"the sleek black monitor with the medium to large size\"? (index of the turn)", "option": [], "answer": "5", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the sleek black monitor with the medium to large size\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_45261143/Validation_45261143_45261143_4635.700.png", "Validation_45261143/Validation_45261143_45261143_4637.699.png", "Validation_45261143/Validation_45261143_45261143_4639.698.png", "Validation_45261143/Validation_45261143_45261143_4641.698.png" ], "turn_id": 8, "scan_id": "Validation_45261143", "origin_question": "", "option": [], "answer": "", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47334381/Validation_47334381_47334381_64379.419.png", "Validation_47334381/Validation_47334381_47334381_64381.385.png", "Validation_47334381/Validation_47334381_47334381_64383.401.png", "Validation_47334381/Validation_47334381_47334381_64385.400.png", "Validation_47334381/Validation_47334381_47334381_64387.399.png" ], "turn_id": 1, "scan_id": "Validation_47334381", "origin_question": "Remember, have you seen any shampoo(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any shampoo(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_47334381/Validation_47334381_47334381_64389.398.png", "Validation_47334381/Validation_47334381_47334381_64391.397.png", "Validation_47334381/Validation_47334381_47334381_64393.397.png", "Validation_47334381/Validation_47334381_47334381_64395.396.png", "Validation_47334381/Validation_47334381_47334381_64397.395.png" ], "turn_id": 2, "scan_id": "Validation_47334381", "origin_question": "Please recall \"the compact, transparent bin with the pink tint\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.81", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the compact, transparent bin with the pink tint\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_47334381/Validation_47334381_47334381_64399.394.png", "Validation_47334381/Validation_47334381_47334381_64401.393.png", "Validation_47334381/Validation_47334381_47334381_64403.393.png", "Validation_47334381/Validation_47334381_47334381_64405.392.png", "Validation_47334381/Validation_47334381_47334381_64407.391.png" ], "turn_id": 3, "scan_id": "Validation_47334381", "origin_question": "When did you first discover \"the compact, transparent bin with the pink tint\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the compact, transparent bin with the pink tint\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_47334381/Validation_47334381_47334381_64409.390.png", "Validation_47334381/Validation_47334381_47334381_64411.389.png", "Validation_47334381/Validation_47334381_47334381_64413.389.png", "Validation_47334381/Validation_47334381_47334381_64415.388.png", "Validation_47334381/Validation_47334381_47334381_64417.387.png" ], "turn_id": 4, "scan_id": "Validation_47334381", "origin_question": "When did you first discover \"the sturdy wooden table with the floral-patterned tablecloth\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the sturdy wooden table with the floral-patterned tablecloth\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_47334381/Validation_47334381_47334381_64419.386.png", "Validation_47334381/Validation_47334381_47334381_64421.385.png", "Validation_47334381/Validation_47334381_47334381_64423.401.png", "Validation_47334381/Validation_47334381_47334381_64425.400.png", "Validation_47334381/Validation_47334381_47334381_64427.400.png" ], "turn_id": 5, "scan_id": "Validation_47334381", "origin_question": "When did you first discover \"the white rectangular calendar with black text\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the white rectangular calendar with black text\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_47334381/Validation_47334381_47334381_64429.399.png", "Validation_47334381/Validation_47334381_47334381_64431.398.png", "Validation_47334381/Validation_47334381_47334381_64433.397.png", "Validation_47334381/Validation_47334381_47334381_64435.396.png", "Validation_47334381/Validation_47334381_47334381_64437.395.png" ], "turn_id": 6, "scan_id": "Validation_47334381", "origin_question": "Using your orientation at the end of turn 5 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "41.27", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 5 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_47334381/Validation_47334381_47334381_64439.395.png", "Validation_47334381/Validation_47334381_47334381_64441.394.png", "Validation_47334381/Validation_47334381_47334381_64443.393.png", "Validation_47334381/Validation_47334381_47334381_64445.392.png", "Validation_47334381/Validation_47334381_47334381_64447.391.png" ], "turn_id": 7, "scan_id": "Validation_47334381", "origin_question": "Is the slender picture frame holding the captivating picture to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the slender picture frame holding the captivating picture to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_47334381/Validation_47334381_47334381_64449.391.png", "Validation_47334381/Validation_47334381_47334381_64451.390.png", "Validation_47334381/Validation_47334381_47334381_64453.389.png", "Validation_47334381/Validation_47334381_47334381_64455.388.png", "Validation_47334381/Validation_47334381_47334381_64457.387.png" ], "turn_id": 8, "scan_id": "Validation_47334381", "origin_question": "What will be the first-time appearance order of the following categories: bin, light, statue?", "option": [ "statue, light, bin", "bin, light, statue", "statue, bin, light", "bin, statue, light" ], "answer": "bin, light, statue", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bin, light, statue? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['statue, light, bin', 'bin, light, statue', 'statue, bin, light', 'bin, statue, light'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_47334381/Validation_47334381_47334381_64459.387.png", "Validation_47334381/Validation_47334381_47334381_64461.386.png", "Validation_47334381/Validation_47334381_47334381_64463.402.png", "Validation_47334381/Validation_47334381_47334381_64465.401.png", "Validation_47334381/Validation_47334381_47334381_64467.400.png" ], "turn_id": 9, "scan_id": "Validation_47334381", "origin_question": "What will be the first-time appearance order of the following categories: bin, chair, statue?", "option": [ "chair, bin, statue", "chair, statue, bin", "statue, bin, chair", "bin, chair, statue" ], "answer": "bin, chair, statue", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bin, chair, statue? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['chair, bin, statue', 'chair, statue, bin', 'statue, bin, chair', 'bin, chair, statue'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_47334381/Validation_47334381_47334381_64469.399.png", "Validation_47334381/Validation_47334381_47334381_64471.398.png", "Validation_47334381/Validation_47334381_47334381_64473.397.png", "Validation_47334381/Validation_47334381_47334381_64475.397.png", "Validation_47334381/Validation_47334381_47334381_64477.396.png" ], "turn_id": 10, "scan_id": "Validation_47334381", "origin_question": "Please recall \"the sturdy wooden table with the floral-patterned tablecloth\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.52", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the sturdy wooden table with the floral-patterned tablecloth\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47334256/Validation_47334256_47334256_111995.710.png", "Validation_47334256/Validation_47334256_47334256_111997.693.png", "Validation_47334256/Validation_47334256_47334256_111999.692.png", "Validation_47334256/Validation_47334256_47334256_112001.691.png", "Validation_47334256/Validation_47334256_47334256_112003.690.png" ], "turn_id": 1, "scan_id": "Validation_47334256", "origin_question": "Remember, have you seen any teapot(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any teapot(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47334256/Validation_47334256_47334256_112005.690.png", "Validation_47334256/Validation_47334256_47334256_112007.689.png", "Validation_47334256/Validation_47334256_47334256_112009.688.png", "Validation_47334256/Validation_47334256_47334256_112011.687.png", "Validation_47334256/Validation_47334256_47334256_112013.686.png" ], "turn_id": 2, "scan_id": "Validation_47334256", "origin_question": "In which turn were you farthest from \"the sleek glass table with chrome metal legs\"? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the sleek glass table with chrome metal legs\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47334256/Validation_47334256_47334256_112015.686.png", "Validation_47334256/Validation_47334256_47334256_112017.685.png", "Validation_47334256/Validation_47334256_47334256_112019.684.png", "Validation_47334256/Validation_47334256_47334256_112021.700.png", "Validation_47334256/Validation_47334256_47334256_112023.699.png" ], "turn_id": 3, "scan_id": "Validation_47334256", "origin_question": "", "option": [], "answer": "", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47334256/Validation_47334256_47334256_112025.698.png", "Validation_47334256/Validation_47334256_47334256_112027.697.png", "Validation_47334256/Validation_47334256_47334256_112029.697.png", "Validation_47334256/Validation_47334256_47334256_112031.696.png", "Validation_47334256/Validation_47334256_47334256_112033.695.png" ], "turn_id": 4, "scan_id": "Validation_47334256", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the contemporary stainless steel oven with the dark glass door\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the contemporary stainless steel oven with the dark glass door\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_47334256/Validation_47334256_47334256_112035.694.png", "Validation_47334256/Validation_47334256_47334256_112037.693.png", "Validation_47334256/Validation_47334256_47334256_112039.693.png", "Validation_47334256/Validation_47334256_47334256_112041.692.png", "Validation_47334256/Validation_47334256_47334256_112043.691.png" ], "turn_id": 5, "scan_id": "Validation_47334256", "origin_question": "Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_47334256/Validation_47334256_47334256_112045.690.png", "Validation_47334256/Validation_47334256_47334256_112047.689.png", "Validation_47334256/Validation_47334256_47334256_112049.689.png", "Validation_47334256/Validation_47334256_47334256_112051.688.png", "Validation_47334256/Validation_47334256_47334256_112053.687.png" ], "turn_id": 6, "scan_id": "Validation_47334256", "origin_question": "At the end of which turn were both of \"the contemporary stainless steel oven with the dark glass door\" and \"the glossy black rectangular stove with four burners\" on your right side?", "option": [], "answer": "1", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the contemporary stainless steel oven with the dark glass door\" and \"the glossy black rectangular stove with four burners\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47334256/Validation_47334256_47334256_112055.686.png", "Validation_47334256/Validation_47334256_47334256_112057.685.png", "Validation_47334256/Validation_47334256_47334256_112059.685.png", "Validation_47334256/Validation_47334256_47334256_112061.684.png", "Validation_47334256/Validation_47334256_47334256_112063.700.png" ], "turn_id": 7, "scan_id": "Validation_47334256", "origin_question": "", "option": [], "answer": "", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_47334256/Validation_47334256_47334256_112065.699.png", "Validation_47334256/Validation_47334256_47334256_112067.698.png", "Validation_47334256/Validation_47334256_47334256_112069.697.png", "Validation_47334256/Validation_47334256_47334256_112071.696.png", "Validation_47334256/Validation_47334256_47334256_112073.696.png" ], "turn_id": 8, "scan_id": "Validation_47334256", "origin_question": "When did you first discover \"the glossy black rectangular stove with four burners\"? (index of the turn)", "option": [], "answer": "6", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the glossy black rectangular stove with four burners\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47204566/Validation_47204566_47204566_4076.068.png", "Validation_47204566/Validation_47204566_47204566_4078.050.png", "Validation_47204566/Validation_47204566_47204566_4080.050.png", "Validation_47204566/Validation_47204566_47204566_4082.049.png" ], "turn_id": 1, "scan_id": "Validation_47204566", "origin_question": "Remember, have you seen any stool(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any stool(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_47204566/Validation_47204566_47204566_4084.048.png", "Validation_47204566/Validation_47204566_47204566_4086.047.png", "Validation_47204566/Validation_47204566_47204566_4088.046.png", "Validation_47204566/Validation_47204566_47204566_4090.046.png" ], "turn_id": 2, "scan_id": "Validation_47204566", "origin_question": "Which two objects are on the same side of you now? \"the medium-brown vertical wooden shelf\", \"colorful books with glossy and matte finishes\", \"the rectangular window with white wood framing and blinds\". ", "option": [ "the medium-brown vertical wooden shelf, colorful books with glossy and matte finishes", "the medium-brown vertical wooden shelf, the rectangular window with white wood framing and blinds", "colorful books with glossy and matte finishes, the rectangular window with white wood framing and blinds" ], "answer": "the medium-brown vertical wooden shelf, colorful books with glossy and matte finishes", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the medium-brown vertical wooden shelf\", \"colorful books with glossy and matte finishes\", \"the rectangular window with white wood framing and blinds\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the medium-brown vertical wooden shelf, colorful books with glossy and matte finishes', 'the medium-brown vertical wooden shelf, the rectangular window with white wood framing and blinds', 'colorful books with glossy and matte finishes, the rectangular window with white wood framing and blinds'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_47204566/Validation_47204566_47204566_4092.045.png", "Validation_47204566/Validation_47204566_47204566_4094.044.png", "Validation_47204566/Validation_47204566_47204566_4096.043.png", "Validation_47204566/Validation_47204566_47204566_4098.042.png" ], "turn_id": 3, "scan_id": "Validation_47204566", "origin_question": "Is the glossy white ceramic flowerpot to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the glossy white ceramic flowerpot to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47204566/Validation_47204566_47204566_4100.041.png", "Validation_47204566/Validation_47204566_47204566_4102.041.png", "Validation_47204566/Validation_47204566_47204566_4104.040.png", "Validation_47204566/Validation_47204566_47204566_4106.039.png" ], "turn_id": 4, "scan_id": "Validation_47204566", "origin_question": "Which direction is the smooth, light brown rectangular door in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the smooth, light brown rectangular door in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_47204566/Validation_47204566_47204566_4108.038.png", "Validation_47204566/Validation_47204566_47204566_4110.037.png", "Validation_47204566/Validation_47204566_47204566_4112.036.png", "Validation_47204566/Validation_47204566_47204566_4114.036.png" ], "turn_id": 5, "scan_id": "Validation_47204566", "origin_question": "Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance forward or backward from that position?", "option": [ "Forward", "Backward" ], "answer": "Forward", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance forward or backward from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Forward', 'Backward'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47204566/Validation_47204566_47204566_4116.051.png", "Validation_47204566/Validation_47204566_47204566_4118.051.png", "Validation_47204566/Validation_47204566_47204566_4120.050.png", "Validation_47204566/Validation_47204566_47204566_4122.049.png" ], "turn_id": 6, "scan_id": "Validation_47204566", "origin_question": "Which direction is the smooth, light brown rectangular door in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the smooth, light brown rectangular door in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_47204566/Validation_47204566_47204566_4124.048.png", "Validation_47204566/Validation_47204566_47204566_4126.047.png", "Validation_47204566/Validation_47204566_47204566_4128.046.png", "Validation_47204566/Validation_47204566_47204566_4130.046.png" ], "turn_id": 7, "scan_id": "Validation_47204566", "origin_question": "Is the compact gray square stool with cushioned upholstery to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the compact gray square stool with cushioned upholstery to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47204566/Validation_47204566_47204566_4132.045.png", "Validation_47204566/Validation_47204566_47204566_4134.044.png", "Validation_47204566/Validation_47204566_47204566_4136.043.png", "Validation_47204566/Validation_47204566_47204566_4138.042.png" ], "turn_id": 8, "scan_id": "Validation_47204566", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the smooth, light brown rectangular door\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the smooth, light brown rectangular door\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47430475/Validation_47430475_47430475_310603.176.png", "Validation_47430475/Validation_47430475_47430475_310605.158.png", "Validation_47430475/Validation_47430475_47430475_310607.157.png", "Validation_47430475/Validation_47430475_47430475_310609.173.png" ], "turn_id": 1, "scan_id": "Validation_47430475", "origin_question": "Remember, have you seen any crate(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any crate(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47430475/Validation_47430475_47430475_310611.172.png", "Validation_47430475/Validation_47430475_47430475_310613.171.png", "Validation_47430475/Validation_47430475_47430475_310615.171.png", "Validation_47430475/Validation_47430475_47430475_310617.170.png" ], "turn_id": 2, "scan_id": "Validation_47430475", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_47430475/Validation_47430475_47430475_310619.169.png", "Validation_47430475/Validation_47430475_47430475_310621.168.png", "Validation_47430475/Validation_47430475_47430475_310623.167.png", "Validation_47430475/Validation_47430475_47430475_310625.167.png" ], "turn_id": 3, "scan_id": "Validation_47430475", "origin_question": "When did you first discover \"the book\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the book\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_47430475/Validation_47430475_47430475_310627.166.png", "Validation_47430475/Validation_47430475_47430475_310629.165.png", "Validation_47430475/Validation_47430475_47430475_310631.164.png", "Validation_47430475/Validation_47430475_47430475_310633.163.png" ], "turn_id": 4, "scan_id": "Validation_47430475", "origin_question": "Please recall \"the large vertical window with the dark brown frame\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.42", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the large vertical window with the dark brown frame\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_47430475/Validation_47430475_47430475_310635.163.png", "Validation_47430475/Validation_47430475_47430475_310637.162.png", "Validation_47430475/Validation_47430475_47430475_310639.161.png", "Validation_47430475/Validation_47430475_47430475_310641.160.png" ], "turn_id": 5, "scan_id": "Validation_47430475", "origin_question": "Please recall \"the file\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.53", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the file\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_47430475/Validation_47430475_47430475_310643.159.png", "Validation_47430475/Validation_47430475_47430475_310645.159.png", "Validation_47430475/Validation_47430475_47430475_310647.158.png", "Validation_47430475/Validation_47430475_47430475_310649.174.png" ], "turn_id": 6, "scan_id": "Validation_47430475", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the large vertical window with the dark brown frame\" from your position?", "option": [], "answer": "139.71", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the large vertical window with the dark brown frame\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_47430475/Validation_47430475_47430475_310651.173.png", "Validation_47430475/Validation_47430475_47430475_310653.172.png", "Validation_47430475/Validation_47430475_47430475_310655.171.png", "Validation_47430475/Validation_47430475_47430475_310657.170.png" ], "turn_id": 7, "scan_id": "Validation_47430475", "origin_question": "Which two objects are on the same side of you now? \"the large vertical window with the dark brown frame\", \"the sleek rectangular dark brown door\", \"the shelf\". ", "option": [ "the large vertical window with the dark brown frame, the sleek rectangular dark brown door", "the large vertical window with the dark brown frame, the shelf", "the sleek rectangular dark brown door, the shelf" ], "answer": "the large vertical window with the dark brown frame, the sleek rectangular dark brown door", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the large vertical window with the dark brown frame\", \"the sleek rectangular dark brown door\", \"the shelf\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the large vertical window with the dark brown frame, the sleek rectangular dark brown door', 'the large vertical window with the dark brown frame, the shelf', 'the sleek rectangular dark brown door, the shelf'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47430475/Validation_47430475_47430475_310659.170.png", "Validation_47430475/Validation_47430475_47430475_310661.169.png", "Validation_47430475/Validation_47430475_47430475_310663.168.png", "Validation_47430475/Validation_47430475_47430475_310665.167.png" ], "turn_id": 8, "scan_id": "Validation_47430475", "origin_question": "In which turn were you closest to \"the green round cushion with the textured fabric surface\"? ", "option": [], "answer": "6", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the green round cushion with the textured fabric surface\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_48018960/Validation_48018960_48018960_556539.762.png", "Validation_48018960/Validation_48018960_48018960_556541.245.png", "Validation_48018960/Validation_48018960_48018960_556542.744.png", "Validation_48018960/Validation_48018960_48018960_556544.243.png" ], "turn_id": 1, "scan_id": "Validation_48018960", "origin_question": "Remember, have you seen any mirror(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any mirror(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_48018960/Validation_48018960_48018960_556545.743.png", "Validation_48018960/Validation_48018960_48018960_556547.242.png", "Validation_48018960/Validation_48018960_48018960_556548.758.png", "Validation_48018960/Validation_48018960_48018960_556550.257.png" ], "turn_id": 2, "scan_id": "Validation_48018960", "origin_question": "In which turn did you see both \"the round white ceramic sink with the single faucet\" and \"the modern white rectangular plastic box\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the round white ceramic sink with the single faucet\" and \"the modern white rectangular plastic box\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_48018960/Validation_48018960_48018960_556551.757.png", "Validation_48018960/Validation_48018960_48018960_556553.256.png", "Validation_48018960/Validation_48018960_48018960_556554.756.png", "Validation_48018960/Validation_48018960_48018960_556556.255.png" ], "turn_id": 3, "scan_id": "Validation_48018960", "origin_question": "At the end of which turn were both of \"the modern white rectangular plastic box\" and \"the sleek white oval-shaped porcelain toilet\" on your right side?", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the modern white rectangular plastic box\" and \"the sleek white oval-shaped porcelain toilet\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_48018960/Validation_48018960_48018960_556557.754.png", "Validation_48018960/Validation_48018960_48018960_556559.254.png", "Validation_48018960/Validation_48018960_48018960_556560.753.png", "Validation_48018960/Validation_48018960_48018960_556562.252.png" ], "turn_id": 4, "scan_id": "Validation_48018960", "origin_question": "What will be the first-time appearance order of the following categories: mirror, sink, toilet?", "option": [ "mirror, sink, toilet", "mirror, toilet, sink", "toilet, mirror, sink", "sink, toilet, mirror" ], "answer": "mirror, sink, toilet", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: mirror, sink, toilet? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['mirror, sink, toilet', 'mirror, toilet, sink', 'toilet, mirror, sink', 'sink, toilet, mirror'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_48018960/Validation_48018960_48018960_556563.752.png", "Validation_48018960/Validation_48018960_48018960_556565.251.png", "Validation_48018960/Validation_48018960_48018960_556566.751.png", "Validation_48018960/Validation_48018960_48018960_556568.250.png" ], "turn_id": 5, "scan_id": "Validation_48018960", "origin_question": "Please recall \"the modern white rectangular plastic box\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.88", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the modern white rectangular plastic box\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_48018960/Validation_48018960_48018960_556569.749.png", "Validation_48018960/Validation_48018960_48018960_556571.249.png", "Validation_48018960/Validation_48018960_48018960_556572.748.png", "Validation_48018960/Validation_48018960_48018960_556574.248.png" ], "turn_id": 6, "scan_id": "Validation_48018960", "origin_question": "Which two objects are on the same side of you now? \"the modern white rectangular plastic box\", \"the light\", \"the stylish white wooden shelf with compartments\". ", "option": [ "the modern white rectangular plastic box, the light", "the modern white rectangular plastic box, the stylish white wooden shelf with compartments", "the light, the stylish white wooden shelf with compartments" ], "answer": "the light, the stylish white wooden shelf with compartments", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the modern white rectangular plastic box\", \"the light\", \"the stylish white wooden shelf with compartments\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the modern white rectangular plastic box, the light', 'the modern white rectangular plastic box, the stylish white wooden shelf with compartments', 'the light, the stylish white wooden shelf with compartments'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42897508/Validation_42897508_42897508_269924.104.png", "Validation_42897508/Validation_42897508_42897508_269926.086.png", "Validation_42897508/Validation_42897508_42897508_269928.086.png", "Validation_42897508/Validation_42897508_42897508_269930.085.png" ], "turn_id": 1, "scan_id": "Validation_42897508", "origin_question": "Remember, have you seen any oven(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any oven(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_42897508/Validation_42897508_42897508_269932.084.png", "Validation_42897508/Validation_42897508_42897508_269934.083.png", "Validation_42897508/Validation_42897508_42897508_269936.082.png", "Validation_42897508/Validation_42897508_42897508_269938.082.png" ], "turn_id": 2, "scan_id": "Validation_42897508", "origin_question": "At the end of which turn were both of \"the stainless steel and glass oven with advanced features\" and \"the compact grey leather couch with the modern shape\" on your right side?", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the stainless steel and glass oven with advanced features\" and \"the compact grey leather couch with the modern shape\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_42897508/Validation_42897508_42897508_269940.081.png", "Validation_42897508/Validation_42897508_42897508_269942.080.png", "Validation_42897508/Validation_42897508_42897508_269944.079.png", "Validation_42897508/Validation_42897508_42897508_269946.078.png" ], "turn_id": 3, "scan_id": "Validation_42897508", "origin_question": "Which direction is the stylish stainless steel pyramid-shaped range hood in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the stylish stainless steel pyramid-shaped range hood in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_42897508/Validation_42897508_42897508_269948.078.png", "Validation_42897508/Validation_42897508_42897508_269950.093.png", "Validation_42897508/Validation_42897508_42897508_269952.093.png", "Validation_42897508/Validation_42897508_42897508_269954.092.png" ], "turn_id": 4, "scan_id": "Validation_42897508", "origin_question": "By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the compact grey leather couch with the modern shape\", Y: \"the medium-sized, earthy brown bowl\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to X and farther away from Y", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the compact grey leather couch with the modern shape\", Y: \"the medium-sized, earthy brown bowl\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_42897508/Validation_42897508_42897508_269956.091.png", "Validation_42897508/Validation_42897508_42897508_269958.090.png", "Validation_42897508/Validation_42897508_42897508_269960.089.png", "Validation_42897508/Validation_42897508_42897508_269962.088.png" ], "turn_id": 5, "scan_id": "Validation_42897508", "origin_question": "Which direction is the stainless steel and glass oven with advanced features in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the stainless steel and glass oven with advanced features in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "Validation_42897508/Validation_42897508_42897508_269964.088.png", "Validation_42897508/Validation_42897508_42897508_269966.087.png", "Validation_42897508/Validation_42897508_42897508_269968.086.png", "Validation_42897508/Validation_42897508_42897508_269970.085.png" ], "turn_id": 6, "scan_id": "Validation_42897508", "origin_question": "Among these three objects(\"the stainless steel and glass oven with advanced features\",\"the stylish stainless steel pyramid-shaped range hood\",\"the medium-sized, earthy brown bowl\" ), which one is the closest to you?", "option": [ "the stainless steel and glass oven with advanced features", "the stylish stainless steel pyramid-shaped range hood", "the medium-sized, earthy brown bowl" ], "answer": "the medium-sized, earthy brown bowl", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects(\"the stainless steel and glass oven with advanced features\",\"the stylish stainless steel pyramid-shaped range hood\",\"the medium-sized, earthy brown bowl\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the stainless steel and glass oven with advanced features', 'the stylish stainless steel pyramid-shaped range hood', 'the medium-sized, earthy brown bowl'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_42897508/Validation_42897508_42897508_269972.084.png", "Validation_42897508/Validation_42897508_42897508_269974.084.png", "Validation_42897508/Validation_42897508_42897508_269976.083.png", "Validation_42897508/Validation_42897508_42897508_269978.082.png" ], "turn_id": 7, "scan_id": "Validation_42897508", "origin_question": "Is the medium-sized, earthy brown bowl to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the medium-sized, earthy brown bowl to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_42897508/Validation_42897508_42897508_269980.081.png", "Validation_42897508/Validation_42897508_42897508_269982.080.png", "Validation_42897508/Validation_42897508_42897508_269984.079.png", "Validation_42897508/Validation_42897508_42897508_269986.079.png" ], "turn_id": 8, "scan_id": "Validation_42897508", "origin_question": "How far is your current position from where you were at the end of the turn 6? (in meters)", "option": [], "answer": "2.06", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 6? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47331964/Validation_47331964_47331964_167481.561.png", "Validation_47331964/Validation_47331964_47331964_167483.544.png", "Validation_47331964/Validation_47331964_47331964_167485.543.png", "Validation_47331964/Validation_47331964_47331964_167487.542.png", "Validation_47331964/Validation_47331964_47331964_167489.541.png" ], "turn_id": 1, "scan_id": "Validation_47331964", "origin_question": "Remember, have you seen any case(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any case(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_47331964/Validation_47331964_47331964_167491.541.png", "Validation_47331964/Validation_47331964_47331964_167493.540.png", "Validation_47331964/Validation_47331964_47331964_167495.539.png", "Validation_47331964/Validation_47331964_47331964_167497.538.png", "Validation_47331964/Validation_47331964_47331964_167499.537.png" ], "turn_id": 2, "scan_id": "Validation_47331964", "origin_question": "In which turn did you see both \"the modern stainless steel dishwasher with neutral color\" and \"the contemporary chair with the curved backrest\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the modern stainless steel dishwasher with neutral color\" and \"the contemporary chair with the curved backrest\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_47331964/Validation_47331964_47331964_167501.536.png", "Validation_47331964/Validation_47331964_47331964_167503.552.png", "Validation_47331964/Validation_47331964_47331964_167505.552.png", "Validation_47331964/Validation_47331964_47331964_167507.551.png", "Validation_47331964/Validation_47331964_47331964_167509.550.png" ], "turn_id": 3, "scan_id": "Validation_47331964", "origin_question": "When did you first discover \"the modern stainless steel dishwasher with neutral color\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the modern stainless steel dishwasher with neutral color\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_47331964/Validation_47331964_47331964_167511.549.png", "Validation_47331964/Validation_47331964_47331964_167513.548.png", "Validation_47331964/Validation_47331964_47331964_167515.547.png", "Validation_47331964/Validation_47331964_47331964_167517.547.png", "Validation_47331964/Validation_47331964_47331964_167519.546.png" ], "turn_id": 4, "scan_id": "Validation_47331964", "origin_question": "When did you first discover \"the classic rectangular white metal radiator with grooves\"? (index of the turn)", "option": [], "answer": "4", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the classic rectangular white metal radiator with grooves\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47331964/Validation_47331964_47331964_167521.545.png", "Validation_47331964/Validation_47331964_47331964_167523.544.png", "Validation_47331964/Validation_47331964_47331964_167525.543.png", "Validation_47331964/Validation_47331964_47331964_167527.543.png", "Validation_47331964/Validation_47331964_47331964_167529.542.png" ], "turn_id": 5, "scan_id": "Validation_47331964", "origin_question": "", "option": [], "answer": "", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_47331964/Validation_47331964_47331964_167531.541.png", "Validation_47331964/Validation_47331964_47331964_167533.540.png", "Validation_47331964/Validation_47331964_47331964_167535.539.png", "Validation_47331964/Validation_47331964_47331964_167537.539.png", "Validation_47331964/Validation_47331964_47331964_167539.538.png" ], "turn_id": 6, "scan_id": "Validation_47331964", "origin_question": "When did you first discover \"the round white ceramic plate with the smooth finish\"? (index of the turn)", "option": [], "answer": "5", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the round white ceramic plate with the smooth finish\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47331964/Validation_47331964_47331964_167541.537.png", "Validation_47331964/Validation_47331964_47331964_167543.536.png", "Validation_47331964/Validation_47331964_47331964_167545.552.png", "Validation_47331964/Validation_47331964_47331964_167547.551.png", "Validation_47331964/Validation_47331964_47331964_167549.550.png" ], "turn_id": 7, "scan_id": "Validation_47331964", "origin_question": "Which direction is the modern stainless steel dishwasher with neutral color in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the modern stainless steel dishwasher with neutral color in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "Validation_47331964/Validation_47331964_47331964_167551.550.png", "Validation_47331964/Validation_47331964_47331964_167553.549.png", "Validation_47331964/Validation_47331964_47331964_167555.548.png", "Validation_47331964/Validation_47331964_47331964_167557.547.png", "Validation_47331964/Validation_47331964_47331964_167559.546.png" ], "turn_id": 8, "scan_id": "Validation_47331964", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the modern stainless steel dishwasher with neutral color\"; \"the tall, stainless steel refrigerator with modern design\"; \"the modern rectangular sink\" ", "option": [ "the modern stainless steel dishwasher with neutral color", "the tall, stainless steel refrigerator with modern design", "the modern rectangular sink" ], "answer": "the tall, stainless steel refrigerator with modern design", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the modern stainless steel dishwasher with neutral color\"; \"the tall, stainless steel refrigerator with modern design\"; \"the modern rectangular sink\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the modern stainless steel dishwasher with neutral color', 'the tall, stainless steel refrigerator with modern design', 'the modern rectangular sink'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_47331964/Validation_47331964_47331964_167561.546.png", "Validation_47331964/Validation_47331964_47331964_167563.545.png", "Validation_47331964/Validation_47331964_47331964_167565.544.png", "Validation_47331964/Validation_47331964_47331964_167567.543.png", "Validation_47331964/Validation_47331964_47331964_167569.542.png" ], "turn_id": 9, "scan_id": "Validation_47331964", "origin_question": "Is the round white ceramic plate with the smooth finish to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the round white ceramic plate with the smooth finish to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_47331964/Validation_47331964_47331964_167571.541.png", "Validation_47331964/Validation_47331964_47331964_167573.541.png", "Validation_47331964/Validation_47331964_47331964_167575.540.png", "Validation_47331964/Validation_47331964_47331964_167577.539.png", "Validation_47331964/Validation_47331964_47331964_167579.538.png" ], "turn_id": 10, "scan_id": "Validation_47331964", "origin_question": "Please recall \"the modern stainless steel dishwasher with neutral color\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.69", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the modern stainless steel dishwasher with neutral color\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_48018368/Validation_48018368_48018368_389534.291.png", "Validation_48018368/Validation_48018368_48018368_389536.273.png", "Validation_48018368/Validation_48018368_48018368_389538.273.png", "Validation_48018368/Validation_48018368_48018368_389540.272.png" ], "turn_id": 1, "scan_id": "Validation_48018368", "origin_question": "Remember, have you seen any cabinet(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any cabinet(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_48018368/Validation_48018368_48018368_389542.271.png", "Validation_48018368/Validation_48018368_48018368_389544.270.png", "Validation_48018368/Validation_48018368_48018368_389546.269.png", "Validation_48018368/Validation_48018368_48018368_389548.268.png" ], "turn_id": 2, "scan_id": "Validation_48018368", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the standard upright refrigerator in the neutral color\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the standard upright refrigerator in the neutral color\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "Validation_48018368/Validation_48018368_48018368_389550.268.png", "Validation_48018368/Validation_48018368_48018368_389552.283.png", "Validation_48018368/Validation_48018368_48018368_389554.283.png", "Validation_48018368/Validation_48018368_48018368_389556.282.png" ], "turn_id": 3, "scan_id": "Validation_48018368", "origin_question": "Among these three objects(\"the standard upright refrigerator in the neutral color\",\"the light blue square pillow with the colorful pattern\",\"the sleek rectangular white microwave made of metal-plastic\" ), which one is the farthest from you?", "option": [ "the standard upright refrigerator in the neutral color", "the light blue square pillow with the colorful pattern", "the sleek rectangular white microwave made of metal-plastic" ], "answer": "the light blue square pillow with the colorful pattern", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects(\"the standard upright refrigerator in the neutral color\",\"the light blue square pillow with the colorful pattern\",\"the sleek rectangular white microwave made of metal-plastic\" ), which one is the farthest from you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the standard upright refrigerator in the neutral color', 'the light blue square pillow with the colorful pattern', 'the sleek rectangular white microwave made of metal-plastic'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_48018368/Validation_48018368_48018368_389558.281.png", "Validation_48018368/Validation_48018368_48018368_389560.280.png", "Validation_48018368/Validation_48018368_48018368_389562.279.png", "Validation_48018368/Validation_48018368_48018368_389564.279.png" ], "turn_id": 4, "scan_id": "Validation_48018368", "origin_question": "Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the rectangular stainless steel sink with the metallic sheen\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the rectangular stainless steel sink with the metallic sheen\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_48018368/Validation_48018368_48018368_389566.278.png", "Validation_48018368/Validation_48018368_48018368_389568.277.png", "Validation_48018368/Validation_48018368_48018368_389570.276.png", "Validation_48018368/Validation_48018368_48018368_389572.275.png" ], "turn_id": 5, "scan_id": "Validation_48018368", "origin_question": "What will be the first-time appearance order of the following categories: cabinet, couch, microwave?", "option": [ "microwave, couch, cabinet", "microwave, cabinet, couch", "cabinet, microwave, couch", "couch, cabinet, microwave" ], "answer": "cabinet, microwave, couch", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: cabinet, couch, microwave? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['microwave, couch, cabinet', 'microwave, cabinet, couch', 'cabinet, microwave, couch', 'couch, cabinet, microwave'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_48018368/Validation_48018368_48018368_389574.274.png", "Validation_48018368/Validation_48018368_48018368_389576.274.png", "Validation_48018368/Validation_48018368_48018368_389578.273.png", "Validation_48018368/Validation_48018368_48018368_389580.272.png" ], "turn_id": 6, "scan_id": "Validation_48018368", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the standard upright refrigerator in the neutral color\" from your position?", "option": [], "answer": "167.26", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the standard upright refrigerator in the neutral color\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_48018368/Validation_48018368_48018368_389582.271.png", "Validation_48018368/Validation_48018368_48018368_389584.270.png", "Validation_48018368/Validation_48018368_48018368_389586.269.png", "Validation_48018368/Validation_48018368_48018368_389588.269.png" ], "turn_id": 7, "scan_id": "Validation_48018368", "origin_question": "What will be the first-time appearance order of the following categories: cabinet, microwave, stool?", "option": [ "stool, microwave, cabinet", "cabinet, microwave, stool", "microwave, stool, cabinet", "stool, cabinet, microwave" ], "answer": "cabinet, microwave, stool", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: cabinet, microwave, stool? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['stool, microwave, cabinet', 'cabinet, microwave, stool', 'microwave, stool, cabinet', 'stool, cabinet, microwave'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "Validation_41125700/Validation_41125700_41125700_3707.856.png", "Validation_41125700/Validation_41125700_41125700_3709.839.png", "Validation_41125700/Validation_41125700_41125700_3711.838.png", "Validation_41125700/Validation_41125700_41125700_3713.837.png", "Validation_41125700/Validation_41125700_41125700_3715.836.png" ], "turn_id": 1, "scan_id": "Validation_41125700", "origin_question": "Remember, how many chair(s) have you seen so far? ", "option": [], "answer": "2", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many chair(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_41125700/Validation_41125700_41125700_3717.835.png", "Validation_41125700/Validation_41125700_41125700_3719.851.png", "Validation_41125700/Validation_41125700_41125700_3721.850.png", "Validation_41125700/Validation_41125700_41125700_3723.850.png", "Validation_41125700/Validation_41125700_41125700_3725.849.png" ], "turn_id": 2, "scan_id": "Validation_41125700", "origin_question": "Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Right", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_41125700/Validation_41125700_41125700_3727.848.png", "Validation_41125700/Validation_41125700_41125700_3729.847.png", "Validation_41125700/Validation_41125700_41125700_3731.846.png", "Validation_41125700/Validation_41125700_41125700_3733.845.png", "Validation_41125700/Validation_41125700_41125700_3735.845.png" ], "turn_id": 3, "scan_id": "Validation_41125700", "origin_question": "", "option": [], "answer": "", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_41125700/Validation_41125700_41125700_3737.844.png", "Validation_41125700/Validation_41125700_41125700_3739.843.png", "Validation_41125700/Validation_41125700_41125700_3741.842.png", "Validation_41125700/Validation_41125700_41125700_3743.841.png", "Validation_41125700/Validation_41125700_41125700_3745.841.png" ], "turn_id": 4, "scan_id": "Validation_41125700", "origin_question": "Remember, have you seen any basket(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any basket(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_41125700/Validation_41125700_41125700_3747.840.png", "Validation_41125700/Validation_41125700_41125700_3749.839.png", "Validation_41125700/Validation_41125700_41125700_3751.838.png", "Validation_41125700/Validation_41125700_41125700_3753.837.png", "Validation_41125700/Validation_41125700_41125700_3755.837.png" ], "turn_id": 5, "scan_id": "Validation_41125700", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the door\", Y: \"the sleek, modern rectangular off-white radiator\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to Y and farther away from X", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the door\", Y: \"the sleek, modern rectangular off-white radiator\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_41125700/Validation_41125700_41125700_3757.836.png", "Validation_41125700/Validation_41125700_41125700_3759.835.png", "Validation_41125700/Validation_41125700_41125700_3761.851.png", "Validation_41125700/Validation_41125700_41125700_3763.850.png", "Validation_41125700/Validation_41125700_41125700_3765.849.png" ], "turn_id": 6, "scan_id": "Validation_41125700", "origin_question": "Which direction is the door in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the door in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_41125700/Validation_41125700_41125700_3767.848.png", "Validation_41125700/Validation_41125700_41125700_3769.848.png", "Validation_41125700/Validation_41125700_41125700_3771.847.png", "Validation_41125700/Validation_41125700_41125700_3773.846.png", "Validation_41125700/Validation_41125700_41125700_3775.845.png" ], "turn_id": 7, "scan_id": "Validation_41125700", "origin_question": "Is the sleek, modern rectangular off-white radiator to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the sleek, modern rectangular off-white radiator to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "Validation_41125700/Validation_41125700_41125700_3777.844.png", "Validation_41125700/Validation_41125700_41125700_3779.843.png", "Validation_41125700/Validation_41125700_41125700_3781.843.png", "Validation_41125700/Validation_41125700_41125700_3783.842.png", "Validation_41125700/Validation_41125700_41125700_3785.841.png" ], "turn_id": 8, "scan_id": "Validation_41125700", "origin_question": "Remember, how many bin(s) have you seen so far? ", "option": [], "answer": "2", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many bin(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_41125700/Validation_41125700_41125700_3787.840.png", "Validation_41125700/Validation_41125700_41125700_3789.840.png", "Validation_41125700/Validation_41125700_41125700_3791.839.png", "Validation_41125700/Validation_41125700_41125700_3793.838.png", "Validation_41125700/Validation_41125700_41125700_3795.837.png" ], "turn_id": 9, "scan_id": "Validation_41125700", "origin_question": "By the end of the 5 turn, were you getting closer to or farther away from these two objects: X: \"the sleek rectangular stainless steel sink\", Y: \"the sleek stainless steel range hood in metallic silver\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 5 turn, were you getting closer to or farther away from these two objects: X: \"the sleek rectangular stainless steel sink\", Y: \"the sleek stainless steel range hood in metallic silver\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_41125700/Validation_41125700_41125700_3797.836.png", "Validation_41125700/Validation_41125700_41125700_3799.835.png", "Validation_41125700/Validation_41125700_41125700_3801.851.png", "Validation_41125700/Validation_41125700_41125700_3803.851.png", "Validation_41125700/Validation_41125700_41125700_3805.850.png" ], "turn_id": 10, "scan_id": "Validation_41125700", "origin_question": "Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the sleek stainless steel range hood in metallic silver\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the sleek stainless steel range hood in metallic silver\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47331334/Validation_47331334_47331334_180705.324.png", "Validation_47331334/Validation_47331334_47331334_180706.807.png", "Validation_47331334/Validation_47331334_47331334_180708.306.png", "Validation_47331334/Validation_47331334_47331334_180709.805.png" ], "turn_id": 1, "scan_id": "Validation_47331334", "origin_question": "Remember, have you seen any bed(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any bed(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 3 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47331334/Validation_47331334_47331334_180711.305.png", "Validation_47331334/Validation_47331334_47331334_180712.804.png", "Validation_47331334/Validation_47331334_47331334_180714.304.png", "Validation_47331334/Validation_47331334_47331334_180715.803.png" ], "turn_id": 2, "scan_id": "Validation_47331334", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 3 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47331334/Validation_47331334_47331334_180717.302.png", "Validation_47331334/Validation_47331334_47331334_180718.802.png", "Validation_47331334/Validation_47331334_47331334_180720.301.png", "Validation_47331334/Validation_47331334_47331334_180721.801.png" ], "turn_id": 3, "scan_id": "Validation_47331334", "origin_question": "In which turn were you closest to \"the cozy double bed with dark earthy bedding\"? ", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the cozy double bed with dark earthy bedding\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 3 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47334379/Validation_47334379_47334379_64530.691.png", "Validation_47334379/Validation_47334379_47334379_64532.690.png", "Validation_47334379/Validation_47334379_47334379_64534.689.png", "Validation_47334379/Validation_47334379_47334379_64536.688.png", "Validation_47334379/Validation_47334379_47334379_64538.688.png" ], "turn_id": 1, "scan_id": "Validation_47334379", "origin_question": "Remember, have you seen any pool table(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any pool table(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47334379/Validation_47334379_47334379_64540.687.png", "Validation_47334379/Validation_47334379_47334379_64542.686.png", "Validation_47334379/Validation_47334379_47334379_64544.685.png", "Validation_47334379/Validation_47334379_47334379_64546.701.png", "Validation_47334379/Validation_47334379_47334379_64548.700.png" ], "turn_id": 2, "scan_id": "Validation_47334379", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_47334379/Validation_47334379_47334379_64550.699.png", "Validation_47334379/Validation_47334379_47334379_64552.698.png", "Validation_47334379/Validation_47334379_47334379_64554.698.png", "Validation_47334379/Validation_47334379_47334379_64556.697.png", "Validation_47334379/Validation_47334379_47334379_64558.696.png" ], "turn_id": 3, "scan_id": "Validation_47334379", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the curtain\" from your position?", "option": [], "answer": "56.47", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the curtain\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_47334379/Validation_47334379_47334379_64560.695.png", "Validation_47334379/Validation_47334379_47334379_64562.694.png", "Validation_47334379/Validation_47334379_47334379_64564.694.png", "Validation_47334379/Validation_47334379_47334379_64566.693.png", "Validation_47334379/Validation_47334379_47334379_64568.692.png" ], "turn_id": 4, "scan_id": "Validation_47334379", "origin_question": "By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the sleek orange chair with black metal legs\", Y: \"the tall, metallic refrigerator with adjustable shelves\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the sleek orange chair with black metal legs\", Y: \"the tall, metallic refrigerator with adjustable shelves\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_47334379/Validation_47334379_47334379_64570.691.png", "Validation_47334379/Validation_47334379_47334379_64572.690.png", "Validation_47334379/Validation_47334379_47334379_64574.690.png", "Validation_47334379/Validation_47334379_47334379_64576.689.png", "Validation_47334379/Validation_47334379_47334379_64578.688.png" ], "turn_id": 5, "scan_id": "Validation_47334379", "origin_question": "Which two objects are on the same side of you now? \"the dark woven-textured plastic box\", \"the compact, dual-tone plastic bin\", \"the tall, metallic refrigerator with adjustable shelves\". ", "option": [ "the dark woven-textured plastic box, the compact, dual-tone plastic bin", "the dark woven-textured plastic box, the tall, metallic refrigerator with adjustable shelves", "the compact, dual-tone plastic bin, the tall, metallic refrigerator with adjustable shelves" ], "answer": "the dark woven-textured plastic box, the compact, dual-tone plastic bin", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the dark woven-textured plastic box\", \"the compact, dual-tone plastic bin\", \"the tall, metallic refrigerator with adjustable shelves\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the dark woven-textured plastic box, the compact, dual-tone plastic bin', 'the dark woven-textured plastic box, the tall, metallic refrigerator with adjustable shelves', 'the compact, dual-tone plastic bin, the tall, metallic refrigerator with adjustable shelves'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_47334379/Validation_47334379_47334379_64580.687.png", "Validation_47334379/Validation_47334379_47334379_64582.686.png", "Validation_47334379/Validation_47334379_47334379_64584.685.png", "Validation_47334379/Validation_47334379_47334379_64586.685.png", "Validation_47334379/Validation_47334379_47334379_64588.700.png" ], "turn_id": 6, "scan_id": "Validation_47334379", "origin_question": "How far is your current position from where you were at the end of the turn 5? (in meters)", "option": [], "answer": "1.11", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 5? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_47334379/Validation_47334379_47334379_64590.700.png", "Validation_47334379/Validation_47334379_47334379_64592.699.png", "Validation_47334379/Validation_47334379_47334379_64594.698.png", "Validation_47334379/Validation_47334379_47334379_64596.697.png", "Validation_47334379/Validation_47334379_47334379_64598.696.png" ], "turn_id": 7, "scan_id": "Validation_47334379", "origin_question": "Is the dark woven-textured plastic box to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the dark woven-textured plastic box to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47334379/Validation_47334379_47334379_64600.696.png", "Validation_47334379/Validation_47334379_47334379_64602.695.png", "Validation_47334379/Validation_47334379_47334379_64604.694.png", "Validation_47334379/Validation_47334379_47334379_64606.693.png", "Validation_47334379/Validation_47334379_47334379_64608.692.png" ], "turn_id": 8, "scan_id": "Validation_47334379", "origin_question": "Which direction is the glossy rectangular calendar with red and black text in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the glossy rectangular calendar with red and black text in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_47334379/Validation_47334379_47334379_64610.692.png", "Validation_47334379/Validation_47334379_47334379_64612.691.png", "Validation_47334379/Validation_47334379_47334379_64614.690.png", "Validation_47334379/Validation_47334379_47334379_64616.689.png", "Validation_47334379/Validation_47334379_47334379_64618.688.png" ], "turn_id": 9, "scan_id": "Validation_47334379", "origin_question": "In which turn did you see both \"an off-white statue with intricate detailing\" and \"the soft white glossy kettle with the rounded shape\" simultaneously? ", "option": [], "answer": "9", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"an off-white statue with intricate detailing\" and \"the soft white glossy kettle with the rounded shape\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_47334379/Validation_47334379_47334379_64620.688.png", "Validation_47334379/Validation_47334379_47334379_64622.687.png", "Validation_47334379/Validation_47334379_47334379_64624.686.png", "Validation_47334379/Validation_47334379_47334379_64626.685.png", "Validation_47334379/Validation_47334379_47334379_64628.684.png" ], "turn_id": 10, "scan_id": "Validation_47334379", "origin_question": "In which turn did you see both \"the curtain\" and \"the soft white glossy kettle with the rounded shape\" simultaneously? ", "option": [], "answer": "9", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the curtain\" and \"the soft white glossy kettle with the rounded shape\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42898849/Validation_42898849_42898849_196445.935.png", "Validation_42898849/Validation_42898849_42898849_196447.918.png", "Validation_42898849/Validation_42898849_42898849_196449.917.png", "Validation_42898849/Validation_42898849_42898849_196451.916.png", "Validation_42898849/Validation_42898849_42898849_196453.915.png" ], "turn_id": 1, "scan_id": "Validation_42898849", "origin_question": "Remember, have you seen any garbage(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any garbage(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_42898849/Validation_42898849_42898849_196455.915.png", "Validation_42898849/Validation_42898849_42898849_196457.914.png", "Validation_42898849/Validation_42898849_42898849_196459.913.png", "Validation_42898849/Validation_42898849_42898849_196461.912.png", "Validation_42898849/Validation_42898849_42898849_196463.928.png" ], "turn_id": 2, "scan_id": "Validation_42898849", "origin_question": "At the end of which turn were both of \"the compact square table with the light-colored wood finish\" and \"the classic, rounded lamp with the soft beige shade\" on your left side?", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the compact square table with the light-colored wood finish\" and \"the classic, rounded lamp with the soft beige shade\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "Validation_42898849/Validation_42898849_42898849_196465.927.png", "Validation_42898849/Validation_42898849_42898849_196467.926.png", "Validation_42898849/Validation_42898849_42898849_196469.926.png", "Validation_42898849/Validation_42898849_42898849_196471.925.png", "Validation_42898849/Validation_42898849_42898849_196473.924.png" ], "turn_id": 3, "scan_id": "Validation_42898849", "origin_question": "Remember, how many shampoo(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many shampoo(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_42898849/Validation_42898849_42898849_196475.923.png", "Validation_42898849/Validation_42898849_42898849_196477.922.png", "Validation_42898849/Validation_42898849_42898849_196479.922.png", "Validation_42898849/Validation_42898849_42898849_196481.921.png", "Validation_42898849/Validation_42898849_42898849_196483.920.png" ], "turn_id": 4, "scan_id": "Validation_42898849", "origin_question": "In which turn were you farthest from \"the classic cylindrical ceramic cup with the smooth finish\"? ", "option": [], "answer": "1", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the classic cylindrical ceramic cup with the smooth finish\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "Validation_42898849/Validation_42898849_42898849_196485.919.png", "Validation_42898849/Validation_42898849_42898849_196487.918.png", "Validation_42898849/Validation_42898849_42898849_196489.918.png", "Validation_42898849/Validation_42898849_42898849_196491.917.png", "Validation_42898849/Validation_42898849_42898849_196493.916.png" ], "turn_id": 5, "scan_id": "Validation_42898849", "origin_question": "Remember, how many couch(s) have you seen so far? ", "option": [], "answer": "2", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many couch(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_42898849/Validation_42898849_42898849_196495.915.png", "Validation_42898849/Validation_42898849_42898849_196497.915.png", "Validation_42898849/Validation_42898849_42898849_196499.914.png", "Validation_42898849/Validation_42898849_42898849_196501.913.png", "Validation_42898849/Validation_42898849_42898849_196503.912.png" ], "turn_id": 6, "scan_id": "Validation_42898849", "origin_question": "By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the compact black fireplace with the boxy shape\", Y: \"the compact gray shelf with three open compartments\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the compact black fireplace with the boxy shape\", Y: \"the compact gray shelf with three open compartments\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_42898849/Validation_42898849_42898849_196505.928.png", "Validation_42898849/Validation_42898849_42898849_196507.927.png", "Validation_42898849/Validation_42898849_42898849_196509.927.png", "Validation_42898849/Validation_42898849_42898849_196511.926.png", "Validation_42898849/Validation_42898849_42898849_196513.925.png" ], "turn_id": 7, "scan_id": "Validation_42898849", "origin_question": "Which two objects are on the same side of you now? \"the oven\", \"the rounded ceramic vase in the neutral tone\", \"the compact gray shelf with three open compartments\". ", "option": [ "the oven, the rounded ceramic vase in the neutral tone", "the oven, the compact gray shelf with three open compartments", "the rounded ceramic vase in the neutral tone, the compact gray shelf with three open compartments" ], "answer": "the rounded ceramic vase in the neutral tone, the compact gray shelf with three open compartments", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the oven\", \"the rounded ceramic vase in the neutral tone\", \"the compact gray shelf with three open compartments\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the oven, the rounded ceramic vase in the neutral tone', 'the oven, the compact gray shelf with three open compartments', 'the rounded ceramic vase in the neutral tone, the compact gray shelf with three open compartments'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_42898849/Validation_42898849_42898849_196515.924.png", "Validation_42898849/Validation_42898849_42898849_196517.923.png", "Validation_42898849/Validation_42898849_42898849_196519.922.png", "Validation_42898849/Validation_42898849_42898849_196521.922.png", "Validation_42898849/Validation_42898849_42898849_196523.921.png" ], "turn_id": 8, "scan_id": "Validation_42898849", "origin_question": "Please recall \"the classic cylindrical ceramic cup with the smooth finish\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.9", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the classic cylindrical ceramic cup with the smooth finish\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_42898849/Validation_42898849_42898849_196525.920.png", "Validation_42898849/Validation_42898849_42898849_196527.919.png", "Validation_42898849/Validation_42898849_42898849_196529.918.png", "Validation_42898849/Validation_42898849_42898849_196531.918.png", "Validation_42898849/Validation_42898849_42898849_196533.917.png" ], "turn_id": 9, "scan_id": "Validation_42898849", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the rounded ceramic vase in the neutral tone\" from your position?", "option": [], "answer": "94.2", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the rounded ceramic vase in the neutral tone\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_42898849/Validation_42898849_42898849_196535.916.png", "Validation_42898849/Validation_42898849_42898849_196537.915.png", "Validation_42898849/Validation_42898849_42898849_196539.915.png", "Validation_42898849/Validation_42898849_42898849_196541.914.png", "Validation_42898849/Validation_42898849_42898849_196543.913.png" ], "turn_id": 10, "scan_id": "Validation_42898849", "origin_question": "", "option": [], "answer": "", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42898818/Validation_42898818_42898818_168152.197.png", "Validation_42898818/Validation_42898818_42898818_168153.680.png", "Validation_42898818/Validation_42898818_42898818_168155.179.png", "Validation_42898818/Validation_42898818_42898818_168156.679.png" ], "turn_id": 1, "scan_id": "Validation_42898818", "origin_question": "Remember, have you seen any microwave(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any microwave(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_42898818/Validation_42898818_42898818_168158.178.png", "Validation_42898818/Validation_42898818_42898818_168159.677.png", "Validation_42898818/Validation_42898818_42898818_168161.177.png", "Validation_42898818/Validation_42898818_42898818_168162.676.png" ], "turn_id": 2, "scan_id": "Validation_42898818", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the sleek, oval white acrylic bathtub\", Y: \"the lush, compact green plant with oval leaves\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the sleek, oval white acrylic bathtub\", Y: \"the lush, compact green plant with oval leaves\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_42898818/Validation_42898818_42898818_168164.176.png", "Validation_42898818/Validation_42898818_42898818_168165.675.png", "Validation_42898818/Validation_42898818_42898818_168167.174.png", "Validation_42898818/Validation_42898818_42898818_168168.674.png" ], "turn_id": 3, "scan_id": "Validation_42898818", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the sleek, oval white acrylic bathtub\", Y: \"the compact wooden cabinet with the glass front\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the sleek, oval white acrylic bathtub\", Y: \"the compact wooden cabinet with the glass front\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_42898818/Validation_42898818_42898818_168170.173.png", "Validation_42898818/Validation_42898818_42898818_168171.673.png", "Validation_42898818/Validation_42898818_42898818_168173.172.png", "Validation_42898818/Validation_42898818_42898818_168174.671.png" ], "turn_id": 4, "scan_id": "Validation_42898818", "origin_question": "What will be the first-time appearance order of the following categories: bathtub, sink, towel?", "option": [ "sink, bathtub, towel", "bathtub, sink, towel", "towel, bathtub, sink", "towel, sink, bathtub" ], "answer": "bathtub, sink, towel", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bathtub, sink, towel? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['sink, bathtub, towel', 'bathtub, sink, towel', 'towel, bathtub, sink', 'towel, sink, bathtub'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_42898818/Validation_42898818_42898818_168176.187.png", "Validation_42898818/Validation_42898818_42898818_168177.687.png", "Validation_42898818/Validation_42898818_42898818_168179.186.png", "Validation_42898818/Validation_42898818_42898818_168180.685.png" ], "turn_id": 5, "scan_id": "Validation_42898818", "origin_question": "In which turn did you see both \"the lush, compact green plant with oval leaves\" and \"the modern, rounded white ceramic sink\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the lush, compact green plant with oval leaves\" and \"the modern, rounded white ceramic sink\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_42898818/Validation_42898818_42898818_168182.185.png", "Validation_42898818/Validation_42898818_42898818_168183.684.png", "Validation_42898818/Validation_42898818_42898818_168185.184.png", "Validation_42898818/Validation_42898818_42898818_168186.683.png" ], "turn_id": 6, "scan_id": "Validation_42898818", "origin_question": "Please recall \"the lush, compact green plant with oval leaves\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.79", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the lush, compact green plant with oval leaves\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_42898818/Validation_42898818_42898818_168188.182.png", "Validation_42898818/Validation_42898818_42898818_168189.682.png", "Validation_42898818/Validation_42898818_42898818_168191.181.png", "Validation_42898818/Validation_42898818_42898818_168192.680.png" ], "turn_id": 7, "scan_id": "Validation_42898818", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_48458473/Validation_48458473_48458473_111150.811.png", "Validation_48458473/Validation_48458473_48458473_111152.294.png", "Validation_48458473/Validation_48458473_48458473_111153.794.png", "Validation_48458473/Validation_48458473_48458473_111155.293.png" ], "turn_id": 1, "scan_id": "Validation_48458473", "origin_question": "Remember, have you seen any table(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any table(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_48458473/Validation_48458473_48458473_111156.792.png", "Validation_48458473/Validation_48458473_48458473_111158.292.png", "Validation_48458473/Validation_48458473_48458473_111159.791.png", "Validation_48458473/Validation_48458473_48458473_111161.291.png" ], "turn_id": 2, "scan_id": "Validation_48458473", "origin_question": "When did you first discover \"the rich brown leather couch with plush armrests\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the rich brown leather couch with plush armrests\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_48458473/Validation_48458473_48458473_111162.790.png", "Validation_48458473/Validation_48458473_48458473_111164.289.png", "Validation_48458473/Validation_48458473_48458473_111165.789.png", "Validation_48458473/Validation_48458473_48458473_111167.288.png" ], "turn_id": 3, "scan_id": "Validation_48458473", "origin_question": "Remember, have you seen any mirror(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any mirror(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_48458473/Validation_48458473_48458473_111168.788.png", "Validation_48458473/Validation_48458473_48458473_111170.287.png", "Validation_48458473/Validation_48458473_48458473_111171.786.png", "Validation_48458473/Validation_48458473_48458473_111173.286.png" ], "turn_id": 4, "scan_id": "Validation_48458473", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the dark brown leather-upholstered table\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the dark brown leather-upholstered table\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_44358505/Validation_44358505_44358505_50657.701.png", "Validation_44358505/Validation_44358505_44358505_50659.701.png", "Validation_44358505/Validation_44358505_44358505_50661.700.png", "Validation_44358505/Validation_44358505_44358505_50663.699.png", "Validation_44358505/Validation_44358505_44358505_50665.698.png" ], "turn_id": 1, "scan_id": "Validation_44358505", "origin_question": "Remember, have you seen any pen(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any pen(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "Validation_44358505/Validation_44358505_44358505_50667.697.png", "Validation_44358505/Validation_44358505_44358505_50669.696.png", "Validation_44358505/Validation_44358505_44358505_50671.696.png", "Validation_44358505/Validation_44358505_44358505_50673.695.png", "Validation_44358505/Validation_44358505_44358505_50675.694.png" ], "turn_id": 2, "scan_id": "Validation_44358505", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the rectangular natural wood table with tapered legs\"; \"the modern cylindrical brown lamp with metal construction\"; \"the modern cylindrical fireplace in matte black\" ", "option": [ "the rectangular natural wood table with tapered legs", "the modern cylindrical brown lamp with metal construction", "the modern cylindrical fireplace in matte black" ], "answer": "the modern cylindrical fireplace in matte black", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the rectangular natural wood table with tapered legs\"; \"the modern cylindrical brown lamp with metal construction\"; \"the modern cylindrical fireplace in matte black\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular natural wood table with tapered legs', 'the modern cylindrical brown lamp with metal construction', 'the modern cylindrical fireplace in matte black'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_44358505/Validation_44358505_44358505_50677.693.png", "Validation_44358505/Validation_44358505_44358505_50679.692.png", "Validation_44358505/Validation_44358505_44358505_50681.692.png", "Validation_44358505/Validation_44358505_44358505_50683.691.png", "Validation_44358505/Validation_44358505_44358505_50685.690.png" ], "turn_id": 3, "scan_id": "Validation_44358505", "origin_question": "Please recall \"the modern cylindrical translucent plastic bin with the dark lid\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.35", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the modern cylindrical translucent plastic bin with the dark lid\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_44358505/Validation_44358505_44358505_50687.689.png", "Validation_44358505/Validation_44358505_44358505_50689.688.png", "Validation_44358505/Validation_44358505_44358505_50691.704.png", "Validation_44358505/Validation_44358505_44358505_50693.703.png", "Validation_44358505/Validation_44358505_44358505_50695.702.png" ], "turn_id": 4, "scan_id": "Validation_44358505", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_44358505/Validation_44358505_44358505_50697.702.png", "Validation_44358505/Validation_44358505_44358505_50699.701.png", "Validation_44358505/Validation_44358505_44358505_50701.700.png", "Validation_44358505/Validation_44358505_44358505_50703.699.png", "Validation_44358505/Validation_44358505_44358505_50705.698.png" ], "turn_id": 5, "scan_id": "Validation_44358505", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the modern cylindrical translucent plastic bin with the dark lid\" from your position?", "option": [], "answer": "155.94", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the modern cylindrical translucent plastic bin with the dark lid\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_44358505/Validation_44358505_44358505_50707.697.png", "Validation_44358505/Validation_44358505_44358505_50709.697.png", "Validation_44358505/Validation_44358505_44358505_50711.696.png", "Validation_44358505/Validation_44358505_44358505_50713.695.png", "Validation_44358505/Validation_44358505_44358505_50715.694.png" ], "turn_id": 6, "scan_id": "Validation_44358505", "origin_question": "", "option": [], "answer": "", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_44358505/Validation_44358505_44358505_50717.693.png", "Validation_44358505/Validation_44358505_44358505_50719.692.png", "Validation_44358505/Validation_44358505_44358505_50721.692.png", "Validation_44358505/Validation_44358505_44358505_50723.691.png", "Validation_44358505/Validation_44358505_44358505_50725.690.png" ], "turn_id": 7, "scan_id": "Validation_44358505", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the modern cylindrical brown lamp with metal construction\" from your position?", "option": [], "answer": "155.59", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the modern cylindrical brown lamp with metal construction\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "Validation_44358505/Validation_44358505_44358505_50727.689.png", "Validation_44358505/Validation_44358505_44358505_50729.688.png", "Validation_44358505/Validation_44358505_44358505_50731.704.png", "Validation_44358505/Validation_44358505_44358505_50733.703.png", "Validation_44358505/Validation_44358505_44358505_50735.702.png" ], "turn_id": 8, "scan_id": "Validation_44358505", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the modern cylindrical translucent plastic bin with the dark lid\"; \"the sleek, glossy black monitor\"; \"the compact stainless steel machine with black accents\" ", "option": [ "the modern cylindrical translucent plastic bin with the dark lid", "the sleek, glossy black monitor", "the compact stainless steel machine with black accents" ], "answer": "the compact stainless steel machine with black accents", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the modern cylindrical translucent plastic bin with the dark lid\"; \"the sleek, glossy black monitor\"; \"the compact stainless steel machine with black accents\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the modern cylindrical translucent plastic bin with the dark lid', 'the sleek, glossy black monitor', 'the compact stainless steel machine with black accents'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_44358505/Validation_44358505_44358505_50737.702.png", "Validation_44358505/Validation_44358505_44358505_50739.701.png", "Validation_44358505/Validation_44358505_44358505_50741.700.png", "Validation_44358505/Validation_44358505_44358505_50743.699.png", "Validation_44358505/Validation_44358505_44358505_50745.698.png" ], "turn_id": 9, "scan_id": "Validation_44358505", "origin_question": "How far is your current position from where you were at the end of the turn 4? (in meters)", "option": [], "answer": "3.09", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 4? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_44358505/Validation_44358505_44358505_50747.698.png", "Validation_44358505/Validation_44358505_44358505_50749.697.png", "Validation_44358505/Validation_44358505_44358505_50751.696.png", "Validation_44358505/Validation_44358505_44358505_50753.695.png", "Validation_44358505/Validation_44358505_44358505_50755.694.png" ], "turn_id": 10, "scan_id": "Validation_44358505", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the cylindrical roll of white soft pulp paper\" from your position?", "option": [], "answer": "175.34", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the cylindrical roll of white soft pulp paper\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42899612/Validation_42899612_42899612_228426.539.png", "Validation_42899612/Validation_42899612_42899612_228428.522.png", "Validation_42899612/Validation_42899612_42899612_228430.521.png", "Validation_42899612/Validation_42899612_42899612_228432.520.png", "Validation_42899612/Validation_42899612_42899612_228434.519.png" ], "turn_id": 1, "scan_id": "Validation_42899612", "origin_question": "Remember, have you seen any table(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any table(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_42899612/Validation_42899612_42899612_228436.519.png", "Validation_42899612/Validation_42899612_42899612_228438.518.png", "Validation_42899612/Validation_42899612_42899612_228440.517.png", "Validation_42899612/Validation_42899612_42899612_228442.516.png", "Validation_42899612/Validation_42899612_42899612_228444.516.png" ], "turn_id": 2, "scan_id": "Validation_42899612", "origin_question": "In which turn were you farthest from \"the bed\"? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the bed\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_42899612/Validation_42899612_42899612_228446.515.png", "Validation_42899612/Validation_42899612_42899612_228448.514.png", "Validation_42899612/Validation_42899612_42899612_228450.513.png", "Validation_42899612/Validation_42899612_42899612_228452.512.png", "Validation_42899612/Validation_42899612_42899612_228454.511.png" ], "turn_id": 3, "scan_id": "Validation_42899612", "origin_question": "In which turn were you closest to \"the sleek white plastic stool with the minimalist shape\"? ", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the sleek white plastic stool with the minimalist shape\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_42899612/Validation_42899612_42899612_228456.527.png", "Validation_42899612/Validation_42899612_42899612_228458.527.png", "Validation_42899612/Validation_42899612_42899612_228460.526.png", "Validation_42899612/Validation_42899612_42899612_228462.525.png", "Validation_42899612/Validation_42899612_42899612_228464.524.png" ], "turn_id": 4, "scan_id": "Validation_42899612", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the bed\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the bed\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_42899612/Validation_42899612_42899612_228466.523.png", "Validation_42899612/Validation_42899612_42899612_228468.523.png", "Validation_42899612/Validation_42899612_42899612_228470.522.png", "Validation_42899612/Validation_42899612_42899612_228472.521.png", "Validation_42899612/Validation_42899612_42899612_228474.520.png" ], "turn_id": 5, "scan_id": "Validation_42899612", "origin_question": "Using your orientation at the end of turn 4 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "76.86", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 4 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_42899612/Validation_42899612_42899612_228476.519.png", "Validation_42899612/Validation_42899612_42899612_228478.519.png", "Validation_42899612/Validation_42899612_42899612_228480.518.png", "Validation_42899612/Validation_42899612_42899612_228482.517.png", "Validation_42899612/Validation_42899612_42899612_228484.516.png" ], "turn_id": 6, "scan_id": "Validation_42899612", "origin_question": "What will be the first-time appearance order of the following categories: cabinet, chair, pillow?", "option": [ "chair, cabinet, pillow", "cabinet, pillow, chair", "pillow, cabinet, chair", "pillow, chair, cabinet" ], "answer": "pillow, cabinet, chair", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: cabinet, chair, pillow? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['chair, cabinet, pillow', 'cabinet, pillow, chair', 'pillow, cabinet, chair', 'pillow, chair, cabinet'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_42899612/Validation_42899612_42899612_228486.516.png", "Validation_42899612/Validation_42899612_42899612_228488.515.png", "Validation_42899612/Validation_42899612_42899612_228490.514.png", "Validation_42899612/Validation_42899612_42899612_228492.513.png", "Validation_42899612/Validation_42899612_42899612_228494.512.png" ], "turn_id": 7, "scan_id": "Validation_42899612", "origin_question": "Using your orientation at the end of turn 5 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "54.25", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 5 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_42899612/Validation_42899612_42899612_228496.511.png", "Validation_42899612/Validation_42899612_42899612_228498.527.png", "Validation_42899612/Validation_42899612_42899612_228500.527.png", "Validation_42899612/Validation_42899612_42899612_228502.526.png", "Validation_42899612/Validation_42899612_42899612_228504.525.png" ], "turn_id": 8, "scan_id": "Validation_42899612", "origin_question": "Please recall \"the smooth, glossy cylindrical vase in neutral color\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.12", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the smooth, glossy cylindrical vase in neutral color\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_42899612/Validation_42899612_42899612_228506.524.png", "Validation_42899612/Validation_42899612_42899612_228508.524.png", "Validation_42899612/Validation_42899612_42899612_228510.523.png", "Validation_42899612/Validation_42899612_42899612_228512.522.png", "Validation_42899612/Validation_42899612_42899612_228514.521.png" ], "turn_id": 9, "scan_id": "Validation_42899612", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47430038/Validation_47430038_47430038_20533.818.png", "Validation_47430038/Validation_47430038_47430038_20535.284.png", "Validation_47430038/Validation_47430038_47430038_20536.784.png", "Validation_47430038/Validation_47430038_47430038_20538.283.png" ], "turn_id": 1, "scan_id": "Validation_47430038", "origin_question": "Remember, have you seen any chair(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any chair(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_47430038/Validation_47430038_47430038_20539.783.png", "Validation_47430038/Validation_47430038_47430038_20541.282.png", "Validation_47430038/Validation_47430038_47430038_20542.781.png", "Validation_47430038/Validation_47430038_47430038_20544.297.png" ], "turn_id": 2, "scan_id": "Validation_47430038", "origin_question": "When did you first discover \"the vibrant green plant with dense, glossy leaves\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the vibrant green plant with dense, glossy leaves\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_47430038/Validation_47430038_47430038_20545.797.png", "Validation_47430038/Validation_47430038_47430038_20547.296.png", "Validation_47430038/Validation_47430038_47430038_20548.796.png", "Validation_47430038/Validation_47430038_47430038_20550.295.png" ], "turn_id": 3, "scan_id": "Validation_47430038", "origin_question": "How far is your current position from where you were at the end of the turn 1? (in meters)", "option": [], "answer": "2.08", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 1? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_47430038/Validation_47430038_47430038_20551.794.png", "Validation_47430038/Validation_47430038_47430038_20553.294.png", "Validation_47430038/Validation_47430038_47430038_20554.793.png", "Validation_47430038/Validation_47430038_47430038_20556.292.png" ], "turn_id": 4, "scan_id": "Validation_47430038", "origin_question": "Please recall \"the rectangular, earth-tone cotton towel\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.0", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the rectangular, earth-tone cotton towel\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_47430038/Validation_47430038_47430038_20557.792.png", "Validation_47430038/Validation_47430038_47430038_20559.291.png", "Validation_47430038/Validation_47430038_47430038_20560.790.png", "Validation_47430038/Validation_47430038_47430038_20562.290.png" ], "turn_id": 5, "scan_id": "Validation_47430038", "origin_question": "By the end of the 3 turn, were you getting closer to or farther away from these two objects: X: \"the warm-toned wood table with the minimalist design\", Y: \"the classic round beige flowerpot made of ceramic\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "farther away from both X and Y", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 3 turn, were you getting closer to or farther away from these two objects: X: \"the warm-toned wood table with the minimalist design\", Y: \"the classic round beige flowerpot made of ceramic\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47430038/Validation_47430038_47430038_20563.789.png", "Validation_47430038/Validation_47430038_47430038_20565.289.png", "Validation_47430038/Validation_47430038_47430038_20566.788.png", "Validation_47430038/Validation_47430038_47430038_20568.287.png" ], "turn_id": 6, "scan_id": "Validation_47430038", "origin_question": "Which direction is the medium-sized light brown wooden shelf with open compartments in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the medium-sized light brown wooden shelf with open compartments in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_41254441/Validation_41254441_41254441_96306.727.png", "Validation_41254441/Validation_41254441_41254441_96308.709.png", "Validation_41254441/Validation_41254441_41254441_96310.708.png", "Validation_41254441/Validation_41254441_41254441_96312.708.png", "Validation_41254441/Validation_41254441_41254441_96314.707.png" ], "turn_id": 1, "scan_id": "Validation_41254441", "origin_question": "Remember, have you seen any couch(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any couch(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_41254441/Validation_41254441_41254441_96316.706.png", "Validation_41254441/Validation_41254441_41254441_96318.705.png", "Validation_41254441/Validation_41254441_41254441_96320.705.png", "Validation_41254441/Validation_41254441_41254441_96322.704.png", "Validation_41254441/Validation_41254441_41254441_96324.703.png" ], "turn_id": 2, "scan_id": "Validation_41254441", "origin_question": "In which turn did you see both \"the sleek, white pvc vertical slider window\" and \"the soft off-white rectangular wooden door\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the sleek, white pvc vertical slider window\" and \"the soft off-white rectangular wooden door\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_41254441/Validation_41254441_41254441_96326.702.png", "Validation_41254441/Validation_41254441_41254441_96328.701.png", "Validation_41254441/Validation_41254441_41254441_96330.717.png", "Validation_41254441/Validation_41254441_41254441_96332.716.png", "Validation_41254441/Validation_41254441_41254441_96334.716.png" ], "turn_id": 3, "scan_id": "Validation_41254441", "origin_question": "", "option": [], "answer": "", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_41254441/Validation_41254441_41254441_96336.715.png", "Validation_41254441/Validation_41254441_41254441_96338.714.png", "Validation_41254441/Validation_41254441_41254441_96340.713.png", "Validation_41254441/Validation_41254441_41254441_96342.712.png", "Validation_41254441/Validation_41254441_41254441_96344.712.png" ], "turn_id": 4, "scan_id": "Validation_41254441", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the rectangular stainless steel stove for ventilation\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the rectangular stainless steel stove for ventilation\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "Validation_41254441/Validation_41254441_41254441_96346.711.png", "Validation_41254441/Validation_41254441_41254441_96348.710.png", "Validation_41254441/Validation_41254441_41254441_96350.709.png", "Validation_41254441/Validation_41254441_41254441_96352.709.png", "Validation_41254441/Validation_41254441_41254441_96354.708.png" ], "turn_id": 5, "scan_id": "Validation_41254441", "origin_question": "Among these three objects(\"the rectangular stainless steel stove for ventilation\",\"the stainless steel rectangular sink with the deep basin\",\"the compact, rectangular stainless steel microwave\" ), which one is the closest to you?", "option": [ "the rectangular stainless steel stove for ventilation", "the stainless steel rectangular sink with the deep basin", "the compact, rectangular stainless steel microwave" ], "answer": "the rectangular stainless steel stove for ventilation", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the rectangular stainless steel stove for ventilation\",\"the stainless steel rectangular sink with the deep basin\",\"the compact, rectangular stainless steel microwave\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular stainless steel stove for ventilation', 'the stainless steel rectangular sink with the deep basin', 'the compact, rectangular stainless steel microwave'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_41254441/Validation_41254441_41254441_96356.707.png", "Validation_41254441/Validation_41254441_41254441_96358.706.png", "Validation_41254441/Validation_41254441_41254441_96360.705.png", "Validation_41254441/Validation_41254441_41254441_96362.705.png", "Validation_41254441/Validation_41254441_41254441_96364.704.png" ], "turn_id": 6, "scan_id": "Validation_41254441", "origin_question": "Which two objects are on the same side of you now? \"the sleek, white pvc vertical slider window\", \"the sleek black monitor with the flat screen\", \"the large gray metal bin designed for high-volume use\". ", "option": [ "the sleek, white pvc vertical slider window, the sleek black monitor with the flat screen", "the sleek, white pvc vertical slider window, the large gray metal bin designed for high-volume use", "the sleek black monitor with the flat screen, the large gray metal bin designed for high-volume use" ], "answer": "the sleek black monitor with the flat screen, the large gray metal bin designed for high-volume use", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the sleek, white pvc vertical slider window\", \"the sleek black monitor with the flat screen\", \"the large gray metal bin designed for high-volume use\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the sleek, white pvc vertical slider window, the sleek black monitor with the flat screen', 'the sleek, white pvc vertical slider window, the large gray metal bin designed for high-volume use', 'the sleek black monitor with the flat screen, the large gray metal bin designed for high-volume use'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_41254441/Validation_41254441_41254441_96366.703.png", "Validation_41254441/Validation_41254441_41254441_96368.702.png", "Validation_41254441/Validation_41254441_41254441_96370.701.png", "Validation_41254441/Validation_41254441_41254441_96372.717.png", "Validation_41254441/Validation_41254441_41254441_96374.716.png" ], "turn_id": 7, "scan_id": "Validation_41254441", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_41254441/Validation_41254441_41254441_96376.716.png", "Validation_41254441/Validation_41254441_41254441_96378.715.png", "Validation_41254441/Validation_41254441_41254441_96380.714.png", "Validation_41254441/Validation_41254441_41254441_96382.713.png", "Validation_41254441/Validation_41254441_41254441_96384.713.png" ], "turn_id": 8, "scan_id": "Validation_41254441", "origin_question": "Which direction is the rectangular stainless steel stove for ventilation in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the rectangular stainless steel stove for ventilation in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_41254441/Validation_41254441_41254441_96386.712.png", "Validation_41254441/Validation_41254441_41254441_96388.711.png", "Validation_41254441/Validation_41254441_41254441_96390.710.png", "Validation_41254441/Validation_41254441_41254441_96392.709.png", "Validation_41254441/Validation_41254441_41254441_96394.709.png" ], "turn_id": 9, "scan_id": "Validation_41254441", "origin_question": "Please recall \"the soft off-white rectangular wooden door\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "3.3", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the soft off-white rectangular wooden door\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47430045/Validation_47430045_47430045_20626.413.png", "Validation_47430045/Validation_47430045_47430045_20627.896.png", "Validation_47430045/Validation_47430045_47430045_20629.396.png", "Validation_47430045/Validation_47430045_47430045_20630.895.png" ], "turn_id": 1, "scan_id": "Validation_47430045", "origin_question": "Remember, have you seen any plant(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any plant(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "Validation_47430045/Validation_47430045_47430045_20632.394.png", "Validation_47430045/Validation_47430045_47430045_20633.894.png", "Validation_47430045/Validation_47430045_47430045_20635.393.png", "Validation_47430045/Validation_47430045_47430045_20636.893.png" ], "turn_id": 2, "scan_id": "Validation_47430045", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the moderate-sized plant with lush green foliage\"; \"the sink\"; \"the compact stainless steel dishwasher with the sleek design\" ", "option": [ "the moderate-sized plant with lush green foliage", "the sink", "the compact stainless steel dishwasher with the sleek design" ], "answer": "the sink", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the moderate-sized plant with lush green foliage\"; \"the sink\"; \"the compact stainless steel dishwasher with the sleek design\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the moderate-sized plant with lush green foliage', 'the sink', 'the compact stainless steel dishwasher with the sleek design'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_47430045/Validation_47430045_47430045_20638.392.png", "Validation_47430045/Validation_47430045_47430045_20639.891.png", "Validation_47430045/Validation_47430045_47430045_20641.391.png", "Validation_47430045/Validation_47430045_47430045_20642.890.png" ], "turn_id": 3, "scan_id": "Validation_47430045", "origin_question": "By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the compact stainless steel dishwasher with the sleek design\", Y: \"the moderate-sized plant with lush green foliage\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "farther away from both X and Y", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the compact stainless steel dishwasher with the sleek design\", Y: \"the moderate-sized plant with lush green foliage\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_47430045/Validation_47430045_47430045_20644.390.png", "Validation_47430045/Validation_47430045_47430045_20645.889.png", "Validation_47430045/Validation_47430045_47430045_20647.388.png", "Validation_47430045/Validation_47430045_47430045_20648.888.png" ], "turn_id": 4, "scan_id": "Validation_47430045", "origin_question": "Is the cream-colored glazed ceramic cup to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the cream-colored glazed ceramic cup to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_47430045/Validation_47430045_47430045_20650.387.png", "Validation_47430045/Validation_47430045_47430045_20651.886.png", "Validation_47430045/Validation_47430045_47430045_20653.386.png", "Validation_47430045/Validation_47430045_47430045_20654.885.png" ], "turn_id": 5, "scan_id": "Validation_47430045", "origin_question": "Which two objects are on the same side of you now? \"the compact stainless steel dishwasher with the sleek design\", \"the moderate-sized plant with lush green foliage\", \"the sleek cylindrical white ceramic flowerpot\". ", "option": [ "the compact stainless steel dishwasher with the sleek design, the moderate-sized plant with lush green foliage", "the compact stainless steel dishwasher with the sleek design, the sleek cylindrical white ceramic flowerpot", "the moderate-sized plant with lush green foliage, the sleek cylindrical white ceramic flowerpot" ], "answer": "the compact stainless steel dishwasher with the sleek design, the moderate-sized plant with lush green foliage", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the compact stainless steel dishwasher with the sleek design\", \"the moderate-sized plant with lush green foliage\", \"the sleek cylindrical white ceramic flowerpot\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the compact stainless steel dishwasher with the sleek design, the moderate-sized plant with lush green foliage', 'the compact stainless steel dishwasher with the sleek design, the sleek cylindrical white ceramic flowerpot', 'the moderate-sized plant with lush green foliage, the sleek cylindrical white ceramic flowerpot'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_47430045/Validation_47430045_47430045_20656.385.png", "Validation_47430045/Validation_47430045_47430045_20657.884.png", "Validation_47430045/Validation_47430045_47430045_20659.383.png", "Validation_47430045/Validation_47430045_47430045_20660.883.png" ], "turn_id": 6, "scan_id": "Validation_47430045", "origin_question": "Please recall \"the sleek cylindrical white ceramic flowerpot\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.23", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the sleek cylindrical white ceramic flowerpot\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "Validation_47430045/Validation_47430045_47430045_20662.382.png", "Validation_47430045/Validation_47430045_47430045_20663.881.png", "Validation_47430045/Validation_47430045_47430045_20665.381.png", "Validation_47430045/Validation_47430045_47430045_20666.897.png" ], "turn_id": 7, "scan_id": "Validation_47430045", "origin_question": "Among these three objects(\"the compact stainless steel dishwasher with the sleek design\",\"the moderate-sized plant with lush green foliage\",\"the sleek rectangular table with the light wood finish\" ), which one is the closest to you?", "option": [ "the compact stainless steel dishwasher with the sleek design", "the moderate-sized plant with lush green foliage", "the sleek rectangular table with the light wood finish" ], "answer": "the compact stainless steel dishwasher with the sleek design", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects(\"the compact stainless steel dishwasher with the sleek design\",\"the moderate-sized plant with lush green foliage\",\"the sleek rectangular table with the light wood finish\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the compact stainless steel dishwasher with the sleek design', 'the moderate-sized plant with lush green foliage', 'the sleek rectangular table with the light wood finish'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42897692/Validation_42897692_42897692_480609.614.png", "Validation_42897692/Validation_42897692_42897692_480611.596.png", "Validation_42897692/Validation_42897692_42897692_480613.596.png", "Validation_42897692/Validation_42897692_42897692_480615.595.png", "Validation_42897692/Validation_42897692_42897692_480617.594.png" ], "turn_id": 1, "scan_id": "Validation_42897692", "origin_question": "Remember, have you seen any bed(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any bed(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "Validation_42897692/Validation_42897692_42897692_480619.593.png", "Validation_42897692/Validation_42897692_42897692_480621.592.png", "Validation_42897692/Validation_42897692_42897692_480623.591.png", "Validation_42897692/Validation_42897692_42897692_480625.591.png", "Validation_42897692/Validation_42897692_42897692_480627.590.png" ], "turn_id": 2, "scan_id": "Validation_42897692", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the monochromatic picture of the misty pathway\"; \"the cylindrical lamp with the soft blue lampshade\"; \"the light brown rectangular polished table with the drawer\" ", "option": [ "the monochromatic picture of the misty pathway", "the cylindrical lamp with the soft blue lampshade", "the light brown rectangular polished table with the drawer" ], "answer": "the monochromatic picture of the misty pathway", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the monochromatic picture of the misty pathway\"; \"the cylindrical lamp with the soft blue lampshade\"; \"the light brown rectangular polished table with the drawer\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the monochromatic picture of the misty pathway', 'the cylindrical lamp with the soft blue lampshade', 'the light brown rectangular polished table with the drawer'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_42897692/Validation_42897692_42897692_480629.589.png", "Validation_42897692/Validation_42897692_42897692_480631.588.png", "Validation_42897692/Validation_42897692_42897692_480633.587.png", "Validation_42897692/Validation_42897692_42897692_480635.586.png", "Validation_42897692/Validation_42897692_42897692_480637.586.png" ], "turn_id": 3, "scan_id": "Validation_42897692", "origin_question": "Please recall \"the monochromatic picture of the misty pathway\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.57", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the monochromatic picture of the misty pathway\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_42897692/Validation_42897692_42897692_480639.601.png", "Validation_42897692/Validation_42897692_42897692_480641.601.png", "Validation_42897692/Validation_42897692_42897692_480643.600.png", "Validation_42897692/Validation_42897692_42897692_480645.599.png", "Validation_42897692/Validation_42897692_42897692_480647.598.png" ], "turn_id": 4, "scan_id": "Validation_42897692", "origin_question": "By the end of the 3 turn, were you getting closer to or farther away from these two objects: X: \"the cylindrical lamp with the soft blue lampshade\", Y: \"the monochromatic picture of the misty pathway\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "farther away from both X and Y", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 3 turn, were you getting closer to or farther away from these two objects: X: \"the cylindrical lamp with the soft blue lampshade\", Y: \"the monochromatic picture of the misty pathway\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_42897692/Validation_42897692_42897692_480649.597.png", "Validation_42897692/Validation_42897692_42897692_480651.596.png", "Validation_42897692/Validation_42897692_42897692_480653.596.png", "Validation_42897692/Validation_42897692_42897692_480655.595.png", "Validation_42897692/Validation_42897692_42897692_480657.594.png" ], "turn_id": 5, "scan_id": "Validation_42897692", "origin_question": "Which direction is the medium rectangular shelf with the light wood finish in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the medium rectangular shelf with the light wood finish in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_42897692/Validation_42897692_42897692_480659.593.png", "Validation_42897692/Validation_42897692_42897692_480661.592.png", "Validation_42897692/Validation_42897692_42897692_480663.591.png", "Validation_42897692/Validation_42897692_42897692_480665.591.png", "Validation_42897692/Validation_42897692_42897692_480667.590.png" ], "turn_id": 6, "scan_id": "Validation_42897692", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_42897692/Validation_42897692_42897692_480669.589.png", "Validation_42897692/Validation_42897692_42897692_480671.588.png", "Validation_42897692/Validation_42897692_42897692_480673.587.png", "Validation_42897692/Validation_42897692_42897692_480675.586.png", "Validation_42897692/Validation_42897692_42897692_480677.586.png" ], "turn_id": 7, "scan_id": "Validation_42897692", "origin_question": "Which direction is the cylindrical lamp with the soft blue lampshade in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the cylindrical lamp with the soft blue lampshade in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_42897692/Validation_42897692_42897692_480679.585.png", "Validation_42897692/Validation_42897692_42897692_480681.601.png", "Validation_42897692/Validation_42897692_42897692_480683.600.png", "Validation_42897692/Validation_42897692_42897692_480685.599.png", "Validation_42897692/Validation_42897692_42897692_480687.598.png" ], "turn_id": 8, "scan_id": "Validation_42897692", "origin_question": "Assuming the direction you are facing at the end of the turn 3 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Left", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 3 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_42897692/Validation_42897692_42897692_480689.597.png", "Validation_42897692/Validation_42897692_42897692_480691.596.png", "Validation_42897692/Validation_42897692_42897692_480693.596.png", "Validation_42897692/Validation_42897692_42897692_480695.595.png", "Validation_42897692/Validation_42897692_42897692_480697.594.png" ], "turn_id": 9, "scan_id": "Validation_42897692", "origin_question": "When did you first discover \"the monochromatic picture of the misty pathway\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the monochromatic picture of the misty pathway\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_42897692/Validation_42897692_42897692_480699.593.png", "Validation_42897692/Validation_42897692_42897692_480701.592.png", "Validation_42897692/Validation_42897692_42897692_480703.592.png", "Validation_42897692/Validation_42897692_42897692_480705.591.png", "Validation_42897692/Validation_42897692_42897692_480707.590.png" ], "turn_id": 10, "scan_id": "Validation_42897692", "origin_question": "In which turn were you closest to \"the white rectangular metal radiator\"? ", "option": [], "answer": "10", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the white rectangular metal radiator\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47331311/Validation_47331311_47331311_177568.495.png", "Validation_47331311/Validation_47331311_47331311_177570.478.png", "Validation_47331311/Validation_47331311_47331311_177572.477.png", "Validation_47331311/Validation_47331311_47331311_177574.476.png" ], "turn_id": 1, "scan_id": "Validation_47331311", "origin_question": "Remember, have you seen any door(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any door(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_47331311/Validation_47331311_47331311_177576.475.png", "Validation_47331311/Validation_47331311_47331311_177578.474.png", "Validation_47331311/Validation_47331311_47331311_177580.474.png", "Validation_47331311/Validation_47331311_47331311_177582.473.png" ], "turn_id": 2, "scan_id": "Validation_47331311", "origin_question": "Which two objects are on the same side of you now? \"the medium-sized gray rectangular table made of wood\", \"the sleek light fixture with the white glass cover\", \"the door\". ", "option": [ "the medium-sized gray rectangular table made of wood, the sleek light fixture with the white glass cover", "the medium-sized gray rectangular table made of wood, the door", "the sleek light fixture with the white glass cover, the door" ], "answer": "the medium-sized gray rectangular table made of wood, the sleek light fixture with the white glass cover", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the medium-sized gray rectangular table made of wood\", \"the sleek light fixture with the white glass cover\", \"the door\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the medium-sized gray rectangular table made of wood, the sleek light fixture with the white glass cover', 'the medium-sized gray rectangular table made of wood, the door', 'the sleek light fixture with the white glass cover, the door'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_47331311/Validation_47331311_47331311_177584.472.png", "Validation_47331311/Validation_47331311_47331311_177586.471.png", "Validation_47331311/Validation_47331311_47331311_177588.470.png", "Validation_47331311/Validation_47331311_47331311_177590.470.png" ], "turn_id": 3, "scan_id": "Validation_47331311", "origin_question": "When did you first discover \"the sleek light fixture with the white glass cover\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the sleek light fixture with the white glass cover\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47331311/Validation_47331311_47331311_177592.469.png", "Validation_47331311/Validation_47331311_47331311_177594.468.png", "Validation_47331311/Validation_47331311_47331311_177596.467.png", "Validation_47331311/Validation_47331311_47331311_177598.466.png" ], "turn_id": 4, "scan_id": "Validation_47331311", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the tall, modern, light-colored shelf\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the tall, modern, light-colored shelf\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_47331311/Validation_47331311_47331311_177600.465.png", "Validation_47331311/Validation_47331311_47331311_177602.481.png", "Validation_47331311/Validation_47331311_47331311_177604.481.png", "Validation_47331311/Validation_47331311_47331311_177606.480.png" ], "turn_id": 5, "scan_id": "Validation_47331311", "origin_question": "At the end of which turn were both of \"the door\" and \"the vibrant yellow plastic chair with rounded backrest\" on your right side?", "option": [], "answer": "5", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the door\" and \"the vibrant yellow plastic chair with rounded backrest\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_47331311/Validation_47331311_47331311_177608.479.png", "Validation_47331311/Validation_47331311_47331311_177610.478.png", "Validation_47331311/Validation_47331311_47331311_177612.477.png", "Validation_47331311/Validation_47331311_47331311_177614.476.png" ], "turn_id": 6, "scan_id": "Validation_47331311", "origin_question": "Which two objects are on the same side of you now? \"the saddle-shaped, black leather-like stool with foam cushioning\", \"the door\", \"the tall, modern, light-colored shelf\". ", "option": [ "the saddle-shaped, black leather-like stool with foam cushioning, the door", "the saddle-shaped, black leather-like stool with foam cushioning, the tall, modern, light-colored shelf", "the door, the tall, modern, light-colored shelf" ], "answer": "the saddle-shaped, black leather-like stool with foam cushioning, the door", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the saddle-shaped, black leather-like stool with foam cushioning\", \"the door\", \"the tall, modern, light-colored shelf\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the saddle-shaped, black leather-like stool with foam cushioning, the door', 'the saddle-shaped, black leather-like stool with foam cushioning, the tall, modern, light-colored shelf', 'the door, the tall, modern, light-colored shelf'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_47331311/Validation_47331311_47331311_177616.476.png", "Validation_47331311/Validation_47331311_47331311_177618.475.png", "Validation_47331311/Validation_47331311_47331311_177620.474.png", "Validation_47331311/Validation_47331311_47331311_177622.473.png" ], "turn_id": 7, "scan_id": "Validation_47331311", "origin_question": "Please recall \"the vibrant yellow plastic chair with rounded backrest\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.54", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the vibrant yellow plastic chair with rounded backrest\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42897501/Validation_42897501_42897501_270010.602.png", "Validation_42897501/Validation_42897501_42897501_270012.584.png", "Validation_42897501/Validation_42897501_42897501_270014.584.png", "Validation_42897501/Validation_42897501_42897501_270016.583.png" ], "turn_id": 1, "scan_id": "Validation_42897501", "origin_question": "Remember, have you seen any sink(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any sink(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_42897501/Validation_42897501_42897501_270018.582.png", "Validation_42897501/Validation_42897501_42897501_270020.581.png", "Validation_42897501/Validation_42897501_42897501_270022.580.png", "Validation_42897501/Validation_42897501_42897501_270024.579.png" ], "turn_id": 2, "scan_id": "Validation_42897501", "origin_question": "When did you first discover \"the modern rectangular stainless steel sink\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the modern rectangular stainless steel sink\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_42897501/Validation_42897501_42897501_270026.579.png", "Validation_42897501/Validation_42897501_42897501_270028.578.png", "Validation_42897501/Validation_42897501_42897501_270030.577.png", "Validation_42897501/Validation_42897501_42897501_270032.593.png" ], "turn_id": 3, "scan_id": "Validation_42897501", "origin_question": "At the end of which turn were both of \"the modern rectangular stainless steel sink\" and \"the medium-sized window with the clean white frame\" on your right side?", "option": [], "answer": "2", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the modern rectangular stainless steel sink\" and \"the medium-sized window with the clean white frame\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_42897501/Validation_42897501_42897501_270034.592.png", "Validation_42897501/Validation_42897501_42897501_270036.591.png", "Validation_42897501/Validation_42897501_42897501_270038.590.png", "Validation_42897501/Validation_42897501_42897501_270040.590.png" ], "turn_id": 4, "scan_id": "Validation_42897501", "origin_question": "Which direction is the modern rectangular stainless steel sink in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the modern rectangular stainless steel sink in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_42897501/Validation_42897501_42897501_270042.589.png", "Validation_42897501/Validation_42897501_42897501_270044.588.png", "Validation_42897501/Validation_42897501_42897501_270046.587.png", "Validation_42897501/Validation_42897501_42897501_270048.586.png" ], "turn_id": 5, "scan_id": "Validation_42897501", "origin_question": "In which turn did you see both \"the door\" and \"the sleek stainless steel pyramid-shaped range hood\" simultaneously? ", "option": [], "answer": "4", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the door\" and \"the sleek stainless steel pyramid-shaped range hood\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "Validation_42897501/Validation_42897501_42897501_270050.585.png", "Validation_42897501/Validation_42897501_42897501_270052.585.png", "Validation_42897501/Validation_42897501_42897501_270054.584.png", "Validation_42897501/Validation_42897501_42897501_270056.583.png" ], "turn_id": 6, "scan_id": "Validation_42897501", "origin_question": "Among these three objects(\"the medium-sized window with the clean white frame\",\"the door\",\"the medium-sized yellow absorbent kitchen towel\" ), which one is the farthest from you?", "option": [ "the medium-sized window with the clean white frame", "the door", "the medium-sized yellow absorbent kitchen towel" ], "answer": "the medium-sized window with the clean white frame", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects(\"the medium-sized window with the clean white frame\",\"the door\",\"the medium-sized yellow absorbent kitchen towel\" ), which one is the farthest from you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the medium-sized window with the clean white frame', 'the door', 'the medium-sized yellow absorbent kitchen towel'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_42897501/Validation_42897501_42897501_270058.582.png", "Validation_42897501/Validation_42897501_42897501_270060.581.png", "Validation_42897501/Validation_42897501_42897501_270062.581.png", "Validation_42897501/Validation_42897501_42897501_270064.580.png" ], "turn_id": 7, "scan_id": "Validation_42897501", "origin_question": "Is the modern rectangular stainless steel sink to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the modern rectangular stainless steel sink to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_42897501/Validation_42897501_42897501_270066.579.png", "Validation_42897501/Validation_42897501_42897501_270068.578.png", "Validation_42897501/Validation_42897501_42897501_270070.577.png", "Validation_42897501/Validation_42897501_42897501_270072.593.png" ], "turn_id": 8, "scan_id": "Validation_42897501", "origin_question": "In which turn did you see both \"the door\" and \"the sleek stainless steel pyramid-shaped range hood\" simultaneously? ", "option": [], "answer": "4", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the door\" and \"the sleek stainless steel pyramid-shaped range hood\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "Validation_47115460/Validation_47115460_47115460_13137.784.png", "Validation_47115460/Validation_47115460_47115460_13139.767.png", "Validation_47115460/Validation_47115460_47115460_13141.766.png", "Validation_47115460/Validation_47115460_47115460_13143.782.png" ], "turn_id": 1, "scan_id": "Validation_47115460", "origin_question": "Remember, how many sink(s) have you seen so far? ", "option": [], "answer": "2", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many sink(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_47115460/Validation_47115460_47115460_13145.781.png", "Validation_47115460/Validation_47115460_47115460_13147.780.png", "Validation_47115460/Validation_47115460_47115460_13149.779.png", "Validation_47115460/Validation_47115460_47115460_13151.779.png" ], "turn_id": 2, "scan_id": "Validation_47115460", "origin_question": "In which turn did you see both \"the sleek stainless steel and glass oven\" and \"the compact wooden case with the rustic finish\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the sleek stainless steel and glass oven\" and \"the compact wooden case with the rustic finish\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_47115460/Validation_47115460_47115460_13153.778.png", "Validation_47115460/Validation_47115460_47115460_13155.777.png", "Validation_47115460/Validation_47115460_47115460_13157.776.png", "Validation_47115460/Validation_47115460_47115460_13159.775.png" ], "turn_id": 3, "scan_id": "Validation_47115460", "origin_question": "At the end of which turn were both of \"the stainless steel stove with black accents\" and \"the modern rectangular stainless steel range hood\" on your right side?", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the stainless steel stove with black accents\" and \"the modern rectangular stainless steel range hood\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_47115460/Validation_47115460_47115460_13161.774.png", "Validation_47115460/Validation_47115460_47115460_13163.773.png", "Validation_47115460/Validation_47115460_47115460_13165.773.png", "Validation_47115460/Validation_47115460_47115460_13167.772.png" ], "turn_id": 4, "scan_id": "Validation_47115460", "origin_question": "Is the sleek stainless steel and glass oven to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the sleek stainless steel and glass oven to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_47115460/Validation_47115460_47115460_13169.771.png", "Validation_47115460/Validation_47115460_47115460_13171.770.png", "Validation_47115460/Validation_47115460_47115460_13173.769.png", "Validation_47115460/Validation_47115460_47115460_13175.769.png" ], "turn_id": 5, "scan_id": "Validation_47115460", "origin_question": "Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "109.82", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_47115460/Validation_47115460_47115460_13177.768.png", "Validation_47115460/Validation_47115460_47115460_13179.767.png", "Validation_47115460/Validation_47115460_47115460_13181.766.png", "Validation_47115460/Validation_47115460_47115460_13183.782.png" ], "turn_id": 6, "scan_id": "Validation_47115460", "origin_question": "How far is your current position from where you were at the end of the turn 1? (in meters)", "option": [], "answer": "5.5", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 1? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_47115460/Validation_47115460_47115460_13185.781.png", "Validation_47115460/Validation_47115460_47115460_13187.780.png", "Validation_47115460/Validation_47115460_47115460_13189.779.png", "Validation_47115460/Validation_47115460_47115460_13191.779.png" ], "turn_id": 7, "scan_id": "Validation_47115460", "origin_question": "Is the stainless steel stove with black accents to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the stainless steel stove with black accents to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_47115460/Validation_47115460_47115460_13193.778.png", "Validation_47115460/Validation_47115460_47115460_13195.777.png", "Validation_47115460/Validation_47115460_47115460_13197.776.png", "Validation_47115460/Validation_47115460_47115460_13199.775.png" ], "turn_id": 8, "scan_id": "Validation_47115460", "origin_question": "Which two objects are on the same side of you now? \"the stainless steel stove with black accents\", \"the medium-sized stainless steel cylinder bin\", \"the circular glass table with the chrome pedestal\". ", "option": [ "the stainless steel stove with black accents, the medium-sized stainless steel cylinder bin", "the stainless steel stove with black accents, the circular glass table with the chrome pedestal", "the medium-sized stainless steel cylinder bin, the circular glass table with the chrome pedestal" ], "answer": "the stainless steel stove with black accents, the medium-sized stainless steel cylinder bin", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the stainless steel stove with black accents\", \"the medium-sized stainless steel cylinder bin\", \"the circular glass table with the chrome pedestal\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the stainless steel stove with black accents, the medium-sized stainless steel cylinder bin', 'the stainless steel stove with black accents, the circular glass table with the chrome pedestal', 'the medium-sized stainless steel cylinder bin, the circular glass table with the chrome pedestal'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_47115460/Validation_47115460_47115460_13201.774.png", "Validation_47115460/Validation_47115460_47115460_13203.774.png", "Validation_47115460/Validation_47115460_47115460_13205.773.png", "Validation_47115460/Validation_47115460_47115460_13207.772.png" ], "turn_id": 9, "scan_id": "Validation_47115460", "origin_question": "Is the medium-sized stainless steel cylinder bin to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 9 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the medium-sized stainless steel cylinder bin to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47430424/Validation_47430424_47430424_696778.476.png", "Validation_47430424/Validation_47430424_47430424_696780.442.png", "Validation_47430424/Validation_47430424_47430424_696782.441.png", "Validation_47430424/Validation_47430424_47430424_696784.441.png" ], "turn_id": 1, "scan_id": "Validation_47430424", "origin_question": "Remember, have you seen any bathtub(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any bathtub(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_47430424/Validation_47430424_47430424_696786.456.png", "Validation_47430424/Validation_47430424_47430424_696788.456.png", "Validation_47430424/Validation_47430424_47430424_696790.455.png", "Validation_47430424/Validation_47430424_47430424_696792.454.png" ], "turn_id": 2, "scan_id": "Validation_47430424", "origin_question": "When did you first discover \"the bathtub\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the bathtub\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_47430424/Validation_47430424_47430424_696794.453.png", "Validation_47430424/Validation_47430424_47430424_696796.452.png", "Validation_47430424/Validation_47430424_47430424_696798.451.png", "Validation_47430424/Validation_47430424_47430424_696800.451.png" ], "turn_id": 3, "scan_id": "Validation_47430424", "origin_question": "At the end of which turn were both of \"the bathtub\" and \"the contemporary white porcelain toilet with the closed lid\" on your right side?", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the bathtub\" and \"the contemporary white porcelain toilet with the closed lid\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_47430424/Validation_47430424_47430424_696802.450.png", "Validation_47430424/Validation_47430424_47430424_696804.449.png", "Validation_47430424/Validation_47430424_47430424_696806.448.png", "Validation_47430424/Validation_47430424_47430424_696808.448.png" ], "turn_id": 4, "scan_id": "Validation_47430424", "origin_question": "Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "64.07", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_47430424/Validation_47430424_47430424_696810.447.png", "Validation_47430424/Validation_47430424_47430424_696812.446.png", "Validation_47430424/Validation_47430424_47430424_696814.445.png", "Validation_47430424/Validation_47430424_47430424_696816.444.png" ], "turn_id": 5, "scan_id": "Validation_47430424", "origin_question": "Please recall \"the plush, deep purple towel\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.73", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the plush, deep purple towel\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_47430424/Validation_47430424_47430424_696818.444.png", "Validation_47430424/Validation_47430424_47430424_696820.443.png", "Validation_47430424/Validation_47430424_47430424_696822.442.png", "Validation_47430424/Validation_47430424_47430424_696824.441.png" ], "turn_id": 6, "scan_id": "Validation_47430424", "origin_question": "At the end of which turn were both of \"the plush, deep purple towel\" and \"the large rectangular frameless silver mirror\" on your right side?", "option": [], "answer": "5", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the plush, deep purple towel\" and \"the large rectangular frameless silver mirror\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_48018956/Validation_48018956_48018956_555989.241.png", "Validation_48018956/Validation_48018956_48018956_555990.724.png", "Validation_48018956/Validation_48018956_48018956_555992.223.png", "Validation_48018956/Validation_48018956_48018956_555993.723.png" ], "turn_id": 1, "scan_id": "Validation_48018956", "origin_question": "Remember, have you seen any stool(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any stool(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "Validation_48018956/Validation_48018956_48018956_555995.222.png", "Validation_48018956/Validation_48018956_48018956_555996.721.png", "Validation_48018956/Validation_48018956_48018956_555998.221.png", "Validation_48018956/Validation_48018956_48018956_555999.720.png" ], "turn_id": 2, "scan_id": "Validation_48018956", "origin_question": "Remember, how many chair(s) have you seen so far? ", "option": [], "answer": "5", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many chair(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_48018956/Validation_48018956_48018956_556001.220.png", "Validation_48018956/Validation_48018956_48018956_556002.719.png", "Validation_48018956/Validation_48018956_48018956_556004.218.png", "Validation_48018956/Validation_48018956_48018956_556005.718.png" ], "turn_id": 3, "scan_id": "Validation_48018956", "origin_question": "What will be the first-time appearance order of the following categories: cabinet, couch, table?", "option": [ "cabinet, couch, table", "couch, cabinet, table", "table, cabinet, couch", "couch, table, cabinet" ], "answer": "table, cabinet, couch", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: cabinet, couch, table? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['cabinet, couch, table', 'couch, cabinet, table', 'table, cabinet, couch', 'couch, table, cabinet'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_48018956/Validation_48018956_48018956_556007.217.png", "Validation_48018956/Validation_48018956_48018956_556008.716.png", "Validation_48018956/Validation_48018956_48018956_556010.216.png", "Validation_48018956/Validation_48018956_48018956_556011.715.png" ], "turn_id": 4, "scan_id": "Validation_48018956", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_48018956/Validation_48018956_48018956_556013.215.png", "Validation_48018956/Validation_48018956_48018956_556014.714.png", "Validation_48018956/Validation_48018956_48018956_556016.213.png", "Validation_48018956/Validation_48018956_48018956_556017.713.png" ], "turn_id": 5, "scan_id": "Validation_48018956", "origin_question": "Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Right", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_48018956/Validation_48018956_48018956_556019.212.png", "Validation_48018956/Validation_48018956_48018956_556020.728.png", "Validation_48018956/Validation_48018956_48018956_556022.228.png", "Validation_48018956/Validation_48018956_48018956_556023.727.png" ], "turn_id": 6, "scan_id": "Validation_48018956", "origin_question": "What will be the first-time appearance order of the following categories: couch, monitor, oven?", "option": [ "monitor, oven, couch", "monitor, couch, oven", "couch, monitor, oven", "oven, couch, monitor" ], "answer": "couch, monitor, oven", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: couch, monitor, oven? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['monitor, oven, couch', 'monitor, couch, oven', 'couch, monitor, oven', 'oven, couch, monitor'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42897566/Validation_42897566_42897566_279040.021.png", "Validation_42897566/Validation_42897566_42897566_279042.003.png", "Validation_42897566/Validation_42897566_42897566_279044.003.png", "Validation_42897566/Validation_42897566_42897566_279046.002.png" ], "turn_id": 1, "scan_id": "Validation_42897566", "origin_question": "Remember, have you seen any chair(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any chair(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42897566/Validation_42897566_42897566_279048.001.png", "Validation_42897566/Validation_42897566_42897566_279050.000.png", "Validation_42897566/Validation_42897566_42897566_279052.016.png", "Validation_42897566/Validation_42897566_42897566_279054.015.png" ], "turn_id": 2, "scan_id": "Validation_42897566", "origin_question": "Remember, have you seen any radiator(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any radiator(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_42897566/Validation_42897566_42897566_279056.014.png", "Validation_42897566/Validation_42897566_42897566_279058.014.png", "Validation_42897566/Validation_42897566_42897566_279060.013.png", "Validation_42897566/Validation_42897566_42897566_279062.012.png" ], "turn_id": 3, "scan_id": "Validation_42897566", "origin_question": "In which turn were you closest to \"the sleek rectangular white radiator with vertical fins\"? ", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the sleek rectangular white radiator with vertical fins\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_42897566/Validation_42897566_42897566_279064.011.png", "Validation_42897566/Validation_42897566_42897566_279066.010.png", "Validation_42897566/Validation_42897566_42897566_279068.009.png", "Validation_42897566/Validation_42897566_42897566_279070.008.png" ], "turn_id": 4, "scan_id": "Validation_42897566", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the window\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the window\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_42897566/Validation_42897566_42897566_279072.008.png", "Validation_42897566/Validation_42897566_42897566_279074.007.png", "Validation_42897566/Validation_42897566_42897566_279076.006.png", "Validation_42897566/Validation_42897566_42897566_279078.005.png" ], "turn_id": 5, "scan_id": "Validation_42897566", "origin_question": "Which direction is the sleek rectangular white radiator with vertical fins in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the sleek rectangular white radiator with vertical fins in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_42897566/Validation_42897566_42897566_279080.004.png", "Validation_42897566/Validation_42897566_42897566_279082.004.png", "Validation_42897566/Validation_42897566_42897566_279084.003.png", "Validation_42897566/Validation_42897566_42897566_279086.002.png" ], "turn_id": 6, "scan_id": "Validation_42897566", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "137.38", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_48458456/Validation_48458456_48458456_110449.151.png", "Validation_48458456/Validation_48458456_48458456_110450.651.png", "Validation_48458456/Validation_48458456_48458456_110452.150.png", "Validation_48458456/Validation_48458456_48458456_110453.649.png" ], "turn_id": 1, "scan_id": "Validation_48458456", "origin_question": "Remember, have you seen any pool table(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any pool table(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_48458456/Validation_48458456_48458456_110455.149.png", "Validation_48458456/Validation_48458456_48458456_110456.648.png", "Validation_48458456/Validation_48458456_48458456_110458.148.png", "Validation_48458456/Validation_48458456_48458456_110459.647.png" ], "turn_id": 2, "scan_id": "Validation_48458456", "origin_question": "In which turn did you see both \"the compact, modern lamp with metal and glass\" and \"the warm beige carpet made of soft synthetic fibers\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the compact, modern lamp with metal and glass\" and \"the warm beige carpet made of soft synthetic fibers\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_48458456/Validation_48458456_48458456_110461.146.png", "Validation_48458456/Validation_48458456_48458456_110462.646.png", "Validation_48458456/Validation_48458456_48458456_110464.145.png", "Validation_48458456/Validation_48458456_48458456_110465.645.png" ], "turn_id": 3, "scan_id": "Validation_48458456", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the white ceramic oval plate with curved edges\" from your position?", "option": [], "answer": "78.42", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the white ceramic oval plate with curved edges\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_48458456/Validation_48458456_48458456_110467.144.png", "Validation_48458456/Validation_48458456_48458456_110468.643.png", "Validation_48458456/Validation_48458456_48458456_110470.143.png", "Validation_48458456/Validation_48458456_48458456_110471.642.png" ], "turn_id": 4, "scan_id": "Validation_48458456", "origin_question": "Which direction is the large rectangular mirror with the wooden frame in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the large rectangular mirror with the wooden frame in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_41159534/Validation_41159534_41159534_39020.832.png", "Validation_41159534/Validation_41159534_41159534_39022.815.png", "Validation_41159534/Validation_41159534_41159534_39024.814.png", "Validation_41159534/Validation_41159534_41159534_39026.813.png" ], "turn_id": 1, "scan_id": "Validation_41159534", "origin_question": "Remember, have you seen any door(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any door(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_41159534/Validation_41159534_41159534_39028.812.png", "Validation_41159534/Validation_41159534_41159534_39030.811.png", "Validation_41159534/Validation_41159534_41159534_39032.810.png", "Validation_41159534/Validation_41159534_41159534_39034.810.png" ], "turn_id": 2, "scan_id": "Validation_41159534", "origin_question": "When did you first discover \"the sleek white wooden door with the metal knob\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the sleek white wooden door with the metal knob\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_41159534/Validation_41159534_41159534_39036.809.png", "Validation_41159534/Validation_41159534_41159534_39038.808.png", "Validation_41159534/Validation_41159534_41159534_39040.807.png", "Validation_41159534/Validation_41159534_41159534_39042.806.png" ], "turn_id": 3, "scan_id": "Validation_41159534", "origin_question": "In which turn did you see both \"the sleek black cap with an adjustable strap\" and \"the silver, rounded stainless steel kettle\" simultaneously? ", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the sleek black cap with an adjustable strap\" and \"the silver, rounded stainless steel kettle\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_41159534/Validation_41159534_41159534_39044.805.png", "Validation_41159534/Validation_41159534_41159534_39046.821.png", "Validation_41159534/Validation_41159534_41159534_39048.820.png", "Validation_41159534/Validation_41159534_41159534_39050.820.png" ], "turn_id": 4, "scan_id": "Validation_41159534", "origin_question": "Please recall \"the sleek gray square table for compact spaces\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.19", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the sleek gray square table for compact spaces\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_41159534/Validation_41159534_41159534_39052.819.png", "Validation_41159534/Validation_41159534_41159534_39054.818.png", "Validation_41159534/Validation_41159534_41159534_39056.817.png", "Validation_41159534/Validation_41159534_41159534_39058.816.png" ], "turn_id": 5, "scan_id": "Validation_41159534", "origin_question": "Is the dark brown wooden grid shelf with open compartments to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the dark brown wooden grid shelf with open compartments to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_41159534/Validation_41159534_41159534_39060.815.png", "Validation_41159534/Validation_41159534_41159534_39062.815.png", "Validation_41159534/Validation_41159534_41159534_39064.814.png", "Validation_41159534/Validation_41159534_41159534_39066.813.png" ], "turn_id": 6, "scan_id": "Validation_41159534", "origin_question": "Please recall \"the sleek black cap with an adjustable strap\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.76", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the sleek black cap with an adjustable strap\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_41159534/Validation_41159534_41159534_39068.812.png", "Validation_41159534/Validation_41159534_41159534_39070.811.png", "Validation_41159534/Validation_41159534_41159534_39072.810.png", "Validation_41159534/Validation_41159534_41159534_39074.810.png" ], "turn_id": 7, "scan_id": "Validation_41159534", "origin_question": "How far is your current position from where you were at the end of the turn 4? (in meters)", "option": [], "answer": "1.13", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 4? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47895374/Validation_47895374_47895374_101801.662.png", "Validation_47895374/Validation_47895374_47895374_101803.628.png", "Validation_47895374/Validation_47895374_47895374_101805.644.png", "Validation_47895374/Validation_47895374_47895374_101807.643.png", "Validation_47895374/Validation_47895374_47895374_101809.642.png" ], "turn_id": 1, "scan_id": "Validation_47895374", "origin_question": "Remember, have you seen any chair(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any chair(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_47895374/Validation_47895374_47895374_101811.641.png", "Validation_47895374/Validation_47895374_47895374_101813.640.png", "Validation_47895374/Validation_47895374_47895374_101815.640.png", "Validation_47895374/Validation_47895374_47895374_101817.639.png", "Validation_47895374/Validation_47895374_47895374_101819.638.png" ], "turn_id": 2, "scan_id": "Validation_47895374", "origin_question": "When did you first discover \"the stack of colorful books with glossy or matte covers\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the stack of colorful books with glossy or matte covers\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47895374/Validation_47895374_47895374_101821.637.png", "Validation_47895374/Validation_47895374_47895374_101823.636.png", "Validation_47895374/Validation_47895374_47895374_101825.636.png", "Validation_47895374/Validation_47895374_47895374_101827.635.png", "Validation_47895374/Validation_47895374_47895374_101829.634.png" ], "turn_id": 3, "scan_id": "Validation_47895374", "origin_question": "In which turn were you farthest from \"the light cream ergonomic chair with the metal frame\"? ", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the light cream ergonomic chair with the metal frame\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_47895374/Validation_47895374_47895374_101831.633.png", "Validation_47895374/Validation_47895374_47895374_101833.632.png", "Validation_47895374/Validation_47895374_47895374_101835.631.png", "Validation_47895374/Validation_47895374_47895374_101837.631.png", "Validation_47895374/Validation_47895374_47895374_101839.630.png" ], "turn_id": 4, "scan_id": "Validation_47895374", "origin_question": "When did you first discover \"the large, silver-gray ball\"? (index of the turn)", "option": [], "answer": "3", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the large, silver-gray ball\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_47895374/Validation_47895374_47895374_101841.629.png", "Validation_47895374/Validation_47895374_47895374_101843.628.png", "Validation_47895374/Validation_47895374_47895374_101845.627.png", "Validation_47895374/Validation_47895374_47895374_101847.643.png", "Validation_47895374/Validation_47895374_47895374_101849.642.png" ], "turn_id": 5, "scan_id": "Validation_47895374", "origin_question": "By the end of the 4 turn, were you getting closer to or farther away from these two objects: X: \"the large, silver-gray ball\", Y: \"the soft white cotton pillow in the slanted position\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 4 turn, were you getting closer to or farther away from these two objects: X: \"the large, silver-gray ball\", Y: \"the soft white cotton pillow in the slanted position\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "Validation_47895374/Validation_47895374_47895374_101851.642.png", "Validation_47895374/Validation_47895374_47895374_101853.641.png", "Validation_47895374/Validation_47895374_47895374_101855.640.png", "Validation_47895374/Validation_47895374_47895374_101857.639.png", "Validation_47895374/Validation_47895374_47895374_101859.638.png" ], "turn_id": 6, "scan_id": "Validation_47895374", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the couch\"; \"the light beige stool with dark brown legs\"; \"the soft white cotton pillow in the slanted position\" ", "option": [ "the couch", "the light beige stool with dark brown legs", "the soft white cotton pillow in the slanted position" ], "answer": "the couch", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the couch\"; \"the light beige stool with dark brown legs\"; \"the soft white cotton pillow in the slanted position\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the couch', 'the light beige stool with dark brown legs', 'the soft white cotton pillow in the slanted position'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_47895374/Validation_47895374_47895374_101861.638.png", "Validation_47895374/Validation_47895374_47895374_101863.637.png", "Validation_47895374/Validation_47895374_47895374_101865.636.png", "Validation_47895374/Validation_47895374_47895374_101867.635.png", "Validation_47895374/Validation_47895374_47895374_101869.634.png" ], "turn_id": 7, "scan_id": "Validation_47895374", "origin_question": "How far is your current position from where you were at the end of the turn 4? (in meters)", "option": [], "answer": "1.78", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 4? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_47895374/Validation_47895374_47895374_101871.633.png", "Validation_47895374/Validation_47895374_47895374_101873.633.png", "Validation_47895374/Validation_47895374_47895374_101875.632.png", "Validation_47895374/Validation_47895374_47895374_101877.631.png", "Validation_47895374/Validation_47895374_47895374_101879.630.png" ], "turn_id": 8, "scan_id": "Validation_47895374", "origin_question": "Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Right", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47204575/Validation_47204575_47204575_4835.224.png", "Validation_47204575/Validation_47204575_47204575_4837.223.png", "Validation_47204575/Validation_47204575_47204575_4839.222.png", "Validation_47204575/Validation_47204575_47204575_4841.222.png" ], "turn_id": 1, "scan_id": "Validation_47204575", "origin_question": "Remember, have you seen any carpet(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any carpet(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47204575/Validation_47204575_47204575_4843.221.png", "Validation_47204575/Validation_47204575_47204575_4845.220.png", "Validation_47204575/Validation_47204575_47204575_4847.219.png", "Validation_47204575/Validation_47204575_47204575_4849.218.png" ], "turn_id": 2, "scan_id": "Validation_47204575", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the modern square table with the sleek white surface\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the modern square table with the sleek white surface\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47204575/Validation_47204575_47204575_4851.217.png", "Validation_47204575/Validation_47204575_47204575_4853.217.png", "Validation_47204575/Validation_47204575_47204575_4855.216.png", "Validation_47204575/Validation_47204575_47204575_4857.215.png" ], "turn_id": 3, "scan_id": "Validation_47204575", "origin_question": "In which turn were you closest to \"the rectangular gray and yellow geometric carpet\"? ", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the rectangular gray and yellow geometric carpet\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47204575/Validation_47204575_47204575_4859.214.png", "Validation_47204575/Validation_47204575_47204575_4861.213.png", "Validation_47204575/Validation_47204575_47204575_4863.213.png", "Validation_47204575/Validation_47204575_47204575_4865.212.png" ], "turn_id": 4, "scan_id": "Validation_47204575", "origin_question": "Which direction is the medium clear plastic rectangular box in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the medium clear plastic rectangular box in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47204575/Validation_47204575_47204575_4867.211.png", "Validation_47204575/Validation_47204575_47204575_4869.210.png", "Validation_47204575/Validation_47204575_47204575_4871.209.png", "Validation_47204575/Validation_47204575_47204575_4873.208.png" ], "turn_id": 5, "scan_id": "Validation_47204575", "origin_question": "In which turn were you closest to \"the glossy black monitor with rounded corners\"? ", "option": [], "answer": "5", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the glossy black monitor with rounded corners\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_47204575/Validation_47204575_47204575_4875.208.png", "Validation_47204575/Validation_47204575_47204575_4877.223.png", "Validation_47204575/Validation_47204575_47204575_4879.223.png", "Validation_47204575/Validation_47204575_47204575_4881.222.png" ], "turn_id": 6, "scan_id": "Validation_47204575", "origin_question": "Please recall \"the pastel-colored toy made of plastic\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "3.46", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the pastel-colored toy made of plastic\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42899734/Validation_42899734_42899734_800763.584.png", "Validation_42899734/Validation_42899734_42899734_800765.566.png", "Validation_42899734/Validation_42899734_42899734_800767.565.png", "Validation_42899734/Validation_42899734_42899734_800769.581.png", "Validation_42899734/Validation_42899734_42899734_800771.580.png" ], "turn_id": 1, "scan_id": "Validation_42899734", "origin_question": "Remember, have you seen any bed(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any bed(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_42899734/Validation_42899734_42899734_800773.579.png", "Validation_42899734/Validation_42899734_42899734_800775.579.png", "Validation_42899734/Validation_42899734_42899734_800777.578.png", "Validation_42899734/Validation_42899734_42899734_800779.577.png", "Validation_42899734/Validation_42899734_42899734_800781.576.png" ], "turn_id": 2, "scan_id": "Validation_42899734", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the small cylindrical gray plastic bin\" from your position?", "option": [], "answer": "78.49", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the small cylindrical gray plastic bin\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_42899734/Validation_42899734_42899734_800783.575.png", "Validation_42899734/Validation_42899734_42899734_800785.574.png", "Validation_42899734/Validation_42899734_42899734_800787.574.png", "Validation_42899734/Validation_42899734_42899734_800789.573.png", "Validation_42899734/Validation_42899734_42899734_800791.572.png" ], "turn_id": 3, "scan_id": "Validation_42899734", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the plush, large white cotton towel\" from your position?", "option": [], "answer": "57.55", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the plush, large white cotton towel\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_42899734/Validation_42899734_42899734_800793.571.png", "Validation_42899734/Validation_42899734_42899734_800795.570.png", "Validation_42899734/Validation_42899734_42899734_800797.570.png", "Validation_42899734/Validation_42899734_42899734_800799.569.png", "Validation_42899734/Validation_42899734_42899734_800801.568.png" ], "turn_id": 4, "scan_id": "Validation_42899734", "origin_question": "In which turn did you see both \"the cylindrical lamp with the chrome base\" and \"the plush, large white cotton towel\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the cylindrical lamp with the chrome base\" and \"the plush, large white cotton towel\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_42899734/Validation_42899734_42899734_800803.567.png", "Validation_42899734/Validation_42899734_42899734_800805.566.png", "Validation_42899734/Validation_42899734_42899734_800807.566.png", "Validation_42899734/Validation_42899734_42899734_800809.581.png", "Validation_42899734/Validation_42899734_42899734_800811.581.png" ], "turn_id": 5, "scan_id": "Validation_42899734", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the small cylindrical gray plastic bin\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the small cylindrical gray plastic bin\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_42899734/Validation_42899734_42899734_800813.580.png", "Validation_42899734/Validation_42899734_42899734_800815.579.png", "Validation_42899734/Validation_42899734_42899734_800817.578.png", "Validation_42899734/Validation_42899734_42899734_800819.577.png", "Validation_42899734/Validation_42899734_42899734_800821.576.png" ], "turn_id": 6, "scan_id": "Validation_42899734", "origin_question": "Please recall \"the cylindrical lamp with the chrome base\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.87", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the cylindrical lamp with the chrome base\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_42899734/Validation_42899734_42899734_800823.576.png", "Validation_42899734/Validation_42899734_42899734_800825.575.png", "Validation_42899734/Validation_42899734_42899734_800827.574.png", "Validation_42899734/Validation_42899734_42899734_800829.573.png", "Validation_42899734/Validation_42899734_42899734_800831.572.png" ], "turn_id": 7, "scan_id": "Validation_42899734", "origin_question": "At the end of which turn were both of \"the classic double bed with elegant brown bedding\" and \"the small cylindrical gray plastic bin\" on your right side?", "option": [], "answer": "5", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the classic double bed with elegant brown bedding\" and \"the small cylindrical gray plastic bin\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_42899734/Validation_42899734_42899734_800833.571.png", "Validation_42899734/Validation_42899734_42899734_800835.571.png", "Validation_42899734/Validation_42899734_42899734_800837.570.png", "Validation_42899734/Validation_42899734_42899734_800839.569.png", "Validation_42899734/Validation_42899734_42899734_800841.568.png" ], "turn_id": 8, "scan_id": "Validation_42899734", "origin_question": "Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the sleek, round white ceramic toilet\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the sleek, round white ceramic toilet\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42897559/Validation_42897559_42897559_277431.618.png", "Validation_42897559/Validation_42897559_42897559_277433.601.png", "Validation_42897559/Validation_42897559_42897559_277435.600.png", "Validation_42897559/Validation_42897559_42897559_277437.599.png" ], "turn_id": 1, "scan_id": "Validation_42897559", "origin_question": "Remember, have you seen any door(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any door(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_42897559/Validation_42897559_42897559_277439.598.png", "Validation_42897559/Validation_42897559_42897559_277441.597.png", "Validation_42897559/Validation_42897559_42897559_277443.596.png", "Validation_42897559/Validation_42897559_42897559_277445.596.png" ], "turn_id": 2, "scan_id": "Validation_42897559", "origin_question": "When did you first discover \"the rectangular oak door with the natural wood color\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the rectangular oak door with the natural wood color\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42897559/Validation_42897559_42897559_277447.595.png", "Validation_42897559/Validation_42897559_42897559_277449.594.png", "Validation_42897559/Validation_42897559_42897559_277451.593.png", "Validation_42897559/Validation_42897559_42897559_277453.592.png" ], "turn_id": 3, "scan_id": "Validation_42897559", "origin_question": "Remember, have you seen any cabinet(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any cabinet(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_42897559/Validation_42897559_42897559_277455.608.png", "Validation_42897559/Validation_42897559_42897559_277457.607.png", "Validation_42897559/Validation_42897559_42897559_277459.606.png", "Validation_42897559/Validation_42897559_42897559_277461.606.png" ], "turn_id": 4, "scan_id": "Validation_42897559", "origin_question": "Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_42897559/Validation_42897559_42897559_277463.605.png", "Validation_42897559/Validation_42897559_42897559_277465.604.png", "Validation_42897559/Validation_42897559_42897559_277467.603.png", "Validation_42897559/Validation_42897559_42897559_277469.602.png" ], "turn_id": 5, "scan_id": "Validation_42897559", "origin_question": "By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the bed\", Y: \"the rectangular window with white vertical blinds\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the bed\", Y: \"the rectangular window with white vertical blinds\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_42897559/Validation_42897559_42897559_277471.602.png", "Validation_42897559/Validation_42897559_42897559_277473.601.png", "Validation_42897559/Validation_42897559_42897559_277475.600.png", "Validation_42897559/Validation_42897559_42897559_277477.599.png" ], "turn_id": 6, "scan_id": "Validation_42897559", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the rectangular window with white vertical blinds\" from your position?", "option": [], "answer": "167.17", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the rectangular window with white vertical blinds\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_48018372/Validation_48018372_48018372_389490.692.png", "Validation_48018372/Validation_48018372_48018372_389492.175.png", "Validation_48018372/Validation_48018372_48018372_389493.674.png", "Validation_48018372/Validation_48018372_48018372_389495.174.png" ], "turn_id": 1, "scan_id": "Validation_48018372", "origin_question": "", "option": [], "answer": "", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_48018372/Validation_48018372_48018372_389496.673.png", "Validation_48018372/Validation_48018372_48018372_389498.173.png", "Validation_48018372/Validation_48018372_48018372_389499.672.png", "Validation_48018372/Validation_48018372_48018372_389501.171.png" ], "turn_id": 2, "scan_id": "Validation_48018372", "origin_question": "In which turn were you closest to \"the classic white rectangular metal refrigerator\"? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the classic white rectangular metal refrigerator\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_48018372/Validation_48018372_48018372_389502.671.png", "Validation_48018372/Validation_48018372_48018372_389504.170.png", "Validation_48018372/Validation_48018372_48018372_389505.670.png", "Validation_48018372/Validation_48018372_48018372_389507.169.png" ], "turn_id": 3, "scan_id": "Validation_48018372", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the sleek, white, front-loading washing machine\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the sleek, white, front-loading washing machine\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_48018372/Validation_48018372_48018372_389508.668.png", "Validation_48018372/Validation_48018372_48018372_389510.168.png", "Validation_48018372/Validation_48018372_48018372_389511.684.png", "Validation_48018372/Validation_48018372_48018372_389513.183.png" ], "turn_id": 4, "scan_id": "Validation_48018372", "origin_question": "When did you first discover \"the round, earthy brown stool with metal legs\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the round, earthy brown stool with metal legs\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_48018372/Validation_48018372_48018372_389514.682.png", "Validation_48018372/Validation_48018372_48018372_389516.182.png", "Validation_48018372/Validation_48018372_48018372_389517.681.png", "Validation_48018372/Validation_48018372_48018372_389519.181.png" ], "turn_id": 5, "scan_id": "Validation_48018372", "origin_question": "Using your orientation at the end of turn 4 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "102.68", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 4 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_48018372/Validation_48018372_48018372_389520.680.png", "Validation_48018372/Validation_48018372_48018372_389522.179.png", "Validation_48018372/Validation_48018372_48018372_389523.679.png", "Validation_48018372/Validation_48018372_48018372_389525.178.png" ], "turn_id": 6, "scan_id": "Validation_48018372", "origin_question": "At the end of which turn were both of \"the sleek, white, front-loading washing machine\" and \"the compact white microwave with the sleek design\" on your left side?", "option": [], "answer": "2", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the sleek, white, front-loading washing machine\" and \"the compact white microwave with the sleek design\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_48018372/Validation_48018372_48018372_389526.677.png", "Validation_48018372/Validation_48018372_48018372_389528.177.png", "Validation_48018372/Validation_48018372_48018372_389529.676.png", "Validation_48018372/Validation_48018372_48018372_389531.176.png" ], "turn_id": 7, "scan_id": "Validation_48018372", "origin_question": "Compared to your position at the end of the 5 turn, are you now closer or farther away from \"the sleek, white, front-loading washing machine\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 5 turn, are you now closer or farther away from \"the sleek, white, front-loading washing machine\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_45663115/Validation_45663115_45663115_55429.878.png", "Validation_45663115/Validation_45663115_45663115_55431.860.png", "Validation_45663115/Validation_45663115_45663115_55433.859.png", "Validation_45663115/Validation_45663115_45663115_55435.859.png" ], "turn_id": 1, "scan_id": "Validation_45663115", "origin_question": "Remember, have you seen any bathtub(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any bathtub(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "Validation_45663115/Validation_45663115_45663115_55437.858.png", "Validation_45663115/Validation_45663115_45663115_55439.857.png", "Validation_45663115/Validation_45663115_45663115_55441.856.png", "Validation_45663115/Validation_45663115_45663115_55443.855.png" ], "turn_id": 2, "scan_id": "Validation_45663115", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the white ceramic pedestal sink with the glossy finish\"; \"the modern elongated toilet with the blue whale pattern\"; \"the sleek white rectangular bathtub\" ", "option": [ "the white ceramic pedestal sink with the glossy finish", "the modern elongated toilet with the blue whale pattern", "the sleek white rectangular bathtub" ], "answer": "the white ceramic pedestal sink with the glossy finish", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the white ceramic pedestal sink with the glossy finish\"; \"the modern elongated toilet with the blue whale pattern\"; \"the sleek white rectangular bathtub\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the white ceramic pedestal sink with the glossy finish', 'the modern elongated toilet with the blue whale pattern', 'the sleek white rectangular bathtub'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_45663115/Validation_45663115_45663115_55445.855.png", "Validation_45663115/Validation_45663115_45663115_55447.870.png", "Validation_45663115/Validation_45663115_45663115_55449.870.png", "Validation_45663115/Validation_45663115_45663115_55451.869.png" ], "turn_id": 3, "scan_id": "Validation_45663115", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_45663115/Validation_45663115_45663115_55453.868.png", "Validation_45663115/Validation_45663115_45663115_55455.867.png", "Validation_45663115/Validation_45663115_45663115_55457.866.png", "Validation_45663115/Validation_45663115_45663115_55459.865.png" ], "turn_id": 4, "scan_id": "Validation_45663115", "origin_question": "Which two objects are on the same side of you now? \"the sleek white rectangular bathtub\", \"the modern elongated toilet with the blue whale pattern\", \"the cylindrical, durable plastic bin\". ", "option": [ "the sleek white rectangular bathtub, the modern elongated toilet with the blue whale pattern", "the sleek white rectangular bathtub, the cylindrical, durable plastic bin", "the modern elongated toilet with the blue whale pattern, the cylindrical, durable plastic bin" ], "answer": "the modern elongated toilet with the blue whale pattern, the cylindrical, durable plastic bin", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the sleek white rectangular bathtub\", \"the modern elongated toilet with the blue whale pattern\", \"the cylindrical, durable plastic bin\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the sleek white rectangular bathtub, the modern elongated toilet with the blue whale pattern', 'the sleek white rectangular bathtub, the cylindrical, durable plastic bin', 'the modern elongated toilet with the blue whale pattern, the cylindrical, durable plastic bin'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_45663115/Validation_45663115_45663115_55461.865.png", "Validation_45663115/Validation_45663115_45663115_55463.864.png", "Validation_45663115/Validation_45663115_45663115_55465.863.png", "Validation_45663115/Validation_45663115_45663115_55467.862.png" ], "turn_id": 5, "scan_id": "Validation_45663115", "origin_question": "Which two objects are on the same side of you now? \"the sleek white rectangular bathtub\", \"the white ceramic pedestal sink with the glossy finish\", \"the rectangular window with frosted glass panels\". ", "option": [ "the sleek white rectangular bathtub, the white ceramic pedestal sink with the glossy finish", "the sleek white rectangular bathtub, the rectangular window with frosted glass panels", "the white ceramic pedestal sink with the glossy finish, the rectangular window with frosted glass panels" ], "answer": "the white ceramic pedestal sink with the glossy finish, the rectangular window with frosted glass panels", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the sleek white rectangular bathtub\", \"the white ceramic pedestal sink with the glossy finish\", \"the rectangular window with frosted glass panels\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the sleek white rectangular bathtub, the white ceramic pedestal sink with the glossy finish', 'the sleek white rectangular bathtub, the rectangular window with frosted glass panels', 'the white ceramic pedestal sink with the glossy finish, the rectangular window with frosted glass panels'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_45663115/Validation_45663115_45663115_55469.861.png", "Validation_45663115/Validation_45663115_45663115_55471.860.png", "Validation_45663115/Validation_45663115_45663115_55473.860.png", "Validation_45663115/Validation_45663115_45663115_55475.859.png" ], "turn_id": 6, "scan_id": "Validation_45663115", "origin_question": "Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance forward or backward from that position?", "option": [ "Forward", "Backward" ], "answer": "Forward", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance forward or backward from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Forward', 'Backward'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47204559/Validation_47204559_47204559_3988.554.png", "Validation_47204559/Validation_47204559_47204559_3990.037.png", "Validation_47204559/Validation_47204559_47204559_3991.536.png", "Validation_47204559/Validation_47204559_47204559_3993.052.png" ], "turn_id": 1, "scan_id": "Validation_47204559", "origin_question": "Remember, have you seen any window(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any window(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47204559/Validation_47204559_47204559_3994.552.png", "Validation_47204559/Validation_47204559_47204559_3996.051.png", "Validation_47204559/Validation_47204559_47204559_3997.550.png", "Validation_47204559/Validation_47204559_47204559_3999.050.png" ], "turn_id": 2, "scan_id": "Validation_47204559", "origin_question": "In which turn were you closest to \"the matte black rectangular box\"? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the matte black rectangular box\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_47204559/Validation_47204559_47204559_4000.549.png", "Validation_47204559/Validation_47204559_47204559_4002.049.png", "Validation_47204559/Validation_47204559_47204559_4003.548.png", "Validation_47204559/Validation_47204559_47204559_4005.047.png" ], "turn_id": 3, "scan_id": "Validation_47204559", "origin_question": "Is the matte black rectangular box to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the matte black rectangular box to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47204559/Validation_47204559_47204559_4006.547.png", "Validation_47204559/Validation_47204559_47204559_4008.046.png", "Validation_47204559/Validation_47204559_47204559_4009.546.png", "Validation_47204559/Validation_47204559_47204559_4011.045.png" ], "turn_id": 4, "scan_id": "Validation_47204559", "origin_question": "Which direction is the large rectangular window with the white frame in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the large rectangular window with the white frame in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47204559/Validation_47204559_47204559_4012.544.png", "Validation_47204559/Validation_47204559_47204559_4014.044.png", "Validation_47204559/Validation_47204559_47204559_4015.543.png", "Validation_47204559/Validation_47204559_47204559_4017.042.png" ], "turn_id": 5, "scan_id": "Validation_47204559", "origin_question": "Which direction is the small wooden picture frame with the natural finish in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the small wooden picture frame with the natural finish in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47204559/Validation_47204559_47204559_4018.542.png", "Validation_47204559/Validation_47204559_47204559_4020.041.png", "Validation_47204559/Validation_47204559_47204559_4021.541.png", "Validation_47204559/Validation_47204559_47204559_4023.040.png" ], "turn_id": 6, "scan_id": "Validation_47204559", "origin_question": "In which turn were you closest to \"the large rectangular window with the white frame\"? ", "option": [], "answer": "6", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the large rectangular window with the white frame\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_41254435/Validation_41254435_41254435_96208.033.png", "Validation_41254435/Validation_41254435_41254435_96210.015.png", "Validation_41254435/Validation_41254435_41254435_96212.014.png", "Validation_41254435/Validation_41254435_41254435_96214.014.png", "Validation_41254435/Validation_41254435_41254435_96216.013.png" ], "turn_id": 1, "scan_id": "Validation_41254435", "origin_question": "Remember, have you seen any toothpaste(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any toothpaste(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_41254435/Validation_41254435_41254435_96218.012.png", "Validation_41254435/Validation_41254435_41254435_96220.011.png", "Validation_41254435/Validation_41254435_41254435_96222.010.png", "Validation_41254435/Validation_41254435_41254435_96224.010.png", "Validation_41254435/Validation_41254435_41254435_96226.009.png" ], "turn_id": 2, "scan_id": "Validation_41254435", "origin_question": "In which turn did you see both \"the slender, light-colored door with the smooth finish\" and \"the medium-sized window with the sleek white frame\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the slender, light-colored door with the smooth finish\" and \"the medium-sized window with the sleek white frame\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "Validation_41254435/Validation_41254435_41254435_96228.008.png", "Validation_41254435/Validation_41254435_41254435_96230.007.png", "Validation_41254435/Validation_41254435_41254435_96232.006.png", "Validation_41254435/Validation_41254435_41254435_96234.006.png", "Validation_41254435/Validation_41254435_41254435_96236.005.png" ], "turn_id": 3, "scan_id": "Validation_41254435", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the sleek, metallic bin with the neutral silver finish\"; \"the slender, light-colored door with the smooth finish\"; \"the sleek black monitor with the minimalistic design\" ", "option": [ "the sleek, metallic bin with the neutral silver finish", "the slender, light-colored door with the smooth finish", "the sleek black monitor with the minimalistic design" ], "answer": "the sleek, metallic bin with the neutral silver finish", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the sleek, metallic bin with the neutral silver finish\"; \"the slender, light-colored door with the smooth finish\"; \"the sleek black monitor with the minimalistic design\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the sleek, metallic bin with the neutral silver finish', 'the slender, light-colored door with the smooth finish', 'the sleek black monitor with the minimalistic design'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_41254435/Validation_41254435_41254435_96238.004.png", "Validation_41254435/Validation_41254435_41254435_96240.003.png", "Validation_41254435/Validation_41254435_41254435_96242.002.png", "Validation_41254435/Validation_41254435_41254435_96244.002.png", "Validation_41254435/Validation_41254435_41254435_96246.001.png" ], "turn_id": 4, "scan_id": "Validation_41254435", "origin_question": "In which turn did you see both \"the sleek stainless steel microwave with the glass window\" and \"the sleek stainless steel stove with the metallic silver color\" simultaneously? ", "option": [], "answer": "4", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the sleek stainless steel microwave with the glass window\" and \"the sleek stainless steel stove with the metallic silver color\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_41254435/Validation_41254435_41254435_96248.017.png", "Validation_41254435/Validation_41254435_41254435_96250.016.png", "Validation_41254435/Validation_41254435_41254435_96252.015.png", "Validation_41254435/Validation_41254435_41254435_96254.014.png", "Validation_41254435/Validation_41254435_41254435_96256.014.png" ], "turn_id": 5, "scan_id": "Validation_41254435", "origin_question": "Which two objects are on the same side of you now? \"the sleek, metallic bin with the neutral silver finish\", \"the compact stainless steel rectangular oven\", \"the sleek stainless steel microwave with the glass window\". ", "option": [ "the sleek, metallic bin with the neutral silver finish, the compact stainless steel rectangular oven", "the sleek, metallic bin with the neutral silver finish, the sleek stainless steel microwave with the glass window", "the compact stainless steel rectangular oven, the sleek stainless steel microwave with the glass window" ], "answer": "the sleek, metallic bin with the neutral silver finish, the compact stainless steel rectangular oven", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the sleek, metallic bin with the neutral silver finish\", \"the compact stainless steel rectangular oven\", \"the sleek stainless steel microwave with the glass window\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the sleek, metallic bin with the neutral silver finish, the compact stainless steel rectangular oven', 'the sleek, metallic bin with the neutral silver finish, the sleek stainless steel microwave with the glass window', 'the compact stainless steel rectangular oven, the sleek stainless steel microwave with the glass window'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "Validation_41254435/Validation_41254435_41254435_96258.013.png", "Validation_41254435/Validation_41254435_41254435_96260.012.png", "Validation_41254435/Validation_41254435_41254435_96262.011.png", "Validation_41254435/Validation_41254435_41254435_96264.010.png", "Validation_41254435/Validation_41254435_41254435_96266.010.png" ], "turn_id": 6, "scan_id": "Validation_41254435", "origin_question": "Among these three objects(\"the slender, light-colored door with the smooth finish\",\"the sleek stainless steel microwave with the glass window\",\"the sleek stainless steel stove with the metallic silver color\" ), which one is the closest to you?", "option": [ "the slender, light-colored door with the smooth finish", "the sleek stainless steel microwave with the glass window", "the sleek stainless steel stove with the metallic silver color" ], "answer": "the sleek stainless steel stove with the metallic silver color", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the slender, light-colored door with the smooth finish\",\"the sleek stainless steel microwave with the glass window\",\"the sleek stainless steel stove with the metallic silver color\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the slender, light-colored door with the smooth finish', 'the sleek stainless steel microwave with the glass window', 'the sleek stainless steel stove with the metallic silver color'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_41254435/Validation_41254435_41254435_96268.009.png", "Validation_41254435/Validation_41254435_41254435_96270.008.png", "Validation_41254435/Validation_41254435_41254435_96272.007.png", "Validation_41254435/Validation_41254435_41254435_96274.006.png", "Validation_41254435/Validation_41254435_41254435_96276.006.png" ], "turn_id": 7, "scan_id": "Validation_41254435", "origin_question": "In which turn did you see both \"the slender, light-colored door with the smooth finish\" and \"the sleek, metallic bin with the neutral silver finish\" simultaneously? ", "option": [], "answer": "3", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the slender, light-colored door with the smooth finish\" and \"the sleek, metallic bin with the neutral silver finish\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_41254435/Validation_41254435_41254435_96278.005.png", "Validation_41254435/Validation_41254435_41254435_96280.004.png", "Validation_41254435/Validation_41254435_41254435_96282.003.png", "Validation_41254435/Validation_41254435_41254435_96284.002.png", "Validation_41254435/Validation_41254435_41254435_96286.002.png" ], "turn_id": 8, "scan_id": "Validation_41254435", "origin_question": "Please recall \"the compact stainless steel rectangular oven\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "3.79", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the compact stainless steel rectangular oven\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_41254435/Validation_41254435_41254435_96288.001.png", "Validation_41254435/Validation_41254435_41254435_96290.017.png", "Validation_41254435/Validation_41254435_41254435_96292.016.png", "Validation_41254435/Validation_41254435_41254435_96294.015.png", "Validation_41254435/Validation_41254435_41254435_96296.014.png" ], "turn_id": 9, "scan_id": "Validation_41254435", "origin_question": "Remember, have you seen any cabinet(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any cabinet(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0143_02/00000.jpg", "scene0143_02/00040.jpg", "scene0143_02/00080.jpg", "scene0143_02/00290.jpg" ], "turn_id": 1, "scan_id": "scene0143_02", "origin_question": "Remember, how many pitcher(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many pitcher(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0143_02/00330.jpg", "scene0143_02/00370.jpg", "scene0143_02/00410.jpg", "scene0143_02/00450.jpg" ], "turn_id": 2, "scan_id": "scene0143_02", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the black rectangular corded telephone made of plastic\"; \"the case\"; \"the light-colored fabric curtain hanging vertically\" ", "option": [ "the black rectangular corded telephone made of plastic", "the case", "the light-colored fabric curtain hanging vertically" ], "answer": "the case", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the black rectangular corded telephone made of plastic\"; \"the case\"; \"the light-colored fabric curtain hanging vertically\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the black rectangular corded telephone made of plastic', 'the case', 'the light-colored fabric curtain hanging vertically'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0143_02/00490.jpg", "scene0143_02/00540.jpg", "scene0143_02/00580.jpg", "scene0143_02/00620.jpg" ], "turn_id": 3, "scan_id": "scene0143_02", "origin_question": "", "option": [], "answer": "", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0143_02/00660.jpg", "scene0143_02/00700.jpg", "scene0143_02/00740.jpg", "scene0143_02/00780.jpg" ], "turn_id": 4, "scan_id": "scene0143_02", "origin_question": "Is the table to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the table to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0143_02/00820.jpg", "scene0143_02/01070.jpg", "scene0143_02/01110.jpg", "scene0143_02/01150.jpg" ], "turn_id": 5, "scan_id": "scene0143_02", "origin_question": "What will be the first-time appearance order of the following categories: bin, door, stand?", "option": [ "bin, door, stand", "stand, bin, door", "door, stand, bin", "stand, door, bin" ], "answer": "stand, door, bin", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bin, door, stand? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['bin, door, stand', 'stand, bin, door', 'door, stand, bin', 'stand, door, bin'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0143_02/01510.jpg", "scene0143_02/01560.jpg", "scene0143_02/01600.jpg", "scene0143_02/01640.jpg" ], "turn_id": 6, "scan_id": "scene0143_02", "origin_question": "Which two objects are on the same side of you now? \"then elegant rectangular bed with a deep brown headboard\", \"the mirror\", \"the new rectangular wooden desk in deep brown\". ", "option": [ "then elegant rectangular bed with a deep brown headboard, the mirror", "then elegant rectangular bed with a deep brown headboard, the new rectangular wooden desk in deep brown", "the mirror, the new rectangular wooden desk in deep brown" ], "answer": "then elegant rectangular bed with a deep brown headboard, the mirror", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"then elegant rectangular bed with a deep brown headboard\", \"the mirror\", \"the new rectangular wooden desk in deep brown\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['then elegant rectangular bed with a deep brown headboard, the mirror', 'then elegant rectangular bed with a deep brown headboard, the new rectangular wooden desk in deep brown', 'the mirror, the new rectangular wooden desk in deep brown'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0143_02/01680.jpg", "scene0143_02/01720.jpg", "scene0143_02/01820.jpg", "scene0143_02/02180.jpg" ], "turn_id": 7, "scan_id": "scene0143_02", "origin_question": "", "option": [], "answer": "", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0143_02/02220.jpg", "scene0143_02/02260.jpg", "scene0143_02/02300.jpg", "scene0143_02/02410.jpg" ], "turn_id": 8, "scan_id": "scene0143_02", "origin_question": "", "option": [], "answer": "", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0280_00/00000.jpg", "scene0280_00/00040.jpg", "scene0280_00/00080.jpg", "scene0280_00/00120.jpg" ], "turn_id": 1, "scan_id": "scene0280_00", "origin_question": "Remember, have you seen any basket(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any basket(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0280_00/00160.jpg", "scene0280_00/00200.jpg", "scene0280_00/00240.jpg", "scene0280_00/00280.jpg" ], "turn_id": 2, "scan_id": "scene0280_00", "origin_question": "Please recall \"the wrinkled beige shirt and a green piece of clothes\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.8", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the wrinkled beige shirt and a green piece of clothes\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0280_00/00320.jpg", "scene0280_00/00360.jpg", "scene0280_00/00400.jpg", "scene0280_00/00440.jpg" ], "turn_id": 3, "scan_id": "scene0280_00", "origin_question": "Which two objects are on the same side of you now? \"the wrinkled beige shirt and a green piece of clothes\", \"the black rectangular case with sturdy material\", \"the dresser\". ", "option": [ "the wrinkled beige shirt and a green piece of clothes, the black rectangular case with sturdy material", "the wrinkled beige shirt and a green piece of clothes, the dresser", "the black rectangular case with sturdy material, the dresser" ], "answer": "the black rectangular case with sturdy material, the dresser", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the wrinkled beige shirt and a green piece of clothes\", \"the black rectangular case with sturdy material\", \"the dresser\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the wrinkled beige shirt and a green piece of clothes, the black rectangular case with sturdy material', 'the wrinkled beige shirt and a green piece of clothes, the dresser', 'the black rectangular case with sturdy material, the dresser'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0280_00/00480.jpg", "scene0280_00/00520.jpg", "scene0280_00/00560.jpg", "scene0280_00/00600.jpg" ], "turn_id": 4, "scan_id": "scene0280_00", "origin_question": "Is the black rectangular case with sturdy material to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the black rectangular case with sturdy material to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0280_00/00640.jpg", "scene0280_00/00680.jpg", "scene0280_00/00720.jpg", "scene0280_00/00760.jpg" ], "turn_id": 5, "scan_id": "scene0280_00", "origin_question": "Please recall \"the medium-sized blue cotton towel\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.1", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the medium-sized blue cotton towel\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0280_00/00800.jpg", "scene0280_00/00840.jpg", "scene0280_00/00880.jpg", "scene0280_00/00920.jpg" ], "turn_id": 6, "scan_id": "scene0280_00", "origin_question": "Please recall \"the black rectangular case with sturdy material\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.3", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the black rectangular case with sturdy material\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0280_00/00960.jpg", "scene0280_00/01000.jpg", "scene0280_00/01040.jpg", "scene0280_00/01080.jpg" ], "turn_id": 7, "scan_id": "scene0280_00", "origin_question": "Which direction is the door in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the door in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0280_00/01120.jpg", "scene0280_00/01160.jpg", "scene0280_00/01200.jpg", "scene0280_00/01240.jpg" ], "turn_id": 8, "scan_id": "scene0280_00", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the medium-sized blue cotton towel\" from your position?", "option": [], "answer": "110.22", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the medium-sized blue cotton towel\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0280_00/01280.jpg", "scene0280_00/01320.jpg", "scene0280_00/01360.jpg", "scene0280_00/01400.jpg" ], "turn_id": 9, "scan_id": "scene0280_00", "origin_question": "Which two objects are on the same side of you now? \"the green round ball with a smooth texture\", \"the door\", \"the medium-sized blue cotton towel\". ", "option": [ "the green round ball with a smooth texture, the door", "the green round ball with a smooth texture, the medium-sized blue cotton towel", "the door, the medium-sized blue cotton towel" ], "answer": "the door, the medium-sized blue cotton towel", "user_message": "For the 9 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the green round ball with a smooth texture\", \"the door\", \"the medium-sized blue cotton towel\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the green round ball with a smooth texture, the door', 'the green round ball with a smooth texture, the medium-sized blue cotton towel', 'the door, the medium-sized blue cotton towel'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0231_00/00000.jpg", "scene0231_00/00040.jpg", "scene0231_00/00080.jpg", "scene0231_00/00120.jpg", "scene0231_00/00160.jpg" ], "turn_id": 1, "scan_id": "scene0231_00", "origin_question": "Remember, have you seen any chair(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any chair(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0231_00/00200.jpg", "scene0231_00/00240.jpg", "scene0231_00/00280.jpg", "scene0231_00/00320.jpg", "scene0231_00/00360.jpg" ], "turn_id": 2, "scan_id": "scene0231_00", "origin_question": "Which direction is the backpack in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the backpack in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "scene0231_00/00400.jpg", "scene0231_00/00440.jpg", "scene0231_00/00480.jpg", "scene0231_00/00520.jpg", "scene0231_00/00560.jpg" ], "turn_id": 3, "scan_id": "scene0231_00", "origin_question": "Among these three objects(\"the jacket\",\"the large rectangular polished granite counter in beige\",\"the round white plastic pan\" ), which one is the closest to you?", "option": [ "the jacket", "the large rectangular polished granite counter in beige", "the round white plastic pan" ], "answer": "the round white plastic pan", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the jacket\",\"the large rectangular polished granite counter in beige\",\"the round white plastic pan\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the jacket', 'the large rectangular polished granite counter in beige', 'the round white plastic pan'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0231_00/00600.jpg", "scene0231_00/00640.jpg", "scene0231_00/00680.jpg", "scene0231_00/00720.jpg", "scene0231_00/00760.jpg" ], "turn_id": 4, "scan_id": "scene0231_00", "origin_question": "In which turn did you see both \"the jacket\" and \"the simple white cylindrical cup\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the jacket\" and \"the simple white cylindrical cup\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0231_00/00800.jpg", "scene0231_00/00840.jpg", "scene0231_00/00880.jpg", "scene0231_00/00920.jpg", "scene0231_00/00960.jpg" ], "turn_id": 5, "scan_id": "scene0231_00", "origin_question": "At the end of which turn were both of \"the clock\" and \"the round white plastic pan\" on your left side?", "option": [], "answer": "4", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the clock\" and \"the round white plastic pan\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0231_00/01000.jpg", "scene0231_00/01040.jpg", "scene0231_00/01080.jpg", "scene0231_00/01120.jpg", "scene0231_00/01160.jpg" ], "turn_id": 6, "scan_id": "scene0231_00", "origin_question": "Using your orientation at the end of turn 5 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 5 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0231_00/01200.jpg", "scene0231_00/01240.jpg", "scene0231_00/01280.jpg", "scene0231_00/01320.jpg", "scene0231_00/01360.jpg" ], "turn_id": 7, "scan_id": "scene0231_00", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the small cylindrical blue plastic bucket\"; \"the small rectangular blue tissue box made of smooth material\"; \"the round white plastic pan\" ", "option": [ "the small cylindrical blue plastic bucket", "the small rectangular blue tissue box made of smooth material", "the round white plastic pan" ], "answer": "the small cylindrical blue plastic bucket", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the small cylindrical blue plastic bucket\"; \"the small rectangular blue tissue box made of smooth material\"; \"the round white plastic pan\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the small cylindrical blue plastic bucket', 'the small rectangular blue tissue box made of smooth material', 'the round white plastic pan'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0231_00/01400.jpg", "scene0231_00/01440.jpg", "scene0231_00/01480.jpg", "scene0231_00/01520.jpg", "scene0231_00/01560.jpg" ], "turn_id": 8, "scan_id": "scene0231_00", "origin_question": "In which turn did you see both \"the cylindrical, stainless steel range hood with a smooth surface\" and \"the couch\" simultaneously? ", "option": [], "answer": "5", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the cylindrical, stainless steel range hood with a smooth surface\" and \"the couch\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0231_00/01600.jpg", "scene0231_00/01640.jpg", "scene0231_00/01680.jpg", "scene0231_00/01720.jpg", "scene0231_00/01760.jpg" ], "turn_id": 9, "scan_id": "scene0231_00", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0231_00/01800.jpg", "scene0231_00/01840.jpg", "scene0231_00/01880.jpg", "scene0231_00/01920.jpg", "scene0231_00/01960.jpg" ], "turn_id": 10, "scan_id": "scene0231_00", "origin_question": "Which direction is the potted plant with white flowers and green leaves in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the potted plant with white flowers and green leaves in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0591_02/00000.jpg", "scene0591_02/00040.jpg", "scene0591_02/00080.jpg", "scene0591_02/00120.jpg", "scene0591_02/00160.jpg" ], "turn_id": 1, "scan_id": "scene0591_02", "origin_question": "Remember, have you seen any shelf(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any shelf(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0591_02/00200.jpg", "scene0591_02/00240.jpg", "scene0591_02/00280.jpg", "scene0591_02/00320.jpg", "scene0591_02/00360.jpg" ], "turn_id": 2, "scan_id": "scene0591_02", "origin_question": "At the end of which turn were both of \"the vibrant yellow square pillow with a fluffy texture\" and \"the medium cylindrical plastic bottle with a blue cap\" on your left side?", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the vibrant yellow square pillow with a fluffy texture\" and \"the medium cylindrical plastic bottle with a blue cap\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0591_02/00400.jpg", "scene0591_02/00440.jpg", "scene0591_02/00480.jpg", "scene0591_02/00520.jpg", "scene0591_02/00560.jpg" ], "turn_id": 3, "scan_id": "scene0591_02", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the medium-sized rectangular couch with blue or black fabric\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the medium-sized rectangular couch with blue or black fabric\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0591_02/00600.jpg", "scene0591_02/00640.jpg", "scene0591_02/00680.jpg", "scene0591_02/00720.jpg", "scene0591_02/00760.jpg" ], "turn_id": 4, "scan_id": "scene0591_02", "origin_question": "Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "35.48", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "scene0591_02/00800.jpg", "scene0591_02/00840.jpg", "scene0591_02/00880.jpg", "scene0591_02/00920.jpg", "scene0591_02/00960.jpg" ], "turn_id": 5, "scan_id": "scene0591_02", "origin_question": "Among these three objects(\"the blackboard\",\"the medium cylindrical plastic bottle with a blue cap\",\"the medium-sized round white plastic bin\" ), which one is the farthest from you?", "option": [ "the blackboard", "the medium cylindrical plastic bottle with a blue cap", "the medium-sized round white plastic bin" ], "answer": "the medium cylindrical plastic bottle with a blue cap", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the blackboard\",\"the medium cylindrical plastic bottle with a blue cap\",\"the medium-sized round white plastic bin\" ), which one is the farthest from you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the blackboard', 'the medium cylindrical plastic bottle with a blue cap', 'the medium-sized round white plastic bin'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0591_02/01000.jpg", "scene0591_02/01040.jpg", "scene0591_02/01080.jpg", "scene0591_02/01120.jpg", "scene0591_02/01160.jpg" ], "turn_id": 6, "scan_id": "scene0591_02", "origin_question": "Which direction is the medium-sized rectangular couch with blue or black fabric in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the medium-sized rectangular couch with blue or black fabric in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "scene0591_02/01200.jpg", "scene0591_02/01240.jpg", "scene0591_02/01280.jpg", "scene0591_02/01320.jpg", "scene0591_02/01360.jpg" ], "turn_id": 7, "scan_id": "scene0591_02", "origin_question": "Among these three objects(\"the standard rectangular brown wooden door\",\"the medium-sized round white plastic bin\",\"the medium black and grey nylon backpack\" ), which one is the closest to you?", "option": [ "the standard rectangular brown wooden door", "the medium-sized round white plastic bin", "the medium black and grey nylon backpack" ], "answer": "the medium black and grey nylon backpack", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the standard rectangular brown wooden door\",\"the medium-sized round white plastic bin\",\"the medium black and grey nylon backpack\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the standard rectangular brown wooden door', 'the medium-sized round white plastic bin', 'the medium black and grey nylon backpack'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0591_02/01400.jpg", "scene0591_02/01440.jpg", "scene0591_02/01480.jpg", "scene0591_02/01520.jpg", "scene0591_02/01560.jpg" ], "turn_id": 8, "scan_id": "scene0591_02", "origin_question": "Please recall \"the medium black and grey nylon backpack\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.34", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the medium black and grey nylon backpack\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0591_02/01600.jpg", "scene0591_02/01640.jpg", "scene0591_02/01680.jpg", "scene0591_02/01720.jpg", "scene0591_02/01760.jpg" ], "turn_id": 9, "scan_id": "scene0591_02", "origin_question": "Compared to your position at the end of the 7 turn, are you now closer or farther away from \"the blackboard\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 7 turn, are you now closer or farther away from \"the blackboard\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0591_02/01800.jpg", "scene0591_02/01840.jpg", "scene0591_02/01880.jpg", "scene0591_02/01920.jpg", "scene0591_02/01960.jpg" ], "turn_id": 10, "scan_id": "scene0591_02", "origin_question": "Which direction is the standard rectangular brown wooden door in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the standard rectangular brown wooden door in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0691_01/00000.jpg", "scene0691_01/00040.jpg", "scene0691_01/00100.jpg", "scene0691_01/00140.jpg" ], "turn_id": 1, "scan_id": "scene0691_01", "origin_question": "Remember, have you seen any basket(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any basket(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0691_01/00180.jpg", "scene0691_01/00220.jpg", "scene0691_01/00260.jpg", "scene0691_01/00300.jpg" ], "turn_id": 2, "scan_id": "scene0691_01", "origin_question": "Is the modern black and silver stainless steel microwave to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the modern black and silver stainless steel microwave to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0691_01/00340.jpg", "scene0691_01/00380.jpg", "scene0691_01/00420.jpg", "scene0691_01/00460.jpg" ], "turn_id": 3, "scan_id": "scene0691_01", "origin_question": "What will be the first-time appearance order of the following categories: chair, door, fire extinguisher?", "option": [ "door, chair, fire extinguisher", "chair, door, fire extinguisher", "fire extinguisher, door, chair", "chair, fire extinguisher, door" ], "answer": "chair, door, fire extinguisher", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: chair, door, fire extinguisher? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['door, chair, fire extinguisher', 'chair, door, fire extinguisher', 'fire extinguisher, door, chair', 'chair, fire extinguisher, door'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0691_01/00500.jpg", "scene0691_01/00540.jpg", "scene0691_01/00580.jpg", "scene0691_01/00620.jpg" ], "turn_id": 4, "scan_id": "scene0691_01", "origin_question": "Is the small rectangular dark-colored basket to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the small rectangular dark-colored basket to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0691_01/00660.jpg", "scene0691_01/00700.jpg", "scene0691_01/00740.jpg", "scene0691_01/00780.jpg" ], "turn_id": 5, "scan_id": "scene0691_01", "origin_question": "When did you first discover \"the modern black and silver stainless steel microwave\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the modern black and silver stainless steel microwave\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0691_01/00820.jpg", "scene0691_01/00860.jpg", "scene0691_01/00900.jpg", "scene0691_01/00940.jpg" ], "turn_id": 6, "scan_id": "scene0691_01", "origin_question": "At the end of which turn were both of \"the modern black and silver stainless steel microwave\" and \"the small rectangular dark-colored basket\" on your right side?", "option": [], "answer": "4", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the modern black and silver stainless steel microwave\" and \"the small rectangular dark-colored basket\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0629_02/00000.jpg", "scene0629_02/00040.jpg", "scene0629_02/00080.jpg", "scene0629_02/00120.jpg", "scene0629_02/00160.jpg" ], "turn_id": 1, "scan_id": "scene0629_02", "origin_question": "Remember, have you seen any plant(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any plant(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0629_02/00200.jpg", "scene0629_02/00240.jpg", "scene0629_02/00280.jpg", "scene0629_02/00320.jpg", "scene0629_02/00360.jpg" ], "turn_id": 2, "scan_id": "scene0629_02", "origin_question": "Which direction is the dark brown rectangular wooden desk in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the dark brown rectangular wooden desk in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0629_02/00400.jpg", "scene0629_02/00440.jpg", "scene0629_02/00480.jpg", "scene0629_02/00520.jpg", "scene0629_02/00560.jpg" ], "turn_id": 3, "scan_id": "scene0629_02", "origin_question": "At the end of which turn were both of \"the table\" and \"the small circular white plastic light fixture\" on your left side?", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the table\" and \"the small circular white plastic light fixture\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0629_02/00600.jpg", "scene0629_02/00640.jpg", "scene0629_02/00680.jpg", "scene0629_02/00720.jpg", "scene0629_02/00760.jpg" ], "turn_id": 4, "scan_id": "scene0629_02", "origin_question": "Which two objects are on the same side of you now? \"the dark brown rectangular wooden desk\", \"the plant\", \"the cylindrical gray metal dumbbell\". ", "option": [ "the dark brown rectangular wooden desk, the plant", "the dark brown rectangular wooden desk, the cylindrical gray metal dumbbell", "the plant, the cylindrical gray metal dumbbell" ], "answer": "the dark brown rectangular wooden desk, the plant", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the dark brown rectangular wooden desk\", \"the plant\", \"the cylindrical gray metal dumbbell\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the dark brown rectangular wooden desk, the plant', 'the dark brown rectangular wooden desk, the cylindrical gray metal dumbbell', 'the plant, the cylindrical gray metal dumbbell'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0629_02/00800.jpg", "scene0629_02/00840.jpg", "scene0629_02/00880.jpg", "scene0629_02/00920.jpg", "scene0629_02/00960.jpg" ], "turn_id": 5, "scan_id": "scene0629_02", "origin_question": "By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular white wooden door with six panels\", Y: \"the small circular white plastic light fixture\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular white wooden door with six panels\", Y: \"the small circular white plastic light fixture\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0629_02/01000.jpg", "scene0629_02/01040.jpg", "scene0629_02/01080.jpg", "scene0629_02/01120.jpg", "scene0629_02/01160.jpg" ], "turn_id": 6, "scan_id": "scene0629_02", "origin_question": "Which two objects are on the same side of you now? \"the table\", \"the cylindrical gray metal dumbbell\", \"the rectangular board with flat tops and bottoms\". ", "option": [ "the table, the cylindrical gray metal dumbbell", "the table, the rectangular board with flat tops and bottoms", "the cylindrical gray metal dumbbell, the rectangular board with flat tops and bottoms" ], "answer": "the cylindrical gray metal dumbbell, the rectangular board with flat tops and bottoms", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the table\", \"the cylindrical gray metal dumbbell\", \"the rectangular board with flat tops and bottoms\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the table, the cylindrical gray metal dumbbell', 'the table, the rectangular board with flat tops and bottoms', 'the cylindrical gray metal dumbbell, the rectangular board with flat tops and bottoms'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0629_02/01200.jpg", "scene0629_02/01240.jpg", "scene0629_02/01280.jpg", "scene0629_02/01320.jpg", "scene0629_02/01360.jpg" ], "turn_id": 7, "scan_id": "scene0629_02", "origin_question": "Please recall \"the cylindrical gray metal dumbbell\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.12", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the cylindrical gray metal dumbbell\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0629_02/01400.jpg", "scene0629_02/01440.jpg", "scene0629_02/01480.jpg", "scene0629_02/01520.jpg", "scene0629_02/01560.jpg" ], "turn_id": 8, "scan_id": "scene0629_02", "origin_question": "Using your orientation at the end of turn 5 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "150.2", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 5 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0573_01/00000.jpg", "scene0573_01/00030.jpg", "scene0573_01/00060.jpg", "scene0573_01/00090.jpg" ], "turn_id": 1, "scan_id": "scene0573_01", "origin_question": "Remember, have you seen any sink(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any sink(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0573_01/00120.jpg", "scene0573_01/00150.jpg", "scene0573_01/00180.jpg", "scene0573_01/00210.jpg" ], "turn_id": 2, "scan_id": "scene0573_01", "origin_question": "At the end of which turn were both of \"the counter\" and \"the smooth, curved metallic faucet\" on your left side?", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the counter\" and \"the smooth, curved metallic faucet\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0573_01/00240.jpg", "scene0573_01/00270.jpg", "scene0573_01/00300.jpg", "scene0573_01/00330.jpg" ], "turn_id": 3, "scan_id": "scene0573_01", "origin_question": "Please recall \"the smooth, curved metallic faucet\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.11", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the smooth, curved metallic faucet\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0573_01/00360.jpg", "scene0573_01/00390.jpg", "scene0573_01/00420.jpg", "scene0573_01/00450.jpg" ], "turn_id": 4, "scan_id": "scene0573_01", "origin_question": "Remember, have you seen any refrigerator(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any refrigerator(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0573_01/00480.jpg", "scene0573_01/00510.jpg", "scene0573_01/00540.jpg", "scene0573_01/00570.jpg" ], "turn_id": 5, "scan_id": "scene0573_01", "origin_question": "In which turn were you closest to \"the white rectangular metal stove with four coil burners\"? ", "option": [], "answer": "4", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the white rectangular metal stove with four coil burners\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0573_01/00600.jpg", "scene0573_01/00630.jpg", "scene0573_01/00660.jpg", "scene0573_01/00690.jpg" ], "turn_id": 6, "scan_id": "scene0573_01", "origin_question": "Which two objects are on the same side of you now? \"the smooth, curved metallic faucet\", \"the white rectangular metal stove with four coil burners\", \"the white double-door refrigerator with stainless steel material\". ", "option": [ "the smooth, curved metallic faucet, the white rectangular metal stove with four coil burners", "the smooth, curved metallic faucet, the white double-door refrigerator with stainless steel material", "the white rectangular metal stove with four coil burners, the white double-door refrigerator with stainless steel material" ], "answer": "the white rectangular metal stove with four coil burners, the white double-door refrigerator with stainless steel material", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the smooth, curved metallic faucet\", \"the white rectangular metal stove with four coil burners\", \"the white double-door refrigerator with stainless steel material\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the smooth, curved metallic faucet, the white rectangular metal stove with four coil burners', 'the smooth, curved metallic faucet, the white double-door refrigerator with stainless steel material', 'the white rectangular metal stove with four coil burners, the white double-door refrigerator with stainless steel material'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0568_00/00000.jpg", "scene0568_00/00040.jpg", "scene0568_00/00080.jpg", "scene0568_00/00120.jpg", "scene0568_00/00160.jpg" ], "turn_id": 1, "scan_id": "scene0568_00", "origin_question": "Remember, have you seen any couch(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any couch(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0568_00/00200.jpg", "scene0568_00/00240.jpg", "scene0568_00/00280.jpg", "scene0568_00/00320.jpg", "scene0568_00/00360.jpg" ], "turn_id": 2, "scan_id": "scene0568_00", "origin_question": "In which turn were you closest to \"the tall black wooden shelf with multiple layers\"? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the tall black wooden shelf with multiple layers\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0568_00/00400.jpg", "scene0568_00/00440.jpg", "scene0568_00/00480.jpg", "scene0568_00/00520.jpg", "scene0568_00/00560.jpg" ], "turn_id": 3, "scan_id": "scene0568_00", "origin_question": "In which turn were you farthest from \"the large red couch made of soft, plush material\"? ", "option": [], "answer": "2", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the large red couch made of soft, plush material\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0568_00/00600.jpg", "scene0568_00/00640.jpg", "scene0568_00/00680.jpg", "scene0568_00/00720.jpg", "scene0568_00/00760.jpg" ], "turn_id": 4, "scan_id": "scene0568_00", "origin_question": "Remember, how many shelf(s) have you seen so far? ", "option": [], "answer": "4", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many shelf(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0568_00/00800.jpg", "scene0568_00/00840.jpg", "scene0568_00/00880.jpg", "scene0568_00/00920.jpg", "scene0568_00/00960.jpg" ], "turn_id": 5, "scan_id": "scene0568_00", "origin_question": "Please recall \"the rectangular wooden picture with a yellow frame\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.86", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the rectangular wooden picture with a yellow frame\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0568_00/01000.jpg", "scene0568_00/01040.jpg", "scene0568_00/01080.jpg", "scene0568_00/01120.jpg", "scene0568_00/01160.jpg" ], "turn_id": 6, "scan_id": "scene0568_00", "origin_question": "How far is your current position from where you were at the end of the turn 4? (in meters)", "option": [], "answer": "1.49", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 4? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0568_00/01200.jpg", "scene0568_00/01240.jpg", "scene0568_00/01280.jpg", "scene0568_00/01320.jpg", "scene0568_00/01360.jpg" ], "turn_id": 7, "scan_id": "scene0568_00", "origin_question": "Which direction is the horizontal white metal radiator in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the horizontal white metal radiator in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0568_00/01400.jpg", "scene0568_00/01440.jpg", "scene0568_00/01480.jpg", "scene0568_00/01520.jpg", "scene0568_00/01560.jpg" ], "turn_id": 8, "scan_id": "scene0568_00", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the curtain\" from your position?", "option": [], "answer": "120.43", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the curtain\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0595_00/00000.jpg", "scene0595_00/00030.jpg", "scene0595_00/00060.jpg", "scene0595_00/00090.jpg" ], "turn_id": 1, "scan_id": "scene0595_00", "origin_question": "Remember, have you seen any washing machine(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any washing machine(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0595_00/00120.jpg", "scene0595_00/00150.jpg", "scene0595_00/00180.jpg", "scene0595_00/00210.jpg" ], "turn_id": 2, "scan_id": "scene0595_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0595_00/00240.jpg", "scene0595_00/00270.jpg", "scene0595_00/00300.jpg", "scene0595_00/00330.jpg" ], "turn_id": 3, "scan_id": "scene0595_00", "origin_question": "In which turn did you see both \"the white rectangular stainless steel clothes dryer\" and \"the medium-sized rectangular gray table with a smooth surface\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the white rectangular stainless steel clothes dryer\" and \"the medium-sized rectangular gray table with a smooth surface\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0595_00/00360.jpg", "scene0595_00/00390.jpg", "scene0595_00/00420.jpg", "scene0595_00/00450.jpg" ], "turn_id": 4, "scan_id": "scene0595_00", "origin_question": "How far is your current position from where you were at the end of the turn 1? (in meters)", "option": [], "answer": "4.83", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 1? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0595_00/00480.jpg", "scene0595_00/00510.jpg", "scene0595_00/00540.jpg", "scene0595_00/00570.jpg" ], "turn_id": 5, "scan_id": "scene0595_00", "origin_question": "Which two objects are on the same side of you now? \"the modern white plastic standalone washing machine\", \"the medium-sized rectangular gray table with a smooth surface\", \"the rectangular picture with a white background and wooden frame\". ", "option": [ "the modern white plastic standalone washing machine, the medium-sized rectangular gray table with a smooth surface", "the modern white plastic standalone washing machine, the rectangular picture with a white background and wooden frame", "the medium-sized rectangular gray table with a smooth surface, the rectangular picture with a white background and wooden frame" ], "answer": "the modern white plastic standalone washing machine, the medium-sized rectangular gray table with a smooth surface", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the modern white plastic standalone washing machine\", \"the medium-sized rectangular gray table with a smooth surface\", \"the rectangular picture with a white background and wooden frame\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the modern white plastic standalone washing machine, the medium-sized rectangular gray table with a smooth surface', 'the modern white plastic standalone washing machine, the rectangular picture with a white background and wooden frame', 'the medium-sized rectangular gray table with a smooth surface, the rectangular picture with a white background and wooden frame'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0595_00/00600.jpg", "scene0595_00/00630.jpg", "scene0595_00/00660.jpg", "scene0595_00/00690.jpg" ], "turn_id": 6, "scan_id": "scene0595_00", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the modern white plastic standalone washing machine\"; \"the medium-sized circular white pendant light\"; \"the white rectangular stainless steel clothes dryer\" ", "option": [ "the modern white plastic standalone washing machine", "the medium-sized circular white pendant light", "the white rectangular stainless steel clothes dryer" ], "answer": "the medium-sized circular white pendant light", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the modern white plastic standalone washing machine\"; \"the medium-sized circular white pendant light\"; \"the white rectangular stainless steel clothes dryer\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the modern white plastic standalone washing machine', 'the medium-sized circular white pendant light', 'the white rectangular stainless steel clothes dryer'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "1mp3d_0030_region8/ae0a116c370141909e0ca2bfeeeb323c_i1_3.jpg", "1mp3d_0030_region8/ae0a116c370141909e0ca2bfeeeb323c_i2_3.jpg", "1mp3d_0030_region8/1316450be64b472c9f84f6f71a991757_i1_1.jpg", "1mp3d_0030_region8/1316450be64b472c9f84f6f71a991757_i2_1.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0030_region8", "origin_question": "Remember, how many footstool(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many footstool(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 2 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "1mp3d_0030_region8/ae0a116c370141909e0ca2bfeeeb323c_i1_4.jpg", "1mp3d_0030_region8/ae0a116c370141909e0ca2bfeeeb323c_i2_4.jpg", "1mp3d_0030_region8/1316450be64b472c9f84f6f71a991757_i1_0.jpg", "1mp3d_0030_region8/1316450be64b472c9f84f6f71a991757_i2_0.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0030_region8", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the cylindrical ceramic flowerpot with a deep blue exterior\"; \"the smooth black ergonomic mouse with a curved shape\"; \"the deep brown wooden rectangular desk\" ", "option": [ "the cylindrical ceramic flowerpot with a deep blue exterior", "the smooth black ergonomic mouse with a curved shape", "the deep brown wooden rectangular desk" ], "answer": "the cylindrical ceramic flowerpot with a deep blue exterior", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the cylindrical ceramic flowerpot with a deep blue exterior\"; \"the smooth black ergonomic mouse with a curved shape\"; \"the deep brown wooden rectangular desk\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the cylindrical ceramic flowerpot with a deep blue exterior', 'the smooth black ergonomic mouse with a curved shape', 'the deep brown wooden rectangular desk'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 2 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0382_01/00000.jpg", "scene0382_01/00030.jpg", "scene0382_01/00060.jpg", "scene0382_01/00090.jpg" ], "turn_id": 1, "scan_id": "scene0382_01", "origin_question": "Remember, have you seen any range hood(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any range hood(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0382_01/00120.jpg", "scene0382_01/00150.jpg", "scene0382_01/00180.jpg", "scene0382_01/00210.jpg" ], "turn_id": 2, "scan_id": "scene0382_01", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the light brown rectangular wooden desk with a smooth surface\"; \"the white rectangular metal radiator\"; \"the dresser\" ", "option": [ "the light brown rectangular wooden desk with a smooth surface", "the white rectangular metal radiator", "the dresser" ], "answer": "the white rectangular metal radiator", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the light brown rectangular wooden desk with a smooth surface\"; \"the white rectangular metal radiator\"; \"the dresser\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the light brown rectangular wooden desk with a smooth surface', 'the white rectangular metal radiator', 'the dresser'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0382_01/00240.jpg", "scene0382_01/00300.jpg", "scene0382_01/00330.jpg", "scene0382_01/00360.jpg" ], "turn_id": 3, "scan_id": "scene0382_01", "origin_question": "In which turn did you see both \"the vertical window with a white frame and curtains\" and \"the door\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the vertical window with a white frame and curtains\" and \"the door\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0382_01/00390.jpg", "scene0382_01/00420.jpg", "scene0382_01/00450.jpg", "scene0382_01/00480.jpg" ], "turn_id": 4, "scan_id": "scene0382_01", "origin_question": "Which two objects are on the same side of you now? \"the dresser\", \"the black swivel office chair with textured patterns\", \"the white rectangular metal radiator\". ", "option": [ "the dresser, the black swivel office chair with textured patterns", "the dresser, the white rectangular metal radiator", "the black swivel office chair with textured patterns, the white rectangular metal radiator" ], "answer": "the dresser, the black swivel office chair with textured patterns", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the dresser\", \"the black swivel office chair with textured patterns\", \"the white rectangular metal radiator\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the dresser, the black swivel office chair with textured patterns', 'the dresser, the white rectangular metal radiator', 'the black swivel office chair with textured patterns, the white rectangular metal radiator'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0382_01/00510.jpg", "scene0382_01/00540.jpg", "scene0382_01/00570.jpg", "scene0382_01/00600.jpg" ], "turn_id": 5, "scan_id": "scene0382_01", "origin_question": "Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the dresser\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the dresser\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0382_01/00630.jpg", "scene0382_01/00660.jpg", "scene0382_01/00690.jpg", "scene0382_01/00720.jpg" ], "turn_id": 6, "scan_id": "scene0382_01", "origin_question": "Please recall \"the white rectangular metal radiator\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.66", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the white rectangular metal radiator\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0578_02/00000.jpg", "scene0578_02/00040.jpg", "scene0578_02/00080.jpg", "scene0578_02/00120.jpg" ], "turn_id": 1, "scan_id": "scene0578_02", "origin_question": "Remember, have you seen any desk(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any desk(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0578_02/00160.jpg", "scene0578_02/00200.jpg", "scene0578_02/00240.jpg", "scene0578_02/00280.jpg" ], "turn_id": 2, "scan_id": "scene0578_02", "origin_question": "Please recall \"the mobile wooden desk with a black metal frame\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.38", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the mobile wooden desk with a black metal frame\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0578_02/00320.jpg", "scene0578_02/00360.jpg", "scene0578_02/00400.jpg", "scene0578_02/00440.jpg" ], "turn_id": 3, "scan_id": "scene0578_02", "origin_question": "Please recall \"the mobile wooden desk with a black metal frame\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.71", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the mobile wooden desk with a black metal frame\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0578_02/00480.jpg", "scene0578_02/00520.jpg", "scene0578_02/00560.jpg", "scene0578_02/00600.jpg" ], "turn_id": 4, "scan_id": "scene0578_02", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the black rectangular plastic projector\"; \"the light brown rectangular wooden table with a smooth surface\"; \"the mobile wooden desk with a black metal frame\" ", "option": [ "the black rectangular plastic projector", "the light brown rectangular wooden table with a smooth surface", "the mobile wooden desk with a black metal frame" ], "answer": "the black rectangular plastic projector", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the black rectangular plastic projector\"; \"the light brown rectangular wooden table with a smooth surface\"; \"the mobile wooden desk with a black metal frame\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the black rectangular plastic projector', 'the light brown rectangular wooden table with a smooth surface', 'the mobile wooden desk with a black metal frame'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0578_02/00640.jpg", "scene0578_02/00680.jpg", "scene0578_02/00720.jpg", "scene0578_02/00760.jpg" ], "turn_id": 5, "scan_id": "scene0578_02", "origin_question": "Which two objects are on the same side of you now? \"the mobile wooden desk with a black metal frame\", \"the sleek gray flat monitor made of plastic\", \"the small black cylindrical plastic bin\". ", "option": [ "the mobile wooden desk with a black metal frame, the sleek gray flat monitor made of plastic", "the mobile wooden desk with a black metal frame, the small black cylindrical plastic bin", "the sleek gray flat monitor made of plastic, the small black cylindrical plastic bin" ], "answer": "the mobile wooden desk with a black metal frame, the sleek gray flat monitor made of plastic", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the mobile wooden desk with a black metal frame\", \"the sleek gray flat monitor made of plastic\", \"the small black cylindrical plastic bin\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the mobile wooden desk with a black metal frame, the sleek gray flat monitor made of plastic', 'the mobile wooden desk with a black metal frame, the small black cylindrical plastic bin', 'the sleek gray flat monitor made of plastic, the small black cylindrical plastic bin'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0578_02/00800.jpg", "scene0578_02/00840.jpg", "scene0578_02/00880.jpg", "scene0578_02/00920.jpg" ], "turn_id": 6, "scan_id": "scene0578_02", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the door\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the door\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "scene0578_02/00960.jpg", "scene0578_02/01000.jpg", "scene0578_02/01040.jpg", "scene0578_02/01080.jpg" ], "turn_id": 7, "scan_id": "scene0578_02", "origin_question": "Among these three objects(\"the medium blue backpack with a white logo\",\"the door\",\"the black rectangular plastic projector\" ), which one is the farthest from you?", "option": [ "the medium blue backpack with a white logo", "the door", "the black rectangular plastic projector" ], "answer": "the door", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects(\"the medium blue backpack with a white logo\",\"the door\",\"the black rectangular plastic projector\" ), which one is the farthest from you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the medium blue backpack with a white logo', 'the door', 'the black rectangular plastic projector'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0044_region9/0bd6017fd8b6410a982c2a6775f6e9df_i1_4.jpg", "1mp3d_0044_region9/0bd6017fd8b6410a982c2a6775f6e9df_i2_4.jpg", "1mp3d_0044_region9/5bb6a347aea44050b8aa3c0debe93dfc_i1_3.jpg", "1mp3d_0044_region9/5bb6a347aea44050b8aa3c0debe93dfc_i2_3.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0044_region9", "origin_question": "Remember, have you seen any picture(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any picture(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0044_region9/b95772692ede4e9aa2ce4828447e8419_i2_1.jpg", "1mp3d_0044_region9/9177fe5f923242f0b1ba16c7a058e5ac_i1_0.jpg", "1mp3d_0044_region9/9177fe5f923242f0b1ba16c7a058e5ac_i2_0.jpg", "1mp3d_0044_region9/9177fe5f923242f0b1ba16c7a058e5ac_i2_1.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0044_region9", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "1mp3d_0044_region9/b95772692ede4e9aa2ce4828447e8419_i1_0.jpg", "1mp3d_0044_region9/b95772692ede4e9aa2ce4828447e8419_i2_0.jpg", "1mp3d_0044_region9/5bb6a347aea44050b8aa3c0debe93dfc_i1_4.jpg", "1mp3d_0044_region9/5bb6a347aea44050b8aa3c0debe93dfc_i2_4.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0044_region9", "origin_question": "Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "58.49", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "1mp3d_0044_region9/0bd6017fd8b6410a982c2a6775f6e9df_i1_5.jpg", "1mp3d_0044_region9/0bd6017fd8b6410a982c2a6775f6e9df_i2_5.jpg", "1mp3d_0044_region9/0bd6017fd8b6410a982c2a6775f6e9df_i1_3.jpg", "1mp3d_0044_region9/0bd6017fd8b6410a982c2a6775f6e9df_i2_3.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0044_region9", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "32.46", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0044_region9/0bd6017fd8b6410a982c2a6775f6e9df_i1_2.jpg", "1mp3d_0044_region9/0bd6017fd8b6410a982c2a6775f6e9df_i2_2.jpg", "1mp3d_0044_region9/665f53aa4a8f4f1bbb5de7195154d864_i1_5.jpg", "1mp3d_0044_region9/665f53aa4a8f4f1bbb5de7195154d864_i2_5.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0044_region9", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the medium grey and white floral-patterned curtain\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the medium grey and white floral-patterned curtain\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "1mp3d_0044_region9/195363061eff49f186edea6167f4c993_i1_0.jpg", "1mp3d_0044_region9/195363061eff49f186edea6167f4c993_i2_0.jpg", "1mp3d_0044_region9/3950b23f45a344fcaf115a293d591cb1_i1_4.jpg", "1mp3d_0044_region9/3950b23f45a344fcaf115a293d591cb1_i2_4.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0044_region9", "origin_question": "Which direction is the medium grey and white floral-patterned curtain in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the medium grey and white floral-patterned curtain in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "1mp3d_0044_region9/3950b23f45a344fcaf115a293d591cb1_i1_5.jpg", "1mp3d_0044_region9/3950b23f45a344fcaf115a293d591cb1_i2_5.jpg", "1mp3d_0044_region9/195363061eff49f186edea6167f4c993_i1_5.jpg", "1mp3d_0044_region9/195363061eff49f186edea6167f4c993_i2_5.jpg" ], "turn_id": 7, "scan_id": "1mp3d_0044_region9", "origin_question": "When did you first discover \"the dark brown wooden banister with vertical bars\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the dark brown wooden banister with vertical bars\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "1mp3d_0044_region9/665f53aa4a8f4f1bbb5de7195154d864_i1_0.jpg", "1mp3d_0044_region9/665f53aa4a8f4f1bbb5de7195154d864_i2_0.jpg", "1mp3d_0044_region9/f4ccaa94abf4445cb445dec358641b92_i1_2.jpg", "1mp3d_0044_region9/f4ccaa94abf4445cb445dec358641b92_i2_2.jpg" ], "turn_id": 8, "scan_id": "1mp3d_0044_region9", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the light-colored rectangular stairs with medium-sized steps\" from your position?", "option": [], "answer": "164.55", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the light-colored rectangular stairs with medium-sized steps\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0319_00/00000.jpg", "scene0319_00/00040.jpg", "scene0319_00/00080.jpg", "scene0319_00/00120.jpg", "scene0319_00/00160.jpg" ], "turn_id": 1, "scan_id": "scene0319_00", "origin_question": "Remember, how many bucket(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many bucket(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0319_00/00200.jpg", "scene0319_00/00240.jpg", "scene0319_00/00280.jpg", "scene0319_00/00320.jpg", "scene0319_00/00360.jpg" ], "turn_id": 2, "scan_id": "scene0319_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "scene0319_00/00400.jpg", "scene0319_00/00440.jpg", "scene0319_00/00480.jpg", "scene0319_00/00520.jpg", "scene0319_00/00560.jpg" ], "turn_id": 3, "scan_id": "scene0319_00", "origin_question": "Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Right", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "scene0319_00/00600.jpg", "scene0319_00/00640.jpg", "scene0319_00/00680.jpg", "scene0319_00/00720.jpg", "scene0319_00/00760.jpg" ], "turn_id": 4, "scan_id": "scene0319_00", "origin_question": "Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Right", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0319_00/00800.jpg", "scene0319_00/00840.jpg", "scene0319_00/00880.jpg", "scene0319_00/00920.jpg", "scene0319_00/01010.jpg" ], "turn_id": 5, "scan_id": "scene0319_00", "origin_question": "Remember, have you seen any picture(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any picture(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0319_00/01050.jpg", "scene0319_00/01090.jpg", "scene0319_00/01130.jpg", "scene0319_00/01170.jpg", "scene0319_00/01210.jpg" ], "turn_id": 6, "scan_id": "scene0319_00", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0319_00/01250.jpg", "scene0319_00/01290.jpg", "scene0319_00/01330.jpg", "scene0319_00/01370.jpg", "scene0319_00/01410.jpg" ], "turn_id": 7, "scan_id": "scene0319_00", "origin_question": "Using your orientation at the end of turn 6 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "101.84", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 6 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0319_00/01450.jpg", "scene0319_00/01490.jpg", "scene0319_00/01530.jpg", "scene0319_00/01600.jpg", "scene0319_00/01640.jpg" ], "turn_id": 8, "scan_id": "scene0319_00", "origin_question": "Using your orientation at the end of turn 6 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 6 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0471_00/00000.jpg", "scene0471_00/00030.jpg", "scene0471_00/00060.jpg", "scene0471_00/00090.jpg" ], "turn_id": 1, "scan_id": "scene0471_00", "origin_question": "Remember, have you seen any toilet(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any toilet(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0471_00/00120.jpg", "scene0471_00/00150.jpg", "scene0471_00/00180.jpg", "scene0471_00/00210.jpg" ], "turn_id": 2, "scan_id": "scene0471_00", "origin_question": "In which turn were you farthest from \"the rectangular, light brown wooden stall\"? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the rectangular, light brown wooden stall\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0471_00/00240.jpg", "scene0471_00/00270.jpg", "scene0471_00/00300.jpg", "scene0471_00/00330.jpg" ], "turn_id": 3, "scan_id": "scene0471_00", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "80.75", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0471_00/00360.jpg", "scene0471_00/00390.jpg", "scene0471_00/00420.jpg", "scene0471_00/00450.jpg" ], "turn_id": 4, "scan_id": "scene0471_00", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "1.23", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0471_00/00480.jpg", "scene0471_00/00510.jpg", "scene0471_00/00540.jpg", "scene0471_00/00570.jpg" ], "turn_id": 5, "scan_id": "scene0471_00", "origin_question": "Remember, have you seen any car(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any car(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0004_region25/525eea07ff0b45f49f6a47eb55b6d75e_i1_1.jpg", "1mp3d_0004_region25/525eea07ff0b45f49f6a47eb55b6d75e_i2_1.jpg", "1mp3d_0004_region25/0b06e41938c84b91b41b6b14db0529b9_i1_2.jpg", "1mp3d_0004_region25/0b06e41938c84b91b41b6b14db0529b9_i2_2.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0004_region25", "origin_question": "Remember, have you seen any door(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any door(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 2 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0004_region25/525eea07ff0b45f49f6a47eb55b6d75e_i1_4.jpg", "1mp3d_0004_region25/525eea07ff0b45f49f6a47eb55b6d75e_i2_4.jpg", "1mp3d_0004_region25/525eea07ff0b45f49f6a47eb55b6d75e_i1_3.jpg", "1mp3d_0004_region25/525eea07ff0b45f49f6a47eb55b6d75e_i2_3.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0004_region25", "origin_question": "Remember, have you seen any stove(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any stove(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 2 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0608_00/00000.jpg", "scene0608_00/00040.jpg", "scene0608_00/00080.jpg", "scene0608_00/00120.jpg", "scene0608_00/00160.jpg" ], "turn_id": 1, "scan_id": "scene0608_00", "origin_question": "Remember, have you seen any tv(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any tv(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0608_00/00200.jpg", "scene0608_00/00240.jpg", "scene0608_00/00280.jpg", "scene0608_00/00320.jpg", "scene0608_00/00360.jpg" ], "turn_id": 2, "scan_id": "scene0608_00", "origin_question": "When did you first discover \"then acoustic guitar with a natural wood finish\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"then acoustic guitar with a natural wood finish\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0608_00/00400.jpg", "scene0608_00/00440.jpg", "scene0608_00/00480.jpg", "scene0608_00/00520.jpg", "scene0608_00/00560.jpg" ], "turn_id": 3, "scan_id": "scene0608_00", "origin_question": "At the end of which turn were both of \"then acoustic guitar with a natural wood finish\" and \"the window\" on your right side?", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"then acoustic guitar with a natural wood finish\" and \"the window\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0608_00/00600.jpg", "scene0608_00/00640.jpg", "scene0608_00/00680.jpg", "scene0608_00/00720.jpg", "scene0608_00/00760.jpg" ], "turn_id": 4, "scan_id": "scene0608_00", "origin_question": "Which direction is the curtain in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the curtain in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0608_00/00800.jpg", "scene0608_00/00840.jpg", "scene0608_00/00880.jpg", "scene0608_00/00920.jpg", "scene0608_00/00960.jpg" ], "turn_id": 5, "scan_id": "scene0608_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0608_00/01000.jpg", "scene0608_00/01040.jpg", "scene0608_00/01080.jpg", "scene0608_00/01120.jpg", "scene0608_00/01160.jpg" ], "turn_id": 6, "scan_id": "scene0608_00", "origin_question": "By the end of the 4 turn, were you getting closer to or farther away from these two objects: X: \"the curtain\", Y: \"the sturdy square dark leather ottoman\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to Y and farther away from X", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 4 turn, were you getting closer to or farther away from these two objects: X: \"the curtain\", Y: \"the sturdy square dark leather ottoman\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0608_00/01200.jpg", "scene0608_00/01240.jpg", "scene0608_00/01280.jpg", "scene0608_00/01320.jpg", "scene0608_00/01360.jpg" ], "turn_id": 7, "scan_id": "scene0608_00", "origin_question": "Is the large framed picture of a sunset city view to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the large framed picture of a sunset city view to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0608_00/01400.jpg", "scene0608_00/01440.jpg", "scene0608_00/01480.jpg", "scene0608_00/01520.jpg", "scene0608_00/01560.jpg" ], "turn_id": 8, "scan_id": "scene0608_00", "origin_question": "Which direction is the rectangular, light-colored wooden door with six panels in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the rectangular, light-colored wooden door with six panels in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0608_00/01600.jpg", "scene0608_00/01640.jpg", "scene0608_00/01680.jpg", "scene0608_00/01720.jpg", "scene0608_00/01760.jpg" ], "turn_id": 9, "scan_id": "scene0608_00", "origin_question": "When did you first discover \"the curtain\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the curtain\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0608_00/01800.jpg", "scene0608_00/01840.jpg", "scene0608_00/01880.jpg", "scene0608_00/01920.jpg", "scene0608_00/01960.jpg" ], "turn_id": 10, "scan_id": "scene0608_00", "origin_question": "Please recall \"the sleek black flat-screen television on a glossy stand\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.93", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the sleek black flat-screen television on a glossy stand\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0029_region17/49a229bbd8ad4a1585b13622a0db35f8_i1_5.jpg", "1mp3d_0029_region17/49a229bbd8ad4a1585b13622a0db35f8_i2_5.jpg", "1mp3d_0029_region17/dd9fb6ab085d4d558d5d023bb2dc90ff_i1_0.jpg", "1mp3d_0029_region17/dd9fb6ab085d4d558d5d023bb2dc90ff_i2_0.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0029_region17", "origin_question": "Remember, have you seen any table(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any table(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0029_region17/f3accce1c4e44e1d9e94cff071f22eeb_i1_3.jpg", "1mp3d_0029_region17/f3accce1c4e44e1d9e94cff071f22eeb_i2_3.jpg", "1mp3d_0029_region17/1d5c1b5e0a164fd8860723b2069f9b6a_i1_0.jpg", "1mp3d_0029_region17/1d5c1b5e0a164fd8860723b2069f9b6a_i2_0.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0029_region17", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the white round ceramic vase with a smooth surface\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the white round ceramic vase with a smooth surface\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0029_region17/1d5c1b5e0a164fd8860723b2069f9b6a_i1_1.jpg", "1mp3d_0029_region17/1d5c1b5e0a164fd8860723b2069f9b6a_i2_1.jpg", "1mp3d_0029_region17/f3accce1c4e44e1d9e94cff071f22eeb_i1_4.jpg", "1mp3d_0029_region17/f3accce1c4e44e1d9e94cff071f22eeb_i2_4.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0029_region17", "origin_question": "", "option": [], "answer": "", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "1mp3d_0029_region17/dd9fb6ab085d4d558d5d023bb2dc90ff_i1_5.jpg", "1mp3d_0029_region17/dd9fb6ab085d4d558d5d023bb2dc90ff_i2_5.jpg", "1mp3d_0029_region17/46aef22029c84fc491a34cd12b8c9d0e_i1_2.jpg", "1mp3d_0029_region17/46aef22029c84fc491a34cd12b8c9d0e_i2_2.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0029_region17", "origin_question": "Which direction is the rectangular blue and gray patterned carpet in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the rectangular blue and gray patterned carpet in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0029_region17/d4c7bbd42a9f482c9940cb0074dd5cb0_i1_0.jpg", "1mp3d_0029_region17/d4c7bbd42a9f482c9940cb0074dd5cb0_i2_0.jpg", "1mp3d_0029_region17/d4c7bbd42a9f482c9940cb0074dd5cb0_i1_4.jpg", "1mp3d_0029_region17/d4c7bbd42a9f482c9940cb0074dd5cb0_i2_4.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0029_region17", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the large brown round basin made of clay\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the large brown round basin made of clay\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "1mp3d_0029_region17/d4c7bbd42a9f482c9940cb0074dd5cb0_i1_3.jpg", "1mp3d_0029_region17/d4c7bbd42a9f482c9940cb0074dd5cb0_i2_3.jpg", "1mp3d_0029_region17/46aef22029c84fc491a34cd12b8c9d0e_i1_5.jpg", "1mp3d_0029_region17/46aef22029c84fc491a34cd12b8c9d0e_i2_5.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0029_region17", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the large brown round basin made of clay\" from your position?", "option": [], "answer": "49.11", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the large brown round basin made of clay\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "1mp3d_0029_region17/e3d4d3974e75419cade09114d96cbace_i1_2.jpg", "1mp3d_0029_region17/e3d4d3974e75419cade09114d96cbace_i2_2.jpg", "1mp3d_0029_region17/e3d4d3974e75419cade09114d96cbace_i1_1.jpg", "1mp3d_0029_region17/e3d4d3974e75419cade09114d96cbace_i2_1.jpg" ], "turn_id": 7, "scan_id": "1mp3d_0029_region17", "origin_question": "At the end of which turn were both of \"the white round ceramic vase with a smooth surface\" and \"the rectangular blue and gray patterned carpet\" on your left side?", "option": [], "answer": "3", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the white round ceramic vase with a smooth surface\" and \"the rectangular blue and gray patterned carpet\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0029_region17/d7f5f168d31547ce866159165b644da5_i1_3.jpg", "1mp3d_0029_region17/d7f5f168d31547ce866159165b644da5_i2_3.jpg", "1mp3d_0029_region17/27b26e3f29384e60944fc6151074b172_i1_2.jpg", "1mp3d_0029_region17/27b26e3f29384e60944fc6151074b172_i2_2.jpg" ], "turn_id": 8, "scan_id": "1mp3d_0029_region17", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the large brown round basin made of clay\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the large brown round basin made of clay\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "1mp3d_0029_region17/d7f5f168d31547ce866159165b644da5_i1_4.jpg", "1mp3d_0029_region17/d7f5f168d31547ce866159165b644da5_i2_4.jpg", "1mp3d_0029_region17/27b26e3f29384e60944fc6151074b172_i1_3.jpg", "1mp3d_0029_region17/27b26e3f29384e60944fc6151074b172_i2_3.jpg" ], "turn_id": 9, "scan_id": "1mp3d_0029_region17", "origin_question": "By the end of the 3 turn, were you getting closer to or farther away from these two objects: X: \"the round white marble table with a black metal base\", Y: \"the large brown round basin made of clay\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "farther away from both X and Y", "user_message": "For the 9 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 3 turn, were you getting closer to or farther away from these two objects: X: \"the round white marble table with a black metal base\", Y: \"the large brown round basin made of clay\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0338_02/00000.jpg", "scene0338_02/00030.jpg", "scene0338_02/00060.jpg", "scene0338_02/00090.jpg" ], "turn_id": 1, "scan_id": "scene0338_02", "origin_question": "Remember, have you seen any statue(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any statue(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 3 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0338_02/00120.jpg", "scene0338_02/00150.jpg", "scene0338_02/00180.jpg", "scene0338_02/00210.jpg" ], "turn_id": 2, "scan_id": "scene0338_02", "origin_question": "When did you first discover \"the door\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the door\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 3 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0338_02/00240.jpg", "scene0338_02/00270.jpg", "scene0338_02/00300.jpg", "scene0338_02/00330.jpg" ], "turn_id": 3, "scan_id": "scene0338_02", "origin_question": "In which turn did you see both \"the door\" and \"the large rectangular dark stone counter with compartments\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the door\" and \"the large rectangular dark stone counter with compartments\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 3 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0078_region2/af250884a37d4b608b23d395741ee597_i1_4.jpg", "1mp3d_0078_region2/af250884a37d4b608b23d395741ee597_i2_4.jpg", "1mp3d_0078_region2/fd934e994f6144b3a241fd9fb3a4ace9_i1_4.jpg", "1mp3d_0078_region2/fd934e994f6144b3a241fd9fb3a4ace9_i2_4.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0078_region2", "origin_question": "Remember, have you seen any bread(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any bread(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0078_region2/67e8dbc37f1649a7927ba7fdf60c7bf6_i1_5.jpg", "1mp3d_0078_region2/67e8dbc37f1649a7927ba7fdf60c7bf6_i2_5.jpg", "1mp3d_0078_region2/67e8dbc37f1649a7927ba7fdf60c7bf6_i1_4.jpg", "1mp3d_0078_region2/67e8dbc37f1649a7927ba7fdf60c7bf6_i2_4.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0078_region2", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "1mp3d_0078_region2/fd934e994f6144b3a241fd9fb3a4ace9_i1_3.jpg", "1mp3d_0078_region2/fd934e994f6144b3a241fd9fb3a4ace9_i2_3.jpg", "1mp3d_0078_region2/af250884a37d4b608b23d395741ee597_i1_3.jpg", "1mp3d_0078_region2/af250884a37d4b608b23d395741ee597_i2_3.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0078_region2", "origin_question": "When did you first discover \"the medium-sized white round bowl toilet\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the medium-sized white round bowl toilet\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0078_region2/ffe91becf15c455ca85be73f029db24a_i1_4.jpg", "1mp3d_0078_region2/ffe91becf15c455ca85be73f029db24a_i2_4.jpg", "1mp3d_0078_region2/9490f85349c3481eb7a43e7d5eda5451_i1_4.jpg", "1mp3d_0078_region2/9490f85349c3481eb7a43e7d5eda5451_i2_4.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0078_region2", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the medium-sized white round bowl toilet\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the medium-sized white round bowl toilet\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0078_region2/ffe91becf15c455ca85be73f029db24a_i1_3.jpg", "1mp3d_0078_region2/ffe91becf15c455ca85be73f029db24a_i2_3.jpg", "1mp3d_0078_region2/9490f85349c3481eb7a43e7d5eda5451_i1_3.jpg", "1mp3d_0078_region2/9490f85349c3481eb7a43e7d5eda5451_i2_3.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0078_region2", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the rectangular light-colored wooden window frame\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the rectangular light-colored wooden window frame\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0078_region2/1c0715389b644f578961d420672e3613_i1_3.jpg", "1mp3d_0078_region2/1c0715389b644f578961d420672e3613_i2_3.jpg", "1mp3d_0078_region2/1c0715389b644f578961d420672e3613_i1_4.jpg", "1mp3d_0078_region2/1c0715389b644f578961d420672e3613_i2_4.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0078_region2", "origin_question": "", "option": [], "answer": "", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0357_00/00000.jpg", "scene0357_00/00030.jpg", "scene0357_00/00060.jpg", "scene0357_00/00090.jpg" ], "turn_id": 1, "scan_id": "scene0357_00", "origin_question": "Remember, how many dishwasher(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many dishwasher(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0357_00/00120.jpg", "scene0357_00/00150.jpg", "scene0357_00/00180.jpg", "scene0357_00/00210.jpg" ], "turn_id": 2, "scan_id": "scene0357_00", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the towel\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the towel\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0357_00/00240.jpg", "scene0357_00/00270.jpg", "scene0357_00/00300.jpg", "scene0357_00/00330.jpg" ], "turn_id": 3, "scan_id": "scene0357_00", "origin_question": "At the end of which turn were both of \"the stairs\" and \"the medium-sized white sheer curtain made of light material\" on your right side?", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the stairs\" and \"the medium-sized white sheer curtain made of light material\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0357_00/00360.jpg", "scene0357_00/00390.jpg", "scene0357_00/00420.jpg", "scene0357_00/00450.jpg" ], "turn_id": 4, "scan_id": "scene0357_00", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the small round wooden table with a pink tabletop\" from your position?", "option": [], "answer": "102.02", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the small round wooden table with a pink tabletop\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0357_00/00480.jpg", "scene0357_00/00510.jpg", "scene0357_00/00540.jpg", "scene0357_00/00570.jpg" ], "turn_id": 5, "scan_id": "scene0357_00", "origin_question": "Which direction is the medium-sized white sheer curtain made of light material in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the medium-sized white sheer curtain made of light material in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0357_00/00600.jpg", "scene0357_00/00630.jpg", "scene0357_00/00660.jpg", "scene0357_00/00690.jpg" ], "turn_id": 6, "scan_id": "scene0357_00", "origin_question": "Please recall \"the window\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.97", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the window\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "scene0357_00/00720.jpg", "scene0357_00/00750.jpg", "scene0357_00/00780.jpg", "scene0357_00/00810.jpg" ], "turn_id": 7, "scan_id": "scene0357_00", "origin_question": "Among these three objects(\"the window\",\"the cup\",\"the small round wooden table with a pink tabletop\" ), which one is the closest to you?", "option": [ "the window", "the cup", "the small round wooden table with a pink tabletop" ], "answer": "the small round wooden table with a pink tabletop", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects(\"the window\",\"the cup\",\"the small round wooden table with a pink tabletop\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the window', 'the cup', 'the small round wooden table with a pink tabletop'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0004_region4/7c213023b8a14cf9a77272791de04b9a_i2_4.jpg", "1mp3d_0004_region4/7c213023b8a14cf9a77272791de04b9a_i1_5.jpg", "1mp3d_0004_region4/7c213023b8a14cf9a77272791de04b9a_i2_5.jpg", "1mp3d_0004_region4/7c213023b8a14cf9a77272791de04b9a_i1_0.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0004_region4", "origin_question": "Remember, have you seen any chair(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any chair(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 2 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0004_region4/7c213023b8a14cf9a77272791de04b9a_i2_0.jpg", "1mp3d_0004_region4/525eea07ff0b45f49f6a47eb55b6d75e_i1_2.jpg", "1mp3d_0004_region4/525eea07ff0b45f49f6a47eb55b6d75e_i2_2.jpg", "1mp3d_0004_region4/525eea07ff0b45f49f6a47eb55b6d75e_i1_3.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0004_region4", "origin_question": "Remember, have you seen any picture(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any picture(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 2 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0106_01/00000.jpg", "scene0106_01/00040.jpg", "scene0106_01/00080.jpg", "scene0106_01/00120.jpg", "scene0106_01/00160.jpg" ], "turn_id": 1, "scan_id": "scene0106_01", "origin_question": "Remember, how many arch(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many arch(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "scene0106_01/00200.jpg", "scene0106_01/00240.jpg", "scene0106_01/00280.jpg", "scene0106_01/00320.jpg", "scene0106_01/00360.jpg" ], "turn_id": 2, "scan_id": "scene0106_01", "origin_question": "Among these three objects(\"the white rectangular wall-mounted dispenser\",\"the black plastic pen in a stationary position\",\"the round white plastic mouse\" ), which one is the farthest from you?", "option": [ "the white rectangular wall-mounted dispenser", "the black plastic pen in a stationary position", "the round white plastic mouse" ], "answer": "the white rectangular wall-mounted dispenser", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the white rectangular wall-mounted dispenser\",\"the black plastic pen in a stationary position\",\"the round white plastic mouse\" ), which one is the farthest from you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the white rectangular wall-mounted dispenser', 'the black plastic pen in a stationary position', 'the round white plastic mouse'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0106_01/00400.jpg", "scene0106_01/00440.jpg", "scene0106_01/00480.jpg", "scene0106_01/00520.jpg", "scene0106_01/00560.jpg" ], "turn_id": 3, "scan_id": "scene0106_01", "origin_question": "", "option": [], "answer": "", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "scene0106_01/00600.jpg", "scene0106_01/00640.jpg", "scene0106_01/00680.jpg", "scene0106_01/00720.jpg", "scene0106_01/00760.jpg" ], "turn_id": 4, "scan_id": "scene0106_01", "origin_question": "Assuming the direction you are facing at the end of the turn 3 is forward, did you move a certain distance forward or backward from that position?", "option": [ "Forward", "Backward" ], "answer": "Forward", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 3 is forward, did you move a certain distance forward or backward from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Forward', 'Backward'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0106_01/00800.jpg", "scene0106_01/00840.jpg", "scene0106_01/00880.jpg", "scene0106_01/00920.jpg", "scene0106_01/00960.jpg" ], "turn_id": 5, "scan_id": "scene0106_01", "origin_question": "When did you first discover \"the silver and black modern laptop made of aluminum\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the silver and black modern laptop made of aluminum\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0106_01/01000.jpg", "scene0106_01/01040.jpg", "scene0106_01/01080.jpg", "scene0106_01/01120.jpg", "scene0106_01/01160.jpg" ], "turn_id": 6, "scan_id": "scene0106_01", "origin_question": "What will be the first-time appearance order of the following categories: bag, chair, window?", "option": [ "chair, bag, window", "bag, window, chair", "chair, window, bag", "bag, chair, window" ], "answer": "chair, window, bag", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bag, chair, window? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['chair, bag, window', 'bag, window, chair', 'chair, window, bag', 'bag, chair, window'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0106_01/01200.jpg", "scene0106_01/01240.jpg", "scene0106_01/01280.jpg", "scene0106_01/01320.jpg", "scene0106_01/01360.jpg" ], "turn_id": 7, "scan_id": "scene0106_01", "origin_question": "What will be the first-time appearance order of the following categories: couch, sink, window?", "option": [ "window, sink, couch", "sink, window, couch", "couch, sink, window", "sink, couch, window" ], "answer": "sink, window, couch", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: couch, sink, window? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['window, sink, couch', 'sink, window, couch', 'couch, sink, window', 'sink, couch, window'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0106_01/01400.jpg", "scene0106_01/01440.jpg", "scene0106_01/01480.jpg", "scene0106_01/01520.jpg", "scene0106_01/01560.jpg" ], "turn_id": 8, "scan_id": "scene0106_01", "origin_question": "When did you first discover \"the large round clean white ceramic sink\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the large round clean white ceramic sink\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0106_01/01600.jpg", "scene0106_01/01640.jpg", "scene0106_01/01680.jpg", "scene0106_01/01720.jpg", "scene0106_01/01760.jpg" ], "turn_id": 9, "scan_id": "scene0106_01", "origin_question": "Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0106_01/01800.jpg", "scene0106_01/01840.jpg", "scene0106_01/01880.jpg", "scene0106_01/01920.jpg", "scene0106_01/01960.jpg" ], "turn_id": 10, "scan_id": "scene0106_01", "origin_question": "Which two objects are on the same side of you now? \"the flat-screen TV with a deep black color\", \"the white rectangular wall-mounted dispenser\", \"the cylindrical white plastic bin\". ", "option": [ "the flat-screen TV with a deep black color, the white rectangular wall-mounted dispenser", "the flat-screen TV with a deep black color, the cylindrical white plastic bin", "the white rectangular wall-mounted dispenser, the cylindrical white plastic bin" ], "answer": "the white rectangular wall-mounted dispenser, the cylindrical white plastic bin", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the flat-screen TV with a deep black color\", \"the white rectangular wall-mounted dispenser\", \"the cylindrical white plastic bin\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the flat-screen TV with a deep black color, the white rectangular wall-mounted dispenser', 'the flat-screen TV with a deep black color, the cylindrical white plastic bin', 'the white rectangular wall-mounted dispenser, the cylindrical white plastic bin'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0362_02/00000.jpg", "scene0362_02/00040.jpg", "scene0362_02/00080.jpg", "scene0362_02/00120.jpg", "scene0362_02/00160.jpg" ], "turn_id": 1, "scan_id": "scene0362_02", "origin_question": "Remember, how many bucket(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many bucket(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0362_02/00200.jpg", "scene0362_02/00240.jpg", "scene0362_02/00280.jpg", "scene0362_02/00320.jpg", "scene0362_02/00360.jpg" ], "turn_id": 2, "scan_id": "scene0362_02", "origin_question": "When did you first discover \"the framed picture with abstract shapes in red, yellow, and beige\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the framed picture with abstract shapes in red, yellow, and beige\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0362_02/00400.jpg", "scene0362_02/00440.jpg", "scene0362_02/00480.jpg", "scene0362_02/00520.jpg", "scene0362_02/00560.jpg" ], "turn_id": 3, "scan_id": "scene0362_02", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the full-length beige curtain made of thin fabric\"; \"the rectangular two-seater couch with deep brown fabric\"; \"the white rectangular paper with black text\" ", "option": [ "the full-length beige curtain made of thin fabric", "the rectangular two-seater couch with deep brown fabric", "the white rectangular paper with black text" ], "answer": "the white rectangular paper with black text", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the full-length beige curtain made of thin fabric\"; \"the rectangular two-seater couch with deep brown fabric\"; \"the white rectangular paper with black text\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the full-length beige curtain made of thin fabric', 'the rectangular two-seater couch with deep brown fabric', 'the white rectangular paper with black text'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0362_02/00600.jpg", "scene0362_02/00640.jpg", "scene0362_02/00680.jpg", "scene0362_02/00720.jpg", "scene0362_02/00760.jpg" ], "turn_id": 4, "scan_id": "scene0362_02", "origin_question": "What will be the first-time appearance order of the following categories: curtain, picture, tv?", "option": [ "curtain, tv, picture", "picture, curtain, tv", "picture, tv, curtain", "tv, curtain, picture" ], "answer": "picture, curtain, tv", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: curtain, picture, tv? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['curtain, tv, picture', 'picture, curtain, tv', 'picture, tv, curtain', 'tv, curtain, picture'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0362_02/00800.jpg", "scene0362_02/00840.jpg", "scene0362_02/00880.jpg", "scene0362_02/00920.jpg", "scene0362_02/00960.jpg" ], "turn_id": 5, "scan_id": "scene0362_02", "origin_question": "Remember, how many speaker(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many speaker(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0362_02/01000.jpg", "scene0362_02/01040.jpg", "scene0362_02/01080.jpg", "scene0362_02/01120.jpg", "scene0362_02/01160.jpg" ], "turn_id": 6, "scan_id": "scene0362_02", "origin_question": "At the end of which turn were both of \"the medium-sized white cylindrical plastic cup\" and \"the medium-sized rectangular sink with a white finish\" on your right side?", "option": [], "answer": "2", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the medium-sized white cylindrical plastic cup\" and \"the medium-sized rectangular sink with a white finish\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0362_02/01200.jpg", "scene0362_02/01240.jpg", "scene0362_02/01280.jpg", "scene0362_02/01320.jpg", "scene0362_02/01360.jpg" ], "turn_id": 7, "scan_id": "scene0362_02", "origin_question": "Which two objects are on the same side of you now? \"the rectangular two-seater couch with deep brown fabric\", \"the soft gray oversized winter jacket\", \"the rectangular dark brown wooden stand\". ", "option": [ "the rectangular two-seater couch with deep brown fabric, the soft gray oversized winter jacket", "the rectangular two-seater couch with deep brown fabric, the rectangular dark brown wooden stand", "the soft gray oversized winter jacket, the rectangular dark brown wooden stand" ], "answer": "the soft gray oversized winter jacket, the rectangular dark brown wooden stand", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the rectangular two-seater couch with deep brown fabric\", \"the soft gray oversized winter jacket\", \"the rectangular dark brown wooden stand\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular two-seater couch with deep brown fabric, the soft gray oversized winter jacket', 'the rectangular two-seater couch with deep brown fabric, the rectangular dark brown wooden stand', 'the soft gray oversized winter jacket, the rectangular dark brown wooden stand'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0362_02/01400.jpg", "scene0362_02/01440.jpg", "scene0362_02/01480.jpg", "scene0362_02/01520.jpg", "scene0362_02/01560.jpg" ], "turn_id": 8, "scan_id": "scene0362_02", "origin_question": "Compared to your position at the end of the 7 turn, are you now closer or farther away from \"the standard roll of white toilet paper\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 7 turn, are you now closer or farther away from \"the standard roll of white toilet paper\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0362_02/01600.jpg", "scene0362_02/01640.jpg", "scene0362_02/01680.jpg", "scene0362_02/01720.jpg", "scene0362_02/01760.jpg" ], "turn_id": 9, "scan_id": "scene0362_02", "origin_question": "Which two objects are on the same side of you now? \"the rectangular two-seater couch with deep brown fabric\", \"the black flat-screen TV with a sleek design\", \"the vanity\". ", "option": [ "the rectangular two-seater couch with deep brown fabric, the black flat-screen TV with a sleek design", "the rectangular two-seater couch with deep brown fabric, the vanity", "the black flat-screen TV with a sleek design, the vanity" ], "answer": "the rectangular two-seater couch with deep brown fabric, the black flat-screen TV with a sleek design", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the rectangular two-seater couch with deep brown fabric\", \"the black flat-screen TV with a sleek design\", \"the vanity\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular two-seater couch with deep brown fabric, the black flat-screen TV with a sleek design', 'the rectangular two-seater couch with deep brown fabric, the vanity', 'the black flat-screen TV with a sleek design, the vanity'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "scene0362_02/01800.jpg", "scene0362_02/01840.jpg", "scene0362_02/01880.jpg", "scene0362_02/01920.jpg", "scene0362_02/01960.jpg" ], "turn_id": 10, "scan_id": "scene0362_02", "origin_question": "Among these three objects(\"the horizontal creamy yellow metal radiator\",\"the soft gray oversized winter jacket\",\"the black flat-screen TV with a sleek design\" ), which one is the closest to you?", "option": [ "the horizontal creamy yellow metal radiator", "the soft gray oversized winter jacket", "the black flat-screen TV with a sleek design" ], "answer": "the soft gray oversized winter jacket", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the horizontal creamy yellow metal radiator\",\"the soft gray oversized winter jacket\",\"the black flat-screen TV with a sleek design\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the horizontal creamy yellow metal radiator', 'the soft gray oversized winter jacket', 'the black flat-screen TV with a sleek design'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0164_02/00000.jpg", "scene0164_02/00040.jpg", "scene0164_02/00080.jpg", "scene0164_02/00120.jpg" ], "turn_id": 1, "scan_id": "scene0164_02", "origin_question": "Remember, have you seen any coffee maker(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any coffee maker(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0164_02/00160.jpg", "scene0164_02/00200.jpg", "scene0164_02/00240.jpg", "scene0164_02/00280.jpg" ], "turn_id": 2, "scan_id": "scene0164_02", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the rectangular towel with a checkered pattern\"; \"the sleek black rectangular microwave with a glossy surface\"; \"the counter\" ", "option": [ "the rectangular towel with a checkered pattern", "the sleek black rectangular microwave with a glossy surface", "the counter" ], "answer": "the sleek black rectangular microwave with a glossy surface", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the rectangular towel with a checkered pattern\"; \"the sleek black rectangular microwave with a glossy surface\"; \"the counter\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular towel with a checkered pattern', 'the sleek black rectangular microwave with a glossy surface', 'the counter'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0164_02/00320.jpg", "scene0164_02/00360.jpg", "scene0164_02/00400.jpg", "scene0164_02/00440.jpg" ], "turn_id": 3, "scan_id": "scene0164_02", "origin_question": "Which direction is the rectangular black and silver toaster made of metal in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the rectangular black and silver toaster made of metal in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0164_02/00480.jpg", "scene0164_02/00520.jpg", "scene0164_02/00560.jpg", "scene0164_02/00600.jpg" ], "turn_id": 4, "scan_id": "scene0164_02", "origin_question": "At the end of which turn were both of \"the rectangular black and silver toaster made of metal\" and \"the rectangular silver stainless steel sink\" on your left side?", "option": [], "answer": "1", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the rectangular black and silver toaster made of metal\" and \"the rectangular silver stainless steel sink\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0164_02/00640.jpg", "scene0164_02/00680.jpg", "scene0164_02/00720.jpg", "scene0164_02/00760.jpg" ], "turn_id": 5, "scan_id": "scene0164_02", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the dish rack\"; \"the medium-size blue plastic bottle with a white cap\"; \"the white metal double-door refrigerator\" ", "option": [ "the dish rack", "the medium-size blue plastic bottle with a white cap", "the white metal double-door refrigerator" ], "answer": "the medium-size blue plastic bottle with a white cap", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the dish rack\"; \"the medium-size blue plastic bottle with a white cap\"; \"the white metal double-door refrigerator\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the dish rack', 'the medium-size blue plastic bottle with a white cap', 'the white metal double-door refrigerator'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0164_02/00800.jpg", "scene0164_02/00840.jpg", "scene0164_02/00880.jpg", "scene0164_02/00920.jpg" ], "turn_id": 6, "scan_id": "scene0164_02", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the dark brown and black backpack with a white stripe\"; \"the white metal double-door refrigerator\"; \"the medium-size blue plastic bottle with a white cap\" ", "option": [ "the dark brown and black backpack with a white stripe", "the white metal double-door refrigerator", "the medium-size blue plastic bottle with a white cap" ], "answer": "the dark brown and black backpack with a white stripe", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the dark brown and black backpack with a white stripe\"; \"the white metal double-door refrigerator\"; \"the medium-size blue plastic bottle with a white cap\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the dark brown and black backpack with a white stripe', 'the white metal double-door refrigerator', 'the medium-size blue plastic bottle with a white cap'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0164_02/00960.jpg", "scene0164_02/01000.jpg", "scene0164_02/01040.jpg", "scene0164_02/01080.jpg" ], "turn_id": 7, "scan_id": "scene0164_02", "origin_question": "Please recall \"the medium-size blue plastic bottle with a white cap\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.07", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the medium-size blue plastic bottle with a white cap\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0176_00/00000.jpg", "scene0176_00/00040.jpg", "scene0176_00/00080.jpg", "scene0176_00/00120.jpg" ], "turn_id": 1, "scan_id": "scene0176_00", "origin_question": "Remember, have you seen any picture(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any picture(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0176_00/00160.jpg", "scene0176_00/00200.jpg", "scene0176_00/00240.jpg", "scene0176_00/00280.jpg" ], "turn_id": 2, "scan_id": "scene0176_00", "origin_question": "When did you first discover \"the dark rectangular table with a green top\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the dark rectangular table with a green top\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0176_00/00320.jpg", "scene0176_00/00360.jpg", "scene0176_00/00400.jpg", "scene0176_00/00440.jpg" ], "turn_id": 3, "scan_id": "scene0176_00", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the dark rectangular table with a green top\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the dark rectangular table with a green top\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0176_00/00480.jpg", "scene0176_00/00520.jpg", "scene0176_00/00560.jpg", "scene0176_00/00600.jpg" ], "turn_id": 4, "scan_id": "scene0176_00", "origin_question": "By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the window\", Y: \"the windowsill\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the window\", Y: \"the windowsill\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0176_00/00640.jpg", "scene0176_00/00680.jpg", "scene0176_00/00720.jpg", "scene0176_00/00760.jpg" ], "turn_id": 5, "scan_id": "scene0176_00", "origin_question": "Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0176_00/00800.jpg", "scene0176_00/00840.jpg", "scene0176_00/00880.jpg", "scene0176_00/00920.jpg" ], "turn_id": 6, "scan_id": "scene0176_00", "origin_question": "Which direction is the window in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the window in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0656_02/00000.jpg", "scene0656_02/00040.jpg", "scene0656_02/00080.jpg", "scene0656_02/00120.jpg", "scene0656_02/00160.jpg" ], "turn_id": 1, "scan_id": "scene0656_02", "origin_question": "Remember, have you seen any bin(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any bin(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0656_02/00200.jpg", "scene0656_02/00240.jpg", "scene0656_02/00280.jpg", "scene0656_02/00320.jpg", "scene0656_02/00360.jpg" ], "turn_id": 2, "scan_id": "scene0656_02", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the rectangular, deep brown wooden desk with drawers\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the rectangular, deep brown wooden desk with drawers\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0656_02/00400.jpg", "scene0656_02/00440.jpg", "scene0656_02/00480.jpg", "scene0656_02/00520.jpg", "scene0656_02/00560.jpg" ], "turn_id": 3, "scan_id": "scene0656_02", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the paper\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the paper\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0656_02/00600.jpg", "scene0656_02/00640.jpg", "scene0656_02/00680.jpg", "scene0656_02/00720.jpg", "scene0656_02/00760.jpg" ], "turn_id": 4, "scan_id": "scene0656_02", "origin_question": "What will be the first-time appearance order of the following categories: picture, shelf, tissue?", "option": [ "tissue, shelf, picture", "picture, shelf, tissue", "picture, tissue, shelf", "shelf, tissue, picture" ], "answer": "tissue, shelf, picture", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: picture, shelf, tissue? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['tissue, shelf, picture', 'picture, shelf, tissue', 'picture, tissue, shelf', 'shelf, tissue, picture'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0656_02/00800.jpg", "scene0656_02/00840.jpg", "scene0656_02/00880.jpg", "scene0656_02/00920.jpg", "scene0656_02/00960.jpg" ], "turn_id": 5, "scan_id": "scene0656_02", "origin_question": "Is the standard chair with a deep brown cushioned seat to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the standard chair with a deep brown cushioned seat to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0656_02/01000.jpg", "scene0656_02/01040.jpg", "scene0656_02/01080.jpg", "scene0656_02/01120.jpg", "scene0656_02/01160.jpg" ], "turn_id": 6, "scan_id": "scene0656_02", "origin_question": "Please recall \"the rectangular white pillow with a checkerboard pattern\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.03", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the rectangular white pillow with a checkerboard pattern\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0656_02/01200.jpg", "scene0656_02/01240.jpg", "scene0656_02/01280.jpg", "scene0656_02/01320.jpg", "scene0656_02/01360.jpg" ], "turn_id": 7, "scan_id": "scene0656_02", "origin_question": "In which turn did you see both \"the dark brown rectangular wooden stand with two drawers\" and \"the modern upright vacuum cleaner in silver and blue\" simultaneously? ", "option": [], "answer": "7", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the dark brown rectangular wooden stand with two drawers\" and \"the modern upright vacuum cleaner in silver and blue\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0656_02/01400.jpg", "scene0656_02/01440.jpg", "scene0656_02/01480.jpg", "scene0656_02/01520.jpg", "scene0656_02/01560.jpg" ], "turn_id": 8, "scan_id": "scene0656_02", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the standard chair with a deep brown cushioned seat\"; \"the bag\"; \"the rectangular, deep brown wooden desk with drawers\" ", "option": [ "the standard chair with a deep brown cushioned seat", "the bag", "the rectangular, deep brown wooden desk with drawers" ], "answer": "the bag", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the standard chair with a deep brown cushioned seat\"; \"the bag\"; \"the rectangular, deep brown wooden desk with drawers\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the standard chair with a deep brown cushioned seat', 'the bag', 'the rectangular, deep brown wooden desk with drawers'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0656_02/01600.jpg", "scene0656_02/01640.jpg", "scene0656_02/01680.jpg", "scene0656_02/01720.jpg", "scene0656_02/01760.jpg" ], "turn_id": 9, "scan_id": "scene0656_02", "origin_question": "Which two objects are on the same side of you now? \"the rectangular, deep brown wooden desk with drawers\", \"the paper\", \"the traditional wooden dresser in rich brown color\". ", "option": [ "the rectangular, deep brown wooden desk with drawers, the paper", "the rectangular, deep brown wooden desk with drawers, the traditional wooden dresser in rich brown color", "the paper, the traditional wooden dresser in rich brown color" ], "answer": "the rectangular, deep brown wooden desk with drawers, the paper", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the rectangular, deep brown wooden desk with drawers\", \"the paper\", \"the traditional wooden dresser in rich brown color\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular, deep brown wooden desk with drawers, the paper', 'the rectangular, deep brown wooden desk with drawers, the traditional wooden dresser in rich brown color', 'the paper, the traditional wooden dresser in rich brown color'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0006_00/00000.jpg", "scene0006_00/00040.jpg", "scene0006_00/00080.jpg", "scene0006_00/00120.jpg", "scene0006_00/00160.jpg" ], "turn_id": 1, "scan_id": "scene0006_00", "origin_question": "Remember, how many cushion(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many cushion(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0006_00/00200.jpg", "scene0006_00/00240.jpg", "scene0006_00/00280.jpg", "scene0006_00/00320.jpg", "scene0006_00/00360.jpg" ], "turn_id": 2, "scan_id": "scene0006_00", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the small, white rectangular lamp\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the small, white rectangular lamp\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0006_00/00400.jpg", "scene0006_00/00440.jpg", "scene0006_00/00480.jpg", "scene0006_00/00520.jpg", "scene0006_00/00560.jpg" ], "turn_id": 3, "scan_id": "scene0006_00", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the square-shaped book with a dark glossy cover\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the square-shaped book with a dark glossy cover\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0006_00/00600.jpg", "scene0006_00/00640.jpg", "scene0006_00/00680.jpg", "scene0006_00/00720.jpg", "scene0006_00/00760.jpg" ], "turn_id": 4, "scan_id": "scene0006_00", "origin_question": "Which two objects are on the same side of you now? \"the black rectangular tote bag made of sturdy material\", \"the small circular silver and black coffee maker\", \"the flat-panel TV with a stylish black frame\". ", "option": [ "the black rectangular tote bag made of sturdy material, the small circular silver and black coffee maker", "the black rectangular tote bag made of sturdy material, the flat-panel TV with a stylish black frame", "the small circular silver and black coffee maker, the flat-panel TV with a stylish black frame" ], "answer": "the small circular silver and black coffee maker, the flat-panel TV with a stylish black frame", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the black rectangular tote bag made of sturdy material\", \"the small circular silver and black coffee maker\", \"the flat-panel TV with a stylish black frame\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the black rectangular tote bag made of sturdy material, the small circular silver and black coffee maker', 'the black rectangular tote bag made of sturdy material, the flat-panel TV with a stylish black frame', 'the small circular silver and black coffee maker, the flat-panel TV with a stylish black frame'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0006_00/00800.jpg", "scene0006_00/00840.jpg", "scene0006_00/00880.jpg", "scene0006_00/00920.jpg", "scene0006_00/00960.jpg" ], "turn_id": 5, "scan_id": "scene0006_00", "origin_question": "Which two objects are on the same side of you now? \"the black rectangular tote bag made of sturdy material\", \"the black and gray rectangular fabric backpack\", \"the square-shaped book with a dark glossy cover\". ", "option": [ "the black rectangular tote bag made of sturdy material, the black and gray rectangular fabric backpack", "the black rectangular tote bag made of sturdy material, the square-shaped book with a dark glossy cover", "the black and gray rectangular fabric backpack, the square-shaped book with a dark glossy cover" ], "answer": "the black and gray rectangular fabric backpack, the square-shaped book with a dark glossy cover", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the black rectangular tote bag made of sturdy material\", \"the black and gray rectangular fabric backpack\", \"the square-shaped book with a dark glossy cover\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the black rectangular tote bag made of sturdy material, the black and gray rectangular fabric backpack', 'the black rectangular tote bag made of sturdy material, the square-shaped book with a dark glossy cover', 'the black and gray rectangular fabric backpack, the square-shaped book with a dark glossy cover'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0006_00/01000.jpg", "scene0006_00/01040.jpg", "scene0006_00/01080.jpg", "scene0006_00/01120.jpg", "scene0006_00/01160.jpg" ], "turn_id": 6, "scan_id": "scene0006_00", "origin_question": "What will be the first-time appearance order of the following categories: bed, case, table?", "option": [ "table, bed, case", "bed, case, table", "bed, table, case", "case, bed, table" ], "answer": "bed, case, table", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bed, case, table? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['table, bed, case', 'bed, case, table', 'bed, table, case', 'case, bed, table'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0006_00/01200.jpg", "scene0006_00/01240.jpg", "scene0006_00/01280.jpg", "scene0006_00/01320.jpg", "scene0006_00/01360.jpg" ], "turn_id": 7, "scan_id": "scene0006_00", "origin_question": "When did you first discover \"the small, colorful cylindrical bottle\"? (index of the turn)", "option": [], "answer": "3", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the small, colorful cylindrical bottle\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0006_00/01400.jpg", "scene0006_00/01440.jpg", "scene0006_00/01480.jpg", "scene0006_00/01520.jpg", "scene0006_00/01560.jpg" ], "turn_id": 8, "scan_id": "scene0006_00", "origin_question": "Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the pair of dark leather flat shoes\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the pair of dark leather flat shoes\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0006_00/01600.jpg", "scene0006_00/01640.jpg", "scene0006_00/01680.jpg", "scene0006_00/01720.jpg", "scene0006_00/01760.jpg" ], "turn_id": 9, "scan_id": "scene0006_00", "origin_question": "Which two objects are on the same side of you now? \"the round, deep brown wooden stand\", \"the flat-panel TV with a stylish black frame\", \"the small, colorful cylindrical bottle\". ", "option": [ "the round, deep brown wooden stand, the flat-panel TV with a stylish black frame", "the round, deep brown wooden stand, the small, colorful cylindrical bottle", "the flat-panel TV with a stylish black frame, the small, colorful cylindrical bottle" ], "answer": "the flat-panel TV with a stylish black frame, the small, colorful cylindrical bottle", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the round, deep brown wooden stand\", \"the flat-panel TV with a stylish black frame\", \"the small, colorful cylindrical bottle\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the round, deep brown wooden stand, the flat-panel TV with a stylish black frame', 'the round, deep brown wooden stand, the small, colorful cylindrical bottle', 'the flat-panel TV with a stylish black frame, the small, colorful cylindrical bottle'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0006_00/01800.jpg", "scene0006_00/01840.jpg", "scene0006_00/01880.jpg", "scene0006_00/01920.jpg", "scene0006_00/01960.jpg" ], "turn_id": 10, "scan_id": "scene0006_00", "origin_question": "Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the flat-panel TV with a stylish black frame\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the flat-panel TV with a stylish black frame\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0040_region58/4735b504401b4996bce4281d352eb717_i1_3.jpg", "1mp3d_0040_region58/4735b504401b4996bce4281d352eb717_i2_3.jpg", "1mp3d_0040_region58/4735b504401b4996bce4281d352eb717_i2_0.jpg", "1mp3d_0040_region58/6966032cefef42718e3ee3d98b0101af_i1_0.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0040_region58", "origin_question": "Remember, have you seen any folder(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any folder(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "1mp3d_0040_region58/6966032cefef42718e3ee3d98b0101af_i2_0.jpg", "1mp3d_0040_region58/fb9744e5b14047238a984468f7545341_i1_1.jpg", "1mp3d_0040_region58/fb9744e5b14047238a984468f7545341_i2_1.jpg", "1mp3d_0040_region58/6966032cefef42718e3ee3d98b0101af_i1_5.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0040_region58", "origin_question": "In which turn were you closest to \"the wall with beige ceramic tiles in a diamond pattern\"? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the wall with beige ceramic tiles in a diamond pattern\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0040_region58/6966032cefef42718e3ee3d98b0101af_i2_5.jpg", "1mp3d_0040_region58/4735b504401b4996bce4281d352eb717_i1_4.jpg", "1mp3d_0040_region58/4735b504401b4996bce4281d352eb717_i2_4.jpg", "1mp3d_0040_region58/4735b504401b4996bce4281d352eb717_i1_2.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0040_region58", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the wall with beige ceramic tiles in a diamond pattern\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the wall with beige ceramic tiles in a diamond pattern\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "1mp3d_0040_region58/4735b504401b4996bce4281d352eb717_i2_2.jpg", "1mp3d_0040_region58/4735b504401b4996bce4281d352eb717_i1_1.jpg", "1mp3d_0040_region58/4735b504401b4996bce4281d352eb717_i2_1.jpg", "1mp3d_0040_region58/6966032cefef42718e3ee3d98b0101af_i1_4.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0040_region58", "origin_question": "In which turn were you farthest from \"the switch\"? ", "option": [], "answer": "2", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the switch\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "1mp3d_0040_region58/6966032cefef42718e3ee3d98b0101af_i2_4.jpg", "1mp3d_0040_region58/6966032cefef42718e3ee3d98b0101af_i1_3.jpg", "1mp3d_0040_region58/6966032cefef42718e3ee3d98b0101af_i2_3.jpg", "1mp3d_0040_region58/6966032cefef42718e3ee3d98b0101af_i1_2.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0040_region58", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0426_01/00000.jpg", "scene0426_01/00040.jpg", "scene0426_01/00080.jpg", "scene0426_01/00120.jpg", "scene0426_01/00160.jpg" ], "turn_id": 1, "scan_id": "scene0426_01", "origin_question": "Remember, have you seen any chair(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any chair(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0426_01/00200.jpg", "scene0426_01/00240.jpg", "scene0426_01/00280.jpg", "scene0426_01/00320.jpg", "scene0426_01/00360.jpg" ], "turn_id": 2, "scan_id": "scene0426_01", "origin_question": "In which turn did you see both \"the dark brown rectangular wooden desk\" and \"the rectangular silver metal rack\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the dark brown rectangular wooden desk\" and \"the rectangular silver metal rack\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0426_01/00400.jpg", "scene0426_01/00440.jpg", "scene0426_01/00480.jpg", "scene0426_01/00520.jpg", "scene0426_01/00560.jpg" ], "turn_id": 3, "scan_id": "scene0426_01", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the rectangular silver metal rack\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the rectangular silver metal rack\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0426_01/00600.jpg", "scene0426_01/00640.jpg", "scene0426_01/00680.jpg", "scene0426_01/00720.jpg", "scene0426_01/00760.jpg" ], "turn_id": 4, "scan_id": "scene0426_01", "origin_question": "Is the medium-sized rectangular black backpack made of fabric to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the medium-sized rectangular black backpack made of fabric to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0426_01/00800.jpg", "scene0426_01/00840.jpg", "scene0426_01/00880.jpg", "scene0426_01/00920.jpg", "scene0426_01/00960.jpg" ], "turn_id": 5, "scan_id": "scene0426_01", "origin_question": "Which direction is the rectangular silver metal rack in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the rectangular silver metal rack in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0426_01/01000.jpg", "scene0426_01/01040.jpg", "scene0426_01/01090.jpg", "scene0426_01/01130.jpg", "scene0426_01/01170.jpg" ], "turn_id": 6, "scan_id": "scene0426_01", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0426_01/01210.jpg", "scene0426_01/01250.jpg", "scene0426_01/01290.jpg", "scene0426_01/01330.jpg", "scene0426_01/01370.jpg" ], "turn_id": 7, "scan_id": "scene0426_01", "origin_question": "When did you first discover \"the sleek dark flat-screen TV made of metal and plastic\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the sleek dark flat-screen TV made of metal and plastic\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0426_01/01410.jpg", "scene0426_01/01450.jpg", "scene0426_01/01490.jpg", "scene0426_01/01580.jpg", "scene0426_01/01620.jpg" ], "turn_id": 8, "scan_id": "scene0426_01", "origin_question": "By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the medium-sized rectangular black backpack made of fabric\", Y: \"the rectangular blue tissue box with a textured surface\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to Y and farther away from X", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the medium-sized rectangular black backpack made of fabric\", Y: \"the rectangular blue tissue box with a textured surface\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0072_region7/cc8aaf0543a44f9fb067340d7b4d2d7e_i1_4.jpg", "1mp3d_0072_region7/cc8aaf0543a44f9fb067340d7b4d2d7e_i2_4.jpg", "1mp3d_0072_region7/cc8aaf0543a44f9fb067340d7b4d2d7e_i1_3.jpg", "1mp3d_0072_region7/cc8aaf0543a44f9fb067340d7b4d2d7e_i2_3.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0072_region7", "origin_question": "Remember, have you seen any window(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any window(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 3 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "1mp3d_0072_region7/3d5547ed57e2466bb12264e530574179_i1_0.jpg", "1mp3d_0072_region7/3d5547ed57e2466bb12264e530574179_i2_0.jpg", "1mp3d_0072_region7/bed92083ab4440bab70e7fc5bf0fb42e_i1_3.jpg", "1mp3d_0072_region7/bed92083ab4440bab70e7fc5bf0fb42e_i2_3.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0072_region7", "origin_question": "In which turn were you closest to \"the stairs\"? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the stairs\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 3 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "1mp3d_0072_region7/bed92083ab4440bab70e7fc5bf0fb42e_i2_2.jpg", "1mp3d_0072_region7/bed92083ab4440bab70e7fc5bf0fb42e_i1_1.jpg", "1mp3d_0072_region7/bed92083ab4440bab70e7fc5bf0fb42e_i2_1.jpg", "1mp3d_0072_region7/3d5547ed57e2466bb12264e530574179_i1_4.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0072_region7", "origin_question": "When did you first discover \"the window frame\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the window frame\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 3 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0067_region4/312c145aacbd4014909e033462abdd9e_i2_5.jpg", "1mp3d_0067_region4/61c6596ffc2a4d788d93acf6541c7d95_i1_0.jpg", "1mp3d_0067_region4/61c6596ffc2a4d788d93acf6541c7d95_i2_0.jpg", "1mp3d_0067_region4/312c145aacbd4014909e033462abdd9e_i2_4.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0067_region4", "origin_question": "Remember, have you seen any door(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any door(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 2 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "1mp3d_0067_region4/312c145aacbd4014909e033462abdd9e_i1_3.jpg", "1mp3d_0067_region4/312c145aacbd4014909e033462abdd9e_i2_3.jpg", "1mp3d_0067_region4/61c6596ffc2a4d788d93acf6541c7d95_i1_5.jpg", "1mp3d_0067_region4/61c6596ffc2a4d788d93acf6541c7d95_i2_5.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0067_region4", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the window\", Y: \"the large rectangular white plastic bag with a purple tag\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "farther away from both X and Y", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the window\", Y: \"the large rectangular white plastic bag with a purple tag\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 2 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0040_region51/c929b31b97a14f91b1cadb318a4a8353_i1_2.jpg", "1mp3d_0040_region51/c929b31b97a14f91b1cadb318a4a8353_i2_2.jpg", "1mp3d_0040_region51/d771d6746f274d4d90697765491f308e_i1_2.jpg", "1mp3d_0040_region51/d771d6746f274d4d90697765491f308e_i2_2.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0040_region51", "origin_question": "Remember, have you seen any paper(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any paper(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 2 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "1mp3d_0040_region51/c929b31b97a14f91b1cadb318a4a8353_i1_0.jpg", "1mp3d_0040_region51/c929b31b97a14f91b1cadb318a4a8353_i2_0.jpg", "1mp3d_0040_region51/c929b31b97a14f91b1cadb318a4a8353_i1_5.jpg", "1mp3d_0040_region51/c929b31b97a14f91b1cadb318a4a8353_i2_5.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0040_region51", "origin_question": "In which turn were you closest to \"the rectangular polished sink with beige and brown tones\"? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the rectangular polished sink with beige and brown tones\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 2 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0583_00/00000.jpg", "scene0583_00/00040.jpg", "scene0583_00/00080.jpg", "scene0583_00/00120.jpg", "scene0583_00/00160.jpg" ], "turn_id": 1, "scan_id": "scene0583_00", "origin_question": "Remember, have you seen any picture(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any picture(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0583_00/00200.jpg", "scene0583_00/00240.jpg", "scene0583_00/00280.jpg", "scene0583_00/00320.jpg", "scene0583_00/00360.jpg" ], "turn_id": 2, "scan_id": "scene0583_00", "origin_question": "In which turn were you closest to \"the rectangular brown wooden door made of solid wood\"? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the rectangular brown wooden door made of solid wood\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0583_00/00400.jpg", "scene0583_00/00440.jpg", "scene0583_00/00480.jpg", "scene0583_00/00520.jpg", "scene0583_00/00560.jpg" ], "turn_id": 3, "scan_id": "scene0583_00", "origin_question": "In which turn were you farthest from \"the rectangular brown wooden door made of solid wood\"? ", "option": [], "answer": "2", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the rectangular brown wooden door made of solid wood\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0583_00/00600.jpg", "scene0583_00/00640.jpg", "scene0583_00/00680.jpg", "scene0583_00/00720.jpg", "scene0583_00/00760.jpg" ], "turn_id": 4, "scan_id": "scene0583_00", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the white rectangular notice sheet printed on paper\" from your position?", "option": [], "answer": "153.92", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the white rectangular notice sheet printed on paper\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0583_00/00800.jpg", "scene0583_00/00840.jpg", "scene0583_00/00880.jpg", "scene0583_00/00920.jpg", "scene0583_00/00960.jpg" ], "turn_id": 5, "scan_id": "scene0583_00", "origin_question": "Which two objects are on the same side of you now? \"the rectangular brown wooden door made of solid wood\", \"the cylindrical white plastic bottle with a pump\", \"the window\". ", "option": [ "the rectangular brown wooden door made of solid wood, the cylindrical white plastic bottle with a pump", "the rectangular brown wooden door made of solid wood, the window", "the cylindrical white plastic bottle with a pump, the window" ], "answer": "the cylindrical white plastic bottle with a pump, the window", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the rectangular brown wooden door made of solid wood\", \"the cylindrical white plastic bottle with a pump\", \"the window\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular brown wooden door made of solid wood, the cylindrical white plastic bottle with a pump', 'the rectangular brown wooden door made of solid wood, the window', 'the cylindrical white plastic bottle with a pump, the window'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0583_00/01000.jpg", "scene0583_00/01040.jpg", "scene0583_00/01080.jpg", "scene0583_00/01120.jpg", "scene0583_00/01160.jpg" ], "turn_id": 6, "scan_id": "scene0583_00", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the shower\" from your position?", "option": [], "answer": "166.96", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the shower\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0583_00/01200.jpg", "scene0583_00/01240.jpg", "scene0583_00/01280.jpg", "scene0583_00/01320.jpg", "scene0583_00/01360.jpg" ], "turn_id": 7, "scan_id": "scene0583_00", "origin_question": "At the end of which turn were both of \"the white rectangular notice sheet printed on paper\" and \"the standard white rectangular porcelain bathtub\" on your right side?", "option": [], "answer": "6", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the white rectangular notice sheet printed on paper\" and \"the standard white rectangular porcelain bathtub\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0583_00/01400.jpg", "scene0583_00/01440.jpg", "scene0583_00/01480.jpg", "scene0583_00/01520.jpg", "scene0583_00/01560.jpg" ], "turn_id": 8, "scan_id": "scene0583_00", "origin_question": "Which two objects are on the same side of you now? \"the standard white rectangular porcelain bathtub\", \"the medium white oval-shaped porcelain toilet\", \"the window\". ", "option": [ "the standard white rectangular porcelain bathtub, the medium white oval-shaped porcelain toilet", "the standard white rectangular porcelain bathtub, the window", "the medium white oval-shaped porcelain toilet, the window" ], "answer": "the medium white oval-shaped porcelain toilet, the window", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the standard white rectangular porcelain bathtub\", \"the medium white oval-shaped porcelain toilet\", \"the window\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the standard white rectangular porcelain bathtub, the medium white oval-shaped porcelain toilet', 'the standard white rectangular porcelain bathtub, the window', 'the medium white oval-shaped porcelain toilet, the window'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0583_00/01600.jpg", "scene0583_00/01640.jpg", "scene0583_00/01680.jpg", "scene0583_00/01720.jpg", "scene0583_00/01760.jpg" ], "turn_id": 9, "scan_id": "scene0583_00", "origin_question": "Which direction is the cylindrical white plastic bottle with a pump in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the cylindrical white plastic bottle with a pump in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0432_01/00000.jpg", "scene0432_01/00030.jpg", "scene0432_01/00060.jpg", "scene0432_01/00090.jpg" ], "turn_id": 1, "scan_id": "scene0432_01", "origin_question": "Remember, how many spoon(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many spoon(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 2 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0432_01/00120.jpg", "scene0432_01/00150.jpg", "scene0432_01/00180.jpg", "scene0432_01/00210.jpg" ], "turn_id": 2, "scan_id": "scene0432_01", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 2 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0029_region32/717d780575424720920e3fea97d98842_i1_1.jpg", "1mp3d_0029_region32/717d780575424720920e3fea97d98842_i2_1.jpg", "1mp3d_0029_region32/c9da6aab1a6c4447975c5099ce158f75_i1_0.jpg", "1mp3d_0029_region32/c9da6aab1a6c4447975c5099ce158f75_i2_0.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0029_region32", "origin_question": "Remember, have you seen any carpet(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any carpet(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "1mp3d_0029_region32/717d780575424720920e3fea97d98842_i1_2.jpg", "1mp3d_0029_region32/717d780575424720920e3fea97d98842_i2_2.jpg", "1mp3d_0029_region32/93318cd6a48a4eb59eb59d2481095044_i1_0.jpg", "1mp3d_0029_region32/93318cd6a48a4eb59eb59d2481095044_i2_0.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0029_region32", "origin_question": "Please recall \"the black triangular table with a smooth surface\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "4.19", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the black triangular table with a smooth surface\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "1mp3d_0029_region32/93318cd6a48a4eb59eb59d2481095044_i1_2.jpg", "1mp3d_0029_region32/93318cd6a48a4eb59eb59d2481095044_i2_2.jpg", "1mp3d_0029_region32/717d780575424720920e3fea97d98842_i1_5.jpg", "1mp3d_0029_region32/717d780575424720920e3fea97d98842_i2_5.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0029_region32", "origin_question": "Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Right", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "1mp3d_0029_region32/717d780575424720920e3fea97d98842_i1_4.jpg", "1mp3d_0029_region32/717d780575424720920e3fea97d98842_i2_4.jpg", "1mp3d_0029_region32/717d780575424720920e3fea97d98842_i1_0.jpg", "1mp3d_0029_region32/717d780575424720920e3fea97d98842_i2_0.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0029_region32", "origin_question": "Which two objects are on the same side of you now? \"the textured beige woven blanket made of thick fabric\", \"the large rectangular creamy carpet with a woven texture\", \"the window frame\". ", "option": [ "the textured beige woven blanket made of thick fabric, the large rectangular creamy carpet with a woven texture", "the textured beige woven blanket made of thick fabric, the window frame", "the large rectangular creamy carpet with a woven texture, the window frame" ], "answer": "the textured beige woven blanket made of thick fabric, the large rectangular creamy carpet with a woven texture", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the textured beige woven blanket made of thick fabric\", \"the large rectangular creamy carpet with a woven texture\", \"the window frame\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the textured beige woven blanket made of thick fabric, the large rectangular creamy carpet with a woven texture', 'the textured beige woven blanket made of thick fabric, the window frame', 'the large rectangular creamy carpet with a woven texture, the window frame'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "1mp3d_0029_region32/c9da6aab1a6c4447975c5099ce158f75_i1_4.jpg", "1mp3d_0029_region32/c9da6aab1a6c4447975c5099ce158f75_i2_4.jpg", "1mp3d_0029_region32/c9da6aab1a6c4447975c5099ce158f75_i1_3.jpg", "1mp3d_0029_region32/c9da6aab1a6c4447975c5099ce158f75_i2_3.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0029_region32", "origin_question": "In which turn did you see both \"the textured beige woven blanket made of thick fabric\" and \"the window frame\" simultaneously? ", "option": [], "answer": "5", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the textured beige woven blanket made of thick fabric\" and \"the window frame\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "1mp3d_0029_region32/c9da6aab1a6c4447975c5099ce158f75_i1_2.jpg", "1mp3d_0029_region32/c9da6aab1a6c4447975c5099ce158f75_i2_2.jpg", "1mp3d_0029_region32/c9da6aab1a6c4447975c5099ce158f75_i1_1.jpg", "1mp3d_0029_region32/c9da6aab1a6c4447975c5099ce158f75_i2_1.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0029_region32", "origin_question": "Which direction is the headboard in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the headboard in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0353_02/00000.jpg", "scene0353_02/00040.jpg", "scene0353_02/00080.jpg", "scene0353_02/00120.jpg", "scene0353_02/00160.jpg" ], "turn_id": 1, "scan_id": "scene0353_02", "origin_question": "Remember, have you seen any bin(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any bin(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0353_02/00200.jpg", "scene0353_02/00240.jpg", "scene0353_02/00280.jpg", "scene0353_02/00320.jpg", "scene0353_02/00360.jpg" ], "turn_id": 2, "scan_id": "scene0353_02", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the dark gray, three-seater fabric couch\"; \"the large, smooth black blackboard in good condition\"; \"the rectangular wall calendar with a white background\" ", "option": [ "the dark gray, three-seater fabric couch", "the large, smooth black blackboard in good condition", "the rectangular wall calendar with a white background" ], "answer": "the rectangular wall calendar with a white background", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the dark gray, three-seater fabric couch\"; \"the large, smooth black blackboard in good condition\"; \"the rectangular wall calendar with a white background\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the dark gray, three-seater fabric couch', 'the large, smooth black blackboard in good condition', 'the rectangular wall calendar with a white background'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0353_02/00400.jpg", "scene0353_02/00440.jpg", "scene0353_02/00480.jpg", "scene0353_02/00520.jpg", "scene0353_02/00560.jpg" ], "turn_id": 3, "scan_id": "scene0353_02", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the rectangular wall calendar with a white background\" from your position?", "option": [], "answer": "88.7", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the rectangular wall calendar with a white background\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0353_02/00600.jpg", "scene0353_02/00640.jpg", "scene0353_02/00680.jpg", "scene0353_02/00720.jpg", "scene0353_02/00760.jpg" ], "turn_id": 4, "scan_id": "scene0353_02", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the dark gray, three-seater fabric couch\", Y: \"the round green and white ceramic plate\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the dark gray, three-seater fabric couch\", Y: \"the round green and white ceramic plate\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0353_02/00800.jpg", "scene0353_02/00840.jpg", "scene0353_02/00880.jpg", "scene0353_02/00920.jpg", "scene0353_02/00960.jpg" ], "turn_id": 5, "scan_id": "scene0353_02", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the dark gray, three-seater fabric couch\", Y: \"the pair of dark blue clothes\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the dark gray, three-seater fabric couch\", Y: \"the pair of dark blue clothes\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0353_02/01000.jpg", "scene0353_02/01040.jpg", "scene0353_02/01080.jpg", "scene0353_02/01120.jpg", "scene0353_02/01160.jpg" ], "turn_id": 6, "scan_id": "scene0353_02", "origin_question": "Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0353_02/01200.jpg", "scene0353_02/01240.jpg", "scene0353_02/01280.jpg", "scene0353_02/01320.jpg", "scene0353_02/01360.jpg" ], "turn_id": 7, "scan_id": "scene0353_02", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the bright blue medium-sized soft towel\" from your position?", "option": [], "answer": "86.28", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the bright blue medium-sized soft towel\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0353_02/01400.jpg", "scene0353_02/01440.jpg", "scene0353_02/01480.jpg", "scene0353_02/01520.jpg", "scene0353_02/01560.jpg" ], "turn_id": 8, "scan_id": "scene0353_02", "origin_question": "Please recall \"the sleek black flat-screen monitor with a metal frame\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.0", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the sleek black flat-screen monitor with a metal frame\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0353_02/01600.jpg", "scene0353_02/01640.jpg", "scene0353_02/01680.jpg", "scene0353_02/01720.jpg", "scene0353_02/01760.jpg" ], "turn_id": 9, "scan_id": "scene0353_02", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the bright blue medium-sized soft towel\" from your position?", "option": [], "answer": "105.77", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the bright blue medium-sized soft towel\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0353_02/01800.jpg", "scene0353_02/01840.jpg", "scene0353_02/01880.jpg", "scene0353_02/01920.jpg", "scene0353_02/01960.jpg" ], "turn_id": 10, "scan_id": "scene0353_02", "origin_question": "Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the red and black sturdy bag made of nylon\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the red and black sturdy bag made of nylon\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0040_region6/f83d0032ce684933a4d3e8784a9fd0bf_i1_2.jpg", "1mp3d_0040_region6/f83d0032ce684933a4d3e8784a9fd0bf_i2_2.jpg", "1mp3d_0040_region6/cfdd4f2deb074fbdacf231a19b467767_i1_2.jpg", "1mp3d_0040_region6/cfdd4f2deb074fbdacf231a19b467767_i2_2.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0040_region6", "origin_question": "Remember, have you seen any tv(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any tv(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0040_region6/fdf2c8890ea9424ca8e780e96c84c393_i1_2.jpg", "1mp3d_0040_region6/fdf2c8890ea9424ca8e780e96c84c393_i2_2.jpg", "1mp3d_0040_region6/bff73ad3851642d79ce25063c9c5a71e_i1_3.jpg", "1mp3d_0040_region6/bff73ad3851642d79ce25063c9c5a71e_i2_3.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0040_region6", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the black metal machine with red and white elements\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the black metal machine with red and white elements\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "1mp3d_0040_region6/bff73ad3851642d79ce25063c9c5a71e_i1_4.jpg", "1mp3d_0040_region6/bff73ad3851642d79ce25063c9c5a71e_i2_4.jpg", "1mp3d_0040_region6/468dc41d313746d0a4bcda690794d9bb_i1_4.jpg", "1mp3d_0040_region6/468dc41d313746d0a4bcda690794d9bb_i2_4.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0040_region6", "origin_question": "When did you first discover \"the large canvas picture of a red car\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the large canvas picture of a red car\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "1mp3d_0040_region6/ab9ba8a03d1e494190a7ff9ec6cfb677_i1_5.jpg", "1mp3d_0040_region6/ab9ba8a03d1e494190a7ff9ec6cfb677_i2_5.jpg", "1mp3d_0040_region6/724244a1c621451daf60dac20934ce11_i1_1.jpg", "1mp3d_0040_region6/724244a1c621451daf60dac20934ce11_i2_1.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0040_region6", "origin_question": "Please recall \"the sleek black flat-panel TV with rounded corners\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.59", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the sleek black flat-panel TV with rounded corners\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "1mp3d_0040_region6/724244a1c621451daf60dac20934ce11_i1_2.jpg", "1mp3d_0040_region6/724244a1c621451daf60dac20934ce11_i2_2.jpg", "1mp3d_0040_region6/3ee7f6332cf441d8b240b3613ff2f2b3_i1_1.jpg", "1mp3d_0040_region6/3ee7f6332cf441d8b240b3613ff2f2b3_i2_1.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0040_region6", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "137.78", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "1mp3d_0040_region6/271c2fa0c2e24b29b1f1af665cee8dff_i1_0.jpg", "1mp3d_0040_region6/271c2fa0c2e24b29b1f1af665cee8dff_i2_0.jpg", "1mp3d_0040_region6/271c2fa0c2e24b29b1f1af665cee8dff_i1_1.jpg", "1mp3d_0040_region6/271c2fa0c2e24b29b1f1af665cee8dff_i2_1.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0040_region6", "origin_question": "Assuming the direction you are facing at the end of the turn 3 is forward, did you move a certain distance forward or backward from that position?", "option": [ "Forward", "Backward" ], "answer": "Backward", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 3 is forward, did you move a certain distance forward or backward from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Forward', 'Backward'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "1mp3d_0040_region6/2c7c681379454742bcbf70a2391077d5_i1_0.jpg", "1mp3d_0040_region6/2c7c681379454742bcbf70a2391077d5_i2_0.jpg", "1mp3d_0040_region6/18709d89d73e4ff99f5f1fd55ab4fafd_i1_1.jpg", "1mp3d_0040_region6/18709d89d73e4ff99f5f1fd55ab4fafd_i2_1.jpg" ], "turn_id": 7, "scan_id": "1mp3d_0040_region6", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the sleek black flat-panel TV with rounded corners\", Y: \"the large canvas picture of a red car\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "farther away from both X and Y", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the sleek black flat-panel TV with rounded corners\", Y: \"the large canvas picture of a red car\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "1mp3d_0040_region6/2c7c681379454742bcbf70a2391077d5_i1_5.jpg", "1mp3d_0040_region6/2c7c681379454742bcbf70a2391077d5_i2_5.jpg", "1mp3d_0040_region6/07e0b13e8a3c47f68ed9ac46076f129e_i1_5.jpg", "1mp3d_0040_region6/07e0b13e8a3c47f68ed9ac46076f129e_i2_5.jpg" ], "turn_id": 8, "scan_id": "1mp3d_0040_region6", "origin_question": "In which turn were you closest to \"the rectangular dark brown pool table with elegant legs\"? ", "option": [], "answer": "5", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the rectangular dark brown pool table with elegant legs\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0086_00/00000.jpg", "scene0086_00/00040.jpg", "scene0086_00/00080.jpg", "scene0086_00/00120.jpg" ], "turn_id": 1, "scan_id": "scene0086_00", "origin_question": "Remember, have you seen any stool(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any stool(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0086_00/00160.jpg", "scene0086_00/00200.jpg", "scene0086_00/00240.jpg", "scene0086_00/00280.jpg" ], "turn_id": 2, "scan_id": "scene0086_00", "origin_question": "When did you first discover \"the white plastic bottle with a blue pump\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the white plastic bottle with a blue pump\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0086_00/00320.jpg", "scene0086_00/00360.jpg", "scene0086_00/00400.jpg", "scene0086_00/00440.jpg" ], "turn_id": 3, "scan_id": "scene0086_00", "origin_question": "What will be the first-time appearance order of the following categories: bin, door, mirror?", "option": [ "bin, mirror, door", "mirror, door, bin", "bin, door, mirror", "door, bin, mirror" ], "answer": "bin, mirror, door", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bin, door, mirror? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['bin, mirror, door', 'mirror, door, bin', 'bin, door, mirror', 'door, bin, mirror'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0086_00/00480.jpg", "scene0086_00/00520.jpg", "scene0086_00/00560.jpg", "scene0086_00/00600.jpg" ], "turn_id": 4, "scan_id": "scene0086_00", "origin_question": "At the end of which turn were both of \"the small black soap dispenser with a white interior\" and \"the white plastic bottle with a blue pump\" on your left side?", "option": [], "answer": "2", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the small black soap dispenser with a white interior\" and \"the white plastic bottle with a blue pump\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0086_00/00640.jpg", "scene0086_00/00680.jpg", "scene0086_00/00720.jpg", "scene0086_00/00760.jpg" ], "turn_id": 5, "scan_id": "scene0086_00", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the small, green lush-leaved plant in a brown pot\" from your position?", "option": [], "answer": "77.96", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the small, green lush-leaved plant in a brown pot\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0086_00/00800.jpg", "scene0086_00/00840.jpg", "scene0086_00/00880.jpg", "scene0086_00/00920.jpg" ], "turn_id": 6, "scan_id": "scene0086_00", "origin_question": "By the end of the 3 turn, were you getting closer to or farther away from these two objects: X: \"the small, green lush-leaved plant in a brown pot\", Y: \"the small black soap dispenser with a white interior\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "farther away from both X and Y", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 3 turn, were you getting closer to or farther away from these two objects: X: \"the small, green lush-leaved plant in a brown pot\", Y: \"the small black soap dispenser with a white interior\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0086_00/00960.jpg", "scene0086_00/01000.jpg", "scene0086_00/01040.jpg", "scene0086_00/01080.jpg" ], "turn_id": 7, "scan_id": "scene0086_00", "origin_question": "Is the small, green lush-leaved plant in a brown pot to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the small, green lush-leaved plant in a brown pot to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0086_00/01120.jpg", "scene0086_00/01160.jpg", "scene0086_00/01200.jpg", "scene0086_00/01240.jpg" ], "turn_id": 8, "scan_id": "scene0086_00", "origin_question": "Is the small black soap dispenser with a white interior to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the small black soap dispenser with a white interior to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0044_region17/4a1798eff3714f39b04522a285fe3ede_i1_2.jpg", "1mp3d_0044_region17/4a1798eff3714f39b04522a285fe3ede_i2_2.jpg", "1mp3d_0044_region17/5ae7f77201964f5386990b3ce719df94_i1_3.jpg", "1mp3d_0044_region17/5ae7f77201964f5386990b3ce719df94_i2_3.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0044_region17", "origin_question": "", "option": [], "answer": "", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "1mp3d_0044_region17/6167a976f60f4baebdd66477d83f2fd1_i1_2.jpg", "1mp3d_0044_region17/6167a976f60f4baebdd66477d83f2fd1_i2_2.jpg", "1mp3d_0044_region17/5ae7f77201964f5386990b3ce719df94_i1_2.jpg", "1mp3d_0044_region17/5ae7f77201964f5386990b3ce719df94_i2_2.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0044_region17", "origin_question": "At the end of which turn were both of \"the rectangular dark brown wooden counter with a smooth surface\" and \"the smooth, rounded, deep brown stainless steel sink\" on your left side?", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the rectangular dark brown wooden counter with a smooth surface\" and \"the smooth, rounded, deep brown stainless steel sink\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "1mp3d_0044_region17/dfb1f5674ff1434f8e101bcd9d3cfb20_i1_2.jpg", "1mp3d_0044_region17/dfb1f5674ff1434f8e101bcd9d3cfb20_i2_2.jpg", "1mp3d_0044_region17/dfb1f5674ff1434f8e101bcd9d3cfb20_i1_3.jpg", "1mp3d_0044_region17/dfb1f5674ff1434f8e101bcd9d3cfb20_i2_3.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0044_region17", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the mirror\"; \"the glass vase filled with a plant\"; \"the circular stone arch with a smooth cream surface\" ", "option": [ "the mirror", "the glass vase filled with a plant", "the circular stone arch with a smooth cream surface" ], "answer": "the circular stone arch with a smooth cream surface", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the mirror\"; \"the glass vase filled with a plant\"; \"the circular stone arch with a smooth cream surface\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the mirror', 'the glass vase filled with a plant', 'the circular stone arch with a smooth cream surface'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "1mp3d_0044_region17/dfb1f5674ff1434f8e101bcd9d3cfb20_i1_4.jpg", "1mp3d_0044_region17/dfb1f5674ff1434f8e101bcd9d3cfb20_i2_4.jpg", "1mp3d_0044_region17/9b17f6bce0cf41d9ad7eacc6dfeb73e1_i1_4.jpg", "1mp3d_0044_region17/9b17f6bce0cf41d9ad7eacc6dfeb73e1_i2_4.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0044_region17", "origin_question": "Please recall \"the glass vase filled with a plant\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.66", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the glass vase filled with a plant\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0044_region17/9b17f6bce0cf41d9ad7eacc6dfeb73e1_i1_3.jpg", "1mp3d_0044_region17/9b17f6bce0cf41d9ad7eacc6dfeb73e1_i2_3.jpg", "1mp3d_0044_region17/d9bfec8420e341d1911b9de6ba824f0c_i1_5.jpg", "1mp3d_0044_region17/d9bfec8420e341d1911b9de6ba824f0c_i2_5.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0044_region17", "origin_question": "Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the mirror\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the mirror\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "1mp3d_0044_region17/9b17f6bce0cf41d9ad7eacc6dfeb73e1_i1_2.jpg", "1mp3d_0044_region17/9b17f6bce0cf41d9ad7eacc6dfeb73e1_i2_2.jpg", "1mp3d_0044_region17/d9bfec8420e341d1911b9de6ba824f0c_i1_4.jpg", "1mp3d_0044_region17/d9bfec8420e341d1911b9de6ba824f0c_i2_4.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0044_region17", "origin_question": "Which direction is the smooth, rounded, deep brown stainless steel sink in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the smooth, rounded, deep brown stainless steel sink in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0394_01/00000.jpg", "scene0394_01/00040.jpg", "scene0394_01/00080.jpg", "scene0394_01/00120.jpg", "scene0394_01/00160.jpg" ], "turn_id": 1, "scan_id": "scene0394_01", "origin_question": "Remember, have you seen any pool table(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any pool table(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0394_01/00200.jpg", "scene0394_01/00240.jpg", "scene0394_01/00280.jpg", "scene0394_01/00320.jpg", "scene0394_01/00360.jpg" ], "turn_id": 2, "scan_id": "scene0394_01", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the bed\"; \"the ottoman\"; \"the modern lamp with a conical plastic shade\" ", "option": [ "the bed", "the ottoman", "the modern lamp with a conical plastic shade" ], "answer": "the ottoman", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the bed\"; \"the ottoman\"; \"the modern lamp with a conical plastic shade\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the bed', 'the ottoman', 'the modern lamp with a conical plastic shade'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "scene0394_01/00400.jpg", "scene0394_01/00440.jpg", "scene0394_01/00480.jpg", "scene0394_01/00520.jpg", "scene0394_01/00560.jpg" ], "turn_id": 3, "scan_id": "scene0394_01", "origin_question": "Among these three objects(\"the small dark round wooden table with a smooth surface\",\"the white wall-mounted light with a lampshade\",\"the clear glass box with a metallic lid\" ), which one is the farthest from you?", "option": [ "the small dark round wooden table with a smooth surface", "the white wall-mounted light with a lampshade", "the clear glass box with a metallic lid" ], "answer": "the white wall-mounted light with a lampshade", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the small dark round wooden table with a smooth surface\",\"the white wall-mounted light with a lampshade\",\"the clear glass box with a metallic lid\" ), which one is the farthest from you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the small dark round wooden table with a smooth surface', 'the white wall-mounted light with a lampshade', 'the clear glass box with a metallic lid'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0394_01/00600.jpg", "scene0394_01/00640.jpg", "scene0394_01/00680.jpg", "scene0394_01/00720.jpg", "scene0394_01/00760.jpg" ], "turn_id": 4, "scan_id": "scene0394_01", "origin_question": "Which direction is the modern black wooden dresser with multiple drawers in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the modern black wooden dresser with multiple drawers in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0394_01/00800.jpg", "scene0394_01/00840.jpg", "scene0394_01/00880.jpg", "scene0394_01/00920.jpg", "scene0394_01/00960.jpg" ], "turn_id": 5, "scan_id": "scene0394_01", "origin_question": "Which two objects are on the same side of you now? \"the large gray chair with a white cushion\", \"the rectangular white and red paper tag made of stiff paper\", \"the modern black wooden dresser with multiple drawers\". ", "option": [ "the large gray chair with a white cushion, the rectangular white and red paper tag made of stiff paper", "the large gray chair with a white cushion, the modern black wooden dresser with multiple drawers", "the rectangular white and red paper tag made of stiff paper, the modern black wooden dresser with multiple drawers" ], "answer": "the large gray chair with a white cushion, the rectangular white and red paper tag made of stiff paper", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the large gray chair with a white cushion\", \"the rectangular white and red paper tag made of stiff paper\", \"the modern black wooden dresser with multiple drawers\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the large gray chair with a white cushion, the rectangular white and red paper tag made of stiff paper', 'the large gray chair with a white cushion, the modern black wooden dresser with multiple drawers', 'the rectangular white and red paper tag made of stiff paper, the modern black wooden dresser with multiple drawers'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0394_01/01000.jpg", "scene0394_01/01040.jpg", "scene0394_01/01080.jpg", "scene0394_01/01120.jpg", "scene0394_01/01160.jpg" ], "turn_id": 6, "scan_id": "scene0394_01", "origin_question": "Using your orientation at the end of turn 5 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "41.49", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 5 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0394_01/01200.jpg", "scene0394_01/01240.jpg", "scene0394_01/01280.jpg", "scene0394_01/01320.jpg", "scene0394_01/01360.jpg" ], "turn_id": 7, "scan_id": "scene0394_01", "origin_question": "In which turn did you see both \"the small dark round wooden table with a smooth surface\" and \"the clear glass box with a metallic lid\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the small dark round wooden table with a smooth surface\" and \"the clear glass box with a metallic lid\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0394_01/01400.jpg", "scene0394_01/01440.jpg", "scene0394_01/01480.jpg", "scene0394_01/01520.jpg", "scene0394_01/01560.jpg" ], "turn_id": 8, "scan_id": "scene0394_01", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the clear glass box with a metallic lid\" from your position?", "option": [], "answer": "144.6", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the clear glass box with a metallic lid\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0067_region17/6e5f25022c4d4dbf9a7f35bd77cbf296_i1_1.jpg", "1mp3d_0067_region17/6e5f25022c4d4dbf9a7f35bd77cbf296_i2_1.jpg", "1mp3d_0067_region17/49f26e22f7514331b35d76bc1b8bb9f5_i1_2.jpg", "1mp3d_0067_region17/49f26e22f7514331b35d76bc1b8bb9f5_i2_2.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0067_region17", "origin_question": "Remember, have you seen any picture(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any picture(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "1mp3d_0067_region17/6e5f25022c4d4dbf9a7f35bd77cbf296_i1_3.jpg", "1mp3d_0067_region17/6e5f25022c4d4dbf9a7f35bd77cbf296_i2_3.jpg", "1mp3d_0067_region17/6e5f25022c4d4dbf9a7f35bd77cbf296_i1_2.jpg", "1mp3d_0067_region17/6e5f25022c4d4dbf9a7f35bd77cbf296_i2_2.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0067_region17", "origin_question": "When did you first discover \"the small black metal dumbbell weighing 20 pounds\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the small black metal dumbbell weighing 20 pounds\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "1mp3d_0067_region17/d81930c7b37e4f859430d8dc8e0b8f28_i1_0.jpg", "1mp3d_0067_region17/d81930c7b37e4f859430d8dc8e0b8f28_i2_0.jpg", "1mp3d_0067_region17/5dea2611951e4a908ab308cb165eb241_i1_3.jpg", "1mp3d_0067_region17/5dea2611951e4a908ab308cb165eb241_i2_3.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0067_region17", "origin_question": "Is the small black metal dumbbell weighing 20 pounds to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the small black metal dumbbell weighing 20 pounds to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "1mp3d_0067_region17/d81930c7b37e4f859430d8dc8e0b8f28_i1_1.jpg", "1mp3d_0067_region17/d81930c7b37e4f859430d8dc8e0b8f28_i2_1.jpg", "1mp3d_0067_region17/d81930c7b37e4f859430d8dc8e0b8f28_i1_2.jpg", "1mp3d_0067_region17/d81930c7b37e4f859430d8dc8e0b8f28_i2_2.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0067_region17", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "1mp3d_0067_region17/5dea2611951e4a908ab308cb165eb241_i1_5.jpg", "1mp3d_0067_region17/5dea2611951e4a908ab308cb165eb241_i2_5.jpg", "1mp3d_0067_region17/5dea2611951e4a908ab308cb165eb241_i1_4.jpg", "1mp3d_0067_region17/5dea2611951e4a908ab308cb165eb241_i2_4.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0067_region17", "origin_question": "Is the rectangular picture frame with a creamy white color to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the rectangular picture frame with a creamy white color to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "1mp3d_0067_region17/5dea2611951e4a908ab308cb165eb241_i1_0.jpg", "1mp3d_0067_region17/5dea2611951e4a908ab308cb165eb241_i2_0.jpg", "1mp3d_0067_region17/5dea2611951e4a908ab308cb165eb241_i1_1.jpg", "1mp3d_0067_region17/5dea2611951e4a908ab308cb165eb241_i2_1.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0067_region17", "origin_question": "When did you first discover \"the rectangular picture frame with a creamy white color\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the rectangular picture frame with a creamy white color\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0044_region46/9ea4d4f72dab402a8b82585b0e4f0dda_i1_4.jpg", "1mp3d_0044_region46/9ea4d4f72dab402a8b82585b0e4f0dda_i2_4.jpg", "1mp3d_0044_region46/23b4f824dd6a4eab8dac7282a6623257_i1_5.jpg", "1mp3d_0044_region46/23b4f824dd6a4eab8dac7282a6623257_i2_5.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0044_region46", "origin_question": "Remember, have you seen any mouse(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any mouse(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 2 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0044_region46/23b4f824dd6a4eab8dac7282a6623257_i1_4.jpg", "1mp3d_0044_region46/23b4f824dd6a4eab8dac7282a6623257_i2_4.jpg", "1mp3d_0044_region46/9ea4d4f72dab402a8b82585b0e4f0dda_i1_5.jpg", "1mp3d_0044_region46/9ea4d4f72dab402a8b82585b0e4f0dda_i2_5.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0044_region46", "origin_question": "Remember, have you seen any chair(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any chair(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 2 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0454_00/00000.jpg", "scene0454_00/00030.jpg", "scene0454_00/00060.jpg", "scene0454_00/00090.jpg" ], "turn_id": 1, "scan_id": "scene0454_00", "origin_question": "Remember, have you seen any bag(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any bag(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0454_00/00120.jpg", "scene0454_00/00150.jpg", "scene0454_00/00180.jpg", "scene0454_00/00210.jpg" ], "turn_id": 2, "scan_id": "scene0454_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0454_00/00240.jpg", "scene0454_00/00270.jpg", "scene0454_00/00300.jpg", "scene0454_00/00330.jpg" ], "turn_id": 3, "scan_id": "scene0454_00", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0454_00/00360.jpg", "scene0454_00/00390.jpg", "scene0454_00/00420.jpg", "scene0454_00/00450.jpg" ], "turn_id": 4, "scan_id": "scene0454_00", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "96.42", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0454_00/00480.jpg", "scene0454_00/00510.jpg", "scene0454_00/00540.jpg", "scene0454_00/00570.jpg" ], "turn_id": 5, "scan_id": "scene0454_00", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "1.58", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0454_00/00600.jpg", "scene0454_00/00630.jpg", "scene0454_00/00660.jpg", "scene0454_00/00690.jpg" ], "turn_id": 6, "scan_id": "scene0454_00", "origin_question": "When did you first discover \"the rectangular white sign with black text and graphics\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the rectangular white sign with black text and graphics\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0454_00/00720.jpg", "scene0454_00/00750.jpg", "scene0454_00/00780.jpg", "scene0454_00/00810.jpg" ], "turn_id": 7, "scan_id": "scene0454_00", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0685_01/00000.jpg", "scene0685_01/00040.jpg", "scene0685_01/00080.jpg", "scene0685_01/00120.jpg", "scene0685_01/00160.jpg" ], "turn_id": 1, "scan_id": "scene0685_01", "origin_question": "Remember, have you seen any door(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any door(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0685_01/00200.jpg", "scene0685_01/00240.jpg", "scene0685_01/00280.jpg", "scene0685_01/00320.jpg", "scene0685_01/00360.jpg" ], "turn_id": 2, "scan_id": "scene0685_01", "origin_question": "At the end of which turn were both of \"the rectangular, rich brown wooden door\" and \"the switch\" on your left side?", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the rectangular, rich brown wooden door\" and \"the switch\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0685_01/00400.jpg", "scene0685_01/00440.jpg", "scene0685_01/00480.jpg", "scene0685_01/00520.jpg", "scene0685_01/00560.jpg" ], "turn_id": 3, "scan_id": "scene0685_01", "origin_question": "Is the rectangular, rich brown wooden door to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the rectangular, rich brown wooden door to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0685_01/00600.jpg", "scene0685_01/00640.jpg", "scene0685_01/00680.jpg", "scene0685_01/00720.jpg", "scene0685_01/00760.jpg" ], "turn_id": 4, "scan_id": "scene0685_01", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0685_01/00800.jpg", "scene0685_01/00840.jpg", "scene0685_01/00880.jpg", "scene0685_01/00920.jpg", "scene0685_01/00960.jpg" ], "turn_id": 5, "scan_id": "scene0685_01", "origin_question": "In which turn did you see both \"the switch\" and \"the rectangular wooden cabinet with three drawers\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the switch\" and \"the rectangular wooden cabinet with three drawers\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0685_01/01000.jpg", "scene0685_01/01040.jpg", "scene0685_01/01080.jpg", "scene0685_01/01120.jpg", "scene0685_01/01160.jpg" ], "turn_id": 6, "scan_id": "scene0685_01", "origin_question": "What will be the first-time appearance order of the following categories: bin, cabinet, table?", "option": [ "bin, cabinet, table", "cabinet, bin, table", "bin, table, cabinet", "cabinet, table, bin" ], "answer": "bin, cabinet, table", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bin, cabinet, table? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['bin, cabinet, table', 'cabinet, bin, table', 'bin, table, cabinet', 'cabinet, table, bin'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0685_01/01200.jpg", "scene0685_01/01240.jpg", "scene0685_01/01280.jpg", "scene0685_01/01320.jpg", "scene0685_01/01360.jpg" ], "turn_id": 7, "scan_id": "scene0685_01", "origin_question": "Which two objects are on the same side of you now? \"the switch\", \"the large rectangular white board made of sturdy material\", \"the cylindrical blue roll of adhesive tape with a white lining\". ", "option": [ "the switch, the large rectangular white board made of sturdy material", "the switch, the cylindrical blue roll of adhesive tape with a white lining", "the large rectangular white board made of sturdy material, the cylindrical blue roll of adhesive tape with a white lining" ], "answer": "the large rectangular white board made of sturdy material, the cylindrical blue roll of adhesive tape with a white lining", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the switch\", \"the large rectangular white board made of sturdy material\", \"the cylindrical blue roll of adhesive tape with a white lining\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the switch, the large rectangular white board made of sturdy material', 'the switch, the cylindrical blue roll of adhesive tape with a white lining', 'the large rectangular white board made of sturdy material, the cylindrical blue roll of adhesive tape with a white lining'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0685_01/01400.jpg", "scene0685_01/01440.jpg", "scene0685_01/01480.jpg", "scene0685_01/01520.jpg", "scene0685_01/01560.jpg" ], "turn_id": 8, "scan_id": "scene0685_01", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the rectangular stainless steel shelf with a chrome surface\" from your position?", "option": [], "answer": "141.67", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the rectangular stainless steel shelf with a chrome surface\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0685_01/01600.jpg", "scene0685_01/01640.jpg", "scene0685_01/01680.jpg", "scene0685_01/01720.jpg", "scene0685_01/01760.jpg" ], "turn_id": 9, "scan_id": "scene0685_01", "origin_question": "Using your orientation at the end of turn 7 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 7 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0343_00/00000.jpg", "scene0343_00/00040.jpg", "scene0343_00/00080.jpg", "scene0343_00/00120.jpg" ], "turn_id": 1, "scan_id": "scene0343_00", "origin_question": "Remember, have you seen any dress(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any dress(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0343_00/00160.jpg", "scene0343_00/00200.jpg", "scene0343_00/00240.jpg", "scene0343_00/00280.jpg" ], "turn_id": 2, "scan_id": "scene0343_00", "origin_question": "Remember, have you seen any grab bar(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any grab bar(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0343_00/00320.jpg", "scene0343_00/00360.jpg", "scene0343_00/00400.jpg", "scene0343_00/00440.jpg" ], "turn_id": 3, "scan_id": "scene0343_00", "origin_question": "How far is your current position from where you were at the end of the turn 1? (in meters)", "option": [], "answer": "2.57", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 1? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0343_00/00480.jpg", "scene0343_00/00520.jpg", "scene0343_00/00560.jpg", "scene0343_00/00600.jpg" ], "turn_id": 4, "scan_id": "scene0343_00", "origin_question": "What will be the first-time appearance order of the following categories: chair, shelf, table?", "option": [ "chair, table, shelf", "shelf, table, chair", "table, shelf, chair", "shelf, chair, table" ], "answer": "shelf, table, chair", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: chair, shelf, table? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['chair, table, shelf', 'shelf, table, chair', 'table, shelf, chair', 'shelf, chair, table'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0343_00/00640.jpg", "scene0343_00/00680.jpg", "scene0343_00/00720.jpg", "scene0343_00/00760.jpg" ], "turn_id": 5, "scan_id": "scene0343_00", "origin_question": "How far is your current position from where you were at the end of the turn 4? (in meters)", "option": [], "answer": "2.36", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 4? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0343_00/00800.jpg", "scene0343_00/00840.jpg", "scene0343_00/00880.jpg", "scene0343_00/00920.jpg" ], "turn_id": 6, "scan_id": "scene0343_00", "origin_question": "By the end of the 3 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular yellow sign with black text\", Y: \"the rectangular white paper with black text\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to Y and farther away from X", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 3 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular yellow sign with black text\", Y: \"the rectangular white paper with black text\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0343_00/00960.jpg", "scene0343_00/01000.jpg", "scene0343_00/01040.jpg", "scene0343_00/01080.jpg" ], "turn_id": 7, "scan_id": "scene0343_00", "origin_question": "Is the rectangular yellow sign with black text to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the rectangular yellow sign with black text to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0693_01/00000.jpg", "scene0693_01/00040.jpg", "scene0693_01/00080.jpg", "scene0693_01/00120.jpg", "scene0693_01/00160.jpg" ], "turn_id": 1, "scan_id": "scene0693_01", "origin_question": "Remember, have you seen any tissue box(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any tissue box(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0693_01/00200.jpg", "scene0693_01/00240.jpg", "scene0693_01/00280.jpg", "scene0693_01/00320.jpg", "scene0693_01/00360.jpg" ], "turn_id": 2, "scan_id": "scene0693_01", "origin_question": "At the end of which turn were both of \"the vanity\" and \"the sink\" on your left side?", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the vanity\" and \"the sink\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0693_01/00400.jpg", "scene0693_01/00440.jpg", "scene0693_01/00480.jpg", "scene0693_01/00520.jpg", "scene0693_01/00560.jpg" ], "turn_id": 3, "scan_id": "scene0693_01", "origin_question": "Please recall \"the small rectangular brown wooden tissue box\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.02", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the small rectangular brown wooden tissue box\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0693_01/00600.jpg", "scene0693_01/00640.jpg", "scene0693_01/00680.jpg", "scene0693_01/00720.jpg", "scene0693_01/00760.jpg" ], "turn_id": 4, "scan_id": "scene0693_01", "origin_question": "Please recall \"the sink\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.09", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the sink\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0693_01/00800.jpg", "scene0693_01/00840.jpg", "scene0693_01/00880.jpg", "scene0693_01/00920.jpg", "scene0693_01/00960.jpg" ], "turn_id": 5, "scan_id": "scene0693_01", "origin_question": "Which direction is the sink in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the sink in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0693_01/01000.jpg", "scene0693_01/01040.jpg", "scene0693_01/01080.jpg", "scene0693_01/01120.jpg", "scene0693_01/01160.jpg" ], "turn_id": 6, "scan_id": "scene0693_01", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the small white rectangular plastic bin\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the small white rectangular plastic bin\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0693_01/01200.jpg", "scene0693_01/01240.jpg", "scene0693_01/01280.jpg", "scene0693_01/01320.jpg", "scene0693_01/01360.jpg" ], "turn_id": 7, "scan_id": "scene0693_01", "origin_question": "What will be the first-time appearance order of the following categories: bucket, paper, picture?", "option": [ "picture, bucket, paper", "bucket, paper, picture", "bucket, picture, paper", "paper, bucket, picture" ], "answer": "paper, bucket, picture", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bucket, paper, picture? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['picture, bucket, paper', 'bucket, paper, picture', 'bucket, picture, paper', 'paper, bucket, picture'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0693_01/01400.jpg", "scene0693_01/01440.jpg", "scene0693_01/01480.jpg", "scene0693_01/01520.jpg", "scene0693_01/01560.jpg" ], "turn_id": 8, "scan_id": "scene0693_01", "origin_question": "Please recall \"the small rectangular brown wooden tissue box\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.13", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the small rectangular brown wooden tissue box\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0204_02/00000.jpg", "scene0204_02/00030.jpg", "scene0204_02/00060.jpg", "scene0204_02/00090.jpg" ], "turn_id": 1, "scan_id": "scene0204_02", "origin_question": "Remember, have you seen any chair(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any chair(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0204_02/00120.jpg", "scene0204_02/00150.jpg", "scene0204_02/00180.jpg", "scene0204_02/00210.jpg" ], "turn_id": 2, "scan_id": "scene0204_02", "origin_question": "When did you first discover \"the medium brown rectangular wooden cabinet\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the medium brown rectangular wooden cabinet\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0204_02/00240.jpg", "scene0204_02/00270.jpg", "scene0204_02/00300.jpg", "scene0204_02/00330.jpg" ], "turn_id": 3, "scan_id": "scene0204_02", "origin_question": "Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "117.78", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0204_02/00360.jpg", "scene0204_02/00390.jpg", "scene0204_02/00420.jpg", "scene0204_02/00450.jpg" ], "turn_id": 4, "scan_id": "scene0204_02", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0204_02/00480.jpg", "scene0204_02/00510.jpg", "scene0204_02/00540.jpg", "scene0204_02/00570.jpg" ], "turn_id": 5, "scan_id": "scene0204_02", "origin_question": "", "option": [], "answer": "", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0204_02/00660.jpg", "scene0204_02/00690.jpg", "scene0204_02/00720.jpg", "scene0204_02/00750.jpg" ], "turn_id": 6, "scan_id": "scene0204_02", "origin_question": "How far is your current position from where you were at the end of the turn 3? (in meters)", "option": [], "answer": "2.77", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 3? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0044_region28/2c6873dbf7a64181966d931c0c3d0c1e_i1_0.jpg", "1mp3d_0044_region28/2c6873dbf7a64181966d931c0c3d0c1e_i2_0.jpg", "1mp3d_0044_region28/6f865c1511aa4381b7813ab8dbe71c4f_i1_5.jpg", "1mp3d_0044_region28/6f865c1511aa4381b7813ab8dbe71c4f_i2_5.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0044_region28", "origin_question": "Remember, have you seen any cabinet(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any cabinet(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0044_region28/2c6873dbf7a64181966d931c0c3d0c1e_i1_5.jpg", "1mp3d_0044_region28/2c6873dbf7a64181966d931c0c3d0c1e_i2_5.jpg", "1mp3d_0044_region28/6f865c1511aa4381b7813ab8dbe71c4f_i1_4.jpg", "1mp3d_0044_region28/6f865c1511aa4381b7813ab8dbe71c4f_i2_4.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0044_region28", "origin_question": "Remember, have you seen any keyboard(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any keyboard(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "1mp3d_0044_region28/55f496ac52624c34b6f7a83b596c0e85_i1_1.jpg", "1mp3d_0044_region28/55f496ac52624c34b6f7a83b596c0e85_i2_1.jpg", "1mp3d_0044_region28/55f496ac52624c34b6f7a83b596c0e85_i2_2.jpg", "1mp3d_0044_region28/55f496ac52624c34b6f7a83b596c0e85_i1_3.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0044_region28", "origin_question": "At the end of which turn were both of \"the rectangular four-panel window with a white frame\" and \"the large rectangular light-colored marble vanity\" on your right side?", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the rectangular four-panel window with a white frame\" and \"the large rectangular light-colored marble vanity\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0044_region28/55f496ac52624c34b6f7a83b596c0e85_i2_3.jpg", "1mp3d_0044_region28/73c5fd2eb22f42889baee17f32774cca_i1_4.jpg", "1mp3d_0044_region28/73c5fd2eb22f42889baee17f32774cca_i2_4.jpg", "1mp3d_0044_region28/ffeb70b1d3984c4699d5b9dd97bd9cde_i1_5.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0044_region28", "origin_question": "Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the white rectangular curtain with black and gray design\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the white rectangular curtain with black and gray design\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "1mp3d_0044_region28/ffeb70b1d3984c4699d5b9dd97bd9cde_i2_5.jpg", "1mp3d_0044_region28/ffeb70b1d3984c4699d5b9dd97bd9cde_i1_4.jpg", "1mp3d_0044_region28/ffeb70b1d3984c4699d5b9dd97bd9cde_i2_4.jpg", "1mp3d_0044_region28/ffeb70b1d3984c4699d5b9dd97bd9cde_i1_3.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0044_region28", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "1mp3d_0044_region28/ffeb70b1d3984c4699d5b9dd97bd9cde_i2_3.jpg", "1mp3d_0044_region28/ffeb70b1d3984c4699d5b9dd97bd9cde_i1_2.jpg", "1mp3d_0044_region28/ffeb70b1d3984c4699d5b9dd97bd9cde_i2_2.jpg", "1mp3d_0044_region28/73c5fd2eb22f42889baee17f32774cca_i1_5.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0044_region28", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"then oval-shaped white marble countertop\" from your position?", "option": [], "answer": "174.29", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"then oval-shaped white marble countertop\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0044_region21/905ce9bd3b84466b9da061b4cb46e629_i1_2.jpg", "1mp3d_0044_region21/905ce9bd3b84466b9da061b4cb46e629_i2_2.jpg", "1mp3d_0044_region21/ea12dd63895345369d80a91aacb6e137_i1_2.jpg", "1mp3d_0044_region21/ea12dd63895345369d80a91aacb6e137_i2_2.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0044_region21", "origin_question": "Remember, have you seen any curtain(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any curtain(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "1mp3d_0044_region21/ea12dd63895345369d80a91aacb6e137_i1_1.jpg", "1mp3d_0044_region21/ea12dd63895345369d80a91aacb6e137_i2_1.jpg", "1mp3d_0044_region21/905ce9bd3b84466b9da061b4cb46e629_i1_3.jpg", "1mp3d_0044_region21/905ce9bd3b84466b9da061b4cb46e629_i2_3.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0044_region21", "origin_question": "In which turn did you see both \"the cylindrical roll of white toilet paper\" and \"the shiny stainless steel faucet with gold coating\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the cylindrical roll of white toilet paper\" and \"the shiny stainless steel faucet with gold coating\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "1mp3d_0044_region21/4e978c08440c4b12b4d989f80b22225b_i1_4.jpg", "1mp3d_0044_region21/4e978c08440c4b12b4d989f80b22225b_i2_4.jpg", "1mp3d_0044_region21/4e978c08440c4b12b4d989f80b22225b_i1_5.jpg", "1mp3d_0044_region21/4e978c08440c4b12b4d989f80b22225b_i2_5.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0044_region21", "origin_question": "Please recall \"the cylindrical roll of white toilet paper\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.43", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the cylindrical roll of white toilet paper\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0044_region21/4e978c08440c4b12b4d989f80b22225b_i1_0.jpg", "1mp3d_0044_region21/a3e8df8c66f14855815c39cb50282e07_i1_5.jpg", "1mp3d_0044_region21/a3e8df8c66f14855815c39cb50282e07_i2_5.jpg", "1mp3d_0044_region21/31ed82f1e3b54f87a03a9c5429740665_i1_5.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0044_region21", "origin_question": "", "option": [], "answer": "", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "1mp3d_0044_region21/31ed82f1e3b54f87a03a9c5429740665_i2_5.jpg", "1mp3d_0044_region21/e93071d6a4144f53aa9796e3324ed8c3_i1_5.jpg", "1mp3d_0044_region21/e93071d6a4144f53aa9796e3324ed8c3_i2_5.jpg", "1mp3d_0044_region21/31ed82f1e3b54f87a03a9c5429740665_i1_4.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0044_region21", "origin_question": "What will be the first-time appearance order of the following categories: curtain, door, faucet?", "option": [ "curtain, door, faucet", "faucet, door, curtain", "faucet, curtain, door", "door, curtain, faucet" ], "answer": "curtain, door, faucet", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: curtain, door, faucet? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['curtain, door, faucet', 'faucet, door, curtain', 'faucet, curtain, door', 'door, curtain, faucet'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "1mp3d_0044_region21/31ed82f1e3b54f87a03a9c5429740665_i2_4.jpg", "1mp3d_0044_region21/8fab97869cf5427d94ea2bc3c04d2856_i1_5.jpg", "1mp3d_0044_region21/8fab97869cf5427d94ea2bc3c04d2856_i2_5.jpg", "1mp3d_0044_region21/1bc63da6b7cf473abd0675a6b927e56b_i1_0.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0044_region21", "origin_question": "Among these three objects(\"the cylindrical roll of white toilet paper\",\"the white circular porcelain sink of medium size\",\"the cylindrical stainless steel container\" ), which one is the closest to you?", "option": [ "the cylindrical roll of white toilet paper", "the white circular porcelain sink of medium size", "the cylindrical stainless steel container" ], "answer": "the cylindrical roll of white toilet paper", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects(\"the cylindrical roll of white toilet paper\",\"the white circular porcelain sink of medium size\",\"the cylindrical stainless steel container\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the cylindrical roll of white toilet paper', 'the white circular porcelain sink of medium size', 'the cylindrical stainless steel container'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0044_region21/1bc63da6b7cf473abd0675a6b927e56b_i2_0.jpg", "1mp3d_0044_region21/1bc63da6b7cf473abd0675a6b927e56b_i1_5.jpg", "1mp3d_0044_region21/1bc63da6b7cf473abd0675a6b927e56b_i2_5.jpg", "1mp3d_0044_region21/f07a5e07bf3f4e5a8f74705dc211e55a_i1_3.jpg" ], "turn_id": 7, "scan_id": "1mp3d_0044_region21", "origin_question": "Remember, have you seen any bed(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any bed(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "1mp3d_0044_region21/f07a5e07bf3f4e5a8f74705dc211e55a_i2_3.jpg", "1mp3d_0044_region21/505ef208c8be40f283ade476c7340f3f_i1_4.jpg", "1mp3d_0044_region21/505ef208c8be40f283ade476c7340f3f_i2_4.jpg", "1mp3d_0044_region21/de2c508124c34fb79faa373c8da2ffbe_i1_4.jpg" ], "turn_id": 8, "scan_id": "1mp3d_0044_region21", "origin_question": "By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the counter\", Y: \"the cylindrical stainless steel container\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "farther away from both X and Y", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the counter\", Y: \"the cylindrical stainless steel container\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0153_00/00000.jpg", "scene0153_00/00030.jpg", "scene0153_00/00060.jpg", "scene0153_00/00090.jpg" ], "turn_id": 1, "scan_id": "scene0153_00", "origin_question": "Remember, have you seen any toaster(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any toaster(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0153_00/00120.jpg", "scene0153_00/00150.jpg", "scene0153_00/00180.jpg", "scene0153_00/00210.jpg" ], "turn_id": 2, "scan_id": "scene0153_00", "origin_question": "Which direction is the white oval-shaped glossy sink in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the white oval-shaped glossy sink in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0153_00/00240.jpg", "scene0153_00/00270.jpg", "scene0153_00/00300.jpg", "scene0153_00/00330.jpg" ], "turn_id": 3, "scan_id": "scene0153_00", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the toilet\"; \"the simple, light-colored fabric curtain\"; \"the rectangular, white and silver scale\" ", "option": [ "the toilet", "the simple, light-colored fabric curtain", "the rectangular, white and silver scale" ], "answer": "the rectangular, white and silver scale", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the toilet\"; \"the simple, light-colored fabric curtain\"; \"the rectangular, white and silver scale\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the toilet', 'the simple, light-colored fabric curtain', 'the rectangular, white and silver scale'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0153_00/00360.jpg", "scene0153_00/00390.jpg", "scene0153_00/00420.jpg", "scene0153_00/00460.jpg" ], "turn_id": 4, "scan_id": "scene0153_00", "origin_question": "Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the soap dispenser\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the soap dispenser\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0300_01/00000.jpg", "scene0300_01/00040.jpg", "scene0300_01/00080.jpg", "scene0300_01/00120.jpg", "scene0300_01/00160.jpg" ], "turn_id": 1, "scan_id": "scene0300_01", "origin_question": "Remember, how many commode(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many commode(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0300_01/00200.jpg", "scene0300_01/00240.jpg", "scene0300_01/00280.jpg", "scene0300_01/00320.jpg", "scene0300_01/00360.jpg" ], "turn_id": 2, "scan_id": "scene0300_01", "origin_question": "When did you first discover \"the rectangular gray switch made of plastic\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the rectangular gray switch made of plastic\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0300_01/00400.jpg", "scene0300_01/00440.jpg", "scene0300_01/00480.jpg", "scene0300_01/00520.jpg", "scene0300_01/00560.jpg" ], "turn_id": 3, "scan_id": "scene0300_01", "origin_question": "Is the medium-sized deep blue nylon backpack to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the medium-sized deep blue nylon backpack to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0300_01/00600.jpg", "scene0300_01/00640.jpg", "scene0300_01/00680.jpg", "scene0300_01/00720.jpg", "scene0300_01/00760.jpg" ], "turn_id": 4, "scan_id": "scene0300_01", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the window\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the window\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0300_01/00800.jpg", "scene0300_01/00840.jpg", "scene0300_01/00880.jpg", "scene0300_01/00920.jpg", "scene0300_01/00960.jpg" ], "turn_id": 5, "scan_id": "scene0300_01", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0300_01/01000.jpg", "scene0300_01/01040.jpg", "scene0300_01/01080.jpg", "scene0300_01/01120.jpg", "scene0300_01/01160.jpg" ], "turn_id": 6, "scan_id": "scene0300_01", "origin_question": "Which direction is the medium-sized deep blue nylon backpack in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the medium-sized deep blue nylon backpack in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "scene0300_01/01200.jpg", "scene0300_01/01240.jpg", "scene0300_01/01280.jpg", "scene0300_01/01320.jpg", "scene0300_01/01360.jpg" ], "turn_id": 7, "scan_id": "scene0300_01", "origin_question": "Among these three objects(\"the window\",\"the simple rectangular, semi-transparent vertical divider\",\"the rectangular, shiny bin made of plastic\" ), which one is the farthest from you?", "option": [ "the window", "the simple rectangular, semi-transparent vertical divider", "the rectangular, shiny bin made of plastic" ], "answer": "the window", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the window\",\"the simple rectangular, semi-transparent vertical divider\",\"the rectangular, shiny bin made of plastic\" ), which one is the farthest from you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the window', 'the simple rectangular, semi-transparent vertical divider', 'the rectangular, shiny bin made of plastic'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0300_01/01400.jpg", "scene0300_01/01440.jpg", "scene0300_01/01480.jpg", "scene0300_01/01520.jpg", "scene0300_01/01560.jpg" ], "turn_id": 8, "scan_id": "scene0300_01", "origin_question": "At the end of which turn were both of \"the window\" and \"the rectangular gray switch made of plastic\" on your left side?", "option": [], "answer": "4", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the window\" and \"the rectangular gray switch made of plastic\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0269_00/00000.jpg", "scene0269_00/00040.jpg", "scene0269_00/00080.jpg", "scene0269_00/00120.jpg" ], "turn_id": 1, "scan_id": "scene0269_00", "origin_question": "Remember, have you seen any table(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any table(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0269_00/00160.jpg", "scene0269_00/00200.jpg", "scene0269_00/00240.jpg", "scene0269_00/00280.jpg" ], "turn_id": 2, "scan_id": "scene0269_00", "origin_question": "Remember, have you seen any couch(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any couch(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0269_00/00320.jpg", "scene0269_00/00360.jpg", "scene0269_00/00400.jpg", "scene0269_00/00440.jpg" ], "turn_id": 3, "scan_id": "scene0269_00", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "2.7", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0269_00/00480.jpg", "scene0269_00/00520.jpg", "scene0269_00/00560.jpg", "scene0269_00/00600.jpg" ], "turn_id": 4, "scan_id": "scene0269_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0269_00/00640.jpg", "scene0269_00/00680.jpg", "scene0269_00/00720.jpg", "scene0269_00/00760.jpg" ], "turn_id": 5, "scan_id": "scene0269_00", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the round coffee table with a light wooden surface\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the round coffee table with a light wooden surface\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0269_00/00800.jpg", "scene0269_00/00840.jpg", "scene0269_00/00880.jpg", "scene0269_00/00920.jpg" ], "turn_id": 6, "scan_id": "scene0269_00", "origin_question": "Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "111.92", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0040_region36/a754489b27764d63a334de800a3a6cef_i1_3.jpg", "1mp3d_0040_region36/a754489b27764d63a334de800a3a6cef_i2_3.jpg", "1mp3d_0040_region36/b637b622335f44869ad8cc595cda3d9c_i1_2.jpg", "1mp3d_0040_region36/b637b622335f44869ad8cc595cda3d9c_i2_2.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0040_region36", "origin_question": "Remember, have you seen any window(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any window(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "1mp3d_0040_region36/a754489b27764d63a334de800a3a6cef_i1_1.jpg", "1mp3d_0040_region36/a754489b27764d63a334de800a3a6cef_i2_1.jpg", "1mp3d_0040_region36/b637b622335f44869ad8cc595cda3d9c_i1_1.jpg", "1mp3d_0040_region36/b637b622335f44869ad8cc595cda3d9c_i2_1.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0040_region36", "origin_question": "Remember, how many toy(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many toy(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0040_region36/b637b622335f44869ad8cc595cda3d9c_i1_0.jpg", "1mp3d_0040_region36/b637b622335f44869ad8cc595cda3d9c_i2_0.jpg", "1mp3d_0040_region36/a754489b27764d63a334de800a3a6cef_i1_0.jpg", "1mp3d_0040_region36/a754489b27764d63a334de800a3a6cef_i2_0.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0040_region36", "origin_question": "", "option": [], "answer": "", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0040_region36/59267d1a855e425e85e00a7cad6464b1_i2_5.jpg", "1mp3d_0040_region36/4bfec9453c88441bb88689d0fce2600d_i1_4.jpg", "1mp3d_0040_region36/4bfec9453c88441bb88689d0fce2600d_i2_4.jpg", "1mp3d_0040_region36/4bfec9453c88441bb88689d0fce2600d_i1_5.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0040_region36", "origin_question": "", "option": [], "answer": "", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "1mp3d_0040_region36/4bfec9453c88441bb88689d0fce2600d_i2_5.jpg", "1mp3d_0040_region36/59267d1a855e425e85e00a7cad6464b1_i1_0.jpg", "1mp3d_0040_region36/59267d1a855e425e85e00a7cad6464b1_i2_0.jpg", "1mp3d_0040_region36/59267d1a855e425e85e00a7cad6464b1_i1_1.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0040_region36", "origin_question": "Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0057_region14/63a2a4cb2cba477bba0ad596573efbf9_i2_2.jpg", "1mp3d_0057_region14/b1eaadc1fb0b4bb7b5f4e7afeb0fc9da_i1_1.jpg", "1mp3d_0057_region14/73bf0888b74542d5bea2da959773ae09_i1_0.jpg", "1mp3d_0057_region14/73bf0888b74542d5bea2da959773ae09_i2_0.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0057_region14", "origin_question": "Remember, have you seen any stairs(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any stairs(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 3 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0057_region14/63a2a4cb2cba477bba0ad596573efbf9_i1_5.jpg", "1mp3d_0057_region14/63a2a4cb2cba477bba0ad596573efbf9_i2_5.jpg", "1mp3d_0057_region14/76dd8d8a97814d16b3f5401888e7c1e1_i1_5.jpg", "1mp3d_0057_region14/76dd8d8a97814d16b3f5401888e7c1e1_i2_5.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0057_region14", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 3 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "1mp3d_0057_region14/f2bd01837bb04271a6ec99c31ea850ba_i2_1.jpg", "1mp3d_0057_region14/f2bd01837bb04271a6ec99c31ea850ba_i2_2.jpg", "1mp3d_0057_region14/76dd8d8a97814d16b3f5401888e7c1e1_i1_0.jpg", "1mp3d_0057_region14/76dd8d8a97814d16b3f5401888e7c1e1_i2_0.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0057_region14", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "59.26", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 3 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0025_02/00000.jpg", "scene0025_02/00040.jpg", "scene0025_02/00080.jpg", "scene0025_02/00120.jpg" ], "turn_id": 1, "scan_id": "scene0025_02", "origin_question": "Remember, have you seen any shelf(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any shelf(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0025_02/00160.jpg", "scene0025_02/00200.jpg", "scene0025_02/00240.jpg", "scene0025_02/00280.jpg" ], "turn_id": 2, "scan_id": "scene0025_02", "origin_question": "In which turn did you see both \"the rectangular blackboard with a white background\" and \"the medium-sized black circular fan with multiple blades\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the rectangular blackboard with a white background\" and \"the medium-sized black circular fan with multiple blades\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0025_02/00320.jpg", "scene0025_02/00360.jpg", "scene0025_02/00400.jpg", "scene0025_02/00440.jpg" ], "turn_id": 3, "scan_id": "scene0025_02", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the rectangular, light-colored wooden windowsill\"; \"the vertical light-colored blind\"; \"the standard-sized dark hardcover book made of paper\" ", "option": [ "the rectangular, light-colored wooden windowsill", "the vertical light-colored blind", "the standard-sized dark hardcover book made of paper" ], "answer": "the standard-sized dark hardcover book made of paper", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the rectangular, light-colored wooden windowsill\"; \"the vertical light-colored blind\"; \"the standard-sized dark hardcover book made of paper\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular, light-colored wooden windowsill', 'the vertical light-colored blind', 'the standard-sized dark hardcover book made of paper'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0025_02/00480.jpg", "scene0025_02/00520.jpg", "scene0025_02/00560.jpg", "scene0025_02/00600.jpg" ], "turn_id": 4, "scan_id": "scene0025_02", "origin_question": "Please recall \"the vertical light-colored blind\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.44", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the vertical light-colored blind\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0025_02/00640.jpg", "scene0025_02/00680.jpg", "scene0025_02/00720.jpg", "scene0025_02/00760.jpg" ], "turn_id": 5, "scan_id": "scene0025_02", "origin_question": "What will be the first-time appearance order of the following categories: chair, couch, telephone?", "option": [ "telephone, chair, couch", "telephone, couch, chair", "couch, chair, telephone", "couch, telephone, chair" ], "answer": "couch, chair, telephone", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: chair, couch, telephone? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['telephone, chair, couch', 'telephone, couch, chair', 'couch, chair, telephone', 'couch, telephone, chair'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0025_02/00800.jpg", "scene0025_02/00840.jpg", "scene0025_02/00880.jpg", "scene0025_02/00920.jpg" ], "turn_id": 6, "scan_id": "scene0025_02", "origin_question": "Is the black rectangular chair with a curved backrest to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the black rectangular chair with a curved backrest to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0025_02/00960.jpg", "scene0025_02/01000.jpg", "scene0025_02/01040.jpg", "scene0025_02/01080.jpg" ], "turn_id": 7, "scan_id": "scene0025_02", "origin_question": "Using your orientation at the end of turn 6 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "90.23", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 6 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0025_02/01120.jpg", "scene0025_02/01160.jpg", "scene0025_02/01200.jpg", "scene0025_02/01240.jpg" ], "turn_id": 8, "scan_id": "scene0025_02", "origin_question": "What will be the first-time appearance order of the following categories: blackboard, pillow, window?", "option": [ "blackboard, window, pillow", "pillow, window, blackboard", "window, pillow, blackboard", "pillow, blackboard, window" ], "answer": "blackboard, window, pillow", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: blackboard, pillow, window? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['blackboard, window, pillow', 'pillow, window, blackboard', 'window, pillow, blackboard', 'pillow, blackboard, window'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0025_02/01280.jpg", "scene0025_02/01320.jpg", "scene0025_02/01360.jpg", "scene0025_02/01400.jpg" ], "turn_id": 9, "scan_id": "scene0025_02", "origin_question": "What will be the first-time appearance order of the following categories: cabinet, monitor, telephone?", "option": [ "monitor, telephone, cabinet", "cabinet, monitor, telephone", "cabinet, telephone, monitor", "telephone, monitor, cabinet" ], "answer": "cabinet, monitor, telephone", "user_message": "For the 9 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: cabinet, monitor, telephone? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['monitor, telephone, cabinet', 'cabinet, monitor, telephone', 'cabinet, telephone, monitor', 'telephone, monitor, cabinet'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0678_01/00000.jpg", "scene0678_01/00040.jpg", "scene0678_01/00080.jpg", "scene0678_01/00120.jpg", "scene0678_01/00160.jpg" ], "turn_id": 1, "scan_id": "scene0678_01", "origin_question": "Remember, how many picture(s) have you seen so far? ", "option": [], "answer": "2", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many picture(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0678_01/00200.jpg", "scene0678_01/00240.jpg", "scene0678_01/00280.jpg", "scene0678_01/00320.jpg", "scene0678_01/00360.jpg" ], "turn_id": 2, "scan_id": "scene0678_01", "origin_question": "Please recall \"the small white rectangular plastic basket with curved edges\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "3.57", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the small white rectangular plastic basket with curved edges\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0678_01/00400.jpg", "scene0678_01/00440.jpg", "scene0678_01/00480.jpg", "scene0678_01/00520.jpg", "scene0678_01/00560.jpg" ], "turn_id": 3, "scan_id": "scene0678_01", "origin_question": "In which turn did you see both \"the cylindrical white plastic fan\" and \"the white rectangular cotton towel, neatly folded\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the cylindrical white plastic fan\" and \"the white rectangular cotton towel, neatly folded\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0678_01/00600.jpg", "scene0678_01/00640.jpg", "scene0678_01/00680.jpg", "scene0678_01/00720.jpg", "scene0678_01/00760.jpg" ], "turn_id": 4, "scan_id": "scene0678_01", "origin_question": "Which direction is the cylindrical white plastic fan in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the cylindrical white plastic fan in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0678_01/00800.jpg", "scene0678_01/00840.jpg", "scene0678_01/00880.jpg", "scene0678_01/00920.jpg", "scene0678_01/00960.jpg" ], "turn_id": 5, "scan_id": "scene0678_01", "origin_question": "Please recall \"the durable stainless steel sink with a curved basin\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "3.0", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the durable stainless steel sink with a curved basin\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0678_01/01000.jpg", "scene0678_01/01040.jpg", "scene0678_01/01080.jpg", "scene0678_01/01120.jpg", "scene0678_01/01160.jpg" ], "turn_id": 6, "scan_id": "scene0678_01", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0678_01/01200.jpg", "scene0678_01/01240.jpg", "scene0678_01/01280.jpg", "scene0678_01/01320.jpg", "scene0678_01/01360.jpg" ], "turn_id": 7, "scan_id": "scene0678_01", "origin_question": "Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the rectangular red metal fire extinguisher with a green component\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the rectangular red metal fire extinguisher with a green component\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "scene0678_01/01400.jpg", "scene0678_01/01440.jpg", "scene0678_01/01480.jpg", "scene0678_01/01520.jpg", "scene0678_01/01560.jpg" ], "turn_id": 8, "scan_id": "scene0678_01", "origin_question": "Among these three objects(\"the white rectangular cotton towel, neatly folded\",\"the durable stainless steel sink with a curved basin\",\"the rectangular red metal fire extinguisher with a green component\" ), which one is the closest to you?", "option": [ "the white rectangular cotton towel, neatly folded", "the durable stainless steel sink with a curved basin", "the rectangular red metal fire extinguisher with a green component" ], "answer": "the white rectangular cotton towel, neatly folded", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the white rectangular cotton towel, neatly folded\",\"the durable stainless steel sink with a curved basin\",\"the rectangular red metal fire extinguisher with a green component\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the white rectangular cotton towel, neatly folded', 'the durable stainless steel sink with a curved basin', 'the rectangular red metal fire extinguisher with a green component'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0678_01/01600.jpg", "scene0678_01/01640.jpg", "scene0678_01/01680.jpg", "scene0678_01/01720.jpg", "scene0678_01/01760.jpg" ], "turn_id": 9, "scan_id": "scene0678_01", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the durable stainless steel sink with a curved basin\" from your position?", "option": [], "answer": "70.23", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the durable stainless steel sink with a curved basin\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0678_01/01800.jpg", "scene0678_01/01840.jpg", "scene0678_01/01880.jpg", "scene0678_01/01920.jpg", "scene0678_01/01960.jpg" ], "turn_id": 10, "scan_id": "scene0678_01", "origin_question": "By the end of the 6 turn, were you getting closer to or farther away from these two objects: X: \"the cylindrical white plastic fan\", Y: \"the durable stainless steel sink with a curved basin\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 6 turn, were you getting closer to or farther away from these two objects: X: \"the cylindrical white plastic fan\", Y: \"the durable stainless steel sink with a curved basin\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0644_00/00000.jpg", "scene0644_00/00040.jpg", "scene0644_00/00080.jpg", "scene0644_00/00120.jpg", "scene0644_00/00160.jpg" ], "turn_id": 1, "scan_id": "scene0644_00", "origin_question": "Remember, have you seen any blackboard(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any blackboard(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0644_00/00200.jpg", "scene0644_00/00240.jpg", "scene0644_00/00280.jpg", "scene0644_00/00320.jpg", "scene0644_00/00360.jpg" ], "turn_id": 2, "scan_id": "scene0644_00", "origin_question": "When did you first discover \"the large vertical sliding window with a dark frame\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the large vertical sliding window with a dark frame\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0644_00/00400.jpg", "scene0644_00/00440.jpg", "scene0644_00/00480.jpg", "scene0644_00/00520.jpg", "scene0644_00/00560.jpg" ], "turn_id": 3, "scan_id": "scene0644_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0644_00/00600.jpg", "scene0644_00/00640.jpg", "scene0644_00/00680.jpg", "scene0644_00/00720.jpg", "scene0644_00/00760.jpg" ], "turn_id": 4, "scan_id": "scene0644_00", "origin_question": "Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the rectangular light-colored wooden shelf\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the rectangular light-colored wooden shelf\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0644_00/00800.jpg", "scene0644_00/00840.jpg", "scene0644_00/00880.jpg", "scene0644_00/00920.jpg", "scene0644_00/00960.jpg" ], "turn_id": 5, "scan_id": "scene0644_00", "origin_question": "How far is your current position from where you were at the end of the turn 4? (in meters)", "option": [], "answer": "1.41", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 4? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0644_00/01000.jpg", "scene0644_00/01040.jpg", "scene0644_00/01080.jpg", "scene0644_00/01120.jpg", "scene0644_00/01160.jpg" ], "turn_id": 6, "scan_id": "scene0644_00", "origin_question": "Please recall \"the medium-sized dark cylindrical bottle\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "3.65", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the medium-sized dark cylindrical bottle\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0644_00/01200.jpg", "scene0644_00/01240.jpg", "scene0644_00/01280.jpg", "scene0644_00/01320.jpg", "scene0644_00/01360.jpg" ], "turn_id": 7, "scan_id": "scene0644_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "scene0644_00/01400.jpg", "scene0644_00/01440.jpg", "scene0644_00/01480.jpg", "scene0644_00/01520.jpg", "scene0644_00/01560.jpg" ], "turn_id": 8, "scan_id": "scene0644_00", "origin_question": "Assuming the direction you are facing at the end of the turn 3 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Right", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 3 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0644_00/01600.jpg", "scene0644_00/01640.jpg", "scene0644_00/01680.jpg", "scene0644_00/01720.jpg", "scene0644_00/01760.jpg" ], "turn_id": 9, "scan_id": "scene0644_00", "origin_question": "How far is your current position from where you were at the end of the turn 4? (in meters)", "option": [], "answer": "3.2", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 4? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0644_00/01800.jpg", "scene0644_00/01840.jpg", "scene0644_00/01880.jpg", "scene0644_00/01920.jpg", "scene0644_00/01960.jpg" ], "turn_id": 10, "scan_id": "scene0644_00", "origin_question": "Please recall \"the small cylindrical white plastic bin\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "3.94", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the small cylindrical white plastic bin\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0067_region21/a1ba2d21084a41339e1a52c89ba24f94_i1_0.jpg", "1mp3d_0067_region21/a1ba2d21084a41339e1a52c89ba24f94_i2_0.jpg", "1mp3d_0067_region21/fae83673fc694cd9a18c215ce6d92c58_i1_4.jpg", "1mp3d_0067_region21/fae83673fc694cd9a18c215ce6d92c58_i2_4.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0067_region21", "origin_question": "Remember, have you seen any bed(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any bed(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "1mp3d_0067_region21/dfc786743984401ca9680410c66df136_i1_0.jpg", "1mp3d_0067_region21/dfc786743984401ca9680410c66df136_i2_0.jpg", "1mp3d_0067_region21/7a8fc0425e0c40a69fb216c5345e157c_i1_3.jpg", "1mp3d_0067_region21/7a8fc0425e0c40a69fb216c5345e157c_i2_3.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0067_region21", "origin_question": "Remember, how many tissue(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many tissue(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "1mp3d_0067_region21/7a8fc0425e0c40a69fb216c5345e157c_i1_4.jpg", "1mp3d_0067_region21/7a8fc0425e0c40a69fb216c5345e157c_i2_4.jpg", "1mp3d_0067_region21/dfc786743984401ca9680410c66df136_i1_1.jpg", "1mp3d_0067_region21/dfc786743984401ca9680410c66df136_i2_1.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0067_region21", "origin_question": "Which direction is the small, wooden rack with black straps in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the small, wooden rack with black straps in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "1mp3d_0067_region21/fae83673fc694cd9a18c215ce6d92c58_i1_5.jpg", "1mp3d_0067_region21/fae83673fc694cd9a18c215ce6d92c58_i2_5.jpg", "1mp3d_0067_region21/a1ba2d21084a41339e1a52c89ba24f94_i1_1.jpg", "1mp3d_0067_region21/a1ba2d21084a41339e1a52c89ba24f94_i2_1.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0067_region21", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the small, wooden rack with black straps\" from your position?", "option": [], "answer": "167.35", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the small, wooden rack with black straps\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "1mp3d_0067_region21/e8527069c00e4d929f5eece0d766ebd8_i1_1.jpg", "1mp3d_0067_region21/e8527069c00e4d929f5eece0d766ebd8_i2_1.jpg", "1mp3d_0067_region21/e8527069c00e4d929f5eece0d766ebd8_i1_2.jpg", "1mp3d_0067_region21/e8527069c00e4d929f5eece0d766ebd8_i2_2.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0067_region21", "origin_question": "Which two objects are on the same side of you now? \"the large rectangular window with a deep brown frame\", \"the cylindrical, glossy black rod\", \"the small, wooden rack with black straps\". ", "option": [ "the large rectangular window with a deep brown frame, the cylindrical, glossy black rod", "the large rectangular window with a deep brown frame, the small, wooden rack with black straps", "the cylindrical, glossy black rod, the small, wooden rack with black straps" ], "answer": "the large rectangular window with a deep brown frame, the cylindrical, glossy black rod", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the large rectangular window with a deep brown frame\", \"the cylindrical, glossy black rod\", \"the small, wooden rack with black straps\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the large rectangular window with a deep brown frame, the cylindrical, glossy black rod', 'the large rectangular window with a deep brown frame, the small, wooden rack with black straps', 'the cylindrical, glossy black rod, the small, wooden rack with black straps'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "1mp3d_0067_region21/a9ec9760708f4771b657e254d9ab916e_i1_1.jpg", "1mp3d_0067_region21/a9ec9760708f4771b657e254d9ab916e_i2_1.jpg", "1mp3d_0067_region21/26e495c137994f48835fdf0c5450a681_i1_3.jpg", "1mp3d_0067_region21/26e495c137994f48835fdf0c5450a681_i2_3.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0067_region21", "origin_question": "Among these three objects(\"the large rectangular window with a deep brown frame\",\"the small, wooden rack with black straps\",\"the cylindrical, glossy black rod\" ), which one is the closest to you?", "option": [ "the large rectangular window with a deep brown frame", "the small, wooden rack with black straps", "the cylindrical, glossy black rod" ], "answer": "the large rectangular window with a deep brown frame", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects(\"the large rectangular window with a deep brown frame\",\"the small, wooden rack with black straps\",\"the cylindrical, glossy black rod\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the large rectangular window with a deep brown frame', 'the small, wooden rack with black straps', 'the cylindrical, glossy black rod'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "1mp3d_0067_region21/a9ec9760708f4771b657e254d9ab916e_i1_2.jpg", "1mp3d_0067_region21/a9ec9760708f4771b657e254d9ab916e_i2_2.jpg", "1mp3d_0067_region21/26e495c137994f48835fdf0c5450a681_i1_4.jpg", "1mp3d_0067_region21/26e495c137994f48835fdf0c5450a681_i2_4.jpg" ], "turn_id": 7, "scan_id": "1mp3d_0067_region21", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the rectangular brown wooden door\"; \"the bed\"; \"the large, dark brown leather box with soft fabric lining\" ", "option": [ "the rectangular brown wooden door", "the bed", "the large, dark brown leather box with soft fabric lining" ], "answer": "the large, dark brown leather box with soft fabric lining", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the rectangular brown wooden door\"; \"the bed\"; \"the large, dark brown leather box with soft fabric lining\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular brown wooden door', 'the bed', 'the large, dark brown leather box with soft fabric lining'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0074_region14/8b29770bf446475b814d254d9b2670f6_i1_0.jpg", "1mp3d_0074_region14/8b29770bf446475b814d254d9b2670f6_i2_0.jpg", "1mp3d_0074_region14/17d80d20c75b45e89949666e40f8df7e_i1_3.jpg", "1mp3d_0074_region14/17d80d20c75b45e89949666e40f8df7e_i2_3.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0074_region14", "origin_question": "Remember, have you seen any pot(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any pot(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 2 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "1mp3d_0074_region14/8b29770bf446475b814d254d9b2670f6_i1_5.jpg", "1mp3d_0074_region14/8b29770bf446475b814d254d9b2670f6_i2_5.jpg", "1mp3d_0074_region14/8b29770bf446475b814d254d9b2670f6_i1_4.jpg", "1mp3d_0074_region14/8b29770bf446475b814d254d9b2670f6_i2_4.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0074_region14", "origin_question": "At the end of which turn were both of \"the countertop\" and \"the huge rectangular glossy mirror\" on your left side?", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the countertop\" and \"the huge rectangular glossy mirror\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 2 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0467_00/00000.jpg", "scene0467_00/00040.jpg", "scene0467_00/00080.jpg", "scene0467_00/00120.jpg", "scene0467_00/00160.jpg" ], "turn_id": 1, "scan_id": "scene0467_00", "origin_question": "Remember, have you seen any couch(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any couch(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0467_00/00200.jpg", "scene0467_00/00240.jpg", "scene0467_00/00280.jpg", "scene0467_00/00320.jpg", "scene0467_00/00360.jpg" ], "turn_id": 2, "scan_id": "scene0467_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0467_00/00400.jpg", "scene0467_00/00440.jpg", "scene0467_00/00480.jpg", "scene0467_00/00520.jpg", "scene0467_00/00560.jpg" ], "turn_id": 3, "scan_id": "scene0467_00", "origin_question": "In which turn did you see both \"the transparent round glass cup with a minimalistic design\" and \"the flat-panel TV with a sleek black exterior\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the transparent round glass cup with a minimalistic design\" and \"the flat-panel TV with a sleek black exterior\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "scene0467_00/00600.jpg", "scene0467_00/00640.jpg", "scene0467_00/00680.jpg", "scene0467_00/00720.jpg", "scene0467_00/00760.jpg" ], "turn_id": 4, "scan_id": "scene0467_00", "origin_question": "Among these three objects(\"the curtain\",\"the gray hardcover book in good condition\",\"the paper\" ), which one is the closest to you?", "option": [ "the curtain", "the gray hardcover book in good condition", "the paper" ], "answer": "the curtain", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the curtain\",\"the gray hardcover book in good condition\",\"the paper\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the curtain', 'the gray hardcover book in good condition', 'the paper'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0467_00/00800.jpg", "scene0467_00/00840.jpg", "scene0467_00/00880.jpg", "scene0467_00/00920.jpg", "scene0467_00/00960.jpg" ], "turn_id": 5, "scan_id": "scene0467_00", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the gray hardcover book in good condition\" from your position?", "option": [], "answer": "145.17", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the gray hardcover book in good condition\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0467_00/01000.jpg", "scene0467_00/01040.jpg", "scene0467_00/01080.jpg", "scene0467_00/01120.jpg", "scene0467_00/01160.jpg" ], "turn_id": 6, "scan_id": "scene0467_00", "origin_question": "Please recall \"the curtain\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.68", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the curtain\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0467_00/01200.jpg", "scene0467_00/01240.jpg", "scene0467_00/01280.jpg", "scene0467_00/01320.jpg", "scene0467_00/01360.jpg" ], "turn_id": 7, "scan_id": "scene0467_00", "origin_question": "In which turn were you closest to \"the transparent round glass cup with a minimalistic design\"? ", "option": [], "answer": "5", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the transparent round glass cup with a minimalistic design\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0467_00/01400.jpg", "scene0467_00/01440.jpg", "scene0467_00/01480.jpg", "scene0467_00/01520.jpg", "scene0467_00/01560.jpg" ], "turn_id": 8, "scan_id": "scene0467_00", "origin_question": "At the end of which turn were both of \"the gray hardcover book in good condition\" and \"the flat-panel TV with a sleek black exterior\" on your right side?", "option": [], "answer": "8", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the gray hardcover book in good condition\" and \"the flat-panel TV with a sleek black exterior\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0467_00/01600.jpg", "scene0467_00/01640.jpg", "scene0467_00/01680.jpg", "scene0467_00/01720.jpg", "scene0467_00/01760.jpg" ], "turn_id": 9, "scan_id": "scene0467_00", "origin_question": "When did you first discover \"the creamy soft fabric couch with backrest and armrests\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the creamy soft fabric couch with backrest and armrests\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0697_03/00000.jpg", "scene0697_03/00040.jpg", "scene0697_03/00080.jpg", "scene0697_03/00120.jpg", "scene0697_03/00160.jpg" ], "turn_id": 1, "scan_id": "scene0697_03", "origin_question": "Remember, have you seen any dresser(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any dresser(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0697_03/00200.jpg", "scene0697_03/00240.jpg", "scene0697_03/00280.jpg", "scene0697_03/00320.jpg", "scene0697_03/00360.jpg" ], "turn_id": 2, "scan_id": "scene0697_03", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the medium white rectangular storage box made of plastic\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the medium white rectangular storage box made of plastic\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0697_03/00400.jpg", "scene0697_03/00440.jpg", "scene0697_03/00480.jpg", "scene0697_03/00520.jpg", "scene0697_03/00560.jpg" ], "turn_id": 3, "scan_id": "scene0697_03", "origin_question": "When did you first discover \"the small round silver-gray clock with a white face\"? (index of the turn)", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the small round silver-gray clock with a white face\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "scene0697_03/00600.jpg", "scene0697_03/00640.jpg", "scene0697_03/00680.jpg", "scene0697_03/00720.jpg", "scene0697_03/00760.jpg" ], "turn_id": 4, "scan_id": "scene0697_03", "origin_question": "Among these three objects(\"the flat rectangular white board\",\"the toy\",\"the rectangular paper tissue box with purple patterns\" ), which one is the closest to you?", "option": [ "the flat rectangular white board", "the toy", "the rectangular paper tissue box with purple patterns" ], "answer": "the flat rectangular white board", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the flat rectangular white board\",\"the toy\",\"the rectangular paper tissue box with purple patterns\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the flat rectangular white board', 'the toy', 'the rectangular paper tissue box with purple patterns'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0697_03/00800.jpg", "scene0697_03/00840.jpg", "scene0697_03/00880.jpg", "scene0697_03/00920.jpg", "scene0697_03/00960.jpg" ], "turn_id": 5, "scan_id": "scene0697_03", "origin_question": "Is the small green plastic cylindrical bin to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the small green plastic cylindrical bin to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0697_03/01000.jpg", "scene0697_03/01040.jpg", "scene0697_03/01080.jpg", "scene0697_03/01120.jpg", "scene0697_03/01160.jpg" ], "turn_id": 6, "scan_id": "scene0697_03", "origin_question": "When did you first discover \"the flat rectangular white board\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the flat rectangular white board\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0697_03/01200.jpg", "scene0697_03/01240.jpg", "scene0697_03/01280.jpg", "scene0697_03/01320.jpg", "scene0697_03/01360.jpg" ], "turn_id": 7, "scan_id": "scene0697_03", "origin_question": "When did you first discover \"the small green plastic cylindrical bin\"? (index of the turn)", "option": [], "answer": "4", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the small green plastic cylindrical bin\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0697_03/01400.jpg", "scene0697_03/01440.jpg", "scene0697_03/01480.jpg", "scene0697_03/01520.jpg", "scene0697_03/01560.jpg" ], "turn_id": 8, "scan_id": "scene0697_03", "origin_question": "Which direction is the black rectangular frame with five smaller photos in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the black rectangular frame with five smaller photos in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0697_03/01600.jpg", "scene0697_03/01640.jpg", "scene0697_03/01680.jpg", "scene0697_03/01720.jpg", "scene0697_03/01760.jpg" ], "turn_id": 9, "scan_id": "scene0697_03", "origin_question": "At the end of which turn were both of \"the toy\" and \"the small round silver-gray clock with a white face\" on your left side?", "option": [], "answer": "3", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the toy\" and \"the small round silver-gray clock with a white face\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0697_03/01800.jpg", "scene0697_03/01840.jpg", "scene0697_03/01880.jpg", "scene0697_03/01920.jpg", "scene0697_03/01960.jpg" ], "turn_id": 10, "scan_id": "scene0697_03", "origin_question": "Compared to your position at the end of the 5 turn, are you now closer or farther away from \"the small green plastic cylindrical bin\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 5 turn, are you now closer or farther away from \"the small green plastic cylindrical bin\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0675_00/00000.jpg", "scene0675_00/00030.jpg", "scene0675_00/00060.jpg", "scene0675_00/00090.jpg" ], "turn_id": 1, "scan_id": "scene0675_00", "origin_question": "Remember, how many shelf(s) have you seen so far? ", "option": [], "answer": "2", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many shelf(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0675_00/00120.jpg", "scene0675_00/00150.jpg", "scene0675_00/00180.jpg", "scene0675_00/00210.jpg" ], "turn_id": 2, "scan_id": "scene0675_00", "origin_question": "Please recall \"the window\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "3.86", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the window\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0675_00/00240.jpg", "scene0675_00/00270.jpg", "scene0675_00/00300.jpg", "scene0675_00/00330.jpg" ], "turn_id": 3, "scan_id": "scene0675_00", "origin_question": "Which two objects are on the same side of you now? \"the rectangular single bed made of light-colored wood\", \"the tall rectangular light brown wooden wardrobe\", \"the window\". ", "option": [ "the rectangular single bed made of light-colored wood, the tall rectangular light brown wooden wardrobe", "the rectangular single bed made of light-colored wood, the window", "the tall rectangular light brown wooden wardrobe, the window" ], "answer": "the rectangular single bed made of light-colored wood, the tall rectangular light brown wooden wardrobe", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the rectangular single bed made of light-colored wood\", \"the tall rectangular light brown wooden wardrobe\", \"the window\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular single bed made of light-colored wood, the tall rectangular light brown wooden wardrobe', 'the rectangular single bed made of light-colored wood, the window', 'the tall rectangular light brown wooden wardrobe, the window'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0675_00/00360.jpg", "scene0675_00/00390.jpg", "scene0675_00/00420.jpg", "scene0675_00/00450.jpg" ], "turn_id": 4, "scan_id": "scene0675_00", "origin_question": "When did you first discover \"the tall rectangular light brown wooden wardrobe\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the tall rectangular light brown wooden wardrobe\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0675_00/00480.jpg", "scene0675_00/00510.jpg", "scene0675_00/00540.jpg", "scene0675_00/00570.jpg" ], "turn_id": 5, "scan_id": "scene0675_00", "origin_question": "Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "62.08", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0675_00/00600.jpg", "scene0675_00/00630.jpg", "scene0675_00/00660.jpg", "scene0675_00/00690.jpg" ], "turn_id": 6, "scan_id": "scene0675_00", "origin_question": "How far is your current position from where you were at the end of the turn 1? (in meters)", "option": [], "answer": "1.46", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 1? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0675_00/00720.jpg", "scene0675_00/00750.jpg", "scene0675_00/00780.jpg", "scene0675_00/00810.jpg" ], "turn_id": 7, "scan_id": "scene0675_00", "origin_question": "In which turn did you see both \"the window\" and \"the tall rectangular light brown wooden wardrobe\" simultaneously? ", "option": [], "answer": "4", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the window\" and \"the tall rectangular light brown wooden wardrobe\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0078_region13/85849923dc9e46eda0168f62c0fb43f3_i1_2.jpg", "1mp3d_0078_region13/85849923dc9e46eda0168f62c0fb43f3_i2_2.jpg", "1mp3d_0078_region13/b3b697d0fd2d41f29a861e8411c732bd_i1_4.jpg", "1mp3d_0078_region13/b3b697d0fd2d41f29a861e8411c732bd_i2_4.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0078_region13", "origin_question": "Remember, have you seen any door(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any door(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "1mp3d_0078_region13/b3b697d0fd2d41f29a861e8411c732bd_i1_5.jpg", "1mp3d_0078_region13/b3b697d0fd2d41f29a861e8411c732bd_i2_5.jpg", "1mp3d_0078_region13/15734f1d4789455fabcb6f3c9de5ba5b_i1_3.jpg", "1mp3d_0078_region13/15734f1d4789455fabcb6f3c9de5ba5b_i2_3.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0078_region13", "origin_question": "When did you first discover \"the rectangular, light brown wooden door\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the rectangular, light brown wooden door\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "1mp3d_0078_region13/4de3b63eda49451b86a670c2d665273e_i1_4.jpg", "1mp3d_0078_region13/4de3b63eda49451b86a670c2d665273e_i2_4.jpg", "1mp3d_0078_region13/15734f1d4789455fabcb6f3c9de5ba5b_i1_2.jpg", "1mp3d_0078_region13/15734f1d4789455fabcb6f3c9de5ba5b_i2_2.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0078_region13", "origin_question": "When did you first discover \"the fan\"? (index of the turn)", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the fan\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0078_region13/9aec666f83d8473f8369798bfcaea053_i1_0.jpg", "1mp3d_0078_region13/9aec666f83d8473f8369798bfcaea053_i2_0.jpg", "1mp3d_0078_region13/9aec666f83d8473f8369798bfcaea053_i1_5.jpg", "1mp3d_0078_region13/9aec666f83d8473f8369798bfcaea053_i2_5.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0078_region13", "origin_question": "Remember, have you seen any body loofah(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any body loofah(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "1mp3d_0078_region13/6185f5d7cbd54b82a972211bc6fab3af_i1_0.jpg", "1mp3d_0078_region13/6185f5d7cbd54b82a972211bc6fab3af_i2_0.jpg", "1mp3d_0078_region13/241cec5c35f6452eb638c6da962ccbda_i1_3.jpg", "1mp3d_0078_region13/241cec5c35f6452eb638c6da962ccbda_i2_3.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0078_region13", "origin_question": "At the end of which turn were both of \"the slim round green flowerpot made of plastic or ceramic\" and \"the fan\" on your right side?", "option": [], "answer": "5", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the slim round green flowerpot made of plastic or ceramic\" and \"the fan\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "1mp3d_0078_region13/ae09670efbc049e7b53efec93af06130_i1_0.jpg", "1mp3d_0078_region13/ae09670efbc049e7b53efec93af06130_i2_0.jpg", "1mp3d_0078_region13/241cec5c35f6452eb638c6da962ccbda_i1_5.jpg", "1mp3d_0078_region13/241cec5c35f6452eb638c6da962ccbda_i2_5.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0078_region13", "origin_question": "Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "1mp3d_0078_region13/241cec5c35f6452eb638c6da962ccbda_i1_4.jpg", "1mp3d_0078_region13/241cec5c35f6452eb638c6da962ccbda_i2_4.jpg", "1mp3d_0078_region13/241cec5c35f6452eb638c6da962ccbda_i1_2.jpg", "1mp3d_0078_region13/241cec5c35f6452eb638c6da962ccbda_i2_2.jpg" ], "turn_id": 7, "scan_id": "1mp3d_0078_region13", "origin_question": "Using your orientation at the end of turn 5 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "59.94", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 5 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "1mp3d_0078_region13/7251482d76fe41fca7d8a6959e2351a1_i1_0.jpg", "1mp3d_0078_region13/7251482d76fe41fca7d8a6959e2351a1_i2_0.jpg", "1mp3d_0078_region13/7251482d76fe41fca7d8a6959e2351a1_i1_1.jpg", "1mp3d_0078_region13/7251482d76fe41fca7d8a6959e2351a1_i2_1.jpg" ], "turn_id": 8, "scan_id": "1mp3d_0078_region13", "origin_question": "How far is your current position from where you were at the end of the turn 6? (in meters)", "option": [], "answer": "1.08", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 6? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "1mp3d_0078_region13/2b95618cdb494a4fa4f76a220356415b_i1_0.jpg", "1mp3d_0078_region13/2b95618cdb494a4fa4f76a220356415b_i2_0.jpg", "1mp3d_0078_region13/2b95618cdb494a4fa4f76a220356415b_i1_1.jpg", "1mp3d_0078_region13/2b95618cdb494a4fa4f76a220356415b_i2_1.jpg" ], "turn_id": 9, "scan_id": "1mp3d_0078_region13", "origin_question": "Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Left", "user_message": "For the 9 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0559_00/00000.jpg", "scene0559_00/00030.jpg", "scene0559_00/00060.jpg", "scene0559_00/00090.jpg" ], "turn_id": 1, "scan_id": "scene0559_00", "origin_question": "Remember, have you seen any ottoman(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any ottoman(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0559_00/00120.jpg", "scene0559_00/00150.jpg", "scene0559_00/00180.jpg", "scene0559_00/00210.jpg" ], "turn_id": 2, "scan_id": "scene0559_00", "origin_question": "When did you first discover \"the small rectangular brown cardboard box\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the small rectangular brown cardboard box\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0559_00/00240.jpg", "scene0559_00/00270.jpg", "scene0559_00/00300.jpg", "scene0559_00/00330.jpg" ], "turn_id": 3, "scan_id": "scene0559_00", "origin_question": "When did you first discover \"the small rectangular brown cardboard box\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the small rectangular brown cardboard box\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0559_00/00360.jpg", "scene0559_00/00390.jpg", "scene0559_00/00420.jpg", "scene0559_00/00450.jpg" ], "turn_id": 4, "scan_id": "scene0559_00", "origin_question": "Remember, how many bench(s) have you seen so far? ", "option": [], "answer": "2", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many bench(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0415_01/00000.jpg", "scene0415_01/00040.jpg", "scene0415_01/00080.jpg", "scene0415_01/00120.jpg" ], "turn_id": 1, "scan_id": "scene0415_01", "origin_question": "Remember, have you seen any stand(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any stand(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0415_01/00160.jpg", "scene0415_01/00200.jpg", "scene0415_01/00240.jpg", "scene0415_01/00280.jpg" ], "turn_id": 2, "scan_id": "scene0415_01", "origin_question": "Is the cylindrical stainless steel bin to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the cylindrical stainless steel bin to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0415_01/00320.jpg", "scene0415_01/00360.jpg", "scene0415_01/00400.jpg", "scene0415_01/00440.jpg" ], "turn_id": 3, "scan_id": "scene0415_01", "origin_question": "Please recall \"the framed picture on paper with bright colors\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.76", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the framed picture on paper with bright colors\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0415_01/00480.jpg", "scene0415_01/00520.jpg", "scene0415_01/00560.jpg", "scene0415_01/00600.jpg" ], "turn_id": 4, "scan_id": "scene0415_01", "origin_question": "What will be the first-time appearance order of the following categories: door, picture, toilet?", "option": [ "door, toilet, picture", "picture, door, toilet", "toilet, picture, door", "door, picture, toilet" ], "answer": "toilet, picture, door", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: door, picture, toilet? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['door, toilet, picture', 'picture, door, toilet', 'toilet, picture, door', 'door, picture, toilet'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "scene0415_01/00640.jpg", "scene0415_01/00680.jpg", "scene0415_01/00720.jpg", "scene0415_01/00760.jpg" ], "turn_id": 5, "scan_id": "scene0415_01", "origin_question": "Among these three objects(\"the medium-sized white oval porcelain toilet\",\"the cylindrical stainless steel bin\",\"the framed picture on paper with bright colors\" ), which one is the farthest from you?", "option": [ "the medium-sized white oval porcelain toilet", "the cylindrical stainless steel bin", "the framed picture on paper with bright colors" ], "answer": "the cylindrical stainless steel bin", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects(\"the medium-sized white oval porcelain toilet\",\"the cylindrical stainless steel bin\",\"the framed picture on paper with bright colors\" ), which one is the farthest from you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the medium-sized white oval porcelain toilet', 'the cylindrical stainless steel bin', 'the framed picture on paper with bright colors'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0415_01/00800.jpg", "scene0415_01/00840.jpg", "scene0415_01/00880.jpg", "scene0415_01/00920.jpg" ], "turn_id": 6, "scan_id": "scene0415_01", "origin_question": "Which two objects are on the same side of you now? \"the rectangular, white porcelain bathtub\", \"the vanity\", \"the white rectangular sink with a modern design\". ", "option": [ "the rectangular, white porcelain bathtub, the vanity", "the rectangular, white porcelain bathtub, the white rectangular sink with a modern design", "the vanity, the white rectangular sink with a modern design" ], "answer": "the vanity, the white rectangular sink with a modern design", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the rectangular, white porcelain bathtub\", \"the vanity\", \"the white rectangular sink with a modern design\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular, white porcelain bathtub, the vanity', 'the rectangular, white porcelain bathtub, the white rectangular sink with a modern design', 'the vanity, the white rectangular sink with a modern design'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0067_region28/9d97383beaee4ec78b8b20f44ba72c3b_i1_3.jpg", "1mp3d_0067_region28/9d97383beaee4ec78b8b20f44ba72c3b_i2_3.jpg", "1mp3d_0067_region28/7ecfa7f1ac394e9c94cb3b1b8a22004b_i1_2.jpg", "1mp3d_0067_region28/7ecfa7f1ac394e9c94cb3b1b8a22004b_i2_2.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0067_region28", "origin_question": "Remember, have you seen any window(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any window(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "1mp3d_0067_region28/9d97383beaee4ec78b8b20f44ba72c3b_i1_2.jpg", "1mp3d_0067_region28/9d97383beaee4ec78b8b20f44ba72c3b_i2_2.jpg", "1mp3d_0067_region28/7ecfa7f1ac394e9c94cb3b1b8a22004b_i1_1.jpg", "1mp3d_0067_region28/7ecfa7f1ac394e9c94cb3b1b8a22004b_i2_1.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0067_region28", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the hardcover book with a blue cover and white spine\"; \"the large bed with a black and white pattern\"; \"the rod\" ", "option": [ "the hardcover book with a blue cover and white spine", "the large bed with a black and white pattern", "the rod" ], "answer": "the large bed with a black and white pattern", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the hardcover book with a blue cover and white spine\"; \"the large bed with a black and white pattern\"; \"the rod\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the hardcover book with a blue cover and white spine', 'the large bed with a black and white pattern', 'the rod'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "1mp3d_0067_region28/bb09a1003aa448bbb67af68428ea1103_i1_2.jpg", "1mp3d_0067_region28/bb09a1003aa448bbb67af68428ea1103_i2_2.jpg", "1mp3d_0067_region28/a656d59b1e6d4b1f8056e0d90b85c581_i1_4.jpg", "1mp3d_0067_region28/a656d59b1e6d4b1f8056e0d90b85c581_i2_4.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0067_region28", "origin_question": "At the end of which turn were both of \"the rod\" and \"the large bed with a black and white pattern\" on your left side?", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the rod\" and \"the large bed with a black and white pattern\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0067_region28/ddacc143a5af401fadb60b2e3f2cf473_i1_3.jpg", "1mp3d_0067_region28/ddacc143a5af401fadb60b2e3f2cf473_i2_3.jpg", "1mp3d_0067_region28/a656d59b1e6d4b1f8056e0d90b85c581_i1_0.jpg", "1mp3d_0067_region28/a656d59b1e6d4b1f8056e0d90b85c581_i2_0.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0067_region28", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the large bed with a black and white pattern\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the large bed with a black and white pattern\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "1mp3d_0067_region28/a656d59b1e6d4b1f8056e0d90b85c581_i1_5.jpg", "1mp3d_0067_region28/a656d59b1e6d4b1f8056e0d90b85c581_i2_5.jpg", "1mp3d_0067_region28/a656d59b1e6d4b1f8056e0d90b85c581_i1_2.jpg", "1mp3d_0067_region28/a656d59b1e6d4b1f8056e0d90b85c581_i2_2.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0067_region28", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the hardcover book with a blue cover and white spine\" from your position?", "option": [], "answer": "136.33", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the hardcover book with a blue cover and white spine\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "1mp3d_0067_region28/ddacc143a5af401fadb60b2e3f2cf473_i1_1.jpg", "1mp3d_0067_region28/ddacc143a5af401fadb60b2e3f2cf473_i2_1.jpg", "1mp3d_0067_region28/ddacc143a5af401fadb60b2e3f2cf473_i1_0.jpg", "1mp3d_0067_region28/ddacc143a5af401fadb60b2e3f2cf473_i2_0.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0067_region28", "origin_question": "Please recall \"the hardcover book with a blue cover and white spine\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.67", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the hardcover book with a blue cover and white spine\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0029_region21/1410b021e1c14f529188eb026fbb369a_i2_2.jpg", "1mp3d_0029_region21/0eb1323894e041efa23d2f3e60efbe44_i1_3.jpg", "1mp3d_0029_region21/0eb1323894e041efa23d2f3e60efbe44_i2_3.jpg", "1mp3d_0029_region21/1410b021e1c14f529188eb026fbb369a_i1_3.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0029_region21", "origin_question": "Remember, have you seen any water cooler(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any water cooler(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "1mp3d_0029_region21/1410b021e1c14f529188eb026fbb369a_i2_3.jpg", "1mp3d_0029_region21/1410b021e1c14f529188eb026fbb369a_i1_4.jpg", "1mp3d_0029_region21/1410b021e1c14f529188eb026fbb369a_i2_4.jpg", "1mp3d_0029_region21/2d4a32746fa2462b8e6ee8fe2b56c653_i1_0.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0029_region21", "origin_question": "Which direction is the rectangular window with a white frame and sill in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the rectangular window with a white frame and sill in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0029_region21/2d4a32746fa2462b8e6ee8fe2b56c653_i2_0.jpg", "1mp3d_0029_region21/725d4645400a47ef9b9beaa6e8f56f2b_i1_2.jpg", "1mp3d_0029_region21/725d4645400a47ef9b9beaa6e8f56f2b_i2_2.jpg", "1mp3d_0029_region21/2d4a32746fa2462b8e6ee8fe2b56c653_i1_5.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0029_region21", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the rectangular window with a white frame and sill\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the rectangular window with a white frame and sill\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "1mp3d_0029_region21/2d4a32746fa2462b8e6ee8fe2b56c653_i2_5.jpg", "1mp3d_0029_region21/725d4645400a47ef9b9beaa6e8f56f2b_i2_1.jpg", "1mp3d_0029_region21/725d4645400a47ef9b9beaa6e8f56f2b_i1_0.jpg", "1mp3d_0029_region21/725d4645400a47ef9b9beaa6e8f56f2b_i2_0.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0029_region21", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the rectangular window with a white frame and sill\" from your position?", "option": [], "answer": "80.16", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the rectangular window with a white frame and sill\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0328_00/00000.jpg", "scene0328_00/00040.jpg", "scene0328_00/00080.jpg", "scene0328_00/00120.jpg" ], "turn_id": 1, "scan_id": "scene0328_00", "origin_question": "Remember, have you seen any person(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any person(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0328_00/00160.jpg", "scene0328_00/00200.jpg", "scene0328_00/00240.jpg", "scene0328_00/00280.jpg" ], "turn_id": 2, "scan_id": "scene0328_00", "origin_question": "When did you first discover \"the black and silver metal oven with a closed door\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the black and silver metal oven with a closed door\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0328_00/00320.jpg", "scene0328_00/00360.jpg", "scene0328_00/00400.jpg", "scene0328_00/00440.jpg" ], "turn_id": 3, "scan_id": "scene0328_00", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "1.75", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0328_00/00480.jpg", "scene0328_00/00520.jpg", "scene0328_00/00560.jpg", "scene0328_00/00600.jpg" ], "turn_id": 4, "scan_id": "scene0328_00", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "100.46", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0328_00/00640.jpg", "scene0328_00/00680.jpg", "scene0328_00/00720.jpg", "scene0328_00/00800.jpg" ], "turn_id": 5, "scan_id": "scene0328_00", "origin_question": "How far is your current position from where you were at the end of the turn 1? (in meters)", "option": [], "answer": "2.65", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 1? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0328_00/00840.jpg", "scene0328_00/00880.jpg", "scene0328_00/00920.jpg", "scene0328_00/00960.jpg" ], "turn_id": 6, "scan_id": "scene0328_00", "origin_question": "Is the small rectangular grey glass window in a wooden frame to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the small rectangular grey glass window in a wooden frame to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0328_00/01000.jpg", "scene0328_00/01070.jpg", "scene0328_00/01110.jpg", "scene0328_00/01150.jpg" ], "turn_id": 7, "scan_id": "scene0328_00", "origin_question": "Which two objects are on the same side of you now? \"the black and silver metal oven with a closed door\", \"the moderate-sized lush green plant in a brown pot\", \"the rectangular black and white striped carpet\". ", "option": [ "the black and silver metal oven with a closed door, the moderate-sized lush green plant in a brown pot", "the black and silver metal oven with a closed door, the rectangular black and white striped carpet", "the moderate-sized lush green plant in a brown pot, the rectangular black and white striped carpet" ], "answer": "the black and silver metal oven with a closed door, the moderate-sized lush green plant in a brown pot", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the black and silver metal oven with a closed door\", \"the moderate-sized lush green plant in a brown pot\", \"the rectangular black and white striped carpet\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the black and silver metal oven with a closed door, the moderate-sized lush green plant in a brown pot', 'the black and silver metal oven with a closed door, the rectangular black and white striped carpet', 'the moderate-sized lush green plant in a brown pot, the rectangular black and white striped carpet'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0328_00/01190.jpg", "scene0328_00/01230.jpg", "scene0328_00/01270.jpg", "scene0328_00/01310.jpg" ], "turn_id": 8, "scan_id": "scene0328_00", "origin_question": "Please recall \"the black and silver metal oven with a closed door\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.07", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the black and silver metal oven with a closed door\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0430_01/00000.jpg", "scene0430_01/00040.jpg", "scene0430_01/00080.jpg", "scene0430_01/00120.jpg" ], "turn_id": 1, "scan_id": "scene0430_01", "origin_question": "Remember, have you seen any table(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any table(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0430_01/00160.jpg", "scene0430_01/00200.jpg", "scene0430_01/00240.jpg", "scene0430_01/00280.jpg" ], "turn_id": 2, "scan_id": "scene0430_01", "origin_question": "When did you first discover \"the window\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the window\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0430_01/00320.jpg", "scene0430_01/00360.jpg", "scene0430_01/00400.jpg", "scene0430_01/00440.jpg" ], "turn_id": 3, "scan_id": "scene0430_01", "origin_question": "", "option": [], "answer": "", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0430_01/00480.jpg", "scene0430_01/00520.jpg", "scene0430_01/00560.jpg", "scene0430_01/00600.jpg" ], "turn_id": 4, "scan_id": "scene0430_01", "origin_question": "Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0430_01/00640.jpg", "scene0430_01/00680.jpg", "scene0430_01/00720.jpg", "scene0430_01/00760.jpg" ], "turn_id": 5, "scan_id": "scene0430_01", "origin_question": "What will be the first-time appearance order of the following categories: chair, screen, table?", "option": [ "screen, chair, table", "screen, table, chair", "table, screen, chair", "chair, table, screen" ], "answer": "chair, table, screen", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: chair, screen, table? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['screen, chair, table', 'screen, table, chair', 'table, screen, chair', 'chair, table, screen'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0430_01/00800.jpg", "scene0430_01/00840.jpg", "scene0430_01/00880.jpg", "scene0430_01/00920.jpg" ], "turn_id": 6, "scan_id": "scene0430_01", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "92.19", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0430_01/00960.jpg", "scene0430_01/01000.jpg", "scene0430_01/01040.jpg", "scene0430_01/01080.jpg" ], "turn_id": 7, "scan_id": "scene0430_01", "origin_question": "In which turn were you closest to \"the rectangular light wooden table with wheels\"? ", "option": [], "answer": "4", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the rectangular light wooden table with wheels\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0430_01/01120.jpg", "scene0430_01/01160.jpg", "scene0430_01/01200.jpg", "scene0430_01/01240.jpg" ], "turn_id": 8, "scan_id": "scene0430_01", "origin_question": "Using your orientation at the end of turn 4 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "157.5", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 4 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0430_01/01280.jpg", "scene0430_01/01320.jpg", "scene0430_01/01360.jpg", "scene0430_01/01400.jpg" ], "turn_id": 9, "scan_id": "scene0430_01", "origin_question": "What will be the first-time appearance order of the following categories: screen, table, window?", "option": [ "window, table, screen", "screen, table, window", "table, screen, window", "window, screen, table" ], "answer": "window, table, screen", "user_message": "For the 9 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: screen, table, window? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['window, table, screen', 'screen, table, window', 'table, screen, window', 'window, screen, table'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0605_01/00000.jpg", "scene0605_01/00030.jpg", "scene0605_01/00060.jpg", "scene0605_01/00090.jpg" ], "turn_id": 1, "scan_id": "scene0605_01", "origin_question": "Remember, have you seen any sign(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any sign(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0605_01/00120.jpg", "scene0605_01/00150.jpg", "scene0605_01/00180.jpg", "scene0605_01/00210.jpg" ], "turn_id": 2, "scan_id": "scene0605_01", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the rectangular white storage box with a black handle\"; \"the light grey car with standard dimensions\"; \"the door\" ", "option": [ "the rectangular white storage box with a black handle", "the light grey car with standard dimensions", "the door" ], "answer": "the rectangular white storage box with a black handle", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the rectangular white storage box with a black handle\"; \"the light grey car with standard dimensions\"; \"the door\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular white storage box with a black handle', 'the light grey car with standard dimensions', 'the door'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0605_01/00240.jpg", "scene0605_01/00270.jpg", "scene0605_01/00300.jpg", "scene0605_01/00330.jpg" ], "turn_id": 3, "scan_id": "scene0605_01", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "38.4", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0605_01/00380.jpg", "scene0605_01/00410.jpg", "scene0605_01/00440.jpg", "scene0605_01/00470.jpg" ], "turn_id": 4, "scan_id": "scene0605_01", "origin_question": "", "option": [], "answer": "", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0605_01/00500.jpg", "scene0605_01/00530.jpg", "scene0605_01/00560.jpg", "scene0605_01/00590.jpg" ], "turn_id": 5, "scan_id": "scene0605_01", "origin_question": "Please recall \"the rectangular white storage box with a black handle\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.53", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the rectangular white storage box with a black handle\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0605_01/00620.jpg", "scene0605_01/00650.jpg", "scene0605_01/00680.jpg", "scene0605_01/00710.jpg" ], "turn_id": 6, "scan_id": "scene0605_01", "origin_question": "Which direction is the standard city bicycle with a silver frame in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the standard city bicycle with a silver frame in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0605_01/00740.jpg", "scene0605_01/00770.jpg", "scene0605_01/00800.jpg", "scene0605_01/00830.jpg" ], "turn_id": 7, "scan_id": "scene0605_01", "origin_question": "By the end of the 5 turn, were you getting closer to or farther away from these two objects: X: \"the long, blue waterproof jacket with a shiny texture\", Y: \"the worn-out wooden ladder with a deep brown color\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "farther away from both X and Y", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 5 turn, were you getting closer to or farther away from these two objects: X: \"the long, blue waterproof jacket with a shiny texture\", Y: \"the worn-out wooden ladder with a deep brown color\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0084_00/00000.jpg", "scene0084_00/00040.jpg", "scene0084_00/00080.jpg", "scene0084_00/00120.jpg", "scene0084_00/00160.jpg" ], "turn_id": 1, "scan_id": "scene0084_00", "origin_question": "Remember, have you seen any dustpan(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any dustpan(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0084_00/00200.jpg", "scene0084_00/00240.jpg", "scene0084_00/00280.jpg", "scene0084_00/00320.jpg", "scene0084_00/00360.jpg" ], "turn_id": 2, "scan_id": "scene0084_00", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the rectangular yellow warning sign made of plastic\"; \"the cylindrical plastic bottle with a green lid\"; \"the white rectangular door with a silver handle\" ", "option": [ "the rectangular yellow warning sign made of plastic", "the cylindrical plastic bottle with a green lid", "the white rectangular door with a silver handle" ], "answer": "the white rectangular door with a silver handle", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the rectangular yellow warning sign made of plastic\"; \"the cylindrical plastic bottle with a green lid\"; \"the white rectangular door with a silver handle\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular yellow warning sign made of plastic', 'the cylindrical plastic bottle with a green lid', 'the white rectangular door with a silver handle'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0084_00/00400.jpg", "scene0084_00/00440.jpg", "scene0084_00/00480.jpg", "scene0084_00/00520.jpg", "scene0084_00/00560.jpg" ], "turn_id": 3, "scan_id": "scene0084_00", "origin_question": "Which two objects are on the same side of you now? \"the white wall-mounted soap dispenser with a curved top\", \"then oval milky-white sink with a granite surface\", \"the white rectangular door with a silver handle\". ", "option": [ "the white wall-mounted soap dispenser with a curved top, then oval milky-white sink with a granite surface", "the white wall-mounted soap dispenser with a curved top, the white rectangular door with a silver handle", "then oval milky-white sink with a granite surface, the white rectangular door with a silver handle" ], "answer": "the white wall-mounted soap dispenser with a curved top, then oval milky-white sink with a granite surface", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the white wall-mounted soap dispenser with a curved top\", \"then oval milky-white sink with a granite surface\", \"the white rectangular door with a silver handle\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the white wall-mounted soap dispenser with a curved top, then oval milky-white sink with a granite surface', 'the white wall-mounted soap dispenser with a curved top, the white rectangular door with a silver handle', 'then oval milky-white sink with a granite surface, the white rectangular door with a silver handle'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0084_00/00600.jpg", "scene0084_00/00640.jpg", "scene0084_00/00680.jpg", "scene0084_00/00720.jpg", "scene0084_00/00760.jpg" ], "turn_id": 4, "scan_id": "scene0084_00", "origin_question": "Is the white rectangular door with a silver handle to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the white rectangular door with a silver handle to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "scene0084_00/00800.jpg", "scene0084_00/00840.jpg", "scene0084_00/00880.jpg", "scene0084_00/00920.jpg", "scene0084_00/00960.jpg" ], "turn_id": 5, "scan_id": "scene0084_00", "origin_question": "Assuming the direction you are facing at the end of the turn 3 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Right", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 3 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0084_00/01000.jpg", "scene0084_00/01040.jpg", "scene0084_00/01080.jpg", "scene0084_00/01120.jpg", "scene0084_00/01160.jpg" ], "turn_id": 6, "scan_id": "scene0084_00", "origin_question": "Using your orientation at the end of turn 4 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "28.72", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 4 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "scene0084_00/01200.jpg", "scene0084_00/01240.jpg", "scene0084_00/01280.jpg", "scene0084_00/01320.jpg", "scene0084_00/01360.jpg" ], "turn_id": 7, "scan_id": "scene0084_00", "origin_question": "Assuming the direction you are facing at the end of the turn 4 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Right", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 4 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0084_00/01400.jpg", "scene0084_00/01440.jpg", "scene0084_00/01480.jpg", "scene0084_00/01520.jpg", "scene0084_00/01560.jpg" ], "turn_id": 8, "scan_id": "scene0084_00", "origin_question": "Which direction is the rectangular printed paper with black text in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the rectangular printed paper with black text in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0084_00/01600.jpg", "scene0084_00/01640.jpg", "scene0084_00/01680.jpg", "scene0084_00/01720.jpg", "scene0084_00/01760.jpg" ], "turn_id": 9, "scan_id": "scene0084_00", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the cylindrical plastic bottle with a green lid\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the cylindrical plastic bottle with a green lid\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0084_region11/535a1de14e724c9d8288f2c3e61db742_i1_1.jpg", "1mp3d_0084_region11/535a1de14e724c9d8288f2c3e61db742_i2_1.jpg", "1mp3d_0084_region11/b3f9e188366447e4ba76f5f98f1eaadd_i1_0.jpg", "1mp3d_0084_region11/b3f9e188366447e4ba76f5f98f1eaadd_i2_0.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0084_region11", "origin_question": "Remember, have you seen any bench(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any bench(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "1mp3d_0084_region11/b3f9e188366447e4ba76f5f98f1eaadd_i1_5.jpg", "1mp3d_0084_region11/b3f9e188366447e4ba76f5f98f1eaadd_i2_5.jpg", "1mp3d_0084_region11/76fff8b638e44eabbc34286ef72ee404_i1_2.jpg", "1mp3d_0084_region11/76fff8b638e44eabbc34286ef72ee404_i2_2.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0084_region11", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the bench\"; \"the rectangular yellow picture with black figures\"; \"the small rectangular wooden table with light brown color\" ", "option": [ "the bench", "the rectangular yellow picture with black figures", "the small rectangular wooden table with light brown color" ], "answer": "the small rectangular wooden table with light brown color", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the bench\"; \"the rectangular yellow picture with black figures\"; \"the small rectangular wooden table with light brown color\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the bench', 'the rectangular yellow picture with black figures', 'the small rectangular wooden table with light brown color'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "1mp3d_0084_region11/76fff8b638e44eabbc34286ef72ee404_i1_1.jpg", "1mp3d_0084_region11/76fff8b638e44eabbc34286ef72ee404_i2_1.jpg", "1mp3d_0084_region11/d82d2a30b0bc4d6a8cd5fc526de20a7d_i1_2.jpg", "1mp3d_0084_region11/d82d2a30b0bc4d6a8cd5fc526de20a7d_i2_2.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0084_region11", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "1mp3d_0084_region11/ae71514af4dd481d9b6a00a7d323cfc1_i1_5.jpg", "1mp3d_0084_region11/ae71514af4dd481d9b6a00a7d323cfc1_i2_5.jpg", "1mp3d_0084_region11/ae71514af4dd481d9b6a00a7d323cfc1_i1_0.jpg", "1mp3d_0084_region11/ae71514af4dd481d9b6a00a7d323cfc1_i2_0.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0084_region11", "origin_question": "When did you first discover \"the dark brown rectangular wooden case\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the dark brown rectangular wooden case\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "1mp3d_0084_region11/e706d1515b7449558aada9f826b27d2b_i1_0.jpg", "1mp3d_0084_region11/e706d1515b7449558aada9f826b27d2b_i2_0.jpg", "1mp3d_0084_region11/3d21abc4be0547fd9a91850e871a0b59_i1_0.jpg", "1mp3d_0084_region11/3d21abc4be0547fd9a91850e871a0b59_i2_0.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0084_region11", "origin_question": "Please recall \"the small rectangular brown wooden box\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "5.77", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the small rectangular brown wooden box\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "1mp3d_0084_region11/3d21abc4be0547fd9a91850e871a0b59_i1_3.jpg", "1mp3d_0084_region11/3d21abc4be0547fd9a91850e871a0b59_i2_3.jpg", "1mp3d_0084_region11/3d21abc4be0547fd9a91850e871a0b59_i1_1.jpg", "1mp3d_0084_region11/3d21abc4be0547fd9a91850e871a0b59_i2_1.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0084_region11", "origin_question": "What will be the first-time appearance order of the following categories: bench, ottoman, window?", "option": [ "window, ottoman, bench", "ottoman, bench, window", "bench, window, ottoman", "bench, ottoman, window" ], "answer": "bench, ottoman, window", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bench, ottoman, window? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['window, ottoman, bench', 'ottoman, bench, window', 'bench, window, ottoman', 'bench, ottoman, window'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "1mp3d_0084_region11/f5079094001d43dba71e83d764ec5c87_i1_3.jpg", "1mp3d_0084_region11/f5079094001d43dba71e83d764ec5c87_i2_3.jpg", "1mp3d_0084_region11/f5079094001d43dba71e83d764ec5c87_i1_2.jpg", "1mp3d_0084_region11/f5079094001d43dba71e83d764ec5c87_i2_2.jpg" ], "turn_id": 7, "scan_id": "1mp3d_0084_region11", "origin_question": "What will be the first-time appearance order of the following categories: bench, curtain, table?", "option": [ "table, bench, curtain", "curtain, bench, table", "table, curtain, bench", "bench, table, curtain" ], "answer": "bench, table, curtain", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bench, curtain, table? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['table, bench, curtain', 'curtain, bench, table', 'table, curtain, bench', 'bench, table, curtain'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "1mp3d_0084_region11/9c385eb7acad4e95ad472ed7c92924e7_i1_2.jpg", "1mp3d_0084_region11/9c385eb7acad4e95ad472ed7c92924e7_i2_2.jpg", "1mp3d_0084_region11/9c385eb7acad4e95ad472ed7c92924e7_i1_3.jpg", "1mp3d_0084_region11/9c385eb7acad4e95ad472ed7c92924e7_i2_3.jpg" ], "turn_id": 8, "scan_id": "1mp3d_0084_region11", "origin_question": "Assuming the direction you are facing at the end of the turn 4 is forward, did you move a certain distance forward or backward from that position?", "option": [ "Forward", "Backward" ], "answer": "Backward", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 4 is forward, did you move a certain distance forward or backward from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Forward', 'Backward'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "1mp3d_0084_region11/1cee4c97015545f597400825b252e091_i1_5.jpg", "1mp3d_0084_region11/1cee4c97015545f597400825b252e091_i2_5.jpg", "1mp3d_0084_region11/6333e669c4174c5dbcbc70e6da1c08f6_i1_2.jpg", "1mp3d_0084_region11/6333e669c4174c5dbcbc70e6da1c08f6_i2_2.jpg" ], "turn_id": 9, "scan_id": "1mp3d_0084_region11", "origin_question": "By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the medium-sized round light-colored vase\", Y: \"the white ceiling fan with three curved metal blades\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "farther away from both X and Y", "user_message": "For the 9 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the medium-sized round light-colored vase\", Y: \"the white ceiling fan with three curved metal blades\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0044_region32/9bd3272032c94c049f8d10f68490667f_i1_4.jpg", "1mp3d_0044_region32/9bd3272032c94c049f8d10f68490667f_i2_4.jpg", "1mp3d_0044_region32/4f554d08841b43dd839efedde670b9dd_i1_3.jpg", "1mp3d_0044_region32/4f554d08841b43dd839efedde670b9dd_i2_3.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0044_region32", "origin_question": "Remember, have you seen any bowl(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any bowl(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "1mp3d_0044_region32/a9d6e52246374e2da13d559eafe70aeb_i1_4.jpg", "1mp3d_0044_region32/a9d6e52246374e2da13d559eafe70aeb_i2_4.jpg", "1mp3d_0044_region32/4e98757ce6564f3abb37ccedf18d240e_i1_3.jpg", "1mp3d_0044_region32/4e98757ce6564f3abb37ccedf18d240e_i2_3.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0044_region32", "origin_question": "When did you first discover \"the door\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the door\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "1mp3d_0044_region32/a9d6e52246374e2da13d559eafe70aeb_i1_3.jpg", "1mp3d_0044_region32/a9d6e52246374e2da13d559eafe70aeb_i2_3.jpg", "1mp3d_0044_region32/4f554d08841b43dd839efedde670b9dd_i1_1.jpg", "1mp3d_0044_region32/4f554d08841b43dd839efedde670b9dd_i2_1.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0044_region32", "origin_question": "In which turn were you closest to \"the toilet\"? ", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the toilet\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "1mp3d_0044_region32/9bd3272032c94c049f8d10f68490667f_i1_5.jpg", "1mp3d_0044_region32/9bd3272032c94c049f8d10f68490667f_i2_5.jpg", "1mp3d_0044_region32/9bd3272032c94c049f8d10f68490667f_i1_0.jpg", "1mp3d_0044_region32/9bd3272032c94c049f8d10f68490667f_i2_0.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0044_region32", "origin_question": "By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the toilet\", Y: \"the standard white paper roll of toilet paper\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the toilet\", Y: \"the standard white paper roll of toilet paper\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "1mp3d_0044_region32/4f554d08841b43dd839efedde670b9dd_i1_0.jpg", "1mp3d_0044_region32/4f554d08841b43dd839efedde670b9dd_i2_0.jpg", "1mp3d_0044_region32/a9d6e52246374e2da13d559eafe70aeb_i1_1.jpg", "1mp3d_0044_region32/a9d6e52246374e2da13d559eafe70aeb_i2_1.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0044_region32", "origin_question": "Which direction is the standard white paper roll of toilet paper in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the standard white paper roll of toilet paper in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "1mp3d_0044_region32/a9d6e52246374e2da13d559eafe70aeb_i1_2.jpg", "1mp3d_0044_region32/a9d6e52246374e2da13d559eafe70aeb_i2_2.jpg", "1mp3d_0044_region32/4e98757ce6564f3abb37ccedf18d240e_i1_1.jpg", "1mp3d_0044_region32/4e98757ce6564f3abb37ccedf18d240e_i2_1.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0044_region32", "origin_question": "Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0057_region22/f0fea79c1a36425b9979f781ecd4d092_i1_5.jpg", "1mp3d_0057_region22/f0fea79c1a36425b9979f781ecd4d092_i2_5.jpg", "1mp3d_0057_region22/d1f02cad0f0740478b4665ad7716755a_i1_4.jpg", "1mp3d_0057_region22/d1f02cad0f0740478b4665ad7716755a_i2_4.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0057_region22", "origin_question": "Remember, have you seen any towel(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any towel(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0057_region22/7ed66f570e664d7fa5a01c2308d36991_i1_1.jpg", "1mp3d_0057_region22/7ed66f570e664d7fa5a01c2308d36991_i2_1.jpg", "1mp3d_0057_region22/a5419eec514b4262b45b20f6f231222b_i1_1.jpg", "1mp3d_0057_region22/a5419eec514b4262b45b20f6f231222b_i2_1.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0057_region22", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the neatly folded white towel of soft material\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the neatly folded white towel of soft material\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "1mp3d_0057_region22/a5419eec514b4262b45b20f6f231222b_i1_2.jpg", "1mp3d_0057_region22/a5419eec514b4262b45b20f6f231222b_i2_2.jpg", "1mp3d_0057_region22/7ed66f570e664d7fa5a01c2308d36991_i1_4.jpg", "1mp3d_0057_region22/7ed66f570e664d7fa5a01c2308d36991_i2_4.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0057_region22", "origin_question": "At the end of which turn were both of \"the blinds\" and \"the toilet\" on your right side?", "option": [], "answer": "2", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the blinds\" and \"the toilet\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "1mp3d_0057_region22/a5419eec514b4262b45b20f6f231222b_i1_3.jpg", "1mp3d_0057_region22/a5419eec514b4262b45b20f6f231222b_i2_3.jpg", "1mp3d_0057_region22/7ed66f570e664d7fa5a01c2308d36991_i1_3.jpg", "1mp3d_0057_region22/7ed66f570e664d7fa5a01c2308d36991_i2_3.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0057_region22", "origin_question": "When did you first discover \"the toilet\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the toilet\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "1mp3d_0057_region22/7ed66f570e664d7fa5a01c2308d36991_i1_2.jpg", "1mp3d_0057_region22/7ed66f570e664d7fa5a01c2308d36991_i2_2.jpg", "1mp3d_0057_region22/d1f02cad0f0740478b4665ad7716755a_i1_0.jpg", "1mp3d_0057_region22/d1f02cad0f0740478b4665ad7716755a_i2_0.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0057_region22", "origin_question": "Which direction is the neatly folded white towel of soft material in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the neatly folded white towel of soft material in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0607_01/00000.jpg", "scene0607_01/00030.jpg", "scene0607_01/00060.jpg", "scene0607_01/00090.jpg" ], "turn_id": 1, "scan_id": "scene0607_01", "origin_question": "Remember, have you seen any cabinet(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any cabinet(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0607_01/00120.jpg", "scene0607_01/00150.jpg", "scene0607_01/00180.jpg", "scene0607_01/00210.jpg" ], "turn_id": 2, "scan_id": "scene0607_01", "origin_question": "When did you first discover \"the rectangular smooth counter with a matte grey finish\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the rectangular smooth counter with a matte grey finish\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0607_01/00240.jpg", "scene0607_01/00270.jpg", "scene0607_01/00300.jpg", "scene0607_01/00330.jpg" ], "turn_id": 3, "scan_id": "scene0607_01", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the rectangular wooden table with a smooth surface\"; \"the cylindrical white plastic soap dispenser\"; \"the rectangular smooth counter with a matte grey finish\" ", "option": [ "the rectangular wooden table with a smooth surface", "the cylindrical white plastic soap dispenser", "the rectangular smooth counter with a matte grey finish" ], "answer": "the cylindrical white plastic soap dispenser", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the rectangular wooden table with a smooth surface\"; \"the cylindrical white plastic soap dispenser\"; \"the rectangular smooth counter with a matte grey finish\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular wooden table with a smooth surface', 'the cylindrical white plastic soap dispenser', 'the rectangular smooth counter with a matte grey finish'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0607_01/00360.jpg", "scene0607_01/00390.jpg", "scene0607_01/00420.jpg", "scene0607_01/00450.jpg" ], "turn_id": 4, "scan_id": "scene0607_01", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the cylindrical white plastic soap dispenser\" from your position?", "option": [], "answer": "71.74", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the cylindrical white plastic soap dispenser\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0355_00/00000.jpg", "scene0355_00/00040.jpg", "scene0355_00/00080.jpg", "scene0355_00/00120.jpg" ], "turn_id": 1, "scan_id": "scene0355_00", "origin_question": "Remember, have you seen any bed(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any bed(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0355_00/00160.jpg", "scene0355_00/00200.jpg", "scene0355_00/00240.jpg", "scene0355_00/00280.jpg" ], "turn_id": 2, "scan_id": "scene0355_00", "origin_question": "When did you first discover \"the tall green plant with long narrow leaves\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the tall green plant with long narrow leaves\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0355_00/00320.jpg", "scene0355_00/00360.jpg", "scene0355_00/00400.jpg", "scene0355_00/00440.jpg" ], "turn_id": 3, "scan_id": "scene0355_00", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the tall green plant with long narrow leaves\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the tall green plant with long narrow leaves\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0355_00/00480.jpg", "scene0355_00/00520.jpg", "scene0355_00/00560.jpg", "scene0355_00/00600.jpg" ], "turn_id": 4, "scan_id": "scene0355_00", "origin_question": "How far is your current position from where you were at the end of the turn 1? (in meters)", "option": [], "answer": "4.4", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 1? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0355_00/00640.jpg", "scene0355_00/00680.jpg", "scene0355_00/00720.jpg", "scene0355_00/00760.jpg" ], "turn_id": 5, "scan_id": "scene0355_00", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the tall green plant with long narrow leaves\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the tall green plant with long narrow leaves\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0355_00/00800.jpg", "scene0355_00/00840.jpg", "scene0355_00/00880.jpg", "scene0355_00/00930.jpg" ], "turn_id": 6, "scan_id": "scene0355_00", "origin_question": "When did you first discover \"the tall green plant with long narrow leaves\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the tall green plant with long narrow leaves\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0079_00/00000.jpg", "scene0079_00/00040.jpg", "scene0079_00/00080.jpg", "scene0079_00/00120.jpg" ], "turn_id": 1, "scan_id": "scene0079_00", "origin_question": "Remember, have you seen any monitor(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any monitor(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0079_00/00160.jpg", "scene0079_00/00200.jpg", "scene0079_00/00240.jpg", "scene0079_00/00280.jpg" ], "turn_id": 2, "scan_id": "scene0079_00", "origin_question": "In which turn were you farthest from \"the flat screen monitor with a dark matte surface\"? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the flat screen monitor with a dark matte surface\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "scene0079_00/00320.jpg", "scene0079_00/00360.jpg", "scene0079_00/00400.jpg", "scene0079_00/00440.jpg" ], "turn_id": 3, "scan_id": "scene0079_00", "origin_question": "Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance forward or backward from that position?", "option": [ "Forward", "Backward" ], "answer": "Forward", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance forward or backward from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Forward', 'Backward'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0079_00/00480.jpg", "scene0079_00/00520.jpg", "scene0079_00/00560.jpg", "scene0079_00/00600.jpg" ], "turn_id": 4, "scan_id": "scene0079_00", "origin_question": "When did you first discover \"the tall, white rectangular machine with a digital display\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the tall, white rectangular machine with a digital display\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0079_00/00640.jpg", "scene0079_00/00680.jpg", "scene0079_00/00720.jpg", "scene0079_00/00760.jpg" ], "turn_id": 5, "scan_id": "scene0079_00", "origin_question": "At the end of which turn were both of \"the light gray multifunctional copier with advanced features\" and \"the tall, white rectangular machine with a digital display\" on your right side?", "option": [], "answer": "1", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the light gray multifunctional copier with advanced features\" and \"the tall, white rectangular machine with a digital display\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0079_00/00800.jpg", "scene0079_00/00840.jpg", "scene0079_00/00880.jpg", "scene0079_00/00920.jpg" ], "turn_id": 6, "scan_id": "scene0079_00", "origin_question": "Is the dark fluffy cotton jacket with a synthetic blend to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the dark fluffy cotton jacket with a synthetic blend to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0079_00/00960.jpg", "scene0079_00/01000.jpg", "scene0079_00/01040.jpg", "scene0079_00/01080.jpg" ], "turn_id": 7, "scan_id": "scene0079_00", "origin_question": "In which turn did you see both \"the flat screen monitor with a dark matte surface\" and \"the dark fluffy cotton jacket with a synthetic blend\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the flat screen monitor with a dark matte surface\" and \"the dark fluffy cotton jacket with a synthetic blend\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0047_region22/0e31d2cf016445c5b4adf1aceeed25c4_i1_4.jpg", "1mp3d_0047_region22/0e31d2cf016445c5b4adf1aceeed25c4_i2_4.jpg", "1mp3d_0047_region22/960977315a044216b593b236d429d0e3_i1_1.jpg", "1mp3d_0047_region22/960977315a044216b593b236d429d0e3_i2_1.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0047_region22", "origin_question": "Remember, have you seen any candle(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any candle(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 3 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "1mp3d_0047_region22/960977315a044216b593b236d429d0e3_i1_2.jpg", "1mp3d_0047_region22/960977315a044216b593b236d429d0e3_i2_2.jpg", "1mp3d_0047_region22/960977315a044216b593b236d429d0e3_i1_3.jpg", "1mp3d_0047_region22/960977315a044216b593b236d429d0e3_i2_3.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0047_region22", "origin_question": "When did you first discover \"the sleek black chair with a modern design\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the sleek black chair with a modern design\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 3 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "1mp3d_0047_region22/960977315a044216b593b236d429d0e3_i1_4.jpg", "1mp3d_0047_region22/960977315a044216b593b236d429d0e3_i2_4.jpg", "1mp3d_0047_region22/0e31d2cf016445c5b4adf1aceeed25c4_i1_5.jpg", "1mp3d_0047_region22/0e31d2cf016445c5b4adf1aceeed25c4_i2_5.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0047_region22", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the tall, sleek white glass cabinet with a modern design\"; \"the medium-sized hardcover book with a blue cover\"; \"the laptop\" ", "option": [ "the tall, sleek white glass cabinet with a modern design", "the medium-sized hardcover book with a blue cover", "the laptop" ], "answer": "the laptop", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the tall, sleek white glass cabinet with a modern design\"; \"the medium-sized hardcover book with a blue cover\"; \"the laptop\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the tall, sleek white glass cabinet with a modern design', 'the medium-sized hardcover book with a blue cover', 'the laptop'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 3 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0331_01/00000.jpg", "scene0331_01/00040.jpg", "scene0331_01/00080.jpg", "scene0331_01/00120.jpg", "scene0331_01/00160.jpg" ], "turn_id": 1, "scan_id": "scene0331_01", "origin_question": "Remember, have you seen any light(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any light(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0331_01/00200.jpg", "scene0331_01/00240.jpg", "scene0331_01/00280.jpg", "scene0331_01/00320.jpg", "scene0331_01/00360.jpg" ], "turn_id": 2, "scan_id": "scene0331_01", "origin_question": "Which direction is the wooden rocking chair with a gray seat cushion in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the wooden rocking chair with a gray seat cushion in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0331_01/00400.jpg", "scene0331_01/00440.jpg", "scene0331_01/00480.jpg", "scene0331_01/00520.jpg", "scene0331_01/00560.jpg" ], "turn_id": 3, "scan_id": "scene0331_01", "origin_question": "When did you first discover \"the small, white round tabletop fan made of plastic\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the small, white round tabletop fan made of plastic\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0331_01/00600.jpg", "scene0331_01/00640.jpg", "scene0331_01/00680.jpg", "scene0331_01/00720.jpg", "scene0331_01/00760.jpg" ], "turn_id": 4, "scan_id": "scene0331_01", "origin_question": "Please recall \"the desk\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.77", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the desk\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0331_01/00800.jpg", "scene0331_01/00840.jpg", "scene0331_01/00880.jpg", "scene0331_01/00920.jpg", "scene0331_01/00960.jpg" ], "turn_id": 5, "scan_id": "scene0331_01", "origin_question": "Is the cozy white blanket with snowflake patterns to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the cozy white blanket with snowflake patterns to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0331_01/01000.jpg", "scene0331_01/01040.jpg", "scene0331_01/01080.jpg", "scene0331_01/01120.jpg", "scene0331_01/01160.jpg" ], "turn_id": 6, "scan_id": "scene0331_01", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the wooden rocking chair with a gray seat cushion\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the wooden rocking chair with a gray seat cushion\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0331_01/01200.jpg", "scene0331_01/01240.jpg", "scene0331_01/01280.jpg", "scene0331_01/01320.jpg", "scene0331_01/01360.jpg" ], "turn_id": 7, "scan_id": "scene0331_01", "origin_question": "In which turn were you closest to \"the red and black nylon backpack with multiple compartments\"? ", "option": [], "answer": "1", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the red and black nylon backpack with multiple compartments\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0331_01/01400.jpg", "scene0331_01/01440.jpg", "scene0331_01/01480.jpg", "scene0331_01/01520.jpg", "scene0331_01/01560.jpg" ], "turn_id": 8, "scan_id": "scene0331_01", "origin_question": "Please recall \"the decoration\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.54", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the decoration\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0331_01/01600.jpg", "scene0331_01/01640.jpg", "scene0331_01/01680.jpg", "scene0331_01/01720.jpg", "scene0331_01/01760.jpg" ], "turn_id": 9, "scan_id": "scene0331_01", "origin_question": "Please recall \"the red and black nylon backpack with multiple compartments\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.67", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the red and black nylon backpack with multiple compartments\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0331_01/01800.jpg", "scene0331_01/01840.jpg", "scene0331_01/01880.jpg", "scene0331_01/01920.jpg", "scene0331_01/01960.jpg" ], "turn_id": 10, "scan_id": "scene0331_01", "origin_question": "Is the decoration to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the decoration to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0047_region14/badeec48e1324c82844cc3005a7f7788_i1_1.jpg", "1mp3d_0047_region14/badeec48e1324c82844cc3005a7f7788_i2_1.jpg", "1mp3d_0047_region14/a5e189644ebf47fb9b8932d4de1fb41d_i1_0.jpg", "1mp3d_0047_region14/a5e189644ebf47fb9b8932d4de1fb41d_i2_0.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0047_region14", "origin_question": "Remember, have you seen any table(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any table(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "1mp3d_0047_region14/eec1a7de0fb347b885cb947afce0a171_i1_3.jpg", "1mp3d_0047_region14/eec1a7de0fb347b885cb947afce0a171_i2_3.jpg", "1mp3d_0047_region14/a5e189644ebf47fb9b8932d4de1fb41d_i1_5.jpg", "1mp3d_0047_region14/a5e189644ebf47fb9b8932d4de1fb41d_i2_5.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0047_region14", "origin_question": "In which turn were you farthest from \"the white modern couch with a curved shape\"? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the white modern couch with a curved shape\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "1mp3d_0047_region14/a5e189644ebf47fb9b8932d4de1fb41d_i1_4.jpg", "1mp3d_0047_region14/a5e189644ebf47fb9b8932d4de1fb41d_i2_4.jpg", "1mp3d_0047_region14/badeec48e1324c82844cc3005a7f7788_i1_3.jpg", "1mp3d_0047_region14/badeec48e1324c82844cc3005a7f7788_i2_3.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0047_region14", "origin_question": "What will be the first-time appearance order of the following categories: cabinet, chair, fireplace?", "option": [ "cabinet, chair, fireplace", "fireplace, cabinet, chair", "chair, cabinet, fireplace", "fireplace, chair, cabinet" ], "answer": "chair, cabinet, fireplace", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: cabinet, chair, fireplace? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['cabinet, chair, fireplace', 'fireplace, cabinet, chair', 'chair, cabinet, fireplace', 'fireplace, chair, cabinet'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "1mp3d_0047_region14/6e1f2997c6644be3a3a6bf853b45956d_i1_5.jpg", "1mp3d_0047_region14/6e1f2997c6644be3a3a6bf853b45956d_i2_5.jpg", "1mp3d_0047_region14/704a0258ae694d188fb4fbebed04ba93_i1_1.jpg", "1mp3d_0047_region14/704a0258ae694d188fb4fbebed04ba93_i2_1.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0047_region14", "origin_question": "When did you first discover \"the white modern couch with a curved shape\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the white modern couch with a curved shape\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0047_region14/6e1f2997c6644be3a3a6bf853b45956d_i1_1.jpg", "1mp3d_0047_region14/6e1f2997c6644be3a3a6bf853b45956d_i2_1.jpg", "1mp3d_0047_region14/4b0d49c9ea0547a78f0ac97c54c40799_i1_0.jpg", "1mp3d_0047_region14/4b0d49c9ea0547a78f0ac97c54c40799_i2_0.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0047_region14", "origin_question": "Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the white modern couch with a curved shape\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the white modern couch with a curved shape\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "1mp3d_0047_region14/a92caa72325040bea8093cd576dd9a8d_i1_1.jpg", "1mp3d_0047_region14/a92caa72325040bea8093cd576dd9a8d_i2_1.jpg", "1mp3d_0047_region14/a92caa72325040bea8093cd576dd9a8d_i1_0.jpg", "1mp3d_0047_region14/a92caa72325040bea8093cd576dd9a8d_i2_0.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0047_region14", "origin_question": "Using your orientation at the end of turn 5 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 5 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "1mp3d_0047_region14/4b0d49c9ea0547a78f0ac97c54c40799_i1_4.jpg", "1mp3d_0047_region14/4b0d49c9ea0547a78f0ac97c54c40799_i2_4.jpg", "1mp3d_0047_region14/7fd28d87e981467fa842ed19b5069fd5_i1_4.jpg", "1mp3d_0047_region14/7fd28d87e981467fa842ed19b5069fd5_i2_4.jpg" ], "turn_id": 7, "scan_id": "1mp3d_0047_region14", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the white modern couch with a curved shape\" from your position?", "option": [], "answer": "160.2", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the white modern couch with a curved shape\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "1mp3d_0047_region14/560710839eb64225a5e4dda527556aba_i1_4.jpg", "1mp3d_0047_region14/560710839eb64225a5e4dda527556aba_i2_4.jpg", "1mp3d_0047_region14/5ab92ff9c0ea493e847aeba98e8e9f23_i1_2.jpg", "1mp3d_0047_region14/5ab92ff9c0ea493e847aeba98e8e9f23_i2_2.jpg" ], "turn_id": 8, "scan_id": "1mp3d_0047_region14", "origin_question": "By the end of the 4 turn, were you getting closer to or farther away from these two objects: X: \"the white modern couch with a curved shape\", Y: \"the small, rectangular glossy red cabinet\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to Y and farther away from X", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 4 turn, were you getting closer to or farther away from these two objects: X: \"the white modern couch with a curved shape\", Y: \"the small, rectangular glossy red cabinet\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "1mp3d_0047_region14/560710839eb64225a5e4dda527556aba_i1_3.jpg", "1mp3d_0047_region14/560710839eb64225a5e4dda527556aba_i2_3.jpg", "1mp3d_0047_region14/7fd28d87e981467fa842ed19b5069fd5_i1_1.jpg", "1mp3d_0047_region14/7fd28d87e981467fa842ed19b5069fd5_i2_1.jpg" ], "turn_id": 9, "scan_id": "1mp3d_0047_region14", "origin_question": "Using your orientation at the end of turn 6 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 9 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 6 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0264_01/00000.jpg", "scene0264_01/00040.jpg", "scene0264_01/00080.jpg", "scene0264_01/00120.jpg", "scene0264_01/00160.jpg" ], "turn_id": 1, "scan_id": "scene0264_01", "origin_question": "Remember, have you seen any keyboard(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any keyboard(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0264_01/00200.jpg", "scene0264_01/00240.jpg", "scene0264_01/00280.jpg", "scene0264_01/00320.jpg", "scene0264_01/00360.jpg" ], "turn_id": 2, "scan_id": "scene0264_01", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the computer\"; \"the rectangular, white hardcover book made of paper\"; \"the black and white metal printer with a sleek design\" ", "option": [ "the computer", "the rectangular, white hardcover book made of paper", "the black and white metal printer with a sleek design" ], "answer": "the rectangular, white hardcover book made of paper", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the computer\"; \"the rectangular, white hardcover book made of paper\"; \"the black and white metal printer with a sleek design\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the computer', 'the rectangular, white hardcover book made of paper', 'the black and white metal printer with a sleek design'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0264_01/00400.jpg", "scene0264_01/00440.jpg", "scene0264_01/00480.jpg", "scene0264_01/00520.jpg", "scene0264_01/00560.jpg" ], "turn_id": 3, "scan_id": "scene0264_01", "origin_question": "Which direction is the black and white metal printer with a sleek design in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the black and white metal printer with a sleek design in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0264_01/00600.jpg", "scene0264_01/00640.jpg", "scene0264_01/00680.jpg", "scene0264_01/00720.jpg", "scene0264_01/00760.jpg" ], "turn_id": 4, "scan_id": "scene0264_01", "origin_question": "Which two objects are on the same side of you now? \"the black and white metal printer with a sleek design\", \"the white plastic computer mouse with a scroll wheel\", \"the white metal copier with a blue front panel\". ", "option": [ "the black and white metal printer with a sleek design, the white plastic computer mouse with a scroll wheel", "the black and white metal printer with a sleek design, the white metal copier with a blue front panel", "the white plastic computer mouse with a scroll wheel, the white metal copier with a blue front panel" ], "answer": "the black and white metal printer with a sleek design, the white plastic computer mouse with a scroll wheel", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the black and white metal printer with a sleek design\", \"the white plastic computer mouse with a scroll wheel\", \"the white metal copier with a blue front panel\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the black and white metal printer with a sleek design, the white plastic computer mouse with a scroll wheel', 'the black and white metal printer with a sleek design, the white metal copier with a blue front panel', 'the white plastic computer mouse with a scroll wheel, the white metal copier with a blue front panel'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0264_01/00800.jpg", "scene0264_01/00840.jpg", "scene0264_01/00880.jpg", "scene0264_01/00920.jpg", "scene0264_01/00960.jpg" ], "turn_id": 5, "scan_id": "scene0264_01", "origin_question": "Which two objects are on the same side of you now? \"the black and white metal printer with a sleek design\", \"the computer\", \"the rectangular, white hardcover book made of paper\". ", "option": [ "the black and white metal printer with a sleek design, the computer", "the black and white metal printer with a sleek design, the rectangular, white hardcover book made of paper", "the computer, the rectangular, white hardcover book made of paper" ], "answer": "the black and white metal printer with a sleek design, the computer", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the black and white metal printer with a sleek design\", \"the computer\", \"the rectangular, white hardcover book made of paper\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the black and white metal printer with a sleek design, the computer', 'the black and white metal printer with a sleek design, the rectangular, white hardcover book made of paper', 'the computer, the rectangular, white hardcover book made of paper'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0264_01/01000.jpg", "scene0264_01/01040.jpg", "scene0264_01/01080.jpg", "scene0264_01/01120.jpg", "scene0264_01/01160.jpg" ], "turn_id": 6, "scan_id": "scene0264_01", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the computer\" from your position?", "option": [], "answer": "70.62", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the computer\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0264_01/01200.jpg", "scene0264_01/01240.jpg", "scene0264_01/01280.jpg", "scene0264_01/01320.jpg", "scene0264_01/01360.jpg" ], "turn_id": 7, "scan_id": "scene0264_01", "origin_question": "Which two objects are on the same side of you now? \"the rectangular, white hardcover book made of paper\", \"the rectangular white door made of wood\", \"the white metal copier with a blue front panel\". ", "option": [ "the rectangular, white hardcover book made of paper, the rectangular white door made of wood", "the rectangular, white hardcover book made of paper, the white metal copier with a blue front panel", "the rectangular white door made of wood, the white metal copier with a blue front panel" ], "answer": "the rectangular white door made of wood, the white metal copier with a blue front panel", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the rectangular, white hardcover book made of paper\", \"the rectangular white door made of wood\", \"the white metal copier with a blue front panel\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular, white hardcover book made of paper, the rectangular white door made of wood', 'the rectangular, white hardcover book made of paper, the white metal copier with a blue front panel', 'the rectangular white door made of wood, the white metal copier with a blue front panel'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "scene0264_01/01400.jpg", "scene0264_01/01440.jpg", "scene0264_01/01480.jpg", "scene0264_01/01520.jpg", "scene0264_01/01560.jpg" ], "turn_id": 8, "scan_id": "scene0264_01", "origin_question": "Among these three objects(\"the white plastic computer mouse with a scroll wheel\",\"the rectangular, white hardcover book made of paper\",\"the white metal copier with a blue front panel\" ), which one is the closest to you?", "option": [ "the white plastic computer mouse with a scroll wheel", "the rectangular, white hardcover book made of paper", "the white metal copier with a blue front panel" ], "answer": "the white metal copier with a blue front panel", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the white plastic computer mouse with a scroll wheel\",\"the rectangular, white hardcover book made of paper\",\"the white metal copier with a blue front panel\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the white plastic computer mouse with a scroll wheel', 'the rectangular, white hardcover book made of paper', 'the white metal copier with a blue front panel'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0663_00/00000.jpg", "scene0663_00/00040.jpg", "scene0663_00/00080.jpg", "scene0663_00/00120.jpg", "scene0663_00/00160.jpg" ], "turn_id": 1, "scan_id": "scene0663_00", "origin_question": "Remember, how many cup(s) have you seen so far? ", "option": [], "answer": "2", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many cup(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0663_00/00200.jpg", "scene0663_00/00240.jpg", "scene0663_00/00280.jpg", "scene0663_00/00320.jpg", "scene0663_00/00360.jpg" ], "turn_id": 2, "scan_id": "scene0663_00", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the matte black oval-shaped mouse made of plastic\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the matte black oval-shaped mouse made of plastic\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0663_00/00400.jpg", "scene0663_00/00440.jpg", "scene0663_00/00480.jpg", "scene0663_00/00520.jpg", "scene0663_00/00560.jpg" ], "turn_id": 3, "scan_id": "scene0663_00", "origin_question": "Remember, how many monitor(s) have you seen so far? ", "option": [], "answer": "2", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many monitor(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0663_00/00600.jpg", "scene0663_00/00640.jpg", "scene0663_00/00680.jpg", "scene0663_00/00720.jpg", "scene0663_00/00760.jpg" ], "turn_id": 4, "scan_id": "scene0663_00", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the desk\"; \"the large, smooth, rectangular blackboard\"; \"the modern chair with a gray cushioned seat\" ", "option": [ "the desk", "the large, smooth, rectangular blackboard", "the modern chair with a gray cushioned seat" ], "answer": "the large, smooth, rectangular blackboard", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the desk\"; \"the large, smooth, rectangular blackboard\"; \"the modern chair with a gray cushioned seat\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the desk', 'the large, smooth, rectangular blackboard', 'the modern chair with a gray cushioned seat'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0663_00/00800.jpg", "scene0663_00/00840.jpg", "scene0663_00/00880.jpg", "scene0663_00/00920.jpg", "scene0663_00/00960.jpg" ], "turn_id": 5, "scan_id": "scene0663_00", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the sleek rectangular telephone in gray or silver\"; \"the glossy, vertical wooden door with a glass panel\"; \"the large, smooth, rectangular blackboard\" ", "option": [ "the sleek rectangular telephone in gray or silver", "the glossy, vertical wooden door with a glass panel", "the large, smooth, rectangular blackboard" ], "answer": "the glossy, vertical wooden door with a glass panel", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the sleek rectangular telephone in gray or silver\"; \"the glossy, vertical wooden door with a glass panel\"; \"the large, smooth, rectangular blackboard\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the sleek rectangular telephone in gray or silver', 'the glossy, vertical wooden door with a glass panel', 'the large, smooth, rectangular blackboard'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0663_00/01000.jpg", "scene0663_00/01040.jpg", "scene0663_00/01080.jpg", "scene0663_00/01120.jpg", "scene0663_00/01160.jpg" ], "turn_id": 6, "scan_id": "scene0663_00", "origin_question": "Remember, have you seen any monitor(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any monitor(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0663_00/01200.jpg", "scene0663_00/01240.jpg", "scene0663_00/01280.jpg", "scene0663_00/01320.jpg", "scene0663_00/01360.jpg" ], "turn_id": 7, "scan_id": "scene0663_00", "origin_question": "Which direction is the matte black oval-shaped mouse made of plastic in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the matte black oval-shaped mouse made of plastic in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "scene0663_00/01400.jpg", "scene0663_00/01440.jpg", "scene0663_00/01480.jpg", "scene0663_00/01520.jpg", "scene0663_00/01560.jpg" ], "turn_id": 8, "scan_id": "scene0663_00", "origin_question": "Among these three objects(\"the matte black oval-shaped mouse made of plastic\",\"the large, smooth, rectangular blackboard\",\"the cylindrical blue adhesive tape in standard size\" ), which one is the farthest from you?", "option": [ "the matte black oval-shaped mouse made of plastic", "the large, smooth, rectangular blackboard", "the cylindrical blue adhesive tape in standard size" ], "answer": "the cylindrical blue adhesive tape in standard size", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the matte black oval-shaped mouse made of plastic\",\"the large, smooth, rectangular blackboard\",\"the cylindrical blue adhesive tape in standard size\" ), which one is the farthest from you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the matte black oval-shaped mouse made of plastic', 'the large, smooth, rectangular blackboard', 'the cylindrical blue adhesive tape in standard size'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0663_00/01600.jpg", "scene0663_00/01640.jpg", "scene0663_00/01680.jpg", "scene0663_00/01720.jpg", "scene0663_00/01760.jpg" ], "turn_id": 9, "scan_id": "scene0663_00", "origin_question": "In which turn did you see both \"the glossy, vertical wooden door with a glass panel\" and \"the cylindrical blue adhesive tape in standard size\" simultaneously? ", "option": [], "answer": "6", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the glossy, vertical wooden door with a glass panel\" and \"the cylindrical blue adhesive tape in standard size\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0663_00/01800.jpg", "scene0663_00/01840.jpg", "scene0663_00/01880.jpg", "scene0663_00/01920.jpg", "scene0663_00/01960.jpg" ], "turn_id": 10, "scan_id": "scene0663_00", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0030_region1/860f7e4e9a944bbd856dfcc5f2d1ad2b_i1_3.jpg", "1mp3d_0030_region1/860f7e4e9a944bbd856dfcc5f2d1ad2b_i2_3.jpg", "1mp3d_0030_region1/860f7e4e9a944bbd856dfcc5f2d1ad2b_i1_4.jpg", "1mp3d_0030_region1/860f7e4e9a944bbd856dfcc5f2d1ad2b_i2_4.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0030_region1", "origin_question": "Remember, have you seen any toilet(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any toilet(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 1 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0652_00/00000.jpg", "scene0652_00/00040.jpg", "scene0652_00/00080.jpg", "scene0652_00/00120.jpg" ], "turn_id": 1, "scan_id": "scene0652_00", "origin_question": "Remember, have you seen any cabinet(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any cabinet(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0652_00/00160.jpg", "scene0652_00/00200.jpg", "scene0652_00/00240.jpg", "scene0652_00/00280.jpg" ], "turn_id": 2, "scan_id": "scene0652_00", "origin_question": "Is the wooden picture in a brown frame with trees to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the wooden picture in a brown frame with trees to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0652_00/00320.jpg", "scene0652_00/00360.jpg", "scene0652_00/00400.jpg", "scene0652_00/00440.jpg" ], "turn_id": 3, "scan_id": "scene0652_00", "origin_question": "Which direction is the hardcover, rectangular white book in good condition in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the hardcover, rectangular white book in good condition in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0652_00/00480.jpg", "scene0652_00/00520.jpg", "scene0652_00/00560.jpg", "scene0652_00/00600.jpg" ], "turn_id": 4, "scan_id": "scene0652_00", "origin_question": "Please recall \"the curtain\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.34", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the curtain\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "scene0652_00/00640.jpg", "scene0652_00/00680.jpg", "scene0652_00/00720.jpg", "scene0652_00/00760.jpg" ], "turn_id": 5, "scan_id": "scene0652_00", "origin_question": "Among these three objects(\"the black leather couch with white and gray cushions\",\"the rectangular glass window with a light-colored frame\",\"the rectangular stand with a clear glass top and black metal frame\" ), which one is the closest to you?", "option": [ "the black leather couch with white and gray cushions", "the rectangular glass window with a light-colored frame", "the rectangular stand with a clear glass top and black metal frame" ], "answer": "the black leather couch with white and gray cushions", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects(\"the black leather couch with white and gray cushions\",\"the rectangular glass window with a light-colored frame\",\"the rectangular stand with a clear glass top and black metal frame\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the black leather couch with white and gray cushions', 'the rectangular glass window with a light-colored frame', 'the rectangular stand with a clear glass top and black metal frame'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0652_00/00800.jpg", "scene0652_00/00840.jpg", "scene0652_00/00880.jpg", "scene0652_00/00920.jpg" ], "turn_id": 6, "scan_id": "scene0652_00", "origin_question": "What will be the first-time appearance order of the following categories: curtain, paper, pillow?", "option": [ "curtain, paper, pillow", "paper, pillow, curtain", "paper, curtain, pillow", "curtain, pillow, paper" ], "answer": "curtain, pillow, paper", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: curtain, paper, pillow? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['curtain, paper, pillow', 'paper, pillow, curtain', 'paper, curtain, pillow', 'curtain, pillow, paper'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0652_00/00960.jpg", "scene0652_00/01000.jpg", "scene0652_00/01040.jpg", "scene0652_00/01080.jpg" ], "turn_id": 7, "scan_id": "scene0652_00", "origin_question": "By the end of the 5 turn, were you getting closer to or farther away from these two objects: X: \"the flat-screen TV with a deep black color\", Y: \"the rectangular glass window with a light-colored frame\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to X and farther away from Y", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 5 turn, were you getting closer to or farther away from these two objects: X: \"the flat-screen TV with a deep black color\", Y: \"the rectangular glass window with a light-colored frame\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0652_00/01120.jpg", "scene0652_00/01160.jpg", "scene0652_00/01200.jpg", "scene0652_00/01240.jpg" ], "turn_id": 8, "scan_id": "scene0652_00", "origin_question": "At the end of which turn were both of \"the rectangular glass window with a light-colored frame\" and \"the rectangular stand with a clear glass top and black metal frame\" on your left side?", "option": [], "answer": "1", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the rectangular glass window with a light-colored frame\" and \"the rectangular stand with a clear glass top and black metal frame\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0652_00/01280.jpg", "scene0652_00/01320.jpg", "scene0652_00/01360.jpg", "scene0652_00/01400.jpg" ], "turn_id": 9, "scan_id": "scene0652_00", "origin_question": "Compared to your position at the end of the 7 turn, are you now closer or farther away from \"the rectangular white bed made of sturdy materials\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 9 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 7 turn, are you now closer or farther away from \"the rectangular white bed made of sturdy materials\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0044_region0/fb147ffd72b848efa087731fd61dceb0_i1_3.jpg", "1mp3d_0044_region0/fb147ffd72b848efa087731fd61dceb0_i2_3.jpg", "1mp3d_0044_region0/53633459fd384c8082d663d8968590bd_i1_2.jpg", "1mp3d_0044_region0/53633459fd384c8082d663d8968590bd_i2_2.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0044_region0", "origin_question": "Remember, have you seen any curtain(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any curtain(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0044_region0/149f4021008f4bd282b87b11d782cdf0_i1_5.jpg", "1mp3d_0044_region0/149f4021008f4bd282b87b11d782cdf0_i2_5.jpg", "1mp3d_0044_region0/314e186ad1ea4f80892affce5ccaa715_i1_1.jpg", "1mp3d_0044_region0/314e186ad1ea4f80892affce5ccaa715_i2_1.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0044_region0", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the sleek black rectangular flat-panel TV\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the sleek black rectangular flat-panel TV\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "1mp3d_0044_region0/149f4021008f4bd282b87b11d782cdf0_i1_0.jpg", "1mp3d_0044_region0/149f4021008f4bd282b87b11d782cdf0_i2_0.jpg", "1mp3d_0044_region0/149f4021008f4bd282b87b11d782cdf0_i1_1.jpg", "1mp3d_0044_region0/149f4021008f4bd282b87b11d782cdf0_i2_1.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0044_region0", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "1.57", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "1mp3d_0044_region0/53633459fd384c8082d663d8968590bd_i1_4.jpg", "1mp3d_0044_region0/53633459fd384c8082d663d8968590bd_i2_4.jpg", "1mp3d_0044_region0/53633459fd384c8082d663d8968590bd_i1_3.jpg", "1mp3d_0044_region0/53633459fd384c8082d663d8968590bd_i2_3.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0044_region0", "origin_question": "By the end of the 3 turn, were you getting closer to or farther away from these two objects: X: \"the mirror\", Y: \"the black and silver metal rack for dumbbells\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to Y and farther away from X", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 3 turn, were you getting closer to or farther away from these two objects: X: \"the mirror\", Y: \"the black and silver metal rack for dumbbells\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "1mp3d_0044_region0/fb147ffd72b848efa087731fd61dceb0_i1_5.jpg", "1mp3d_0044_region0/fb147ffd72b848efa087731fd61dceb0_i2_5.jpg", "1mp3d_0044_region0/fb147ffd72b848efa087731fd61dceb0_i1_1.jpg", "1mp3d_0044_region0/fb147ffd72b848efa087731fd61dceb0_i2_1.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0044_region0", "origin_question": "At the end of which turn were both of \"the mirror\" and \"the large rectangular dark brown wooden door\" on your right side?", "option": [], "answer": "5", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the mirror\" and \"the large rectangular dark brown wooden door\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "1mp3d_0044_region0/71f815ce8c124839a2875c75d92bb1d5_i2_1.jpg", "1mp3d_0044_region0/b5c89126d17d47fea581e6db97107874_i1_0.jpg", "1mp3d_0044_region0/b5c89126d17d47fea581e6db97107874_i2_0.jpg", "1mp3d_0044_region0/71f815ce8c124839a2875c75d92bb1d5_i1_0.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0044_region0", "origin_question": "Is the large rectangular dark brown wooden door to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the large rectangular dark brown wooden door to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0040_region25/e5440b494f3d483d91d857f8eb126b1d_i1_3.jpg", "1mp3d_0040_region25/e5440b494f3d483d91d857f8eb126b1d_i2_3.jpg", "1mp3d_0040_region25/ceae9606b23d4b31a157923b09655baa_i1_2.jpg", "1mp3d_0040_region25/ceae9606b23d4b31a157923b09655baa_i2_2.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0040_region25", "origin_question": "Remember, have you seen any balcony(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any balcony(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "1mp3d_0040_region25/e5440b494f3d483d91d857f8eb126b1d_i1_1.jpg", "1mp3d_0040_region25/e5440b494f3d483d91d857f8eb126b1d_i2_1.jpg", "1mp3d_0040_region25/e5440b494f3d483d91d857f8eb126b1d_i1_0.jpg", "1mp3d_0040_region25/e5440b494f3d483d91d857f8eb126b1d_i2_0.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0040_region25", "origin_question": "Remember, how many candlestick(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many candlestick(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "1mp3d_0040_region25/e5440b494f3d483d91d857f8eb126b1d_i1_5.jpg", "1mp3d_0040_region25/e5440b494f3d483d91d857f8eb126b1d_i2_5.jpg", "1mp3d_0040_region25/e5440b494f3d483d91d857f8eb126b1d_i1_4.jpg", "1mp3d_0040_region25/e5440b494f3d483d91d857f8eb126b1d_i2_4.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0040_region25", "origin_question": "In which turn did you see both \"the large square silver stainless steel sink\" and \"the stylish silver metal faucet with a modern design\" simultaneously? ", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the large square silver stainless steel sink\" and \"the stylish silver metal faucet with a modern design\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "1mp3d_0040_region25/ceae9606b23d4b31a157923b09655baa_i1_3.jpg", "1mp3d_0040_region25/ceae9606b23d4b31a157923b09655baa_i2_3.jpg", "1mp3d_0040_region25/ea4e40a346ce4542978c7df62e3080da_i1_2.jpg", "1mp3d_0040_region25/ea4e40a346ce4542978c7df62e3080da_i2_2.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0040_region25", "origin_question": "How far is your current position from where you were at the end of the turn 3? (in meters)", "option": [], "answer": "2.19", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 3? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "1mp3d_0040_region25/ea4e40a346ce4542978c7df62e3080da_i1_1.jpg", "1mp3d_0040_region25/ea4e40a346ce4542978c7df62e3080da_i2_1.jpg", "1mp3d_0040_region25/ceae9606b23d4b31a157923b09655baa_i1_1.jpg", "1mp3d_0040_region25/ceae9606b23d4b31a157923b09655baa_i2_1.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0040_region25", "origin_question": "By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the spherical green plant in a rectangular pot\", Y: \"the stylish silver metal faucet with a modern design\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the spherical green plant in a rectangular pot\", Y: \"the stylish silver metal faucet with a modern design\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "1mp3d_0040_region25/ceae9606b23d4b31a157923b09655baa_i1_0.jpg", "1mp3d_0040_region25/ceae9606b23d4b31a157923b09655baa_i2_0.jpg", "1mp3d_0040_region25/ceae9606b23d4b31a157923b09655baa_i1_4.jpg", "1mp3d_0040_region25/ceae9606b23d4b31a157923b09655baa_i2_4.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0040_region25", "origin_question": "Please recall \"the spherical green plant in a rectangular pot\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.85", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the spherical green plant in a rectangular pot\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0047_region7/bf2cce0aae8f4c81b8dd4a167da03ebf_i1_3.jpg", "1mp3d_0047_region7/bf2cce0aae8f4c81b8dd4a167da03ebf_i2_3.jpg", "1mp3d_0047_region7/78aa4d3dfb8c4c8581a8709e7b9bcde0_i1_1.jpg", "1mp3d_0047_region7/78aa4d3dfb8c4c8581a8709e7b9bcde0_i2_1.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0047_region7", "origin_question": "Remember, have you seen any window(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any window(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "1mp3d_0047_region7/78aa4d3dfb8c4c8581a8709e7b9bcde0_i1_0.jpg", "1mp3d_0047_region7/78aa4d3dfb8c4c8581a8709e7b9bcde0_i2_0.jpg", "1mp3d_0047_region7/bf2cce0aae8f4c81b8dd4a167da03ebf_i1_0.jpg", "1mp3d_0047_region7/bf2cce0aae8f4c81b8dd4a167da03ebf_i2_0.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0047_region7", "origin_question": "Please recall \"the tall rectangular deep blue door with a silver handle\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.85", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the tall rectangular deep blue door with a silver handle\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "1mp3d_0047_region7/bf2cce0aae8f4c81b8dd4a167da03ebf_i1_1.jpg", "1mp3d_0047_region7/bf2cce0aae8f4c81b8dd4a167da03ebf_i2_1.jpg", "1mp3d_0047_region7/bf2cce0aae8f4c81b8dd4a167da03ebf_i1_2.jpg", "1mp3d_0047_region7/bf2cce0aae8f4c81b8dd4a167da03ebf_i2_2.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0047_region7", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "1mp3d_0047_region7/626d2523eff1429f815cdf7431f25ad9_i1_0.jpg", "1mp3d_0047_region7/626d2523eff1429f815cdf7431f25ad9_i2_0.jpg", "1mp3d_0047_region7/5e336d4bd3eb4328ad02aaccc4f3dbb7_i1_4.jpg", "1mp3d_0047_region7/5e336d4bd3eb4328ad02aaccc4f3dbb7_i2_4.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0047_region7", "origin_question": "Which two objects are on the same side of you now? \"the cushion\", \"the tall rectangular deep blue door with a silver handle\", \"then orange, modern chair with a curved backrest\". ", "option": [ "the cushion, the tall rectangular deep blue door with a silver handle", "the cushion, then orange, modern chair with a curved backrest", "the tall rectangular deep blue door with a silver handle, then orange, modern chair with a curved backrest" ], "answer": "the tall rectangular deep blue door with a silver handle, then orange, modern chair with a curved backrest", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the cushion\", \"the tall rectangular deep blue door with a silver handle\", \"then orange, modern chair with a curved backrest\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the cushion, the tall rectangular deep blue door with a silver handle', 'the cushion, then orange, modern chair with a curved backrest', 'the tall rectangular deep blue door with a silver handle, then orange, modern chair with a curved backrest'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "1mp3d_0047_region7/5e336d4bd3eb4328ad02aaccc4f3dbb7_i1_3.jpg", "1mp3d_0047_region7/5e336d4bd3eb4328ad02aaccc4f3dbb7_i2_3.jpg", "1mp3d_0047_region7/5e336d4bd3eb4328ad02aaccc4f3dbb7_i1_2.jpg", "1mp3d_0047_region7/5e336d4bd3eb4328ad02aaccc4f3dbb7_i2_2.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0047_region7", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the tall rectangular deep blue door with a silver handle\", Y: \"the cushion\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the tall rectangular deep blue door with a silver handle\", Y: \"the cushion\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "1mp3d_0047_region7/626d2523eff1429f815cdf7431f25ad9_i1_4.jpg", "1mp3d_0047_region7/626d2523eff1429f815cdf7431f25ad9_i2_4.jpg", "1mp3d_0047_region7/cafb18e8c7b041a8b4f85bf34cf4609d_i1_1.jpg", "1mp3d_0047_region7/cafb18e8c7b041a8b4f85bf34cf4609d_i2_1.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0047_region7", "origin_question": "Which two objects are on the same side of you now? \"the tall rectangular deep blue door with a silver handle\", \"the cushion\", \"the table\". ", "option": [ "the tall rectangular deep blue door with a silver handle, the cushion", "the tall rectangular deep blue door with a silver handle, the table", "the cushion, the table" ], "answer": "the tall rectangular deep blue door with a silver handle, the cushion", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the tall rectangular deep blue door with a silver handle\", \"the cushion\", \"the table\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the tall rectangular deep blue door with a silver handle, the cushion', 'the tall rectangular deep blue door with a silver handle, the table', 'the cushion, the table'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0029_region3/39d7ad2d721e47aa9fe29090663265e8_i1_3.jpg", "1mp3d_0029_region3/39d7ad2d721e47aa9fe29090663265e8_i2_3.jpg", "1mp3d_0029_region3/0b7767a8959a48719e014e9ab7d3597b_i1_5.jpg", "1mp3d_0029_region3/0b7767a8959a48719e014e9ab7d3597b_i2_5.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0029_region3", "origin_question": "Remember, have you seen any light(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any light(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "1mp3d_0029_region3/39d7ad2d721e47aa9fe29090663265e8_i1_4.jpg", "1mp3d_0029_region3/39d7ad2d721e47aa9fe29090663265e8_i2_4.jpg", "1mp3d_0029_region3/39d7ad2d721e47aa9fe29090663265e8_i1_5.jpg", "1mp3d_0029_region3/39d7ad2d721e47aa9fe29090663265e8_i2_5.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0029_region3", "origin_question": "When did you first discover \"the rectangular dark brown wooden table\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the rectangular dark brown wooden table\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "1mp3d_0029_region3/0b7767a8959a48719e014e9ab7d3597b_i1_1.jpg", "1mp3d_0029_region3/0b7767a8959a48719e014e9ab7d3597b_i2_1.jpg", "1mp3d_0029_region3/0b7767a8959a48719e014e9ab7d3597b_i1_0.jpg", "1mp3d_0029_region3/0b7767a8959a48719e014e9ab7d3597b_i2_0.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0029_region3", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "27.59", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "1mp3d_0029_region3/e3b44e490fdb4632a07618be3c52f7a8_i1_1.jpg", "1mp3d_0029_region3/e3b44e490fdb4632a07618be3c52f7a8_i2_1.jpg", "1mp3d_0029_region3/fd4752f551d04e22be492adea0d9fb04_i1_0.jpg", "1mp3d_0029_region3/fd4752f551d04e22be492adea0d9fb04_i2_0.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0029_region3", "origin_question": "In which turn did you see both \"the vent\" and \"the rectangular dark brown wooden table\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the vent\" and \"the rectangular dark brown wooden table\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "1mp3d_0029_region3/e3b44e490fdb4632a07618be3c52f7a8_i1_2.jpg", "1mp3d_0029_region3/e3b44e490fdb4632a07618be3c52f7a8_i2_2.jpg", "1mp3d_0029_region3/9ee37c1bce0a41e8bd50821540e80713_i1_2.jpg", "1mp3d_0029_region3/9ee37c1bce0a41e8bd50821540e80713_i2_2.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0029_region3", "origin_question": "Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Left", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "1mp3d_0029_region3/cf5dbac616f94d65845bb25be5078cb8_i1_3.jpg", "1mp3d_0029_region3/cf5dbac616f94d65845bb25be5078cb8_i2_3.jpg", "1mp3d_0029_region3/af5f1942b4374015823bd32e31b31c64_i1_2.jpg", "1mp3d_0029_region3/af5f1942b4374015823bd32e31b31c64_i2_2.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0029_region3", "origin_question": "Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance forward or backward from that position?", "option": [ "Forward", "Backward" ], "answer": "Forward", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance forward or backward from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Forward', 'Backward'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0029_region3/cf5dbac616f94d65845bb25be5078cb8_i1_2.jpg", "1mp3d_0029_region3/cf5dbac616f94d65845bb25be5078cb8_i2_2.jpg", "1mp3d_0029_region3/9ee37c1bce0a41e8bd50821540e80713_i1_0.jpg", "1mp3d_0029_region3/9ee37c1bce0a41e8bd50821540e80713_i2_0.jpg" ], "turn_id": 7, "scan_id": "1mp3d_0029_region3", "origin_question": "Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the rectangular dark brown wooden table\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the rectangular dark brown wooden table\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "1mp3d_0029_region3/9ee37c1bce0a41e8bd50821540e80713_i1_5.jpg", "1mp3d_0029_region3/9ee37c1bce0a41e8bd50821540e80713_i2_5.jpg", "1mp3d_0029_region3/204fc459fd244be1b08fe56aa9caea9d_i1_4.jpg", "1mp3d_0029_region3/204fc459fd244be1b08fe56aa9caea9d_i2_4.jpg" ], "turn_id": 8, "scan_id": "1mp3d_0029_region3", "origin_question": "Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "86.24", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "1mp3d_0057_region7/06a5647528f6448f819428c191d8df72_i1_4.jpg", "1mp3d_0057_region7/06a5647528f6448f819428c191d8df72_i2_4.jpg", "1mp3d_0057_region7/51711e23a5bc47c79ca35538e3ffcddd_i1_1.jpg", "1mp3d_0057_region7/51711e23a5bc47c79ca35538e3ffcddd_i2_1.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0057_region7", "origin_question": "Remember, how many person(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many person(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "1mp3d_0057_region7/06a5647528f6448f819428c191d8df72_i1_1.jpg", "1mp3d_0057_region7/06a5647528f6448f819428c191d8df72_i2_1.jpg", "1mp3d_0057_region7/06a5647528f6448f819428c191d8df72_i1_2.jpg", "1mp3d_0057_region7/06a5647528f6448f819428c191d8df72_i2_2.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0057_region7", "origin_question": "At the end of which turn were both of \"the medium-sized fireplace made of stacked gray stones\" and \"the medium-sized rectangular flat-panel television with a matte finish\" on your right side?", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the medium-sized fireplace made of stacked gray stones\" and \"the medium-sized rectangular flat-panel television with a matte finish\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "1mp3d_0057_region7/06a5647528f6448f819428c191d8df72_i1_3.jpg", "1mp3d_0057_region7/06a5647528f6448f819428c191d8df72_i2_3.jpg", "1mp3d_0057_region7/43c2c264bfdc44a680ebb2841bf9d8c9_i1_0.jpg", "1mp3d_0057_region7/43c2c264bfdc44a680ebb2841bf9d8c9_i2_0.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0057_region7", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the medium-sized fireplace made of stacked gray stones\" from your position?", "option": [], "answer": "165.6", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the medium-sized fireplace made of stacked gray stones\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "1mp3d_0057_region7/8c0424b4b69c4ff8956cb074e1488a73_i1_1.jpg", "1mp3d_0057_region7/8c0424b4b69c4ff8956cb074e1488a73_i2_1.jpg", "1mp3d_0057_region7/8c0424b4b69c4ff8956cb074e1488a73_i1_0.jpg", "1mp3d_0057_region7/8c0424b4b69c4ff8956cb074e1488a73_i2_0.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0057_region7", "origin_question": "When did you first discover \"the medium wooden fan with dark blades\"? (index of the turn)", "option": [], "answer": "4", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the medium wooden fan with dark blades\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "1mp3d_0057_region7/43c2c264bfdc44a680ebb2841bf9d8c9_i1_2.jpg", "1mp3d_0057_region7/43c2c264bfdc44a680ebb2841bf9d8c9_i2_2.jpg", "1mp3d_0057_region7/43c2c264bfdc44a680ebb2841bf9d8c9_i1_3.jpg", "1mp3d_0057_region7/43c2c264bfdc44a680ebb2841bf9d8c9_i2_3.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0057_region7", "origin_question": "By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the medium-sized rectangular flat-panel television with a matte finish\", Y: \"the medium wooden fan with dark blades\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to X and farther away from Y", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the medium-sized rectangular flat-panel television with a matte finish\", Y: \"the medium wooden fan with dark blades\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "1mp3d_0057_region7/43c2c264bfdc44a680ebb2841bf9d8c9_i1_4.jpg", "1mp3d_0057_region7/43c2c264bfdc44a680ebb2841bf9d8c9_i2_4.jpg", "1mp3d_0057_region7/60af61d9d92f4d9cbac7e4e2aed4e2b6_i2_3.jpg", "1mp3d_0057_region7/a5ed9e39260e4273b0cdbda78d3f917d_i1_4.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0057_region7", "origin_question": "What will be the first-time appearance order of the following categories: couch, fan, tv?", "option": [ "couch, fan, tv", "tv, fan, couch", "couch, tv, fan", "fan, tv, couch" ], "answer": "couch, tv, fan", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: couch, fan, tv? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['couch, fan, tv', 'tv, fan, couch', 'couch, tv, fan', 'fan, tv, couch'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "1mp3d_0057_region7/a5ed9e39260e4273b0cdbda78d3f917d_i2_4.jpg", "1mp3d_0057_region7/60af61d9d92f4d9cbac7e4e2aed4e2b6_i1_2.jpg", "1mp3d_0057_region7/60af61d9d92f4d9cbac7e4e2aed4e2b6_i2_2.jpg", "1mp3d_0057_region7/60af61d9d92f4d9cbac7e4e2aed4e2b6_i1_1.jpg" ], "turn_id": 7, "scan_id": "1mp3d_0057_region7", "origin_question": "In which turn were you farthest from \"the medium wooden fan with dark blades\"? ", "option": [], "answer": "4", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the medium wooden fan with dark blades\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "1mp3d_0057_region7/60af61d9d92f4d9cbac7e4e2aed4e2b6_i2_1.jpg", "1mp3d_0057_region7/60af61d9d92f4d9cbac7e4e2aed4e2b6_i1_0.jpg", "1mp3d_0057_region7/60af61d9d92f4d9cbac7e4e2aed4e2b6_i2_0.jpg", "1mp3d_0057_region7/60af61d9d92f4d9cbac7e4e2aed4e2b6_i2_5.jpg" ], "turn_id": 8, "scan_id": "1mp3d_0057_region7", "origin_question": "Among these three objects(\"the medium-sized rectangular flat-panel television with a matte finish\",\"the medium-sized lamp with a square light brown shade\",\"the medium wooden fan with dark blades\" ), which one is the closest to you?", "option": [ "the medium-sized rectangular flat-panel television with a matte finish", "the medium-sized lamp with a square light brown shade", "the medium wooden fan with dark blades" ], "answer": "the medium wooden fan with dark blades", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects(\"the medium-sized rectangular flat-panel television with a matte finish\",\"the medium-sized lamp with a square light brown shade\",\"the medium wooden fan with dark blades\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the medium-sized rectangular flat-panel television with a matte finish', 'the medium-sized lamp with a square light brown shade', 'the medium wooden fan with dark blades'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0695_03/00000.jpg", "scene0695_03/00040.jpg", "scene0695_03/00080.jpg", "scene0695_03/00120.jpg", "scene0695_03/00160.jpg" ], "turn_id": 1, "scan_id": "scene0695_03", "origin_question": "Remember, how many chair(s) have you seen so far? ", "option": [], "answer": "2", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many chair(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0695_03/00200.jpg", "scene0695_03/00240.jpg", "scene0695_03/00280.jpg", "scene0695_03/00320.jpg", "scene0695_03/00360.jpg" ], "turn_id": 2, "scan_id": "scene0695_03", "origin_question": "Please recall \"the rectangular soft gray mat with a simple pattern\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.34", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the rectangular soft gray mat with a simple pattern\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0695_03/00400.jpg", "scene0695_03/00440.jpg", "scene0695_03/00480.jpg", "scene0695_03/00520.jpg", "scene0695_03/00560.jpg" ], "turn_id": 3, "scan_id": "scene0695_03", "origin_question": "At the end of which turn were both of \"the curtain\" and \"the medium-sized white flannel blanket with gray spots\" on your left side?", "option": [], "answer": "2", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the curtain\" and \"the medium-sized white flannel blanket with gray spots\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0695_03/00600.jpg", "scene0695_03/00640.jpg", "scene0695_03/00680.jpg", "scene0695_03/00720.jpg", "scene0695_03/00760.jpg" ], "turn_id": 4, "scan_id": "scene0695_03", "origin_question": "Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0695_03/00800.jpg", "scene0695_03/00840.jpg", "scene0695_03/00880.jpg", "scene0695_03/00920.jpg", "scene0695_03/00960.jpg" ], "turn_id": 5, "scan_id": "scene0695_03", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the black rectangular monitor with a modern design\" from your position?", "option": [], "answer": "119.32", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the black rectangular monitor with a modern design\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0695_03/01000.jpg", "scene0695_03/01040.jpg", "scene0695_03/01080.jpg", "scene0695_03/01120.jpg", "scene0695_03/01160.jpg" ], "turn_id": 6, "scan_id": "scene0695_03", "origin_question": "Please recall \"the rectangular wooden desk made of oak\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.75", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the rectangular wooden desk made of oak\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0695_03/01200.jpg", "scene0695_03/01240.jpg", "scene0695_03/01280.jpg", "scene0695_03/01320.jpg", "scene0695_03/01360.jpg" ], "turn_id": 7, "scan_id": "scene0695_03", "origin_question": "In which turn did you see both \"the small round transparent plastic hamper with a lid\" and \"the slim, white remote control with black buttons\" simultaneously? ", "option": [], "answer": "7", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the small round transparent plastic hamper with a lid\" and \"the slim, white remote control with black buttons\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0695_03/01400.jpg", "scene0695_03/01440.jpg", "scene0695_03/01480.jpg", "scene0695_03/01520.jpg", "scene0695_03/01560.jpg" ], "turn_id": 8, "scan_id": "scene0695_03", "origin_question": "When did you first discover \"the cabinet\"? (index of the turn)", "option": [], "answer": "8", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the cabinet\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "scene0695_03/01600.jpg", "scene0695_03/01760.jpg", "scene0695_03/01800.jpg", "scene0695_03/01840.jpg", "scene0695_03/01880.jpg" ], "turn_id": 9, "scan_id": "scene0695_03", "origin_question": "Among these three objects(\"the silver metal laptop with a black keyboard\",\"the rectangular white pillow with grey edges\",\"the slim, white remote control with black buttons\" ), which one is the farthest from you?", "option": [ "the silver metal laptop with a black keyboard", "the rectangular white pillow with grey edges", "the slim, white remote control with black buttons" ], "answer": "the rectangular white pillow with grey edges", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the silver metal laptop with a black keyboard\",\"the rectangular white pillow with grey edges\",\"the slim, white remote control with black buttons\" ), which one is the farthest from you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the silver metal laptop with a black keyboard', 'the rectangular white pillow with grey edges', 'the slim, white remote control with black buttons'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0695_03/01920.jpg", "scene0695_03/01960.jpg", "scene0695_03/02000.jpg", "scene0695_03/02040.jpg", "scene0695_03/02080.jpg" ], "turn_id": 10, "scan_id": "scene0695_03", "origin_question": "Which direction is the silver metal laptop with a black keyboard in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the silver metal laptop with a black keyboard in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "1mp3d_0084_region7/5bc41a6e3b7748149e0e8592c5b4d142_i1_2.jpg", "1mp3d_0084_region7/5bc41a6e3b7748149e0e8592c5b4d142_i2_2.jpg", "1mp3d_0084_region7/5bc41a6e3b7748149e0e8592c5b4d142_i1_3.jpg", "1mp3d_0084_region7/5bc41a6e3b7748149e0e8592c5b4d142_i2_3.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0084_region7", "origin_question": "Remember, how many countertop(s) have you seen so far? ", "option": [], "answer": "2", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many countertop(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "1mp3d_0084_region7/1a6418a306b8448a8529ffc6cb508d8a_i1_4.jpg", "1mp3d_0084_region7/1a6418a306b8448a8529ffc6cb508d8a_i2_4.jpg", "1mp3d_0084_region7/1a6418a306b8448a8529ffc6cb508d8a_i1_5.jpg", "1mp3d_0084_region7/1a6418a306b8448a8529ffc6cb508d8a_i2_5.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0084_region7", "origin_question": "Which two objects are on the same side of you now? \"the white rectangular microwave with a digital display\", \"the black metal stool with a round wooden seat\", \"the white rectangular sink with a black faucet\". ", "option": [ "the white rectangular microwave with a digital display, the black metal stool with a round wooden seat", "the white rectangular microwave with a digital display, the white rectangular sink with a black faucet", "the black metal stool with a round wooden seat, the white rectangular sink with a black faucet" ], "answer": "the black metal stool with a round wooden seat, the white rectangular sink with a black faucet", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the white rectangular microwave with a digital display\", \"the black metal stool with a round wooden seat\", \"the white rectangular sink with a black faucet\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the white rectangular microwave with a digital display, the black metal stool with a round wooden seat', 'the white rectangular microwave with a digital display, the white rectangular sink with a black faucet', 'the black metal stool with a round wooden seat, the white rectangular sink with a black faucet'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0084_region7/3f432ddd169d4433979e004d1237d029_i1_1.jpg", "1mp3d_0084_region7/3f432ddd169d4433979e004d1237d029_i2_1.jpg", "1mp3d_0084_region7/3f432ddd169d4433979e004d1237d029_i1_2.jpg", "1mp3d_0084_region7/3f432ddd169d4433979e004d1237d029_i2_2.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0084_region7", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the white rectangular gas stove with a black burner grate\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the white rectangular gas stove with a black burner grate\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0084_region7/3f432ddd169d4433979e004d1237d029_i1_3.jpg", "1mp3d_0084_region7/3f432ddd169d4433979e004d1237d029_i2_3.jpg", "1mp3d_0084_region7/3f432ddd169d4433979e004d1237d029_i1_4.jpg", "1mp3d_0084_region7/3f432ddd169d4433979e004d1237d029_i2_4.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0084_region7", "origin_question": "", "option": [], "answer": "", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "1mp3d_0084_region7/1a6418a306b8448a8529ffc6cb508d8a_i1_2.jpg", "1mp3d_0084_region7/1a6418a306b8448a8529ffc6cb508d8a_i2_2.jpg", "1mp3d_0084_region7/1a6418a306b8448a8529ffc6cb508d8a_i1_1.jpg", "1mp3d_0084_region7/1a6418a306b8448a8529ffc6cb508d8a_i2_1.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0084_region7", "origin_question": "Please recall \"the white rectangular sink with a black faucet\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.18", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the white rectangular sink with a black faucet\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0084_region7/5bc41a6e3b7748149e0e8592c5b4d142_i1_0.jpg", "1mp3d_0084_region7/5bc41a6e3b7748149e0e8592c5b4d142_i2_0.jpg", "1mp3d_0084_region7/5bc41a6e3b7748149e0e8592c5b4d142_i1_1.jpg", "1mp3d_0084_region7/5bc41a6e3b7748149e0e8592c5b4d142_i2_1.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0084_region7", "origin_question": "Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the black metal stool with a round wooden seat\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the black metal stool with a round wooden seat\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0598_01/00000.jpg", "scene0598_01/00030.jpg", "scene0598_01/00060.jpg", "scene0598_01/00090.jpg" ], "turn_id": 1, "scan_id": "scene0598_01", "origin_question": "Remember, how many chair(s) have you seen so far? ", "option": [], "answer": "3", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many chair(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0598_01/00120.jpg", "scene0598_01/00150.jpg", "scene0598_01/00180.jpg", "scene0598_01/00210.jpg" ], "turn_id": 2, "scan_id": "scene0598_01", "origin_question": "In which turn did you see both \"the rectangular light brown wooden door with a glass panel\" and \"the stylish white rectangular keyboard with a clean design\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the rectangular light brown wooden door with a glass panel\" and \"the stylish white rectangular keyboard with a clean design\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0598_01/00240.jpg", "scene0598_01/00270.jpg", "scene0598_01/00300.jpg", "scene0598_01/00330.jpg" ], "turn_id": 3, "scan_id": "scene0598_01", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0598_01/00360.jpg", "scene0598_01/00390.jpg", "scene0598_01/00420.jpg", "scene0598_01/00450.jpg" ], "turn_id": 4, "scan_id": "scene0598_01", "origin_question": "Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "88.83", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0598_01/00480.jpg", "scene0598_01/00510.jpg", "scene0598_01/00540.jpg", "scene0598_01/00570.jpg" ], "turn_id": 5, "scan_id": "scene0598_01", "origin_question": "At the end of which turn were both of \"Light brown vertical wooden blinds of medium size\" and \"the rectangular light brown wooden door with a glass panel\" on your right side?", "option": [], "answer": "1", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"Light brown vertical wooden blinds of medium size\" and \"the rectangular light brown wooden door with a glass panel\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0598_01/00600.jpg", "scene0598_01/00630.jpg", "scene0598_01/00660.jpg", "scene0598_01/00690.jpg" ], "turn_id": 6, "scan_id": "scene0598_01", "origin_question": "Which direction is the rectangular light brown wooden door with a glass panel in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the rectangular light brown wooden door with a glass panel in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0598_01/00720.jpg", "scene0598_01/00750.jpg", "scene0598_01/00780.jpg", "scene0598_01/00810.jpg" ], "turn_id": 7, "scan_id": "scene0598_01", "origin_question": "How far is your current position from where you were at the end of the turn 3? (in meters)", "option": [], "answer": "1.27", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 3? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0030_region14/2f20259787124d7e8a85408cd5767c65_i1_1.jpg", "1mp3d_0030_region14/2f20259787124d7e8a85408cd5767c65_i2_1.jpg", "1mp3d_0030_region14/e3beaf62164442018d421c9825039736_i1_1.jpg", "1mp3d_0030_region14/e3beaf62164442018d421c9825039736_i2_1.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0030_region14", "origin_question": "Remember, have you seen any towel(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any towel(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "1mp3d_0030_region14/9d10a60eac3346f085b0df039ab14d17_i1_2.jpg", "1mp3d_0030_region14/9d10a60eac3346f085b0df039ab14d17_i2_2.jpg", "1mp3d_0030_region14/9d10a60eac3346f085b0df039ab14d17_i1_1.jpg", "1mp3d_0030_region14/9d10a60eac3346f085b0df039ab14d17_i2_1.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0030_region14", "origin_question": "At the end of which turn were both of \"the toilet\" and \"the sink\" on your right side?", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the toilet\" and \"the sink\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "1mp3d_0030_region14/e3beaf62164442018d421c9825039736_i1_0.jpg", "1mp3d_0030_region14/e3beaf62164442018d421c9825039736_i2_0.jpg", "1mp3d_0030_region14/9d10a60eac3346f085b0df039ab14d17_i1_4.jpg", "1mp3d_0030_region14/9d10a60eac3346f085b0df039ab14d17_i2_4.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0030_region14", "origin_question": "Is the sink to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the sink to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "1mp3d_0030_region14/e3beaf62164442018d421c9825039736_i2_2.jpg", "1mp3d_0030_region14/e3beaf62164442018d421c9825039736_i1_3.jpg", "1mp3d_0030_region14/e3beaf62164442018d421c9825039736_i2_3.jpg", "1mp3d_0030_region14/2f20259787124d7e8a85408cd5767c65_i1_2.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0030_region14", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the large rectangular mirror with a white frame\" from your position?", "option": [], "answer": "98.47", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the large rectangular mirror with a white frame\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "1mp3d_0030_region14/2f20259787124d7e8a85408cd5767c65_i2_2.jpg", "1mp3d_0030_region14/2f20259787124d7e8a85408cd5767c65_i1_0.jpg", "1mp3d_0030_region14/2f20259787124d7e8a85408cd5767c65_i2_0.jpg", "1mp3d_0030_region14/a9366b89465a49028782a840a165a7e8_i1_0.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0030_region14", "origin_question": "Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance forward or backward from that position?", "option": [ "Forward", "Backward" ], "answer": "Backward", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance forward or backward from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Forward', 'Backward'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0701_02/00000.jpg", "scene0701_02/00040.jpg", "scene0701_02/00080.jpg", "scene0701_02/00120.jpg" ], "turn_id": 1, "scan_id": "scene0701_02", "origin_question": "Remember, how many couch(s) have you seen so far? ", "option": [], "answer": "2", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many couch(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0701_02/00160.jpg", "scene0701_02/00200.jpg", "scene0701_02/00240.jpg", "scene0701_02/00280.jpg" ], "turn_id": 2, "scan_id": "scene0701_02", "origin_question": "In which turn did you see both \"the circular leather table with a deep brown color\" and \"the vibrant red backpack with black straps\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the circular leather table with a deep brown color\" and \"the vibrant red backpack with black straps\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0701_02/00320.jpg", "scene0701_02/00360.jpg", "scene0701_02/00400.jpg", "scene0701_02/00440.jpg" ], "turn_id": 3, "scan_id": "scene0701_02", "origin_question": "When did you first discover \"the vibrant red backpack with black straps\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the vibrant red backpack with black straps\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0701_02/00480.jpg", "scene0701_02/00520.jpg", "scene0701_02/00560.jpg", "scene0701_02/00600.jpg" ], "turn_id": 4, "scan_id": "scene0701_02", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "1.07", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0701_02/00640.jpg", "scene0701_02/00680.jpg", "scene0701_02/00720.jpg", "scene0701_02/00760.jpg" ], "turn_id": 5, "scan_id": "scene0701_02", "origin_question": "How far is your current position from where you were at the end of the turn 4? (in meters)", "option": [], "answer": "2.87", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 4? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0701_02/00800.jpg", "scene0701_02/00840.jpg", "scene0701_02/00880.jpg", "scene0701_02/00920.jpg" ], "turn_id": 6, "scan_id": "scene0701_02", "origin_question": "In which turn were you closest to \"the circular leather table with a deep brown color\"? ", "option": [], "answer": "3", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the circular leather table with a deep brown color\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0701_02/00960.jpg", "scene0701_02/01000.jpg", "scene0701_02/01040.jpg", "scene0701_02/01080.jpg" ], "turn_id": 7, "scan_id": "scene0701_02", "origin_question": "Using your orientation at the end of turn 6 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 6 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0498_00/00000.jpg", "scene0498_00/00030.jpg", "scene0498_00/00060.jpg", "scene0498_00/00090.jpg" ], "turn_id": 1, "scan_id": "scene0498_00", "origin_question": "Remember, have you seen any bag(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any bag(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0498_00/00120.jpg", "scene0498_00/00150.jpg", "scene0498_00/00180.jpg", "scene0498_00/00210.jpg" ], "turn_id": 2, "scan_id": "scene0498_00", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the rectangular stainless steel microwave with a shiny surface\"; \"the window\"; \"the medium-sized white round ceramic plate\" ", "option": [ "the rectangular stainless steel microwave with a shiny surface", "the window", "the medium-sized white round ceramic plate" ], "answer": "the medium-sized white round ceramic plate", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the rectangular stainless steel microwave with a shiny surface\"; \"the window\"; \"the medium-sized white round ceramic plate\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular stainless steel microwave with a shiny surface', 'the window', 'the medium-sized white round ceramic plate'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0498_00/00240.jpg", "scene0498_00/00270.jpg", "scene0498_00/00300.jpg", "scene0498_00/00330.jpg" ], "turn_id": 3, "scan_id": "scene0498_00", "origin_question": "In which turn did you see both \"the medium-sized rectangular brown wooden table\" and \"the slightly wrinkled, irregularly shaped white plastic bag\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the medium-sized rectangular brown wooden table\" and \"the slightly wrinkled, irregularly shaped white plastic bag\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0498_00/00360.jpg", "scene0498_00/00390.jpg", "scene0498_00/00420.jpg", "scene0498_00/00450.jpg" ], "turn_id": 4, "scan_id": "scene0498_00", "origin_question": "Which direction is the slightly wrinkled, irregularly shaped white plastic bag in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the slightly wrinkled, irregularly shaped white plastic bag in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0498_00/00480.jpg", "scene0498_00/00510.jpg", "scene0498_00/00540.jpg", "scene0498_00/00580.jpg" ], "turn_id": 5, "scan_id": "scene0498_00", "origin_question": "What will be the first-time appearance order of the following categories: bin, counter, cup?", "option": [ "bin, cup, counter", "cup, counter, bin", "bin, counter, cup", "counter, cup, bin" ], "answer": "counter, cup, bin", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bin, counter, cup? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['bin, cup, counter', 'cup, counter, bin', 'bin, counter, cup', 'counter, cup, bin'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0498_00/00610.jpg", "scene0498_00/00640.jpg", "scene0498_00/00670.jpg", "scene0498_00/00700.jpg" ], "turn_id": 6, "scan_id": "scene0498_00", "origin_question": "Please recall \"the small rectangular white paper made of thin material\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.96", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the small rectangular white paper made of thin material\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "Validation_41159503/Validation_41159503_41159503_34934.696.png", "Validation_41159503/Validation_41159503_41159503_34936.679.png", "Validation_41159503/Validation_41159503_41159503_34938.678.png", "Validation_41159503/Validation_41159503_41159503_34940.677.png", "Validation_41159503/Validation_41159503_41159503_34942.676.png" ], "turn_id": 1, "scan_id": "Validation_41159503", "origin_question": "Remember, how many carpet(s) have you seen so far? ", "option": [], "answer": "2", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many carpet(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_41159503/Validation_41159503_41159503_34944.676.png", "Validation_41159503/Validation_41159503_41159503_34946.675.png", "Validation_41159503/Validation_41159503_41159503_34948.674.png", "Validation_41159503/Validation_41159503_41159503_34950.673.png", "Validation_41159503/Validation_41159503_41159503_34952.672.png" ], "turn_id": 2, "scan_id": "Validation_41159503", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_41159503/Validation_41159503_41159503_34954.671.png", "Validation_41159503/Validation_41159503_41159503_34956.687.png", "Validation_41159503/Validation_41159503_41159503_34958.687.png", "Validation_41159503/Validation_41159503_41159503_34960.686.png", "Validation_41159503/Validation_41159503_41159503_34962.685.png" ], "turn_id": 3, "scan_id": "Validation_41159503", "origin_question": "Is the sleek oval white bathtub with the modern design to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the sleek oval white bathtub with the modern design to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_41159503/Validation_41159503_41159503_34964.684.png", "Validation_41159503/Validation_41159503_41159503_34966.683.png", "Validation_41159503/Validation_41159503_41159503_34968.682.png", "Validation_41159503/Validation_41159503_41159503_34970.681.png", "Validation_41159503/Validation_41159503_41159503_34972.681.png" ], "turn_id": 4, "scan_id": "Validation_41159503", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the white rounded ceramic toilet\" from your position?", "option": [], "answer": "38.55", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the white rounded ceramic toilet\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_41159503/Validation_41159503_41159503_34974.680.png", "Validation_41159503/Validation_41159503_41159503_34976.679.png", "Validation_41159503/Validation_41159503_41159503_34978.678.png", "Validation_41159503/Validation_41159503_41159503_34980.677.png", "Validation_41159503/Validation_41159503_41159503_34982.677.png" ], "turn_id": 5, "scan_id": "Validation_41159503", "origin_question": "Which direction is the mirror in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the mirror in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_41159503/Validation_41159503_41159503_34984.676.png", "Validation_41159503/Validation_41159503_41159503_34986.675.png", "Validation_41159503/Validation_41159503_41159503_34988.674.png", "Validation_41159503/Validation_41159503_41159503_34990.673.png", "Validation_41159503/Validation_41159503_41159503_34992.672.png" ], "turn_id": 6, "scan_id": "Validation_41159503", "origin_question": "How far is your current position from where you were at the end of the turn 3? (in meters)", "option": [], "answer": "1.55", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 3? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_41159503/Validation_41159503_41159503_34994.671.png", "Validation_41159503/Validation_41159503_41159503_34996.687.png", "Validation_41159503/Validation_41159503_41159503_34998.687.png", "Validation_41159503/Validation_41159503_41159503_35000.686.png", "Validation_41159503/Validation_41159503_41159503_35002.685.png" ], "turn_id": 7, "scan_id": "Validation_41159503", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the polished chrome faucet with the sleek cylindrical design\" from your position?", "option": [], "answer": "125.09", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the polished chrome faucet with the sleek cylindrical design\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_41159503/Validation_41159503_41159503_35004.684.png", "Validation_41159503/Validation_41159503_41159503_35006.683.png", "Validation_41159503/Validation_41159503_41159503_35008.682.png", "Validation_41159503/Validation_41159503_41159503_35010.681.png", "Validation_41159503/Validation_41159503_41159503_35012.681.png" ], "turn_id": 8, "scan_id": "Validation_41159503", "origin_question": "When did you first discover \"the medium-sized woven rectangular basket in natural tones\"? (index of the turn)", "option": [], "answer": "7", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the medium-sized woven rectangular basket in natural tones\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_41159503/Validation_41159503_41159503_35014.680.png", "Validation_41159503/Validation_41159503_41159503_35016.679.png", "Validation_41159503/Validation_41159503_41159503_35018.678.png", "Validation_41159503/Validation_41159503_41159503_35020.677.png", "Validation_41159503/Validation_41159503_41159503_35022.677.png" ], "turn_id": 9, "scan_id": "Validation_41159503", "origin_question": "In which turn were you closest to \"the white rounded ceramic toilet\"? ", "option": [], "answer": "6", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the white rounded ceramic toilet\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_41159503/Validation_41159503_41159503_35024.676.png", "Validation_41159503/Validation_41159503_41159503_35026.675.png", "Validation_41159503/Validation_41159503_41159503_35028.674.png", "Validation_41159503/Validation_41159503_41159503_35030.673.png", "Validation_41159503/Validation_41159503_41159503_35032.672.png" ], "turn_id": 10, "scan_id": "Validation_41159503", "origin_question": "Which direction is the ceramic oval soap dish in the neutral hue in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the ceramic oval soap dish in the neutral hue in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42897542/Validation_42897542_42897542_275456.136.png", "Validation_42897542/Validation_42897542_42897542_275457.635.png", "Validation_42897542/Validation_42897542_42897542_275459.134.png", "Validation_42897542/Validation_42897542_42897542_275460.634.png" ], "turn_id": 1, "scan_id": "Validation_42897542", "origin_question": "Remember, have you seen any flowerpot(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any flowerpot(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_42897542/Validation_42897542_42897542_275462.133.png", "Validation_42897542/Validation_42897542_42897542_275463.633.png", "Validation_42897542/Validation_42897542_42897542_275465.132.png", "Validation_42897542/Validation_42897542_42897542_275466.631.png" ], "turn_id": 2, "scan_id": "Validation_42897542", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the large rectangular mirror with the wooden frame\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the large rectangular mirror with the wooden frame\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_42897542/Validation_42897542_42897542_275468.131.png", "Validation_42897542/Validation_42897542_42897542_275469.630.png", "Validation_42897542/Validation_42897542_42897542_275471.129.png", "Validation_42897542/Validation_42897542_42897542_275472.629.png" ], "turn_id": 3, "scan_id": "Validation_42897542", "origin_question": "", "option": [], "answer": "", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_42897542/Validation_42897542_42897542_275474.128.png", "Validation_42897542/Validation_42897542_42897542_275475.628.png", "Validation_42897542/Validation_42897542_42897542_275477.127.png", "Validation_42897542/Validation_42897542_42897542_275478.627.png" ], "turn_id": 4, "scan_id": "Validation_42897542", "origin_question": "Which two objects are on the same side of you now? \"the round white ceramic plate with the minimalist design\", \"the beige conical lampshade with the metallic body\", \"the metal rectangular radiator with vertical grooves\". ", "option": [ "the round white ceramic plate with the minimalist design, the beige conical lampshade with the metallic body", "the round white ceramic plate with the minimalist design, the metal rectangular radiator with vertical grooves", "the beige conical lampshade with the metallic body, the metal rectangular radiator with vertical grooves" ], "answer": "the round white ceramic plate with the minimalist design, the beige conical lampshade with the metallic body", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the round white ceramic plate with the minimalist design\", \"the beige conical lampshade with the metallic body\", \"the metal rectangular radiator with vertical grooves\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the round white ceramic plate with the minimalist design, the beige conical lampshade with the metallic body', 'the round white ceramic plate with the minimalist design, the metal rectangular radiator with vertical grooves', 'the beige conical lampshade with the metallic body, the metal rectangular radiator with vertical grooves'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_41159519/Validation_41159519_41159519_38238.829.png", "Validation_41159519/Validation_41159519_41159519_38240.812.png", "Validation_41159519/Validation_41159519_41159519_38242.811.png", "Validation_41159519/Validation_41159519_41159519_38244.810.png", "Validation_41159519/Validation_41159519_41159519_38246.810.png" ], "turn_id": 1, "scan_id": "Validation_41159519", "origin_question": "Remember, have you seen any boots(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any boots(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_41159519/Validation_41159519_41159519_38248.809.png", "Validation_41159519/Validation_41159519_41159519_38250.808.png", "Validation_41159519/Validation_41159519_41159519_38252.807.png", "Validation_41159519/Validation_41159519_41159519_38254.806.png", "Validation_41159519/Validation_41159519_41159519_38256.805.png" ], "turn_id": 2, "scan_id": "Validation_41159519", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_41159519/Validation_41159519_41159519_38258.804.png", "Validation_41159519/Validation_41159519_41159519_38260.804.png", "Validation_41159519/Validation_41159519_41159519_38262.803.png", "Validation_41159519/Validation_41159519_41159519_38264.802.png", "Validation_41159519/Validation_41159519_41159519_38266.801.png" ], "turn_id": 3, "scan_id": "Validation_41159519", "origin_question": "Please recall \"the sturdy rectangular wooden table with the natural tone\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.87", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the sturdy rectangular wooden table with the natural tone\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_41159519/Validation_41159519_41159519_38268.800.png", "Validation_41159519/Validation_41159519_41159519_38270.800.png", "Validation_41159519/Validation_41159519_41159519_38272.799.png", "Validation_41159519/Validation_41159519_41159519_38274.815.png", "Validation_41159519/Validation_41159519_41159519_38276.814.png" ], "turn_id": 4, "scan_id": "Validation_41159519", "origin_question": "In which turn were you farthest from \"the modern, polished chrome cylindrical faucet\"? ", "option": [], "answer": "1", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the modern, polished chrome cylindrical faucet\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_41159519/Validation_41159519_41159519_38278.813.png", "Validation_41159519/Validation_41159519_41159519_38280.812.png", "Validation_41159519/Validation_41159519_41159519_38282.811.png", "Validation_41159519/Validation_41159519_41159519_38284.810.png", "Validation_41159519/Validation_41159519_41159519_38286.810.png" ], "turn_id": 5, "scan_id": "Validation_41159519", "origin_question": "Please recall \"the curved pale blue vintage chair\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.99", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the curved pale blue vintage chair\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_41159519/Validation_41159519_41159519_38288.809.png", "Validation_41159519/Validation_41159519_41159519_38290.808.png", "Validation_41159519/Validation_41159519_41159519_38292.807.png", "Validation_41159519/Validation_41159519_41159519_38294.806.png", "Validation_41159519/Validation_41159519_41159519_38296.805.png" ], "turn_id": 6, "scan_id": "Validation_41159519", "origin_question": "Which two objects are on the same side of you now? \"the curved pale blue vintage chair\", \"the sleek black glossy stove with metal construction\", \"the natural wood board with rounded edges\". ", "option": [ "the curved pale blue vintage chair, the sleek black glossy stove with metal construction", "the curved pale blue vintage chair, the natural wood board with rounded edges", "the sleek black glossy stove with metal construction, the natural wood board with rounded edges" ], "answer": "the sleek black glossy stove with metal construction, the natural wood board with rounded edges", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the curved pale blue vintage chair\", \"the sleek black glossy stove with metal construction\", \"the natural wood board with rounded edges\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the curved pale blue vintage chair, the sleek black glossy stove with metal construction', 'the curved pale blue vintage chair, the natural wood board with rounded edges', 'the sleek black glossy stove with metal construction, the natural wood board with rounded edges'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_41159519/Validation_41159519_41159519_38298.804.png", "Validation_41159519/Validation_41159519_41159519_38300.804.png", "Validation_41159519/Validation_41159519_41159519_38302.803.png", "Validation_41159519/Validation_41159519_41159519_38304.802.png", "Validation_41159519/Validation_41159519_41159519_38306.801.png" ], "turn_id": 7, "scan_id": "Validation_41159519", "origin_question": "Which two objects are on the same side of you now? \"the sturdy rectangular wooden table with the natural tone\", \"the black cotton blend clothes with adjustable straps\", \"the sleek black glossy stove with metal construction\". ", "option": [ "the sturdy rectangular wooden table with the natural tone, the black cotton blend clothes with adjustable straps", "the sturdy rectangular wooden table with the natural tone, the sleek black glossy stove with metal construction", "the black cotton blend clothes with adjustable straps, the sleek black glossy stove with metal construction" ], "answer": "the sturdy rectangular wooden table with the natural tone, the black cotton blend clothes with adjustable straps", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the sturdy rectangular wooden table with the natural tone\", \"the black cotton blend clothes with adjustable straps\", \"the sleek black glossy stove with metal construction\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the sturdy rectangular wooden table with the natural tone, the black cotton blend clothes with adjustable straps', 'the sturdy rectangular wooden table with the natural tone, the sleek black glossy stove with metal construction', 'the black cotton blend clothes with adjustable straps, the sleek black glossy stove with metal construction'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "Validation_41159519/Validation_41159519_41159519_38308.800.png", "Validation_41159519/Validation_41159519_41159519_38310.800.png", "Validation_41159519/Validation_41159519_41159519_38312.799.png", "Validation_41159519/Validation_41159519_41159519_38314.798.png", "Validation_41159519/Validation_41159519_41159519_38316.814.png" ], "turn_id": 8, "scan_id": "Validation_41159519", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the large, vibrant map with colorful representations\"; \"the sturdy rectangular wooden table with the natural tone\"; \"the semi-circular stainless steel bin with the black lid\" ", "option": [ "the large, vibrant map with colorful representations", "the sturdy rectangular wooden table with the natural tone", "the semi-circular stainless steel bin with the black lid" ], "answer": "the semi-circular stainless steel bin with the black lid", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the large, vibrant map with colorful representations\"; \"the sturdy rectangular wooden table with the natural tone\"; \"the semi-circular stainless steel bin with the black lid\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the large, vibrant map with colorful representations', 'the sturdy rectangular wooden table with the natural tone', 'the semi-circular stainless steel bin with the black lid'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_41159519/Validation_41159519_41159519_38318.813.png", "Validation_41159519/Validation_41159519_41159519_38320.812.png", "Validation_41159519/Validation_41159519_41159519_38322.811.png", "Validation_41159519/Validation_41159519_41159519_38324.810.png", "Validation_41159519/Validation_41159519_41159519_38326.810.png" ], "turn_id": 9, "scan_id": "Validation_41159519", "origin_question": "Please recall \"the compact lime green plastic case with the glossy finish\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "3.34", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the compact lime green plastic case with the glossy finish\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_41159519/Validation_41159519_41159519_38328.809.png", "Validation_41159519/Validation_41159519_41159519_38330.808.png", "Validation_41159519/Validation_41159519_41159519_38332.807.png", "Validation_41159519/Validation_41159519_41159519_38334.806.png", "Validation_41159519/Validation_41159519_41159519_38336.805.png" ], "turn_id": 10, "scan_id": "Validation_41159519", "origin_question": "How far is your current position from where you were at the end of the turn 3? (in meters)", "option": [], "answer": "2.12", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 3? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42897647/Validation_42897647_42897647_478130.191.png", "Validation_42897647/Validation_42897647_42897647_478132.174.png", "Validation_42897647/Validation_42897647_42897647_478134.189.png", "Validation_42897647/Validation_42897647_42897647_478136.189.png", "Validation_42897647/Validation_42897647_42897647_478138.188.png" ], "turn_id": 1, "scan_id": "Validation_42897647", "origin_question": "Remember, have you seen any table(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any table(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_42897647/Validation_42897647_42897647_478140.187.png", "Validation_42897647/Validation_42897647_42897647_478142.186.png", "Validation_42897647/Validation_42897647_42897647_478144.185.png", "Validation_42897647/Validation_42897647_42897647_478146.184.png", "Validation_42897647/Validation_42897647_42897647_478148.184.png" ], "turn_id": 2, "scan_id": "Validation_42897647", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_42897647/Validation_42897647_42897647_478150.183.png", "Validation_42897647/Validation_42897647_42897647_478152.182.png", "Validation_42897647/Validation_42897647_42897647_478154.181.png", "Validation_42897647/Validation_42897647_42897647_478156.180.png", "Validation_42897647/Validation_42897647_42897647_478158.180.png" ], "turn_id": 3, "scan_id": "Validation_42897647", "origin_question": "What will be the first-time appearance order of the following categories: cabinet, couch, pillow?", "option": [ "cabinet, pillow, couch", "couch, cabinet, pillow", "pillow, couch, cabinet", "cabinet, couch, pillow" ], "answer": "couch, cabinet, pillow", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: cabinet, couch, pillow? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['cabinet, pillow, couch', 'couch, cabinet, pillow', 'pillow, couch, cabinet', 'cabinet, couch, pillow'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_42897647/Validation_42897647_42897647_478160.179.png", "Validation_42897647/Validation_42897647_42897647_478162.178.png", "Validation_42897647/Validation_42897647_42897647_478164.177.png", "Validation_42897647/Validation_42897647_42897647_478166.176.png", "Validation_42897647/Validation_42897647_42897647_478168.176.png" ], "turn_id": 4, "scan_id": "Validation_42897647", "origin_question": "What will be the first-time appearance order of the following categories: cabinet, pillow, table?", "option": [ "table, cabinet, pillow", "table, pillow, cabinet", "cabinet, table, pillow", "pillow, table, cabinet" ], "answer": "table, cabinet, pillow", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: cabinet, pillow, table? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['table, cabinet, pillow', 'table, pillow, cabinet', 'cabinet, table, pillow', 'pillow, table, cabinet'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_42897647/Validation_42897647_42897647_478170.175.png", "Validation_42897647/Validation_42897647_42897647_478172.174.png", "Validation_42897647/Validation_42897647_42897647_478174.190.png", "Validation_42897647/Validation_42897647_42897647_478176.189.png", "Validation_42897647/Validation_42897647_42897647_478178.188.png" ], "turn_id": 5, "scan_id": "Validation_42897647", "origin_question": "At the end of which turn were both of \"the large, dark-colored monitor with the sleek design\" and \"the standard-sized woven pillow with the tribal pattern\" on your left side?", "option": [], "answer": "2", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the large, dark-colored monitor with the sleek design\" and \"the standard-sized woven pillow with the tribal pattern\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_42897647/Validation_42897647_42897647_478180.187.png", "Validation_42897647/Validation_42897647_42897647_478182.186.png", "Validation_42897647/Validation_42897647_42897647_478184.186.png", "Validation_42897647/Validation_42897647_42897647_478186.185.png", "Validation_42897647/Validation_42897647_42897647_478188.184.png" ], "turn_id": 6, "scan_id": "Validation_42897647", "origin_question": "Assuming the direction you are facing at the end of the turn 4 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Left", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 4 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_42897647/Validation_42897647_42897647_478190.183.png", "Validation_42897647/Validation_42897647_42897647_478192.182.png", "Validation_42897647/Validation_42897647_42897647_478194.182.png", "Validation_42897647/Validation_42897647_42897647_478196.181.png", "Validation_42897647/Validation_42897647_42897647_478198.180.png" ], "turn_id": 7, "scan_id": "Validation_42897647", "origin_question": "Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "130.49", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_42897647/Validation_42897647_42897647_478200.179.png", "Validation_42897647/Validation_42897647_42897647_478202.178.png", "Validation_42897647/Validation_42897647_42897647_478204.177.png", "Validation_42897647/Validation_42897647_42897647_478206.177.png", "Validation_42897647/Validation_42897647_42897647_478208.176.png" ], "turn_id": 8, "scan_id": "Validation_42897647", "origin_question": "What will be the first-time appearance order of the following categories: couch, pillow, plant?", "option": [ "pillow, couch, plant", "plant, couch, pillow", "couch, plant, pillow", "couch, pillow, plant" ], "answer": "couch, plant, pillow", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: couch, pillow, plant? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['pillow, couch, plant', 'plant, couch, pillow', 'couch, plant, pillow', 'couch, pillow, plant'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_42897647/Validation_42897647_42897647_478210.175.png", "Validation_42897647/Validation_42897647_42897647_478212.174.png", "Validation_42897647/Validation_42897647_42897647_478214.173.png", "Validation_42897647/Validation_42897647_42897647_478216.189.png", "Validation_42897647/Validation_42897647_42897647_478218.188.png" ], "turn_id": 9, "scan_id": "Validation_42897647", "origin_question": "Please recall \"the large, dark-colored monitor with the sleek design\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.92", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the large, dark-colored monitor with the sleek design\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_42897647/Validation_42897647_42897647_478220.188.png", "Validation_42897647/Validation_42897647_42897647_478222.187.png", "Validation_42897647/Validation_42897647_42897647_478224.186.png", "Validation_42897647/Validation_42897647_42897647_478226.185.png", "Validation_42897647/Validation_42897647_42897647_478228.184.png" ], "turn_id": 10, "scan_id": "Validation_42897647", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the medium-sized indoor plant with lush green leaves\" from your position?", "option": [], "answer": "145.24", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the medium-sized indoor plant with lush green leaves\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47334380/Validation_47334380_47334380_64245.907.png", "Validation_47334380/Validation_47334380_47334380_64247.889.png", "Validation_47334380/Validation_47334380_47334380_64249.888.png", "Validation_47334380/Validation_47334380_47334380_64251.888.png", "Validation_47334380/Validation_47334380_47334380_64253.887.png" ], "turn_id": 1, "scan_id": "Validation_47334380", "origin_question": "Remember, have you seen any cabinet(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any cabinet(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_47334380/Validation_47334380_47334380_64255.886.png", "Validation_47334380/Validation_47334380_47334380_64257.902.png", "Validation_47334380/Validation_47334380_47334380_64259.901.png", "Validation_47334380/Validation_47334380_47334380_64261.900.png", "Validation_47334380/Validation_47334380_47334380_64263.899.png" ], "turn_id": 2, "scan_id": "Validation_47334380", "origin_question": "When did you first discover \"the rounded stainless steel kettle with the sleek dark color\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the rounded stainless steel kettle with the sleek dark color\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "Validation_47334380/Validation_47334380_47334380_64265.899.png", "Validation_47334380/Validation_47334380_47334380_64267.898.png", "Validation_47334380/Validation_47334380_47334380_64269.897.png", "Validation_47334380/Validation_47334380_47334380_64271.896.png", "Validation_47334380/Validation_47334380_47334380_64273.895.png" ], "turn_id": 3, "scan_id": "Validation_47334380", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the sleek rectangular white stove with four burners\"; \"the sleek cylindrical stainless-steel pan\"; \"the seasoning\" ", "option": [ "the sleek rectangular white stove with four burners", "the sleek cylindrical stainless-steel pan", "the seasoning" ], "answer": "the sleek cylindrical stainless-steel pan", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the sleek rectangular white stove with four burners\"; \"the sleek cylindrical stainless-steel pan\"; \"the seasoning\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the sleek rectangular white stove with four burners', 'the sleek cylindrical stainless-steel pan', 'the seasoning'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_47334380/Validation_47334380_47334380_64275.895.png", "Validation_47334380/Validation_47334380_47334380_64277.894.png", "Validation_47334380/Validation_47334380_47334380_64279.893.png", "Validation_47334380/Validation_47334380_47334380_64281.892.png", "Validation_47334380/Validation_47334380_47334380_64283.891.png" ], "turn_id": 4, "scan_id": "Validation_47334380", "origin_question": "What will be the first-time appearance order of the following categories: cabinet, picture, seasoning?", "option": [ "picture, seasoning, cabinet", "seasoning, cabinet, picture", "picture, cabinet, seasoning", "cabinet, seasoning, picture" ], "answer": "cabinet, seasoning, picture", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: cabinet, picture, seasoning? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['picture, seasoning, cabinet', 'seasoning, cabinet, picture', 'picture, cabinet, seasoning', 'cabinet, seasoning, picture'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_47334380/Validation_47334380_47334380_64285.891.png", "Validation_47334380/Validation_47334380_47334380_64287.890.png", "Validation_47334380/Validation_47334380_47334380_64289.889.png", "Validation_47334380/Validation_47334380_47334380_64291.888.png", "Validation_47334380/Validation_47334380_47334380_64293.887.png" ], "turn_id": 5, "scan_id": "Validation_47334380", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47334380/Validation_47334380_47334380_64295.886.png", "Validation_47334380/Validation_47334380_47334380_64297.902.png", "Validation_47334380/Validation_47334380_47334380_64299.901.png", "Validation_47334380/Validation_47334380_47334380_64301.901.png", "Validation_47334380/Validation_47334380_47334380_64303.900.png" ], "turn_id": 6, "scan_id": "Validation_47334380", "origin_question": "Which direction is the rounded stainless steel kettle with the sleek dark color in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the rounded stainless steel kettle with the sleek dark color in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_47334380/Validation_47334380_47334380_64305.899.png", "Validation_47334380/Validation_47334380_47334380_64307.898.png", "Validation_47334380/Validation_47334380_47334380_64309.897.png", "Validation_47334380/Validation_47334380_47334380_64311.897.png", "Validation_47334380/Validation_47334380_47334380_64313.896.png" ], "turn_id": 7, "scan_id": "Validation_47334380", "origin_question": "When did you first discover \"the tall silver stainless steel refrigerator with adjustable shelves\"? (index of the turn)", "option": [], "answer": "5", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the tall silver stainless steel refrigerator with adjustable shelves\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "Validation_47334380/Validation_47334380_47334380_64315.895.png", "Validation_47334380/Validation_47334380_47334380_64317.894.png", "Validation_47334380/Validation_47334380_47334380_64319.893.png", "Validation_47334380/Validation_47334380_47334380_64321.893.png", "Validation_47334380/Validation_47334380_47334380_64323.892.png" ], "turn_id": 8, "scan_id": "Validation_47334380", "origin_question": "Among these three objects(\"the stainless steel oven with the white exterior\",\"the sleek rectangular white stove with four burners\",\"the black rectangular box with white graphics\" ), which one is the closest to you?", "option": [ "the stainless steel oven with the white exterior", "the sleek rectangular white stove with four burners", "the black rectangular box with white graphics" ], "answer": "the black rectangular box with white graphics", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the stainless steel oven with the white exterior\",\"the sleek rectangular white stove with four burners\",\"the black rectangular box with white graphics\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the stainless steel oven with the white exterior', 'the sleek rectangular white stove with four burners', 'the black rectangular box with white graphics'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_47334380/Validation_47334380_47334380_64325.891.png", "Validation_47334380/Validation_47334380_47334380_64327.890.png", "Validation_47334380/Validation_47334380_47334380_64329.889.png", "Validation_47334380/Validation_47334380_47334380_64331.888.png", "Validation_47334380/Validation_47334380_47334380_64333.888.png" ], "turn_id": 9, "scan_id": "Validation_47334380", "origin_question": "When did you first discover \"the stainless steel oven with the white exterior\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the stainless steel oven with the white exterior\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_47334380/Validation_47334380_47334380_64335.887.png", "Validation_47334380/Validation_47334380_47334380_64337.886.png", "Validation_47334380/Validation_47334380_47334380_64339.902.png", "Validation_47334380/Validation_47334380_47334380_64341.901.png", "Validation_47334380/Validation_47334380_47334380_64343.900.png" ], "turn_id": 10, "scan_id": "Validation_47334380", "origin_question": "How far is your current position from where you were at the end of the turn 6? (in meters)", "option": [], "answer": "1.62", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 6? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47331661/Validation_47331661_47331661_23586.583.png", "Validation_47331661/Validation_47331661_47331661_23588.582.png", "Validation_47331661/Validation_47331661_47331661_23590.581.png", "Validation_47331661/Validation_47331661_47331661_23592.580.png", "Validation_47331661/Validation_47331661_47331661_23594.579.png" ], "turn_id": 1, "scan_id": "Validation_47331661", "origin_question": "Remember, have you seen any cleanser(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any cleanser(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "Validation_47331661/Validation_47331661_47331661_23596.579.png", "Validation_47331661/Validation_47331661_47331661_23598.578.png", "Validation_47331661/Validation_47331661_47331661_23600.594.png", "Validation_47331661/Validation_47331661_47331661_23602.593.png", "Validation_47331661/Validation_47331661_47331661_23604.592.png" ], "turn_id": 2, "scan_id": "Validation_47331661", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the light-colored door with the panel design\"; \"the medium-sized, rounded dark backpack\"; \"the sleek rectangular glossy light wooden table\" ", "option": [ "the light-colored door with the panel design", "the medium-sized, rounded dark backpack", "the sleek rectangular glossy light wooden table" ], "answer": "the sleek rectangular glossy light wooden table", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the light-colored door with the panel design\"; \"the medium-sized, rounded dark backpack\"; \"the sleek rectangular glossy light wooden table\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the light-colored door with the panel design', 'the medium-sized, rounded dark backpack', 'the sleek rectangular glossy light wooden table'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_47331661/Validation_47331661_47331661_23606.591.png", "Validation_47331661/Validation_47331661_47331661_23608.590.png", "Validation_47331661/Validation_47331661_47331661_23610.589.png", "Validation_47331661/Validation_47331661_47331661_23612.589.png", "Validation_47331661/Validation_47331661_47331661_23614.588.png" ], "turn_id": 3, "scan_id": "Validation_47331661", "origin_question": "In which turn did you see both \"the light-colored door with the panel design\" and \"the sleek rectangular glossy light wooden table\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the light-colored door with the panel design\" and \"the sleek rectangular glossy light wooden table\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_47331661/Validation_47331661_47331661_23616.587.png", "Validation_47331661/Validation_47331661_47331661_23618.586.png", "Validation_47331661/Validation_47331661_47331661_23620.586.png", "Validation_47331661/Validation_47331661_47331661_23622.585.png", "Validation_47331661/Validation_47331661_47331661_23624.584.png" ], "turn_id": 4, "scan_id": "Validation_47331661", "origin_question": "By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the medium-sized, rounded dark backpack\", Y: \"the sleek rectangular glossy light wooden table\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to Y and farther away from X", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the medium-sized, rounded dark backpack\", Y: \"the sleek rectangular glossy light wooden table\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_47331661/Validation_47331661_47331661_23626.583.png", "Validation_47331661/Validation_47331661_47331661_23628.582.png", "Validation_47331661/Validation_47331661_47331661_23630.581.png", "Validation_47331661/Validation_47331661_47331661_23632.581.png", "Validation_47331661/Validation_47331661_47331661_23634.580.png" ], "turn_id": 5, "scan_id": "Validation_47331661", "origin_question": "Please recall \"the light-colored door with the panel design\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.7", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the light-colored door with the panel design\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_47331661/Validation_47331661_47331661_23636.579.png", "Validation_47331661/Validation_47331661_47331661_23638.578.png", "Validation_47331661/Validation_47331661_47331661_23640.594.png", "Validation_47331661/Validation_47331661_47331661_23642.593.png", "Validation_47331661/Validation_47331661_47331661_23644.592.png" ], "turn_id": 6, "scan_id": "Validation_47331661", "origin_question": "By the end of the 5 turn, were you getting closer to or farther away from these two objects: X: \"the sleek rectangular window with the white frame\", Y: \"the stool\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to X and farther away from Y", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 5 turn, were you getting closer to or farther away from these two objects: X: \"the sleek rectangular window with the white frame\", Y: \"the stool\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_47331661/Validation_47331661_47331661_23646.591.png", "Validation_47331661/Validation_47331661_47331661_23648.591.png", "Validation_47331661/Validation_47331661_47331661_23650.590.png", "Validation_47331661/Validation_47331661_47331661_23652.589.png", "Validation_47331661/Validation_47331661_47331661_23654.588.png" ], "turn_id": 7, "scan_id": "Validation_47331661", "origin_question": "Which two objects are on the same side of you now? \"the light-colored door with the panel design\", \"the vibrant green floral print chair with hardwood frame\", \"the stool\". ", "option": [ "the light-colored door with the panel design, the vibrant green floral print chair with hardwood frame", "the light-colored door with the panel design, the stool", "the vibrant green floral print chair with hardwood frame, the stool" ], "answer": "the vibrant green floral print chair with hardwood frame, the stool", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the light-colored door with the panel design\", \"the vibrant green floral print chair with hardwood frame\", \"the stool\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the light-colored door with the panel design, the vibrant green floral print chair with hardwood frame', 'the light-colored door with the panel design, the stool', 'the vibrant green floral print chair with hardwood frame, the stool'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_47331661/Validation_47331661_47331661_23656.587.png", "Validation_47331661/Validation_47331661_47331661_23658.587.png", "Validation_47331661/Validation_47331661_47331661_23660.586.png", "Validation_47331661/Validation_47331661_47331661_23662.585.png", "Validation_47331661/Validation_47331661_47331661_23664.584.png" ], "turn_id": 8, "scan_id": "Validation_47331661", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the sleek rectangular window with the white frame\" from your position?", "option": [], "answer": "157.77", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the sleek rectangular window with the white frame\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "Validation_47331661/Validation_47331661_47331661_23666.583.png", "Validation_47331661/Validation_47331661_47331661_23668.583.png", "Validation_47331661/Validation_47331661_47331661_23670.582.png", "Validation_47331661/Validation_47331661_47331661_23672.581.png", "Validation_47331661/Validation_47331661_47331661_23674.580.png" ], "turn_id": 9, "scan_id": "Validation_47331661", "origin_question": "Among these three objects(\"the medium-sized, rounded dark backpack\",\"the light-colored door with the panel design\",\"the stool\" ), which one is the closest to you?", "option": [ "the medium-sized, rounded dark backpack", "the light-colored door with the panel design", "the stool" ], "answer": "the stool", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the medium-sized, rounded dark backpack\",\"the light-colored door with the panel design\",\"the stool\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the medium-sized, rounded dark backpack', 'the light-colored door with the panel design', 'the stool'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_47331661/Validation_47331661_47331661_23676.579.png", "Validation_47331661/Validation_47331661_47331661_23678.578.png", "Validation_47331661/Validation_47331661_47331661_23680.578.png", "Validation_47331661/Validation_47331661_47331661_23682.593.png", "Validation_47331661/Validation_47331661_47331661_23684.593.png" ], "turn_id": 10, "scan_id": "Validation_47331661", "origin_question": "Which two objects are on the same side of you now? \"the sleek rectangular window with the white frame\", \"the vibrant green floral print chair with hardwood frame\", \"the bed\". ", "option": [ "the sleek rectangular window with the white frame, the vibrant green floral print chair with hardwood frame", "the sleek rectangular window with the white frame, the bed", "the vibrant green floral print chair with hardwood frame, the bed" ], "answer": "the sleek rectangular window with the white frame, the vibrant green floral print chair with hardwood frame", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the sleek rectangular window with the white frame\", \"the vibrant green floral print chair with hardwood frame\", \"the bed\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the sleek rectangular window with the white frame, the vibrant green floral print chair with hardwood frame', 'the sleek rectangular window with the white frame, the bed', 'the vibrant green floral print chair with hardwood frame, the bed'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47333436/Validation_47333436_47333436_52306.524.png", "Validation_47333436/Validation_47333436_47333436_52308.507.png", "Validation_47333436/Validation_47333436_47333436_52310.506.png", "Validation_47333436/Validation_47333436_47333436_52312.505.png" ], "turn_id": 1, "scan_id": "Validation_47333436", "origin_question": "Remember, have you seen any cabinet(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any cabinet(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47333436/Validation_47333436_47333436_52314.504.png", "Validation_47333436/Validation_47333436_47333436_52316.503.png", "Validation_47333436/Validation_47333436_47333436_52318.503.png", "Validation_47333436/Validation_47333436_47333436_52320.502.png" ], "turn_id": 2, "scan_id": "Validation_47333436", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the flowerpot\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the flowerpot\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_47333436/Validation_47333436_47333436_52322.501.png", "Validation_47333436/Validation_47333436_47333436_52324.500.png", "Validation_47333436/Validation_47333436_47333436_52326.499.png", "Validation_47333436/Validation_47333436_47333436_52328.515.png" ], "turn_id": 3, "scan_id": "Validation_47333436", "origin_question": "Please recall \"the monitor\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.99", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the monitor\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_47333436/Validation_47333436_47333436_52330.514.png", "Validation_47333436/Validation_47333436_47333436_52332.514.png", "Validation_47333436/Validation_47333436_47333436_52334.513.png", "Validation_47333436/Validation_47333436_47333436_52336.512.png" ], "turn_id": 4, "scan_id": "Validation_47333436", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_47333436/Validation_47333436_47333436_52338.511.png", "Validation_47333436/Validation_47333436_47333436_52340.510.png", "Validation_47333436/Validation_47333436_47333436_52342.510.png", "Validation_47333436/Validation_47333436_47333436_52344.509.png" ], "turn_id": 5, "scan_id": "Validation_47333436", "origin_question": "What will be the first-time appearance order of the following categories: bed, monitor, window?", "option": [ "monitor, window, bed", "bed, window, monitor", "window, bed, monitor", "window, monitor, bed" ], "answer": "monitor, window, bed", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bed, monitor, window? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['monitor, window, bed', 'bed, window, monitor', 'window, bed, monitor', 'window, monitor, bed'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_47333436/Validation_47333436_47333436_52346.508.png", "Validation_47333436/Validation_47333436_47333436_52348.507.png", "Validation_47333436/Validation_47333436_47333436_52350.506.png", "Validation_47333436/Validation_47333436_47333436_52352.505.png" ], "turn_id": 6, "scan_id": "Validation_47333436", "origin_question": "By the end of the 5 turn, were you getting closer to or farther away from these two objects: X: \"the monitor\", Y: \"the rectangular modern picture frame for multiple pictures\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 5 turn, were you getting closer to or farther away from these two objects: X: \"the monitor\", Y: \"the rectangular modern picture frame for multiple pictures\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_48458430/Validation_48458430_48458430_116441.262.png", "Validation_48458430/Validation_48458430_48458430_116442.745.png", "Validation_48458430/Validation_48458430_48458430_116444.244.png", "Validation_48458430/Validation_48458430_48458430_116445.743.png" ], "turn_id": 1, "scan_id": "Validation_48458430", "origin_question": "Remember, have you seen any towel(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any towel(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_48458430/Validation_48458430_48458430_116447.243.png", "Validation_48458430/Validation_48458430_48458430_116448.742.png", "Validation_48458430/Validation_48458430_48458430_116450.242.png", "Validation_48458430/Validation_48458430_48458430_116451.741.png" ], "turn_id": 2, "scan_id": "Validation_48458430", "origin_question": "In which turn did you see both \"the body loofah\" and \"the small rectangular window with the white pvc frame\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the body loofah\" and \"the small rectangular window with the white pvc frame\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_48458430/Validation_48458430_48458430_116453.240.png", "Validation_48458430/Validation_48458430_48458430_116454.740.png", "Validation_48458430/Validation_48458430_48458430_116456.239.png", "Validation_48458430/Validation_48458430_48458430_116457.738.png" ], "turn_id": 3, "scan_id": "Validation_48458430", "origin_question": "Which two objects are on the same side of you now? \"the soft white rectangular door with recessed panels\", \"the small rectangular window with the white pvc frame\", \"the cylindrical ceramic flowerpot in the neutral color\". ", "option": [ "the soft white rectangular door with recessed panels, the small rectangular window with the white pvc frame", "the soft white rectangular door with recessed panels, the cylindrical ceramic flowerpot in the neutral color", "the small rectangular window with the white pvc frame, the cylindrical ceramic flowerpot in the neutral color" ], "answer": "the soft white rectangular door with recessed panels, the small rectangular window with the white pvc frame", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the soft white rectangular door with recessed panels\", \"the small rectangular window with the white pvc frame\", \"the cylindrical ceramic flowerpot in the neutral color\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the soft white rectangular door with recessed panels, the small rectangular window with the white pvc frame', 'the soft white rectangular door with recessed panels, the cylindrical ceramic flowerpot in the neutral color', 'the small rectangular window with the white pvc frame, the cylindrical ceramic flowerpot in the neutral color'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_48458430/Validation_48458430_48458430_116459.238.png", "Validation_48458430/Validation_48458430_48458430_116460.737.png", "Validation_48458430/Validation_48458430_48458430_116462.253.png", "Validation_48458430/Validation_48458430_48458430_116463.753.png" ], "turn_id": 4, "scan_id": "Validation_48458430", "origin_question": "By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the soft white rectangular door with recessed panels\", Y: \"the small rectangular window with the white pvc frame\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to X and farther away from Y", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the soft white rectangular door with recessed panels\", Y: \"the small rectangular window with the white pvc frame\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_41069048/Validation_41069048_41069048_5002.435.png", "Validation_41069048/Validation_41069048_41069048_5003.917.png", "Validation_41069048/Validation_41069048_41069048_5005.417.png", "Validation_41069048/Validation_41069048_41069048_5006.916.png" ], "turn_id": 1, "scan_id": "Validation_41069048", "origin_question": "Remember, have you seen any bin(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any bin(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_41069048/Validation_41069048_41069048_5008.415.png", "Validation_41069048/Validation_41069048_41069048_5009.915.png", "Validation_41069048/Validation_41069048_41069048_5011.414.png", "Validation_41069048/Validation_41069048_41069048_5012.914.png" ], "turn_id": 2, "scan_id": "Validation_41069048", "origin_question": "In which turn did you see both \"the large, plush white towel with decorative patterns\" and \"the bathtub\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the large, plush white towel with decorative patterns\" and \"the bathtub\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_41069048/Validation_41069048_41069048_5014.413.png", "Validation_41069048/Validation_41069048_41069048_5015.912.png", "Validation_41069048/Validation_41069048_41069048_5017.412.png", "Validation_41069048/Validation_41069048_41069048_5018.911.png" ], "turn_id": 3, "scan_id": "Validation_41069048", "origin_question": "By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the sleek, white porcelain toilet with the rounded shape\", Y: \"the compact chrome cylindrical bin with the closed lid\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "farther away from both X and Y", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the sleek, white porcelain toilet with the rounded shape\", Y: \"the compact chrome cylindrical bin with the closed lid\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_41069048/Validation_41069048_41069048_5020.410.png", "Validation_41069048/Validation_41069048_41069048_5021.910.png", "Validation_41069048/Validation_41069048_41069048_5023.409.png", "Validation_41069048/Validation_41069048_41069048_5024.909.png" ], "turn_id": 4, "scan_id": "Validation_41069048", "origin_question": "At the end of which turn were both of \"the sleek, white porcelain toilet with the rounded shape\" and \"the compact chrome cylindrical bin with the closed lid\" on your right side?", "option": [], "answer": "3", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the sleek, white porcelain toilet with the rounded shape\" and \"the compact chrome cylindrical bin with the closed lid\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_41069048/Validation_41069048_41069048_5026.408.png", "Validation_41069048/Validation_41069048_41069048_5027.907.png", "Validation_41069048/Validation_41069048_41069048_5029.407.png", "Validation_41069048/Validation_41069048_41069048_5030.906.png" ], "turn_id": 5, "scan_id": "Validation_41069048", "origin_question": "By the end of the 3 turn, were you getting closer to or farther away from these two objects: X: \"the large, plush white towel with decorative patterns\", Y: \"an elegant oval glossy porcelain sink with curved edges\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to X and farther away from Y", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 3 turn, were you getting closer to or farther away from these two objects: X: \"the large, plush white towel with decorative patterns\", Y: \"an elegant oval glossy porcelain sink with curved edges\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42898826/Validation_42898826_42898826_168098.203.png", "Validation_42898826/Validation_42898826_42898826_168099.686.png", "Validation_42898826/Validation_42898826_42898826_168101.185.png", "Validation_42898826/Validation_42898826_42898826_168102.685.png" ], "turn_id": 1, "scan_id": "Validation_42898826", "origin_question": "Remember, have you seen any bathtub(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any bathtub(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_42898826/Validation_42898826_42898826_168104.184.png", "Validation_42898826/Validation_42898826_42898826_168105.683.png", "Validation_42898826/Validation_42898826_42898826_168107.183.png", "Validation_42898826/Validation_42898826_42898826_168108.682.png" ], "turn_id": 2, "scan_id": "Validation_42898826", "origin_question": "In which turn were you closest to \"the sleek oval pink bathtub made of acrylic\"? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the sleek oval pink bathtub made of acrylic\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_42898826/Validation_42898826_42898826_168110.181.png", "Validation_42898826/Validation_42898826_42898826_168111.681.png", "Validation_42898826/Validation_42898826_42898826_168113.180.png", "Validation_42898826/Validation_42898826_42898826_168114.679.png" ], "turn_id": 3, "scan_id": "Validation_42898826", "origin_question": "At the end of which turn were both of \"the sleek oval pink bathtub made of acrylic\" and \"the lush plant with deep green irregular foliage\" on your left side?", "option": [], "answer": "2", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the sleek oval pink bathtub made of acrylic\" and \"the lush plant with deep green irregular foliage\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_42898826/Validation_42898826_42898826_168116.179.png", "Validation_42898826/Validation_42898826_42898826_168117.678.png", "Validation_42898826/Validation_42898826_42898826_168119.178.png", "Validation_42898826/Validation_42898826_42898826_168120.677.png" ], "turn_id": 4, "scan_id": "Validation_42898826", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42898826/Validation_42898826_42898826_168122.176.png", "Validation_42898826/Validation_42898826_42898826_168123.676.png", "Validation_42898826/Validation_42898826_42898826_168125.175.png", "Validation_42898826/Validation_42898826_42898826_168126.674.png" ], "turn_id": 5, "scan_id": "Validation_42898826", "origin_question": "Remember, have you seen any towel(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any towel(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_42898826/Validation_42898826_42898826_168128.174.png", "Validation_42898826/Validation_42898826_42898826_168129.673.png", "Validation_42898826/Validation_42898826_42898826_168131.173.png", "Validation_42898826/Validation_42898826_42898826_168132.672.png" ], "turn_id": 6, "scan_id": "Validation_42898826", "origin_question": "Please recall \"the compact modern toilet in off-white plastic\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.12", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the compact modern toilet in off-white plastic\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_42898826/Validation_42898826_42898826_168134.188.png", "Validation_42898826/Validation_42898826_42898826_168135.688.png", "Validation_42898826/Validation_42898826_42898826_168137.187.png", "Validation_42898826/Validation_42898826_42898826_168138.686.png" ], "turn_id": 7, "scan_id": "Validation_42898826", "origin_question": "", "option": [], "answer": "", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "Validation_47429995/Validation_47429995_47429995_16686.705.png", "Validation_47429995/Validation_47429995_47429995_16688.188.png", "Validation_47429995/Validation_47429995_47429995_16689.687.png", "Validation_47429995/Validation_47429995_47429995_16691.187.png" ], "turn_id": 1, "scan_id": "Validation_47429995", "origin_question": "Remember, how many flowerpot(s) have you seen so far? ", "option": [], "answer": "4", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many flowerpot(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_47429995/Validation_47429995_47429995_16692.686.png", "Validation_47429995/Validation_47429995_47429995_16694.185.png", "Validation_47429995/Validation_47429995_47429995_16695.685.png", "Validation_47429995/Validation_47429995_47429995_16697.184.png" ], "turn_id": 2, "scan_id": "Validation_47429995", "origin_question": "At the end of which turn were both of \"the table\" and \"the medium-sized rectangular window with the white frame\" on your left side?", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the table\" and \"the medium-sized rectangular window with the white frame\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "Validation_47429995/Validation_47429995_47429995_16698.684.png", "Validation_47429995/Validation_47429995_47429995_16700.183.png", "Validation_47429995/Validation_47429995_47429995_16701.682.png", "Validation_47429995/Validation_47429995_47429995_16703.182.png" ], "turn_id": 3, "scan_id": "Validation_47429995", "origin_question": "Among these three objects(\"the table\",\"the medium-sized rectangular window with the white frame\",\"the soft pink rectangular canvas bag\" ), which one is the farthest from you?", "option": [ "the table", "the medium-sized rectangular window with the white frame", "the soft pink rectangular canvas bag" ], "answer": "the soft pink rectangular canvas bag", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects(\"the table\",\"the medium-sized rectangular window with the white frame\",\"the soft pink rectangular canvas bag\" ), which one is the farthest from you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the table', 'the medium-sized rectangular window with the white frame', 'the soft pink rectangular canvas bag'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47429995/Validation_47429995_47429995_16704.698.png", "Validation_47429995/Validation_47429995_47429995_16706.197.png", "Validation_47429995/Validation_47429995_47429995_16707.696.png", "Validation_47429995/Validation_47429995_47429995_16709.196.png" ], "turn_id": 4, "scan_id": "Validation_47429995", "origin_question": "Which direction is the slim black rectangular monitor with the sleek design in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the slim black rectangular monitor with the sleek design in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47429995/Validation_47429995_47429995_16710.695.png", "Validation_47429995/Validation_47429995_47429995_16712.195.png", "Validation_47429995/Validation_47429995_47429995_16713.694.png", "Validation_47429995/Validation_47429995_47429995_16715.194.png" ], "turn_id": 5, "scan_id": "Validation_47429995", "origin_question": "In which turn were you closest to \"the natural brown woven box with textured surface\"? ", "option": [], "answer": "5", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the natural brown woven box with textured surface\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_45663114/Validation_45663114_45663114_55376.883.png", "Validation_45663114/Validation_45663114_45663114_55378.366.png", "Validation_45663114/Validation_45663114_45663114_55379.865.png", "Validation_45663114/Validation_45663114_45663114_55381.365.png" ], "turn_id": 1, "scan_id": "Validation_45663114", "origin_question": "Remember, have you seen any magazine(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any magazine(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_45663114/Validation_45663114_45663114_55382.864.png", "Validation_45663114/Validation_45663114_45663114_55384.363.png", "Validation_45663114/Validation_45663114_45663114_55385.863.png", "Validation_45663114/Validation_45663114_45663114_55387.362.png" ], "turn_id": 2, "scan_id": "Validation_45663114", "origin_question": "Which two objects are on the same side of you now? \"the door\", \"the white porcelain pedestal sink with the rounded basin\", \"the plush, absorbent gray towel\". ", "option": [ "the door, the white porcelain pedestal sink with the rounded basin", "the door, the plush, absorbent gray towel", "the white porcelain pedestal sink with the rounded basin, the plush, absorbent gray towel" ], "answer": "the door, the white porcelain pedestal sink with the rounded basin", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the door\", \"the white porcelain pedestal sink with the rounded basin\", \"the plush, absorbent gray towel\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the door, the white porcelain pedestal sink with the rounded basin', 'the door, the plush, absorbent gray towel', 'the white porcelain pedestal sink with the rounded basin, the plush, absorbent gray towel'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_45663114/Validation_45663114_45663114_55388.861.png", "Validation_45663114/Validation_45663114_45663114_55390.361.png", "Validation_45663114/Validation_45663114_45663114_55391.860.png", "Validation_45663114/Validation_45663114_45663114_55393.360.png" ], "turn_id": 3, "scan_id": "Validation_45663114", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the plush, absorbent gray towel\" from your position?", "option": [], "answer": "140.17", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the plush, absorbent gray towel\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "Validation_45663114/Validation_45663114_45663114_55394.859.png", "Validation_45663114/Validation_45663114_45663114_55396.358.png", "Validation_45663114/Validation_45663114_45663114_55397.858.png", "Validation_45663114/Validation_45663114_45663114_55399.357.png" ], "turn_id": 4, "scan_id": "Validation_45663114", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the compact cylindrical white plastic bin\"; \"the door\"; \"the bathtub\" ", "option": [ "the compact cylindrical white plastic bin", "the door", "the bathtub" ], "answer": "the compact cylindrical white plastic bin", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the compact cylindrical white plastic bin\"; \"the door\"; \"the bathtub\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the compact cylindrical white plastic bin', 'the door', 'the bathtub'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_45663114/Validation_45663114_45663114_55400.857.png", "Validation_45663114/Validation_45663114_45663114_55402.356.png", "Validation_45663114/Validation_45663114_45663114_55403.855.png", "Validation_45663114/Validation_45663114_45663114_55405.355.png" ], "turn_id": 5, "scan_id": "Validation_45663114", "origin_question": "Which direction is the ceramic toilet with the blue and teal whale pattern in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the ceramic toilet with the blue and teal whale pattern in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_45663114/Validation_45663114_45663114_55406.871.png", "Validation_45663114/Validation_45663114_45663114_55408.370.png", "Validation_45663114/Validation_45663114_45663114_55409.869.png", "Validation_45663114/Validation_45663114_45663114_55411.369.png" ], "turn_id": 6, "scan_id": "Validation_45663114", "origin_question": "When did you first discover \"the bathtub\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the bathtub\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_45663114/Validation_45663114_45663114_55412.868.png", "Validation_45663114/Validation_45663114_45663114_55414.368.png", "Validation_45663114/Validation_45663114_45663114_55415.867.png", "Validation_45663114/Validation_45663114_45663114_55417.366.png" ], "turn_id": 7, "scan_id": "Validation_45663114", "origin_question": "Assuming the direction you are facing at the end of the turn 4 is forward, did you move a certain distance forward or backward from that position?", "option": [ "Forward", "Backward" ], "answer": "Forward", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 4 is forward, did you move a certain distance forward or backward from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Forward', 'Backward'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_41254402/Validation_41254402_41254402_95895.205.png", "Validation_41254402/Validation_41254402_41254402_95897.188.png", "Validation_41254402/Validation_41254402_41254402_95899.187.png", "Validation_41254402/Validation_41254402_41254402_95901.186.png", "Validation_41254402/Validation_41254402_41254402_95903.185.png" ], "turn_id": 1, "scan_id": "Validation_41254402", "origin_question": "Remember, have you seen any bed(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any bed(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_41254402/Validation_41254402_41254402_95905.185.png", "Validation_41254402/Validation_41254402_41254402_95907.184.png", "Validation_41254402/Validation_41254402_41254402_95909.183.png", "Validation_41254402/Validation_41254402_41254402_95911.182.png", "Validation_41254402/Validation_41254402_41254402_95913.181.png" ], "turn_id": 2, "scan_id": "Validation_41254402", "origin_question": "When did you first discover \"the standard-sized rectangular beige cotton pillow\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the standard-sized rectangular beige cotton pillow\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_41254402/Validation_41254402_41254402_95915.181.png", "Validation_41254402/Validation_41254402_41254402_95917.196.png", "Validation_41254402/Validation_41254402_41254402_95919.196.png", "Validation_41254402/Validation_41254402_41254402_95921.195.png", "Validation_41254402/Validation_41254402_41254402_95923.194.png" ], "turn_id": 3, "scan_id": "Validation_41254402", "origin_question": "In which turn did you see both \"the sleek, modern cabinet with the smooth white finish\" and \"the tall, grey table with the mesh-like design\" simultaneously? ", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the sleek, modern cabinet with the smooth white finish\" and \"the tall, grey table with the mesh-like design\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_41254402/Validation_41254402_41254402_95925.193.png", "Validation_41254402/Validation_41254402_41254402_95927.192.png", "Validation_41254402/Validation_41254402_41254402_95929.192.png", "Validation_41254402/Validation_41254402_41254402_95931.191.png", "Validation_41254402/Validation_41254402_41254402_95933.190.png" ], "turn_id": 4, "scan_id": "Validation_41254402", "origin_question": "By the end of the 3 turn, were you getting closer to or farther away from these two objects: X: \"the warm light-brown wooden picture frame\", Y: \"the tall, grey table with the mesh-like design\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "farther away from both X and Y", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 3 turn, were you getting closer to or farther away from these two objects: X: \"the warm light-brown wooden picture frame\", Y: \"the tall, grey table with the mesh-like design\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_41254402/Validation_41254402_41254402_95935.189.png", "Validation_41254402/Validation_41254402_41254402_95937.189.png", "Validation_41254402/Validation_41254402_41254402_95939.188.png", "Validation_41254402/Validation_41254402_41254402_95941.187.png", "Validation_41254402/Validation_41254402_41254402_95943.186.png" ], "turn_id": 5, "scan_id": "Validation_41254402", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the light neutral wooden door with an elegant panel design\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the light neutral wooden door with an elegant panel design\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_41254402/Validation_41254402_41254402_95945.185.png", "Validation_41254402/Validation_41254402_41254402_95947.185.png", "Validation_41254402/Validation_41254402_41254402_95949.184.png", "Validation_41254402/Validation_41254402_41254402_95951.183.png", "Validation_41254402/Validation_41254402_41254402_95953.182.png" ], "turn_id": 6, "scan_id": "Validation_41254402", "origin_question": "Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "66.02", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_41254402/Validation_41254402_41254402_95955.181.png", "Validation_41254402/Validation_41254402_41254402_95957.181.png", "Validation_41254402/Validation_41254402_41254402_95959.196.png", "Validation_41254402/Validation_41254402_41254402_95961.196.png", "Validation_41254402/Validation_41254402_41254402_95963.195.png" ], "turn_id": 7, "scan_id": "Validation_41254402", "origin_question": "Which direction is the light neutral wooden door with an elegant panel design in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the light neutral wooden door with an elegant panel design in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_41254402/Validation_41254402_41254402_95965.194.png", "Validation_41254402/Validation_41254402_41254402_95967.193.png", "Validation_41254402/Validation_41254402_41254402_95969.193.png", "Validation_41254402/Validation_41254402_41254402_95971.192.png", "Validation_41254402/Validation_41254402_41254402_95973.191.png" ], "turn_id": 8, "scan_id": "Validation_41254402", "origin_question": "Is the sleek, modern cabinet with the smooth white finish to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the sleek, modern cabinet with the smooth white finish to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47895738/Validation_47895738_47895738_384959.070.png", "Validation_47895738/Validation_47895738_47895738_384960.536.png", "Validation_47895738/Validation_47895738_47895738_384962.052.png", "Validation_47895738/Validation_47895738_47895738_384963.552.png" ], "turn_id": 1, "scan_id": "Validation_47895738", "origin_question": "Remember, have you seen any dresser(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any dresser(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47895738/Validation_47895738_47895738_384965.051.png", "Validation_47895738/Validation_47895738_47895738_384966.550.png", "Validation_47895738/Validation_47895738_47895738_384968.050.png", "Validation_47895738/Validation_47895738_47895738_384969.549.png" ], "turn_id": 2, "scan_id": "Validation_47895738", "origin_question": "In which turn were you closest to \"the light\"? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the light\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_47895738/Validation_47895738_47895738_384971.048.png", "Validation_47895738/Validation_47895738_47895738_384972.548.png", "Validation_47895738/Validation_47895738_47895738_384974.047.png", "Validation_47895738/Validation_47895738_47895738_384975.547.png" ], "turn_id": 3, "scan_id": "Validation_47895738", "origin_question": "In which turn did you see both \"the large dark rectangular frameless mirror\" and \"the tall, sleek cabinet with the smooth neutral surface\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the large dark rectangular frameless mirror\" and \"the tall, sleek cabinet with the smooth neutral surface\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47895738/Validation_47895738_47895738_384977.046.png", "Validation_47895738/Validation_47895738_47895738_384978.545.png", "Validation_47895738/Validation_47895738_47895738_384980.045.png", "Validation_47895738/Validation_47895738_47895738_384981.544.png" ], "turn_id": 4, "scan_id": "Validation_47895738", "origin_question": "Remember, have you seen any dvd(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any dvd(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_47895738/Validation_47895738_47895738_384983.044.png", "Validation_47895738/Validation_47895738_47895738_384984.543.png", "Validation_47895738/Validation_47895738_47895738_384986.042.png", "Validation_47895738/Validation_47895738_47895738_384987.542.png" ], "turn_id": 5, "scan_id": "Validation_47895738", "origin_question": "By the end of the 3 turn, were you getting closer to or farther away from these two objects: X: \"the sleek white acrylic bathtub with an elongated shape\", Y: \"the sleek oval white porcelain toilet\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "farther away from both X and Y", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 3 turn, were you getting closer to or farther away from these two objects: X: \"the sleek white acrylic bathtub with an elongated shape\", Y: \"the sleek oval white porcelain toilet\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_47895738/Validation_47895738_47895738_384989.041.png", "Validation_47895738/Validation_47895738_47895738_384990.540.png", "Validation_47895738/Validation_47895738_47895738_384992.040.png", "Validation_47895738/Validation_47895738_47895738_384993.539.png" ], "turn_id": 6, "scan_id": "Validation_47895738", "origin_question": "In which turn did you see both \"the tall, sleek cabinet with the smooth neutral surface\" and \"the small, soft, vibrant teal towel\" simultaneously? ", "option": [], "answer": "5", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the tall, sleek cabinet with the smooth neutral surface\" and \"the small, soft, vibrant teal towel\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42446156/Validation_42446156_42446156_79136.123.png", "Validation_42446156/Validation_42446156_42446156_79138.122.png", "Validation_42446156/Validation_42446156_42446156_79140.121.png", "Validation_42446156/Validation_42446156_42446156_79142.120.png", "Validation_42446156/Validation_42446156_42446156_79144.119.png" ], "turn_id": 1, "scan_id": "Validation_42446156", "origin_question": "Remember, have you seen any monitor(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any monitor(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "Validation_42446156/Validation_42446156_42446156_79146.119.png", "Validation_42446156/Validation_42446156_42446156_79148.118.png", "Validation_42446156/Validation_42446156_42446156_79150.117.png", "Validation_42446156/Validation_42446156_42446156_79152.116.png", "Validation_42446156/Validation_42446156_42446156_79154.115.png" ], "turn_id": 2, "scan_id": "Validation_42446156", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the rectangular, medium-toned wooden table\"; \"the rectangular floral cotton pillow in creamy beige\"; \"the sleek dark square stool with slim legs\" ", "option": [ "the rectangular, medium-toned wooden table", "the rectangular floral cotton pillow in creamy beige", "the sleek dark square stool with slim legs" ], "answer": "the sleek dark square stool with slim legs", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the rectangular, medium-toned wooden table\"; \"the rectangular floral cotton pillow in creamy beige\"; \"the sleek dark square stool with slim legs\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular, medium-toned wooden table', 'the rectangular floral cotton pillow in creamy beige', 'the sleek dark square stool with slim legs'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_42446156/Validation_42446156_42446156_79156.115.png", "Validation_42446156/Validation_42446156_42446156_79158.114.png", "Validation_42446156/Validation_42446156_42446156_79160.113.png", "Validation_42446156/Validation_42446156_42446156_79162.112.png", "Validation_42446156/Validation_42446156_42446156_79164.111.png" ], "turn_id": 3, "scan_id": "Validation_42446156", "origin_question": "Please recall \"the medium-sized cylindrical dark plastic bucket\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.92", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the medium-sized cylindrical dark plastic bucket\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_42446156/Validation_42446156_42446156_79166.111.png", "Validation_42446156/Validation_42446156_42446156_79168.110.png", "Validation_42446156/Validation_42446156_42446156_79170.109.png", "Validation_42446156/Validation_42446156_42446156_79172.108.png", "Validation_42446156/Validation_42446156_42446156_79174.108.png" ], "turn_id": 4, "scan_id": "Validation_42446156", "origin_question": "When did you first discover \"the medium-sized wooden cabinet with the warm brown hue\"? (index of the turn)", "option": [], "answer": "3", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the medium-sized wooden cabinet with the warm brown hue\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_42446156/Validation_42446156_42446156_79176.123.png", "Validation_42446156/Validation_42446156_42446156_79178.123.png", "Validation_42446156/Validation_42446156_42446156_79180.122.png", "Validation_42446156/Validation_42446156_42446156_79182.121.png", "Validation_42446156/Validation_42446156_42446156_79184.120.png" ], "turn_id": 5, "scan_id": "Validation_42446156", "origin_question": "Is the large, sleek rectangular black monitor to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the large, sleek rectangular black monitor to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_42446156/Validation_42446156_42446156_79186.120.png", "Validation_42446156/Validation_42446156_42446156_79188.119.png", "Validation_42446156/Validation_42446156_42446156_79190.118.png", "Validation_42446156/Validation_42446156_42446156_79192.117.png", "Validation_42446156/Validation_42446156_42446156_79194.116.png" ], "turn_id": 6, "scan_id": "Validation_42446156", "origin_question": "Which direction is the medium-sized wooden cabinet with the warm brown hue in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the medium-sized wooden cabinet with the warm brown hue in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_42446156/Validation_42446156_42446156_79196.115.png", "Validation_42446156/Validation_42446156_42446156_79198.115.png", "Validation_42446156/Validation_42446156_42446156_79200.114.png", "Validation_42446156/Validation_42446156_42446156_79202.113.png", "Validation_42446156/Validation_42446156_42446156_79204.112.png" ], "turn_id": 7, "scan_id": "Validation_42446156", "origin_question": "In which turn were you farthest from \"the large, sleek rectangular black monitor\"? ", "option": [], "answer": "1", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the large, sleek rectangular black monitor\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_42446156/Validation_42446156_42446156_79206.112.png", "Validation_42446156/Validation_42446156_42446156_79208.111.png", "Validation_42446156/Validation_42446156_42446156_79210.110.png", "Validation_42446156/Validation_42446156_42446156_79212.109.png", "Validation_42446156/Validation_42446156_42446156_79214.108.png" ], "turn_id": 8, "scan_id": "Validation_42446156", "origin_question": "Using your orientation at the end of turn 4 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "61.13", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 4 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_42446156/Validation_42446156_42446156_79216.108.png", "Validation_42446156/Validation_42446156_42446156_79218.124.png", "Validation_42446156/Validation_42446156_42446156_79220.123.png", "Validation_42446156/Validation_42446156_42446156_79222.122.png", "Validation_42446156/Validation_42446156_42446156_79224.121.png" ], "turn_id": 9, "scan_id": "Validation_42446156", "origin_question": "Which direction is the medium-sized cylindrical dark plastic bucket in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the medium-sized cylindrical dark plastic bucket in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42445988/Validation_42445988_42445988_347214.745.png", "Validation_42445988/Validation_42445988_42445988_347216.727.png", "Validation_42445988/Validation_42445988_42445988_347218.727.png", "Validation_42445988/Validation_42445988_42445988_347220.726.png", "Validation_42445988/Validation_42445988_42445988_347222.725.png" ], "turn_id": 1, "scan_id": "Validation_42445988", "origin_question": "Remember, have you seen any chair(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any chair(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_42445988/Validation_42445988_42445988_347224.741.png", "Validation_42445988/Validation_42445988_42445988_347226.740.png", "Validation_42445988/Validation_42445988_42445988_347228.739.png", "Validation_42445988/Validation_42445988_42445988_347230.738.png", "Validation_42445988/Validation_42445988_42445988_347232.738.png" ], "turn_id": 2, "scan_id": "Validation_42445988", "origin_question": "When did you first discover \"the vibrant green triangular plant\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the vibrant green triangular plant\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_42445988/Validation_42445988_42445988_347234.737.png", "Validation_42445988/Validation_42445988_42445988_347236.736.png", "Validation_42445988/Validation_42445988_42445988_347238.735.png", "Validation_42445988/Validation_42445988_42445988_347240.734.png", "Validation_42445988/Validation_42445988_42445988_347242.733.png" ], "turn_id": 3, "scan_id": "Validation_42445988", "origin_question": "Please recall \"the vibrant green triangular plant\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.95", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the vibrant green triangular plant\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_42445988/Validation_42445988_42445988_347244.733.png", "Validation_42445988/Validation_42445988_42445988_347246.732.png", "Validation_42445988/Validation_42445988_42445988_347248.731.png", "Validation_42445988/Validation_42445988_42445988_347250.730.png", "Validation_42445988/Validation_42445988_42445988_347252.729.png" ], "turn_id": 4, "scan_id": "Validation_42445988", "origin_question": "In which turn were you closest to \"the wooden chair with the warm natural wood tone\"? ", "option": [], "answer": "3", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the wooden chair with the warm natural wood tone\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_42445988/Validation_42445988_42445988_347254.729.png", "Validation_42445988/Validation_42445988_42445988_347256.728.png", "Validation_42445988/Validation_42445988_42445988_347258.727.png", "Validation_42445988/Validation_42445988_42445988_347260.726.png", "Validation_42445988/Validation_42445988_42445988_347262.725.png" ], "turn_id": 5, "scan_id": "Validation_42445988", "origin_question": "Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Right", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_42445988/Validation_42445988_42445988_347264.724.png", "Validation_42445988/Validation_42445988_42445988_347266.740.png", "Validation_42445988/Validation_42445988_42445988_347268.739.png", "Validation_42445988/Validation_42445988_42445988_347270.739.png", "Validation_42445988/Validation_42445988_42445988_347272.738.png" ], "turn_id": 6, "scan_id": "Validation_42445988", "origin_question": "When did you first discover \"the vibrant green triangular plant\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the vibrant green triangular plant\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_42445988/Validation_42445988_42445988_347274.737.png", "Validation_42445988/Validation_42445988_42445988_347276.736.png", "Validation_42445988/Validation_42445988_42445988_347278.735.png", "Validation_42445988/Validation_42445988_42445988_347280.735.png", "Validation_42445988/Validation_42445988_42445988_347282.734.png" ], "turn_id": 7, "scan_id": "Validation_42445988", "origin_question": "When did you first discover \"the modern rectangular monitor with the glossy screen\"? (index of the turn)", "option": [], "answer": "7", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the modern rectangular monitor with the glossy screen\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_42445988/Validation_42445988_42445988_347284.733.png", "Validation_42445988/Validation_42445988_42445988_347286.732.png", "Validation_42445988/Validation_42445988_42445988_347288.731.png", "Validation_42445988/Validation_42445988_42445988_347290.730.png", "Validation_42445988/Validation_42445988_42445988_347292.730.png" ], "turn_id": 8, "scan_id": "Validation_42445988", "origin_question": "In which turn were you closest to \"the modern rectangular monitor with the glossy screen\"? ", "option": [], "answer": "6", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the modern rectangular monitor with the glossy screen\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_42445988/Validation_42445988_42445988_347294.729.png", "Validation_42445988/Validation_42445988_42445988_347296.728.png", "Validation_42445988/Validation_42445988_42445988_347298.727.png", "Validation_42445988/Validation_42445988_42445988_347300.726.png", "Validation_42445988/Validation_42445988_42445988_347302.726.png" ], "turn_id": 9, "scan_id": "Validation_42445988", "origin_question": "What will be the first-time appearance order of the following categories: monitor, plant, window?", "option": [ "plant, monitor, window", "monitor, plant, window", "window, monitor, plant", "window, plant, monitor" ], "answer": "window, plant, monitor", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: monitor, plant, window? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['plant, monitor, window', 'monitor, plant, window', 'window, monitor, plant', 'window, plant, monitor'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_42445988/Validation_42445988_42445988_347304.725.png", "Validation_42445988/Validation_42445988_42445988_347306.741.png", "Validation_42445988/Validation_42445988_42445988_347308.740.png", "Validation_42445988/Validation_42445988_42445988_347310.739.png", "Validation_42445988/Validation_42445988_42445988_347312.738.png" ], "turn_id": 10, "scan_id": "Validation_42445988", "origin_question": "In which turn were you closest to \"the compact grey plastic foldable stool\"? ", "option": [], "answer": "9", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the compact grey plastic foldable stool\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47334362/Validation_47334362_47334362_61826.534.png", "Validation_47334362/Validation_47334362_47334362_61828.517.png", "Validation_47334362/Validation_47334362_47334362_61830.516.png", "Validation_47334362/Validation_47334362_47334362_61832.515.png", "Validation_47334362/Validation_47334362_47334362_61834.531.png" ], "turn_id": 1, "scan_id": "Validation_47334362", "origin_question": "Remember, have you seen any window(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any window(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47334362/Validation_47334362_47334362_61836.530.png", "Validation_47334362/Validation_47334362_47334362_61838.529.png", "Validation_47334362/Validation_47334362_47334362_61840.528.png", "Validation_47334362/Validation_47334362_47334362_61842.528.png", "Validation_47334362/Validation_47334362_47334362_61844.527.png" ], "turn_id": 2, "scan_id": "Validation_47334362", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47334362/Validation_47334362_47334362_61846.526.png", "Validation_47334362/Validation_47334362_47334362_61848.525.png", "Validation_47334362/Validation_47334362_47334362_61850.524.png", "Validation_47334362/Validation_47334362_47334362_61852.524.png", "Validation_47334362/Validation_47334362_47334362_61854.523.png" ], "turn_id": 3, "scan_id": "Validation_47334362", "origin_question": "", "option": [], "answer": "", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_47334362/Validation_47334362_47334362_61856.522.png", "Validation_47334362/Validation_47334362_47334362_61858.521.png", "Validation_47334362/Validation_47334362_47334362_61860.520.png", "Validation_47334362/Validation_47334362_47334362_61862.520.png", "Validation_47334362/Validation_47334362_47334362_61864.519.png" ], "turn_id": 4, "scan_id": "Validation_47334362", "origin_question": "Is the tall rectangular natural wood cabinet to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the tall rectangular natural wood cabinet to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47334362/Validation_47334362_47334362_61866.518.png", "Validation_47334362/Validation_47334362_47334362_61868.517.png", "Validation_47334362/Validation_47334362_47334362_61870.516.png", "Validation_47334362/Validation_47334362_47334362_61872.515.png", "Validation_47334362/Validation_47334362_47334362_61874.515.png" ], "turn_id": 5, "scan_id": "Validation_47334362", "origin_question": "In which turn were you closest to \"the large rectangular window with the white pvc frame\"? ", "option": [], "answer": "5", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the large rectangular window with the white pvc frame\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_47334362/Validation_47334362_47334362_61876.531.png", "Validation_47334362/Validation_47334362_47334362_61878.530.png", "Validation_47334362/Validation_47334362_47334362_61880.529.png", "Validation_47334362/Validation_47334362_47334362_61882.528.png", "Validation_47334362/Validation_47334362_47334362_61884.527.png" ], "turn_id": 6, "scan_id": "Validation_47334362", "origin_question": "Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47334362/Validation_47334362_47334362_61886.526.png", "Validation_47334362/Validation_47334362_47334362_61888.526.png", "Validation_47334362/Validation_47334362_47334362_61890.525.png", "Validation_47334362/Validation_47334362_47334362_61892.524.png", "Validation_47334362/Validation_47334362_47334362_61894.523.png" ], "turn_id": 7, "scan_id": "Validation_47334362", "origin_question": "Compared to your position at the end of the 5 turn, are you now closer or farther away from \"the large rectangular window with the white pvc frame\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 5 turn, are you now closer or farther away from \"the large rectangular window with the white pvc frame\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_47334362/Validation_47334362_47334362_61896.522.png", "Validation_47334362/Validation_47334362_47334362_61898.522.png", "Validation_47334362/Validation_47334362_47334362_61900.521.png", "Validation_47334362/Validation_47334362_47334362_61902.520.png", "Validation_47334362/Validation_47334362_47334362_61904.519.png" ], "turn_id": 8, "scan_id": "Validation_47334362", "origin_question": "By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the sleek light gray rectangular door\", Y: \"the large rectangular window with the white pvc frame\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to Y and farther away from X", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the sleek light gray rectangular door\", Y: \"the large rectangular window with the white pvc frame\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_47334362/Validation_47334362_47334362_61906.518.png", "Validation_47334362/Validation_47334362_47334362_61908.518.png", "Validation_47334362/Validation_47334362_47334362_61910.517.png", "Validation_47334362/Validation_47334362_47334362_61912.516.png", "Validation_47334362/Validation_47334362_47334362_61914.515.png" ], "turn_id": 9, "scan_id": "Validation_47334362", "origin_question": "When did you first discover \"the large, sleek flat-screen monitor with the thin frame\"? (index of the turn)", "option": [], "answer": "8", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the large, sleek flat-screen monitor with the thin frame\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_47334362/Validation_47334362_47334362_61916.531.png", "Validation_47334362/Validation_47334362_47334362_61918.530.png", "Validation_47334362/Validation_47334362_47334362_61920.529.png", "Validation_47334362/Validation_47334362_47334362_61922.529.png", "Validation_47334362/Validation_47334362_47334362_61924.528.png" ], "turn_id": 10, "scan_id": "Validation_47334362", "origin_question": "Assuming the direction you are facing at the end of the turn 3 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Left", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 3 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "Validation_47331644/Validation_47331644_47331644_22107.531.png", "Validation_47331644/Validation_47331644_47331644_22109.514.png", "Validation_47331644/Validation_47331644_47331644_22111.513.png", "Validation_47331644/Validation_47331644_47331644_22113.512.png", "Validation_47331644/Validation_47331644_47331644_22115.511.png" ], "turn_id": 1, "scan_id": "Validation_47331644", "origin_question": "Remember, how many stool(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many stool(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_47331644/Validation_47331644_47331644_22117.511.png", "Validation_47331644/Validation_47331644_47331644_22119.510.png", "Validation_47331644/Validation_47331644_47331644_22121.509.png", "Validation_47331644/Validation_47331644_47331644_22123.508.png", "Validation_47331644/Validation_47331644_47331644_22125.507.png" ], "turn_id": 2, "scan_id": "Validation_47331644", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the colorful, patterned fabric couch with plush cushions\", Y: \"the vintage, cream-colored lamp with the textured shade\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the colorful, patterned fabric couch with plush cushions\", Y: \"the vintage, cream-colored lamp with the textured shade\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47331644/Validation_47331644_47331644_22127.506.png", "Validation_47331644/Validation_47331644_47331644_22129.506.png", "Validation_47331644/Validation_47331644_47331644_22131.505.png", "Validation_47331644/Validation_47331644_47331644_22133.504.png", "Validation_47331644/Validation_47331644_47331644_22135.503.png" ], "turn_id": 3, "scan_id": "Validation_47331644", "origin_question": "Which direction is the colorful, patterned fabric couch with plush cushions in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the colorful, patterned fabric couch with plush cushions in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_47331644/Validation_47331644_47331644_22137.502.png", "Validation_47331644/Validation_47331644_47331644_22139.501.png", "Validation_47331644/Validation_47331644_47331644_22141.501.png", "Validation_47331644/Validation_47331644_47331644_22143.500.png", "Validation_47331644/Validation_47331644_47331644_22145.516.png" ], "turn_id": 4, "scan_id": "Validation_47331644", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the colorful, patterned fabric couch with plush cushions\", Y: \"the vintage, cream-colored lamp with the textured shade\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the colorful, patterned fabric couch with plush cushions\", Y: \"the vintage, cream-colored lamp with the textured shade\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_47331644/Validation_47331644_47331644_22147.515.png", "Validation_47331644/Validation_47331644_47331644_22149.514.png", "Validation_47331644/Validation_47331644_47331644_22151.513.png", "Validation_47331644/Validation_47331644_47331644_22153.513.png", "Validation_47331644/Validation_47331644_47331644_22155.512.png" ], "turn_id": 5, "scan_id": "Validation_47331644", "origin_question": "Is the cylindrical wireframe flowerpot in earthy color to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the cylindrical wireframe flowerpot in earthy color to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_47331644/Validation_47331644_47331644_22157.511.png", "Validation_47331644/Validation_47331644_47331644_22159.510.png", "Validation_47331644/Validation_47331644_47331644_22161.509.png", "Validation_47331644/Validation_47331644_47331644_22163.508.png", "Validation_47331644/Validation_47331644_47331644_22165.508.png" ], "turn_id": 6, "scan_id": "Validation_47331644", "origin_question": "Which two objects are on the same side of you now? \"the cylindrical wireframe flowerpot in earthy color\", \"the vintage, cream-colored lamp with the textured shade\", \"the light rectangular door with frosted glass panels\". ", "option": [ "the cylindrical wireframe flowerpot in earthy color, the vintage, cream-colored lamp with the textured shade", "the cylindrical wireframe flowerpot in earthy color, the light rectangular door with frosted glass panels", "the vintage, cream-colored lamp with the textured shade, the light rectangular door with frosted glass panels" ], "answer": "the vintage, cream-colored lamp with the textured shade, the light rectangular door with frosted glass panels", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the cylindrical wireframe flowerpot in earthy color\", \"the vintage, cream-colored lamp with the textured shade\", \"the light rectangular door with frosted glass panels\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the cylindrical wireframe flowerpot in earthy color, the vintage, cream-colored lamp with the textured shade', 'the cylindrical wireframe flowerpot in earthy color, the light rectangular door with frosted glass panels', 'the vintage, cream-colored lamp with the textured shade, the light rectangular door with frosted glass panels'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_47331644/Validation_47331644_47331644_22167.507.png", "Validation_47331644/Validation_47331644_47331644_22169.506.png", "Validation_47331644/Validation_47331644_47331644_22171.505.png", "Validation_47331644/Validation_47331644_47331644_22173.504.png", "Validation_47331644/Validation_47331644_47331644_22175.503.png" ], "turn_id": 7, "scan_id": "Validation_47331644", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the light rectangular door with frosted glass panels\" from your position?", "option": [], "answer": "55.25", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the light rectangular door with frosted glass panels\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47331644/Validation_47331644_47331644_22177.503.png", "Validation_47331644/Validation_47331644_47331644_22179.502.png", "Validation_47331644/Validation_47331644_47331644_22181.501.png", "Validation_47331644/Validation_47331644_47331644_22183.500.png", "Validation_47331644/Validation_47331644_47331644_22185.499.png" ], "turn_id": 8, "scan_id": "Validation_47331644", "origin_question": "Which direction is the light rectangular door with frosted glass panels in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the light rectangular door with frosted glass panels in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_47331644/Validation_47331644_47331644_22187.515.png", "Validation_47331644/Validation_47331644_47331644_22189.515.png", "Validation_47331644/Validation_47331644_47331644_22191.514.png", "Validation_47331644/Validation_47331644_47331644_22193.513.png", "Validation_47331644/Validation_47331644_47331644_22195.512.png" ], "turn_id": 9, "scan_id": "Validation_47331644", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the round, plush cushion with vibrant colors\" from your position?", "option": [], "answer": "102.5", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the round, plush cushion with vibrant colors\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_47331644/Validation_47331644_47331644_22197.511.png", "Validation_47331644/Validation_47331644_47331644_22199.510.png", "Validation_47331644/Validation_47331644_47331644_22201.510.png", "Validation_47331644/Validation_47331644_47331644_22203.509.png", "Validation_47331644/Validation_47331644_47331644_22205.508.png" ], "turn_id": 10, "scan_id": "Validation_47331644", "origin_question": "At the end of which turn were both of \"the sleek, white metal radiator with vertical fins\" and \"the light rectangular door with frosted glass panels\" on your right side?", "option": [], "answer": "10", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the sleek, white metal radiator with vertical fins\" and \"the light rectangular door with frosted glass panels\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47895745/Validation_47895745_47895745_385175.775.png", "Validation_47895745/Validation_47895745_47895745_385177.257.png", "Validation_47895745/Validation_47895745_47895745_385178.757.png", "Validation_47895745/Validation_47895745_47895745_385180.256.png" ], "turn_id": 1, "scan_id": "Validation_47895745", "origin_question": "", "option": [], "answer": "", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47895745/Validation_47895745_47895745_385181.756.png", "Validation_47895745/Validation_47895745_47895745_385183.255.png", "Validation_47895745/Validation_47895745_47895745_385184.754.png", "Validation_47895745/Validation_47895745_47895745_385186.254.png" ], "turn_id": 2, "scan_id": "Validation_47895745", "origin_question": "Remember, have you seen any pillow(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any pillow(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47895745/Validation_47895745_47895745_385187.753.png", "Validation_47895745/Validation_47895745_47895745_385189.252.png", "Validation_47895745/Validation_47895745_47895745_385190.752.png", "Validation_47895745/Validation_47895745_47895745_385192.251.png" ], "turn_id": 3, "scan_id": "Validation_47895745", "origin_question": "In which turn were you closest to \"the white rectangular steel radiator standing vertically\"? ", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the white rectangular steel radiator standing vertically\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_47895745/Validation_47895745_47895745_385193.751.png", "Validation_47895745/Validation_47895745_47895745_385195.250.png", "Validation_47895745/Validation_47895745_47895745_385196.749.png", "Validation_47895745/Validation_47895745_47895745_385198.249.png" ], "turn_id": 4, "scan_id": "Validation_47895745", "origin_question": "When did you first discover \"the cozy bed with the leafy green patterned bedspread\"? (index of the turn)", "option": [], "answer": "3", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the cozy bed with the leafy green patterned bedspread\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "Validation_47333934/Validation_47333934_47333934_85549.775.png", "Validation_47333934/Validation_47333934_47333934_85551.774.png", "Validation_47333934/Validation_47333934_47333934_85553.773.png", "Validation_47333934/Validation_47333934_47333934_85555.772.png", "Validation_47333934/Validation_47333934_47333934_85557.771.png" ], "turn_id": 1, "scan_id": "Validation_47333934", "origin_question": "Remember, how many can(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many can(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_47333934/Validation_47333934_47333934_85559.771.png", "Validation_47333934/Validation_47333934_47333934_85561.770.png", "Validation_47333934/Validation_47333934_47333934_85563.769.png", "Validation_47333934/Validation_47333934_47333934_85565.768.png", "Validation_47333934/Validation_47333934_47333934_85567.767.png" ], "turn_id": 2, "scan_id": "Validation_47333934", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the small white chair with the straight-backed structure\" from your position?", "option": [], "answer": "130.26", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the small white chair with the straight-backed structure\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47333934/Validation_47333934_47333934_85569.766.png", "Validation_47333934/Validation_47333934_47333934_85571.766.png", "Validation_47333934/Validation_47333934_47333934_85573.765.png", "Validation_47333934/Validation_47333934_47333934_85575.764.png", "Validation_47333934/Validation_47333934_47333934_85577.763.png" ], "turn_id": 3, "scan_id": "Validation_47333934", "origin_question": "Which direction is the modern white radiator with sturdy steel columns in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the modern white radiator with sturdy steel columns in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_47333934/Validation_47333934_47333934_85579.762.png", "Validation_47333934/Validation_47333934_47333934_85581.778.png", "Validation_47333934/Validation_47333934_47333934_85583.777.png", "Validation_47333934/Validation_47333934_47333934_85585.777.png", "Validation_47333934/Validation_47333934_47333934_85587.776.png" ], "turn_id": 4, "scan_id": "Validation_47333934", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the rectangular door with the white frame and frosted glass\" from your position?", "option": [], "answer": "173.56", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the rectangular door with the white frame and frosted glass\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47333934/Validation_47333934_47333934_85589.775.png", "Validation_47333934/Validation_47333934_47333934_85591.774.png", "Validation_47333934/Validation_47333934_47333934_85593.773.png", "Validation_47333934/Validation_47333934_47333934_85595.773.png", "Validation_47333934/Validation_47333934_47333934_85597.772.png" ], "turn_id": 5, "scan_id": "Validation_47333934", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the rectangular door with the white frame and frosted glass\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the rectangular door with the white frame and frosted glass\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_47333934/Validation_47333934_47333934_85599.771.png", "Validation_47333934/Validation_47333934_47333934_85601.770.png", "Validation_47333934/Validation_47333934_47333934_85603.769.png", "Validation_47333934/Validation_47333934_47333934_85605.769.png", "Validation_47333934/Validation_47333934_47333934_85607.768.png" ], "turn_id": 6, "scan_id": "Validation_47333934", "origin_question": "What will be the first-time appearance order of the following categories: couch, door, monitor?", "option": [ "monitor, couch, door", "couch, monitor, door", "monitor, door, couch", "couch, door, monitor" ], "answer": "couch, door, monitor", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: couch, door, monitor? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['monitor, couch, door', 'couch, monitor, door', 'monitor, door, couch', 'couch, door, monitor'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_47333934/Validation_47333934_47333934_85609.767.png", "Validation_47333934/Validation_47333934_47333934_85611.766.png", "Validation_47333934/Validation_47333934_47333934_85613.765.png", "Validation_47333934/Validation_47333934_47333934_85615.765.png", "Validation_47333934/Validation_47333934_47333934_85617.764.png" ], "turn_id": 7, "scan_id": "Validation_47333934", "origin_question": "Using your orientation at the end of turn 6 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "87.9", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 6 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47333934/Validation_47333934_47333934_85619.763.png", "Validation_47333934/Validation_47333934_47333934_85621.762.png", "Validation_47333934/Validation_47333934_47333934_85623.778.png", "Validation_47333934/Validation_47333934_47333934_85625.777.png", "Validation_47333934/Validation_47333934_47333934_85627.776.png" ], "turn_id": 8, "scan_id": "Validation_47333934", "origin_question": "Which direction is the rectangular door with the white frame and frosted glass in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the rectangular door with the white frame and frosted glass in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_47333934/Validation_47333934_47333934_85629.775.png", "Validation_47333934/Validation_47333934_47333934_85631.775.png", "Validation_47333934/Validation_47333934_47333934_85633.774.png", "Validation_47333934/Validation_47333934_47333934_85635.773.png", "Validation_47333934/Validation_47333934_47333934_85637.772.png" ], "turn_id": 9, "scan_id": "Validation_47333934", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "1.82", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47333934/Validation_47333934_47333934_85639.771.png", "Validation_47333934/Validation_47333934_47333934_85641.771.png", "Validation_47333934/Validation_47333934_47333934_85643.770.png", "Validation_47333934/Validation_47333934_47333934_85645.769.png", "Validation_47333934/Validation_47333934_47333934_85647.768.png" ], "turn_id": 10, "scan_id": "Validation_47333934", "origin_question": "Which direction is the rectangular door with the white frame and frosted glass in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the rectangular door with the white frame and frosted glass in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42897629/Validation_42897629_42897629_476908.514.png", "Validation_42897629/Validation_42897629_42897629_476910.496.png", "Validation_42897629/Validation_42897629_42897629_476912.495.png", "Validation_42897629/Validation_42897629_42897629_476914.494.png", "Validation_42897629/Validation_42897629_42897629_476916.494.png" ], "turn_id": 1, "scan_id": "Validation_42897629", "origin_question": "Remember, have you seen any table(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any table(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_42897629/Validation_42897629_42897629_476918.493.png", "Validation_42897629/Validation_42897629_42897629_476920.492.png", "Validation_42897629/Validation_42897629_42897629_476922.508.png", "Validation_42897629/Validation_42897629_42897629_476924.507.png", "Validation_42897629/Validation_42897629_42897629_476926.506.png" ], "turn_id": 2, "scan_id": "Validation_42897629", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_42897629/Validation_42897629_42897629_476928.505.png", "Validation_42897629/Validation_42897629_42897629_476930.504.png", "Validation_42897629/Validation_42897629_42897629_476932.504.png", "Validation_42897629/Validation_42897629_42897629_476934.503.png", "Validation_42897629/Validation_42897629_42897629_476936.502.png" ], "turn_id": 3, "scan_id": "Validation_42897629", "origin_question": "What will be the first-time appearance order of the following categories: chair, oven, plate?", "option": [ "oven, plate, chair", "chair, plate, oven", "oven, chair, plate", "plate, oven, chair" ], "answer": "chair, plate, oven", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: chair, oven, plate? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['oven, plate, chair', 'chair, plate, oven', 'oven, chair, plate', 'plate, oven, chair'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_42897629/Validation_42897629_42897629_476938.501.png", "Validation_42897629/Validation_42897629_42897629_476940.500.png", "Validation_42897629/Validation_42897629_42897629_476942.500.png", "Validation_42897629/Validation_42897629_42897629_476944.499.png", "Validation_42897629/Validation_42897629_42897629_476946.498.png" ], "turn_id": 4, "scan_id": "Validation_42897629", "origin_question": "What will be the first-time appearance order of the following categories: bin, plant, refrigerator?", "option": [ "plant, bin, refrigerator", "bin, plant, refrigerator", "plant, refrigerator, bin", "refrigerator, bin, plant" ], "answer": "plant, refrigerator, bin", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bin, plant, refrigerator? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['plant, bin, refrigerator', 'bin, plant, refrigerator', 'plant, refrigerator, bin', 'refrigerator, bin, plant'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_42897629/Validation_42897629_42897629_476948.497.png", "Validation_42897629/Validation_42897629_42897629_476950.496.png", "Validation_42897629/Validation_42897629_42897629_476952.495.png", "Validation_42897629/Validation_42897629_42897629_476954.495.png", "Validation_42897629/Validation_42897629_42897629_476956.494.png" ], "turn_id": 5, "scan_id": "Validation_42897629", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "2.91", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_42897629/Validation_42897629_42897629_476958.493.png", "Validation_42897629/Validation_42897629_42897629_476960.492.png", "Validation_42897629/Validation_42897629_42897629_476962.508.png", "Validation_42897629/Validation_42897629_42897629_476964.507.png", "Validation_42897629/Validation_42897629_42897629_476966.506.png" ], "turn_id": 6, "scan_id": "Validation_42897629", "origin_question": "In which turn did you see both \"the tall, stainless steel refrigerator with the sleek design\" and \"the sleek black oven with stainless steel trim\" simultaneously? ", "option": [], "answer": "3", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the tall, stainless steel refrigerator with the sleek design\" and \"the sleek black oven with stainless steel trim\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_42897629/Validation_42897629_42897629_476968.506.png", "Validation_42897629/Validation_42897629_42897629_476970.505.png", "Validation_42897629/Validation_42897629_42897629_476972.504.png", "Validation_42897629/Validation_42897629_42897629_476974.503.png", "Validation_42897629/Validation_42897629_42897629_476976.502.png" ], "turn_id": 7, "scan_id": "Validation_42897629", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "136.14", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_42897629/Validation_42897629_42897629_476978.501.png", "Validation_42897629/Validation_42897629_42897629_476980.501.png", "Validation_42897629/Validation_42897629_42897629_476982.500.png", "Validation_42897629/Validation_42897629_42897629_476984.499.png", "Validation_42897629/Validation_42897629_42897629_476986.498.png" ], "turn_id": 8, "scan_id": "Validation_42897629", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the tall, stainless steel refrigerator with the sleek design\" from your position?", "option": [], "answer": "177.4", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the tall, stainless steel refrigerator with the sleek design\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_42897629/Validation_42897629_42897629_476988.497.png", "Validation_42897629/Validation_42897629_42897629_476990.497.png", "Validation_42897629/Validation_42897629_42897629_476992.496.png", "Validation_42897629/Validation_42897629_42897629_476994.495.png", "Validation_42897629/Validation_42897629_42897629_476996.494.png" ], "turn_id": 9, "scan_id": "Validation_42897629", "origin_question": "Compared to your position at the end of the 6 turn, are you now closer or farther away from \"the silver cylindrical bin with the black lid\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 6 turn, are you now closer or farther away from \"the silver cylindrical bin with the black lid\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_42897629/Validation_42897629_42897629_476998.493.png", "Validation_42897629/Validation_42897629_42897629_477000.492.png", "Validation_42897629/Validation_42897629_42897629_477002.508.png", "Validation_42897629/Validation_42897629_42897629_477004.491.png", "Validation_42897629/Validation_42897629_42897629_477006.507.png" ], "turn_id": 10, "scan_id": "Validation_42897629", "origin_question": "Please recall \"the tall, stainless steel refrigerator with the sleek design\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "4.0", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the tall, stainless steel refrigerator with the sleek design\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47429977/Validation_47429977_47429977_16234.691.png", "Validation_47429977/Validation_47429977_47429977_16236.174.png", "Validation_47429977/Validation_47429977_47429977_16237.673.png", "Validation_47429977/Validation_47429977_47429977_16239.172.png" ], "turn_id": 1, "scan_id": "Validation_47429977", "origin_question": "Remember, have you seen any chair(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any chair(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47429977/Validation_47429977_47429977_16240.672.png", "Validation_47429977/Validation_47429977_47429977_16242.171.png", "Validation_47429977/Validation_47429977_47429977_16243.670.png", "Validation_47429977/Validation_47429977_47429977_16245.170.png" ], "turn_id": 2, "scan_id": "Validation_47429977", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the table\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the table\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_47429977/Validation_47429977_47429977_16246.669.png", "Validation_47429977/Validation_47429977_47429977_16248.168.png", "Validation_47429977/Validation_47429977_47429977_16249.668.png", "Validation_47429977/Validation_47429977_47429977_16251.167.png" ], "turn_id": 3, "scan_id": "Validation_47429977", "origin_question": "In which turn did you see both \"the window\" and \"the refrigerator\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the window\" and \"the refrigerator\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_47429977/Validation_47429977_47429977_16252.667.png", "Validation_47429977/Validation_47429977_47429977_16254.166.png", "Validation_47429977/Validation_47429977_47429977_16255.665.png", "Validation_47429977/Validation_47429977_47429977_16257.165.png" ], "turn_id": 4, "scan_id": "Validation_47429977", "origin_question": "Please recall \"the table\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.68", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the table\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47429977/Validation_47429977_47429977_16258.664.png", "Validation_47429977/Validation_47429977_47429977_16260.164.png", "Validation_47429977/Validation_47429977_47429977_16261.663.png", "Validation_47429977/Validation_47429977_47429977_16263.162.png" ], "turn_id": 5, "scan_id": "Validation_47429977", "origin_question": "In which turn were you closest to \"the sleek modern monitor with the dark smooth finish\"? ", "option": [], "answer": "4", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the sleek modern monitor with the dark smooth finish\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_47429977/Validation_47429977_47429977_16264.662.png", "Validation_47429977/Validation_47429977_47429977_16266.161.png", "Validation_47429977/Validation_47429977_47429977_16267.660.png", "Validation_47429977/Validation_47429977_47429977_16269.160.png" ], "turn_id": 6, "scan_id": "Validation_47429977", "origin_question": "When did you first discover \"the sleek cylindrical light with the dark metal finish\"? (index of the turn)", "option": [], "answer": "5", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the sleek cylindrical light with the dark metal finish\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47429977/Validation_47429977_47429977_16270.659.png", "Validation_47429977/Validation_47429977_47429977_16272.175.png", "Validation_47429977/Validation_47429977_47429977_16273.675.png", "Validation_47429977/Validation_47429977_47429977_16275.174.png" ], "turn_id": 7, "scan_id": "Validation_47429977", "origin_question": "Which direction is the refrigerator in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the refrigerator in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_45662981/Validation_45662981_45662981_27937.116.png", "Validation_45662981/Validation_45662981_45662981_27939.099.png", "Validation_45662981/Validation_45662981_45662981_27941.098.png", "Validation_45662981/Validation_45662981_45662981_27943.097.png", "Validation_45662981/Validation_45662981_45662981_27945.096.png" ], "turn_id": 1, "scan_id": "Validation_45662981", "origin_question": "Remember, have you seen any window(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any window(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "Validation_45662981/Validation_45662981_45662981_27947.096.png", "Validation_45662981/Validation_45662981_45662981_27949.095.png", "Validation_45662981/Validation_45662981_45662981_27951.094.png", "Validation_45662981/Validation_45662981_45662981_27953.093.png", "Validation_45662981/Validation_45662981_45662981_27955.092.png" ], "turn_id": 2, "scan_id": "Validation_45662981", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the large, white-framed window with clear glass\"; \"the large, rectangular black monitor with the flat screen\"; \"the white rectangular cabinet with closed doors\" ", "option": [ "the large, white-framed window with clear glass", "the large, rectangular black monitor with the flat screen", "the white rectangular cabinet with closed doors" ], "answer": "the large, rectangular black monitor with the flat screen", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the large, white-framed window with clear glass\"; \"the large, rectangular black monitor with the flat screen\"; \"the white rectangular cabinet with closed doors\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the large, white-framed window with clear glass', 'the large, rectangular black monitor with the flat screen', 'the white rectangular cabinet with closed doors'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_45662981/Validation_45662981_45662981_27957.092.png", "Validation_45662981/Validation_45662981_45662981_27959.091.png", "Validation_45662981/Validation_45662981_45662981_27961.090.png", "Validation_45662981/Validation_45662981_45662981_27963.089.png", "Validation_45662981/Validation_45662981_45662981_27965.089.png" ], "turn_id": 3, "scan_id": "Validation_45662981", "origin_question": "", "option": [], "answer": "", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_45662981/Validation_45662981_45662981_27967.088.png", "Validation_45662981/Validation_45662981_45662981_27969.087.png", "Validation_45662981/Validation_45662981_45662981_27971.086.png", "Validation_45662981/Validation_45662981_45662981_27973.085.png", "Validation_45662981/Validation_45662981_45662981_27975.085.png" ], "turn_id": 4, "scan_id": "Validation_45662981", "origin_question": "What will be the first-time appearance order of the following categories: chair, couch, monitor?", "option": [ "monitor, chair, couch", "couch, monitor, chair", "couch, chair, monitor", "chair, monitor, couch" ], "answer": "couch, chair, monitor", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: chair, couch, monitor? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['monitor, chair, couch', 'couch, monitor, chair', 'couch, chair, monitor', 'chair, monitor, couch'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_45662981/Validation_45662981_45662981_27977.084.png", "Validation_45662981/Validation_45662981_45662981_27979.083.png", "Validation_45662981/Validation_45662981_45662981_27981.099.png", "Validation_45662981/Validation_45662981_45662981_27983.098.png", "Validation_45662981/Validation_45662981_45662981_27985.097.png" ], "turn_id": 5, "scan_id": "Validation_45662981", "origin_question": "Please recall \"the large, white-framed window with clear glass\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "4.41", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the large, white-framed window with clear glass\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_45662981/Validation_45662981_45662981_27987.096.png", "Validation_45662981/Validation_45662981_45662981_27989.096.png", "Validation_45662981/Validation_45662981_45662981_27991.095.png", "Validation_45662981/Validation_45662981_45662981_27993.094.png", "Validation_45662981/Validation_45662981_45662981_27995.093.png" ], "turn_id": 6, "scan_id": "Validation_45662981", "origin_question": "Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "30.04", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_45662981/Validation_45662981_45662981_27997.092.png", "Validation_45662981/Validation_45662981_45662981_27999.092.png", "Validation_45662981/Validation_45662981_45662981_28001.091.png", "Validation_45662981/Validation_45662981_45662981_28003.090.png", "Validation_45662981/Validation_45662981_45662981_28005.089.png" ], "turn_id": 7, "scan_id": "Validation_45662981", "origin_question": "Which direction is the large, rectangular black monitor with the flat screen in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the large, rectangular black monitor with the flat screen in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_45662981/Validation_45662981_45662981_28007.089.png", "Validation_45662981/Validation_45662981_45662981_28009.088.png", "Validation_45662981/Validation_45662981_45662981_28011.087.png", "Validation_45662981/Validation_45662981_45662981_28013.086.png", "Validation_45662981/Validation_45662981_45662981_28015.085.png" ], "turn_id": 8, "scan_id": "Validation_45662981", "origin_question": "At the end of which turn were both of \"the large, white-framed window with clear glass\" and \"the matte black metal and glass fireplace\" on your left side?", "option": [], "answer": "7", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the large, white-framed window with clear glass\" and \"the matte black metal and glass fireplace\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_45662981/Validation_45662981_45662981_28017.085.png", "Validation_45662981/Validation_45662981_45662981_28019.084.png", "Validation_45662981/Validation_45662981_45662981_28021.083.png", "Validation_45662981/Validation_45662981_45662981_28023.099.png", "Validation_45662981/Validation_45662981_45662981_28025.098.png" ], "turn_id": 9, "scan_id": "Validation_45662981", "origin_question": "Which two objects are on the same side of you now? \"the large, white-framed window with clear glass\", \"the matte black metal and glass fireplace\", \"the large, rectangular black monitor with the flat screen\". ", "option": [ "the large, white-framed window with clear glass, the matte black metal and glass fireplace", "the large, white-framed window with clear glass, the large, rectangular black monitor with the flat screen", "the matte black metal and glass fireplace, the large, rectangular black monitor with the flat screen" ], "answer": "the matte black metal and glass fireplace, the large, rectangular black monitor with the flat screen", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the large, white-framed window with clear glass\", \"the matte black metal and glass fireplace\", \"the large, rectangular black monitor with the flat screen\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the large, white-framed window with clear glass, the matte black metal and glass fireplace', 'the large, white-framed window with clear glass, the large, rectangular black monitor with the flat screen', 'the matte black metal and glass fireplace, the large, rectangular black monitor with the flat screen'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_45662981/Validation_45662981_45662981_28027.097.png", "Validation_45662981/Validation_45662981_45662981_28029.096.png", "Validation_45662981/Validation_45662981_45662981_28031.096.png", "Validation_45662981/Validation_45662981_45662981_28033.095.png", "Validation_45662981/Validation_45662981_45662981_28035.094.png" ], "turn_id": 10, "scan_id": "Validation_45662981", "origin_question": "When did you first discover \"the large, white-framed window with clear glass\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the large, white-framed window with clear glass\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47331061/Validation_47331061_47331061_129520.164.png", "Validation_47331061/Validation_47331061_47331061_129522.147.png", "Validation_47331061/Validation_47331061_47331061_129524.146.png", "Validation_47331061/Validation_47331061_47331061_129526.145.png", "Validation_47331061/Validation_47331061_47331061_129528.145.png" ], "turn_id": 1, "scan_id": "Validation_47331061", "origin_question": "", "option": [], "answer": "", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_47331061/Validation_47331061_47331061_129530.144.png", "Validation_47331061/Validation_47331061_47331061_129532.143.png", "Validation_47331061/Validation_47331061_47331061_129534.142.png", "Validation_47331061/Validation_47331061_47331061_129536.141.png", "Validation_47331061/Validation_47331061_47331061_129538.157.png" ], "turn_id": 2, "scan_id": "Validation_47331061", "origin_question": "In which turn did you see both \"the compact, white glossy refrigerator with the rectangular shape\" and \"the modern black electric stove with the flat design\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the compact, white glossy refrigerator with the rectangular shape\" and \"the modern black electric stove with the flat design\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47331061/Validation_47331061_47331061_129540.156.png", "Validation_47331061/Validation_47331061_47331061_129542.156.png", "Validation_47331061/Validation_47331061_47331061_129544.155.png", "Validation_47331061/Validation_47331061_47331061_129546.154.png", "Validation_47331061/Validation_47331061_47331061_129548.153.png" ], "turn_id": 3, "scan_id": "Validation_47331061", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the compact, white glossy refrigerator with the rectangular shape\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the compact, white glossy refrigerator with the rectangular shape\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47331061/Validation_47331061_47331061_129550.152.png", "Validation_47331061/Validation_47331061_47331061_129552.152.png", "Validation_47331061/Validation_47331061_47331061_129554.151.png", "Validation_47331061/Validation_47331061_47331061_129556.150.png", "Validation_47331061/Validation_47331061_47331061_129558.149.png" ], "turn_id": 4, "scan_id": "Validation_47331061", "origin_question": "", "option": [], "answer": "", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_47331061/Validation_47331061_47331061_129560.148.png", "Validation_47331061/Validation_47331061_47331061_129562.148.png", "Validation_47331061/Validation_47331061_47331061_129564.147.png", "Validation_47331061/Validation_47331061_47331061_129566.146.png", "Validation_47331061/Validation_47331061_47331061_129568.145.png" ], "turn_id": 5, "scan_id": "Validation_47331061", "origin_question": "Please recall \"the modern stainless steel oven with the glass door\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.86", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the modern stainless steel oven with the glass door\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47331061/Validation_47331061_47331061_129570.144.png", "Validation_47331061/Validation_47331061_47331061_129572.143.png", "Validation_47331061/Validation_47331061_47331061_129574.143.png", "Validation_47331061/Validation_47331061_47331061_129576.142.png", "Validation_47331061/Validation_47331061_47331061_129578.141.png" ], "turn_id": 6, "scan_id": "Validation_47331061", "origin_question": "Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the sleek white ceramic sink with the rectangular basin\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the sleek white ceramic sink with the rectangular basin\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_47331061/Validation_47331061_47331061_129580.157.png", "Validation_47331061/Validation_47331061_47331061_129582.156.png", "Validation_47331061/Validation_47331061_47331061_129584.155.png", "Validation_47331061/Validation_47331061_47331061_129586.154.png", "Validation_47331061/Validation_47331061_47331061_129588.154.png" ], "turn_id": 7, "scan_id": "Validation_47331061", "origin_question": "Is the modern mop with the metallic handle and blue head to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the modern mop with the metallic handle and blue head to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47331061/Validation_47331061_47331061_129590.153.png", "Validation_47331061/Validation_47331061_47331061_129592.152.png", "Validation_47331061/Validation_47331061_47331061_129594.151.png", "Validation_47331061/Validation_47331061_47331061_129596.150.png", "Validation_47331061/Validation_47331061_47331061_129598.150.png" ], "turn_id": 8, "scan_id": "Validation_47331061", "origin_question": "Remember, have you seen any microwave(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any microwave(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47331061/Validation_47331061_47331061_129600.149.png", "Validation_47331061/Validation_47331061_47331061_129602.148.png", "Validation_47331061/Validation_47331061_47331061_129604.147.png", "Validation_47331061/Validation_47331061_47331061_129606.146.png", "Validation_47331061/Validation_47331061_47331061_129608.145.png" ], "turn_id": 9, "scan_id": "Validation_47331061", "origin_question": "", "option": [], "answer": "", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_47331061/Validation_47331061_47331061_129610.145.png", "Validation_47331061/Validation_47331061_47331061_129612.144.png", "Validation_47331061/Validation_47331061_47331061_129614.143.png", "Validation_47331061/Validation_47331061_47331061_129616.142.png", "Validation_47331061/Validation_47331061_47331061_129618.141.png" ], "turn_id": 10, "scan_id": "Validation_47331061", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_48458735/Validation_48458735_48458735_139170.110.png", "Validation_48458735/Validation_48458735_48458735_139172.093.png", "Validation_48458735/Validation_48458735_48458735_139174.092.png", "Validation_48458735/Validation_48458735_48458735_139176.091.png", "Validation_48458735/Validation_48458735_48458735_139178.091.png" ], "turn_id": 1, "scan_id": "Validation_48458735", "origin_question": "Remember, have you seen any door(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any door(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_48458735/Validation_48458735_48458735_139180.090.png", "Validation_48458735/Validation_48458735_48458735_139182.089.png", "Validation_48458735/Validation_48458735_48458735_139184.088.png", "Validation_48458735/Validation_48458735_48458735_139186.087.png", "Validation_48458735/Validation_48458735_48458735_139188.087.png" ], "turn_id": 2, "scan_id": "Validation_48458735", "origin_question": "In which turn did you see both \"the rectangular white-framed window with sliding panes\" and \"the white compact microwave with the black front window\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the rectangular white-framed window with sliding panes\" and \"the white compact microwave with the black front window\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_48458735/Validation_48458735_48458735_139190.086.png", "Validation_48458735/Validation_48458735_48458735_139192.085.png", "Validation_48458735/Validation_48458735_48458735_139194.084.png", "Validation_48458735/Validation_48458735_48458735_139196.084.png", "Validation_48458735/Validation_48458735_48458735_139198.099.png" ], "turn_id": 3, "scan_id": "Validation_48458735", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the white compact microwave with the black front window\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the white compact microwave with the black front window\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_48458735/Validation_48458735_48458735_139200.099.png", "Validation_48458735/Validation_48458735_48458735_139202.098.png", "Validation_48458735/Validation_48458735_48458735_139204.097.png", "Validation_48458735/Validation_48458735_48458735_139206.096.png", "Validation_48458735/Validation_48458735_48458735_139208.095.png" ], "turn_id": 4, "scan_id": "Validation_48458735", "origin_question": "Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_48458735/Validation_48458735_48458735_139210.095.png", "Validation_48458735/Validation_48458735_48458735_139212.094.png", "Validation_48458735/Validation_48458735_48458735_139214.093.png", "Validation_48458735/Validation_48458735_48458735_139216.092.png", "Validation_48458735/Validation_48458735_48458735_139218.091.png" ], "turn_id": 5, "scan_id": "Validation_48458735", "origin_question": "Which two objects are on the same side of you now? \"the white paneled door made of wood composite\", \"the stainless steel oven with the sleek design\", \"the sleek metallic kitchen case with minimalist handles\". ", "option": [ "the white paneled door made of wood composite, the stainless steel oven with the sleek design", "the white paneled door made of wood composite, the sleek metallic kitchen case with minimalist handles", "the stainless steel oven with the sleek design, the sleek metallic kitchen case with minimalist handles" ], "answer": "the stainless steel oven with the sleek design, the sleek metallic kitchen case with minimalist handles", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the white paneled door made of wood composite\", \"the stainless steel oven with the sleek design\", \"the sleek metallic kitchen case with minimalist handles\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the white paneled door made of wood composite, the stainless steel oven with the sleek design', 'the white paneled door made of wood composite, the sleek metallic kitchen case with minimalist handles', 'the stainless steel oven with the sleek design, the sleek metallic kitchen case with minimalist handles'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_48458735/Validation_48458735_48458735_139220.091.png", "Validation_48458735/Validation_48458735_48458735_139222.090.png", "Validation_48458735/Validation_48458735_48458735_139224.089.png", "Validation_48458735/Validation_48458735_48458735_139226.088.png", "Validation_48458735/Validation_48458735_48458735_139228.087.png" ], "turn_id": 6, "scan_id": "Validation_48458735", "origin_question": "In which turn were you farthest from \"the stainless steel oven with the sleek design\"? ", "option": [], "answer": "1", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the stainless steel oven with the sleek design\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_48458735/Validation_48458735_48458735_139230.087.png", "Validation_48458735/Validation_48458735_48458735_139232.086.png", "Validation_48458735/Validation_48458735_48458735_139234.085.png", "Validation_48458735/Validation_48458735_48458735_139236.084.png", "Validation_48458735/Validation_48458735_48458735_139238.084.png" ], "turn_id": 7, "scan_id": "Validation_48458735", "origin_question": "Is the range hood to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the range hood to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_48458735/Validation_48458735_48458735_139240.099.png", "Validation_48458735/Validation_48458735_48458735_139242.099.png", "Validation_48458735/Validation_48458735_48458735_139244.098.png", "Validation_48458735/Validation_48458735_48458735_139246.097.png", "Validation_48458735/Validation_48458735_48458735_139248.096.png" ], "turn_id": 8, "scan_id": "Validation_48458735", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the rectangular white-framed window with sliding panes\" from your position?", "option": [], "answer": "31.03", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the rectangular white-framed window with sliding panes\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_48458735/Validation_48458735_48458735_139250.095.png", "Validation_48458735/Validation_48458735_48458735_139252.095.png", "Validation_48458735/Validation_48458735_48458735_139254.094.png", "Validation_48458735/Validation_48458735_48458735_139256.093.png", "Validation_48458735/Validation_48458735_48458735_139258.092.png" ], "turn_id": 9, "scan_id": "Validation_48458735", "origin_question": "Assuming the direction you are facing at the end of the turn 6 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Right", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 6 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_48018382/Validation_48018382_48018382_390464.155.png", "Validation_48018382/Validation_48018382_48018382_390465.638.png", "Validation_48018382/Validation_48018382_48018382_390467.137.png", "Validation_48018382/Validation_48018382_48018382_390468.636.png" ], "turn_id": 1, "scan_id": "Validation_48018382", "origin_question": "Remember, have you seen any handle(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any handle(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_48018382/Validation_48018382_48018382_390470.136.png", "Validation_48018382/Validation_48018382_48018382_390471.635.png", "Validation_48018382/Validation_48018382_48018382_390473.134.png", "Validation_48018382/Validation_48018382_48018382_390474.634.png" ], "turn_id": 2, "scan_id": "Validation_48018382", "origin_question": "When did you first discover \"the compact, oval-shaped toilet with the glossy white finish\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the compact, oval-shaped toilet with the glossy white finish\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_48018382/Validation_48018382_48018382_390476.133.png", "Validation_48018382/Validation_48018382_48018382_390477.633.png", "Validation_48018382/Validation_48018382_48018382_390479.149.png", "Validation_48018382/Validation_48018382_48018382_390480.648.png" ], "turn_id": 3, "scan_id": "Validation_48018382", "origin_question": "Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "42.97", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_48018382/Validation_48018382_48018382_390482.147.png", "Validation_48018382/Validation_48018382_48018382_390483.647.png", "Validation_48018382/Validation_48018382_48018382_390485.146.png", "Validation_48018382/Validation_48018382_48018382_390486.645.png" ], "turn_id": 4, "scan_id": "Validation_48018382", "origin_question": "What will be the first-time appearance order of the following categories: body loofah, cabinet, sink?", "option": [ "sink, body loofah, cabinet", "cabinet, sink, body loofah", "body loofah, sink, cabinet", "sink, cabinet, body loofah" ], "answer": "cabinet, sink, body loofah", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: body loofah, cabinet, sink? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['sink, body loofah, cabinet', 'cabinet, sink, body loofah', 'body loofah, sink, cabinet', 'sink, cabinet, body loofah'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_48018382/Validation_48018382_48018382_390488.145.png", "Validation_48018382/Validation_48018382_48018382_390489.644.png", "Validation_48018382/Validation_48018382_48018382_390491.144.png", "Validation_48018382/Validation_48018382_48018382_390492.643.png" ], "turn_id": 5, "scan_id": "Validation_48018382", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_48018382/Validation_48018382_48018382_390494.142.png", "Validation_48018382/Validation_48018382_48018382_390495.642.png", "Validation_48018382/Validation_48018382_48018382_390497.141.png", "Validation_48018382/Validation_48018382_48018382_390498.640.png" ], "turn_id": 6, "scan_id": "Validation_48018382", "origin_question": "In which turn did you see both \"the compact, dark brown minimalist stool\" and \"the sleek, modern, glossy white sink\" simultaneously? ", "option": [], "answer": "3", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the compact, dark brown minimalist stool\" and \"the sleek, modern, glossy white sink\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_48018382/Validation_48018382_48018382_390500.140.png", "Validation_48018382/Validation_48018382_48018382_390501.639.png", "Validation_48018382/Validation_48018382_48018382_390503.139.png", "Validation_48018382/Validation_48018382_48018382_390504.638.png" ], "turn_id": 7, "scan_id": "Validation_48018382", "origin_question": "In which turn were you closest to \"the compact, oval-shaped toilet with the glossy white finish\"? ", "option": [], "answer": "7", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the compact, oval-shaped toilet with the glossy white finish\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_44358532/Validation_44358532_44358532_53586.481.png", "Validation_44358532/Validation_44358532_44358532_53588.464.png", "Validation_44358532/Validation_44358532_44358532_53590.479.png", "Validation_44358532/Validation_44358532_44358532_53592.479.png", "Validation_44358532/Validation_44358532_44358532_53594.478.png" ], "turn_id": 1, "scan_id": "Validation_44358532", "origin_question": "Remember, have you seen any sink(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any sink(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_44358532/Validation_44358532_44358532_53596.477.png", "Validation_44358532/Validation_44358532_44358532_53598.476.png", "Validation_44358532/Validation_44358532_44358532_53600.475.png", "Validation_44358532/Validation_44358532_44358532_53602.474.png", "Validation_44358532/Validation_44358532_44358532_53604.474.png" ], "turn_id": 2, "scan_id": "Validation_44358532", "origin_question": "When did you first discover \"the natural light brown basket\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the natural light brown basket\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_44358532/Validation_44358532_44358532_53606.473.png", "Validation_44358532/Validation_44358532_44358532_53608.472.png", "Validation_44358532/Validation_44358532_44358532_53610.471.png", "Validation_44358532/Validation_44358532_44358532_53612.470.png", "Validation_44358532/Validation_44358532_44358532_53614.469.png" ], "turn_id": 3, "scan_id": "Validation_44358532", "origin_question": "Please recall \"the compact, cylindrical white bin\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.77", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the compact, cylindrical white bin\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_44358532/Validation_44358532_44358532_53616.469.png", "Validation_44358532/Validation_44358532_44358532_53618.468.png", "Validation_44358532/Validation_44358532_44358532_53620.467.png", "Validation_44358532/Validation_44358532_44358532_53622.466.png", "Validation_44358532/Validation_44358532_44358532_53624.465.png" ], "turn_id": 4, "scan_id": "Validation_44358532", "origin_question": "Which two objects are on the same side of you now? \"the sleek, triangular white acrylic bathtub\", \"the natural light brown basket\", \"the modern rectangular dark wood cabinet\". ", "option": [ "the sleek, triangular white acrylic bathtub, the natural light brown basket", "the sleek, triangular white acrylic bathtub, the modern rectangular dark wood cabinet", "the natural light brown basket, the modern rectangular dark wood cabinet" ], "answer": "the natural light brown basket, the modern rectangular dark wood cabinet", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the sleek, triangular white acrylic bathtub\", \"the natural light brown basket\", \"the modern rectangular dark wood cabinet\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the sleek, triangular white acrylic bathtub, the natural light brown basket', 'the sleek, triangular white acrylic bathtub, the modern rectangular dark wood cabinet', 'the natural light brown basket, the modern rectangular dark wood cabinet'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_44358532/Validation_44358532_44358532_53626.465.png", "Validation_44358532/Validation_44358532_44358532_53628.464.png", "Validation_44358532/Validation_44358532_44358532_53630.480.png", "Validation_44358532/Validation_44358532_44358532_53632.479.png", "Validation_44358532/Validation_44358532_44358532_53634.478.png" ], "turn_id": 5, "scan_id": "Validation_44358532", "origin_question": "", "option": [], "answer": "", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_44358532/Validation_44358532_44358532_53636.477.png", "Validation_44358532/Validation_44358532_44358532_53638.476.png", "Validation_44358532/Validation_44358532_44358532_53640.475.png", "Validation_44358532/Validation_44358532_44358532_53642.475.png", "Validation_44358532/Validation_44358532_44358532_53644.474.png" ], "turn_id": 6, "scan_id": "Validation_44358532", "origin_question": "Please recall \"the natural light brown basket\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.53", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the natural light brown basket\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_44358532/Validation_44358532_44358532_53646.473.png", "Validation_44358532/Validation_44358532_44358532_53648.472.png", "Validation_44358532/Validation_44358532_44358532_53650.471.png", "Validation_44358532/Validation_44358532_44358532_53652.470.png", "Validation_44358532/Validation_44358532_44358532_53654.470.png" ], "turn_id": 7, "scan_id": "Validation_44358532", "origin_question": "Compared to your position at the end of the 6 turn, are you now closer or farther away from \"the polished white oval porcelain sink\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 6 turn, are you now closer or farther away from \"the polished white oval porcelain sink\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_44358532/Validation_44358532_44358532_53656.469.png", "Validation_44358532/Validation_44358532_44358532_53658.468.png", "Validation_44358532/Validation_44358532_44358532_53660.467.png", "Validation_44358532/Validation_44358532_44358532_53662.466.png", "Validation_44358532/Validation_44358532_44358532_53664.465.png" ], "turn_id": 8, "scan_id": "Validation_44358532", "origin_question": "Please recall \"the sleek, triangular white acrylic bathtub\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.9", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the sleek, triangular white acrylic bathtub\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_44358532/Validation_44358532_44358532_53666.465.png", "Validation_44358532/Validation_44358532_44358532_53668.464.png", "Validation_44358532/Validation_44358532_44358532_53670.463.png", "Validation_44358532/Validation_44358532_44358532_53672.462.png", "Validation_44358532/Validation_44358532_44358532_53674.478.png" ], "turn_id": 9, "scan_id": "Validation_44358532", "origin_question": "Compared to your position at the end of the 6 turn, are you now closer or farther away from \"the modern rectangular dark wood cabinet\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 6 turn, are you now closer or farther away from \"the modern rectangular dark wood cabinet\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_45662970/Validation_45662970_45662970_26107.552.png", "Validation_45662970/Validation_45662970_45662970_26109.534.png", "Validation_45662970/Validation_45662970_45662970_26111.533.png", "Validation_45662970/Validation_45662970_45662970_26113.532.png" ], "turn_id": 1, "scan_id": "Validation_45662970", "origin_question": "Remember, have you seen any washing machine(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any washing machine(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_45662970/Validation_45662970_45662970_26115.532.png", "Validation_45662970/Validation_45662970_45662970_26117.531.png", "Validation_45662970/Validation_45662970_45662970_26119.530.png", "Validation_45662970/Validation_45662970_45662970_26121.529.png" ], "turn_id": 2, "scan_id": "Validation_45662970", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the light brown rectangular wicker basket\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the light brown rectangular wicker basket\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_45662970/Validation_45662970_45662970_26123.529.png", "Validation_45662970/Validation_45662970_45662970_26125.528.png", "Validation_45662970/Validation_45662970_45662970_26127.527.png", "Validation_45662970/Validation_45662970_45662970_26129.526.png" ], "turn_id": 3, "scan_id": "Validation_45662970", "origin_question": "When did you first discover \"the light brown rectangular wicker basket\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the light brown rectangular wicker basket\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_45662970/Validation_45662970_45662970_26131.525.png", "Validation_45662970/Validation_45662970_45662970_26133.524.png", "Validation_45662970/Validation_45662970_45662970_26135.524.png", "Validation_45662970/Validation_45662970_45662970_26137.540.png" ], "turn_id": 4, "scan_id": "Validation_45662970", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the glossy white rectangular sink made of ceramic\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the glossy white rectangular sink made of ceramic\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_45662970/Validation_45662970_45662970_26139.539.png", "Validation_45662970/Validation_45662970_45662970_26141.538.png", "Validation_45662970/Validation_45662970_45662970_26143.537.png", "Validation_45662970/Validation_45662970_45662970_26145.536.png" ], "turn_id": 5, "scan_id": "Validation_45662970", "origin_question": "Which direction is the glossy white rectangular sink made of ceramic in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the glossy white rectangular sink made of ceramic in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_45662970/Validation_45662970_45662970_26147.536.png", "Validation_45662970/Validation_45662970_45662970_26149.535.png", "Validation_45662970/Validation_45662970_45662970_26151.534.png", "Validation_45662970/Validation_45662970_45662970_26153.533.png" ], "turn_id": 6, "scan_id": "Validation_45662970", "origin_question": "In which turn did you see both \"the glossy white rectangular sink made of ceramic\" and \"the modern door with the frosted glass panel\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the glossy white rectangular sink made of ceramic\" and \"the modern door with the frosted glass panel\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_45662970/Validation_45662970_45662970_26155.533.png", "Validation_45662970/Validation_45662970_45662970_26157.532.png", "Validation_45662970/Validation_45662970_45662970_26159.531.png", "Validation_45662970/Validation_45662970_45662970_26161.530.png" ], "turn_id": 7, "scan_id": "Validation_45662970", "origin_question": "At the end of which turn were both of \"the modern door with the frosted glass panel\" and \"the window\" on your right side?", "option": [], "answer": "4", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the modern door with the frosted glass panel\" and \"the window\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_45662970/Validation_45662970_45662970_26163.529.png", "Validation_45662970/Validation_45662970_45662970_26165.529.png", "Validation_45662970/Validation_45662970_45662970_26167.528.png", "Validation_45662970/Validation_45662970_45662970_26169.527.png" ], "turn_id": 8, "scan_id": "Validation_45662970", "origin_question": "Which two objects are on the same side of you now? \"the glossy white rectangular sink made of ceramic\", \"the light brown rectangular wicker basket\", \"the modern door with the frosted glass panel\". ", "option": [ "the glossy white rectangular sink made of ceramic, the light brown rectangular wicker basket", "the glossy white rectangular sink made of ceramic, the modern door with the frosted glass panel", "the light brown rectangular wicker basket, the modern door with the frosted glass panel" ], "answer": "the glossy white rectangular sink made of ceramic, the light brown rectangular wicker basket", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the glossy white rectangular sink made of ceramic\", \"the light brown rectangular wicker basket\", \"the modern door with the frosted glass panel\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the glossy white rectangular sink made of ceramic, the light brown rectangular wicker basket', 'the glossy white rectangular sink made of ceramic, the modern door with the frosted glass panel', 'the light brown rectangular wicker basket, the modern door with the frosted glass panel'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42899685/Validation_42899685_42899685_790482.237.png", "Validation_42899685/Validation_42899685_42899685_790484.219.png", "Validation_42899685/Validation_42899685_42899685_790486.218.png", "Validation_42899685/Validation_42899685_42899685_790488.218.png", "Validation_42899685/Validation_42899685_42899685_790490.217.png" ], "turn_id": 1, "scan_id": "Validation_42899685", "origin_question": "Remember, have you seen any cabinet(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any cabinet(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_42899685/Validation_42899685_42899685_790492.216.png", "Validation_42899685/Validation_42899685_42899685_790494.215.png", "Validation_42899685/Validation_42899685_42899685_790496.214.png", "Validation_42899685/Validation_42899685_42899685_790498.213.png", "Validation_42899685/Validation_42899685_42899685_790500.213.png" ], "turn_id": 2, "scan_id": "Validation_42899685", "origin_question": "At the end of which turn were both of \"the sleek white dishwasher made of durable steel\" and \"the deep blue, cylindrical upholstered stool\" on your right side?", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the sleek white dishwasher made of durable steel\" and \"the deep blue, cylindrical upholstered stool\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_42899685/Validation_42899685_42899685_790502.212.png", "Validation_42899685/Validation_42899685_42899685_790504.211.png", "Validation_42899685/Validation_42899685_42899685_790506.210.png", "Validation_42899685/Validation_42899685_42899685_790508.209.png", "Validation_42899685/Validation_42899685_42899685_790510.208.png" ], "turn_id": 3, "scan_id": "Validation_42899685", "origin_question": "How far is your current position from where you were at the end of the turn 1? (in meters)", "option": [], "answer": "2.21", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 1? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_42899685/Validation_42899685_42899685_790512.208.png", "Validation_42899685/Validation_42899685_42899685_790514.207.png", "Validation_42899685/Validation_42899685_42899685_790516.206.png", "Validation_42899685/Validation_42899685_42899685_790518.222.png", "Validation_42899685/Validation_42899685_42899685_790520.204.png" ], "turn_id": 4, "scan_id": "Validation_42899685", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the stainless steel rectangular oven with the black front panel\", Y: \"the sleek white dishwasher made of durable steel\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "farther away from both X and Y", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the stainless steel rectangular oven with the black front panel\", Y: \"the sleek white dishwasher made of durable steel\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_42899685/Validation_42899685_42899685_790522.220.png", "Validation_42899685/Validation_42899685_42899685_790524.219.png", "Validation_42899685/Validation_42899685_42899685_790526.218.png", "Validation_42899685/Validation_42899685_42899685_790528.218.png", "Validation_42899685/Validation_42899685_42899685_790530.217.png" ], "turn_id": 5, "scan_id": "Validation_42899685", "origin_question": "In which turn did you see both \"the compact, white front-loading washing machine\" and \"the deep blue, cylindrical upholstered stool\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the compact, white front-loading washing machine\" and \"the deep blue, cylindrical upholstered stool\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_42899685/Validation_42899685_42899685_790532.216.png", "Validation_42899685/Validation_42899685_42899685_790534.215.png", "Validation_42899685/Validation_42899685_42899685_790536.214.png", "Validation_42899685/Validation_42899685_42899685_790538.213.png", "Validation_42899685/Validation_42899685_42899685_790540.213.png" ], "turn_id": 6, "scan_id": "Validation_42899685", "origin_question": "In which turn were you farthest from \"the compact, white front-loading washing machine\"? ", "option": [], "answer": "6", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the compact, white front-loading washing machine\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_42899685/Validation_42899685_42899685_790542.212.png", "Validation_42899685/Validation_42899685_42899685_790544.211.png", "Validation_42899685/Validation_42899685_42899685_790546.210.png", "Validation_42899685/Validation_42899685_42899685_790548.209.png", "Validation_42899685/Validation_42899685_42899685_790550.208.png" ], "turn_id": 7, "scan_id": "Validation_42899685", "origin_question": "Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "79.07", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_42899685/Validation_42899685_42899685_790552.208.png", "Validation_42899685/Validation_42899685_42899685_790554.207.png", "Validation_42899685/Validation_42899685_42899685_790556.206.png", "Validation_42899685/Validation_42899685_42899685_790558.205.png", "Validation_42899685/Validation_42899685_42899685_790560.204.png" ], "turn_id": 8, "scan_id": "Validation_42899685", "origin_question": "Which direction is the compact, white front-loading washing machine in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the compact, white front-loading washing machine in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_42899685/Validation_42899685_42899685_790562.220.png", "Validation_42899685/Validation_42899685_42899685_790564.219.png", "Validation_42899685/Validation_42899685_42899685_790566.219.png", "Validation_42899685/Validation_42899685_42899685_790568.218.png", "Validation_42899685/Validation_42899685_42899685_790570.217.png" ], "turn_id": 9, "scan_id": "Validation_42899685", "origin_question": "Which direction is the sleek white dishwasher made of durable steel in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the sleek white dishwasher made of durable steel in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "Validation_41159541/Validation_41159541_41159541_39778.241.png", "Validation_41159541/Validation_41159541_41159541_39780.224.png", "Validation_41159541/Validation_41159541_41159541_39782.223.png", "Validation_41159541/Validation_41159541_41159541_39784.222.png", "Validation_41159541/Validation_41159541_41159541_39786.221.png" ], "turn_id": 1, "scan_id": "Validation_41159541", "origin_question": "Remember, how many toothbrush(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many toothbrush(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_41159541/Validation_41159541_41159541_39788.237.png", "Validation_41159541/Validation_41159541_41159541_39790.236.png", "Validation_41159541/Validation_41159541_41159541_39792.235.png", "Validation_41159541/Validation_41159541_41159541_39794.234.png", "Validation_41159541/Validation_41159541_41159541_39796.234.png" ], "turn_id": 2, "scan_id": "Validation_41159541", "origin_question": "When did you first discover \"the low-profile, off-white wooden cabinet\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the low-profile, off-white wooden cabinet\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_41159541/Validation_41159541_41159541_39798.233.png", "Validation_41159541/Validation_41159541_41159541_39800.232.png", "Validation_41159541/Validation_41159541_41159541_39802.231.png", "Validation_41159541/Validation_41159541_41159541_39804.230.png", "Validation_41159541/Validation_41159541_41159541_39806.230.png" ], "turn_id": 3, "scan_id": "Validation_41159541", "origin_question": "", "option": [], "answer": "", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_41159541/Validation_41159541_41159541_39808.229.png", "Validation_41159541/Validation_41159541_41159541_39810.228.png", "Validation_41159541/Validation_41159541_41159541_39812.227.png", "Validation_41159541/Validation_41159541_41159541_39814.226.png", "Validation_41159541/Validation_41159541_41159541_39816.225.png" ], "turn_id": 4, "scan_id": "Validation_41159541", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the sleek, dark rectangular laptop made of durable materials\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the sleek, dark rectangular laptop made of durable materials\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_41159541/Validation_41159541_41159541_39818.225.png", "Validation_41159541/Validation_41159541_41159541_39820.224.png", "Validation_41159541/Validation_41159541_41159541_39822.223.png", "Validation_41159541/Validation_41159541_41159541_39824.222.png", "Validation_41159541/Validation_41159541_41159541_39826.221.png" ], "turn_id": 5, "scan_id": "Validation_41159541", "origin_question": "Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the large, sleek black monitor with the matte finish\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the large, sleek black monitor with the matte finish\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_41159541/Validation_41159541_41159541_39828.237.png", "Validation_41159541/Validation_41159541_41159541_39830.236.png", "Validation_41159541/Validation_41159541_41159541_39832.235.png", "Validation_41159541/Validation_41159541_41159541_39834.235.png", "Validation_41159541/Validation_41159541_41159541_39836.234.png" ], "turn_id": 6, "scan_id": "Validation_41159541", "origin_question": "Is the spherical globe featuring blue oceans and colored landmasses to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the spherical globe featuring blue oceans and colored landmasses to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_41159541/Validation_41159541_41159541_39838.233.png", "Validation_41159541/Validation_41159541_41159541_39840.232.png", "Validation_41159541/Validation_41159541_41159541_39842.231.png", "Validation_41159541/Validation_41159541_41159541_39844.230.png", "Validation_41159541/Validation_41159541_41159541_39846.230.png" ], "turn_id": 7, "scan_id": "Validation_41159541", "origin_question": "How far is your current position from where you were at the end of the turn 4? (in meters)", "option": [], "answer": "2.77", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 4? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_41159541/Validation_41159541_41159541_39848.229.png", "Validation_41159541/Validation_41159541_41159541_39850.228.png", "Validation_41159541/Validation_41159541_41159541_39852.227.png", "Validation_41159541/Validation_41159541_41159541_39854.226.png", "Validation_41159541/Validation_41159541_41159541_39856.225.png" ], "turn_id": 8, "scan_id": "Validation_41159541", "origin_question": "Is the low-profile, off-white wooden cabinet to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the low-profile, off-white wooden cabinet to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47115529/Validation_47115529_47115529_14417.404.png", "Validation_47115529/Validation_47115529_47115529_14419.386.png", "Validation_47115529/Validation_47115529_47115529_14421.386.png", "Validation_47115529/Validation_47115529_47115529_14423.385.png" ], "turn_id": 1, "scan_id": "Validation_47115529", "origin_question": "Remember, have you seen any bathtub(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any bathtub(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_47115529/Validation_47115529_47115529_14425.384.png", "Validation_47115529/Validation_47115529_47115529_14427.383.png", "Validation_47115529/Validation_47115529_47115529_14429.382.png", "Validation_47115529/Validation_47115529_47115529_14431.382.png" ], "turn_id": 2, "scan_id": "Validation_47115529", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the fireplace\", Y: \"the medium to dark brown wooden table with the glass top\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the fireplace\", Y: \"the medium to dark brown wooden table with the glass top\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_47115529/Validation_47115529_47115529_14433.397.png", "Validation_47115529/Validation_47115529_47115529_14435.397.png", "Validation_47115529/Validation_47115529_47115529_14437.396.png", "Validation_47115529/Validation_47115529_47115529_14439.395.png" ], "turn_id": 3, "scan_id": "Validation_47115529", "origin_question": "Is the modern curved glass vase with the smooth surface to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the modern curved glass vase with the smooth surface to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_47115529/Validation_47115529_47115529_14441.394.png", "Validation_47115529/Validation_47115529_47115529_14443.394.png", "Validation_47115529/Validation_47115529_47115529_14445.393.png", "Validation_47115529/Validation_47115529_47115529_14447.392.png" ], "turn_id": 4, "scan_id": "Validation_47115529", "origin_question": "Which two objects are on the same side of you now? \"the fireplace\", \"the rectangular light wood door with glass panels\", \"the medium to dark brown wooden table with the glass top\". ", "option": [ "the fireplace, the rectangular light wood door with glass panels", "the fireplace, the medium to dark brown wooden table with the glass top", "the rectangular light wood door with glass panels, the medium to dark brown wooden table with the glass top" ], "answer": "the rectangular light wood door with glass panels, the medium to dark brown wooden table with the glass top", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the fireplace\", \"the rectangular light wood door with glass panels\", \"the medium to dark brown wooden table with the glass top\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the fireplace, the rectangular light wood door with glass panels', 'the fireplace, the medium to dark brown wooden table with the glass top', 'the rectangular light wood door with glass panels, the medium to dark brown wooden table with the glass top'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47115529/Validation_47115529_47115529_14449.391.png", "Validation_47115529/Validation_47115529_47115529_14451.390.png", "Validation_47115529/Validation_47115529_47115529_14453.390.png", "Validation_47115529/Validation_47115529_47115529_14455.389.png" ], "turn_id": 5, "scan_id": "Validation_47115529", "origin_question": "", "option": [], "answer": "", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_47115529/Validation_47115529_47115529_14457.388.png", "Validation_47115529/Validation_47115529_47115529_14459.387.png", "Validation_47115529/Validation_47115529_47115529_14461.386.png", "Validation_47115529/Validation_47115529_47115529_14463.386.png" ], "turn_id": 6, "scan_id": "Validation_47115529", "origin_question": "At the end of which turn were both of \"the fireplace\" and \"the large rectangular window with the white frame\" on your left side?", "option": [], "answer": "1", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the fireplace\" and \"the large rectangular window with the white frame\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "Validation_48458415/Validation_48458415_48458415_115180.178.png", "Validation_48458415/Validation_48458415_48458415_115182.161.png", "Validation_48458415/Validation_48458415_48458415_115184.160.png", "Validation_48458415/Validation_48458415_48458415_115186.159.png" ], "turn_id": 1, "scan_id": "Validation_48458415", "origin_question": "Remember, how many chair(s) have you seen so far? ", "option": [], "answer": "5", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many chair(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "Validation_48458415/Validation_48458415_48458415_115188.158.png", "Validation_48458415/Validation_48458415_48458415_115190.158.png", "Validation_48458415/Validation_48458415_48458415_115192.157.png", "Validation_48458415/Validation_48458415_48458415_115194.156.png" ], "turn_id": 2, "scan_id": "Validation_48458415", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the cylindrical metallic silver bin with black accents\"; \"the compact rectangular box\"; \"the rectangular wooden table with wood-grain legs\" ", "option": [ "the cylindrical metallic silver bin with black accents", "the compact rectangular box", "the rectangular wooden table with wood-grain legs" ], "answer": "the compact rectangular box", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the cylindrical metallic silver bin with black accents\"; \"the compact rectangular box\"; \"the rectangular wooden table with wood-grain legs\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the cylindrical metallic silver bin with black accents', 'the compact rectangular box', 'the rectangular wooden table with wood-grain legs'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_48458415/Validation_48458415_48458415_115196.155.png", "Validation_48458415/Validation_48458415_48458415_115198.154.png", "Validation_48458415/Validation_48458415_48458415_115200.153.png", "Validation_48458415/Validation_48458415_48458415_115202.152.png" ], "turn_id": 3, "scan_id": "Validation_48458415", "origin_question": "Please recall \"the compact rectangular box\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.88", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the compact rectangular box\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_48458415/Validation_48458415_48458415_115204.152.png", "Validation_48458415/Validation_48458415_48458415_115206.151.png", "Validation_48458415/Validation_48458415_48458415_115208.150.png", "Validation_48458415/Validation_48458415_48458415_115210.149.png" ], "turn_id": 4, "scan_id": "Validation_48458415", "origin_question": "Which direction is the sleek rectangular stainless steel sink in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the sleek rectangular stainless steel sink in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "Validation_48458415/Validation_48458415_48458415_115212.148.png", "Validation_48458415/Validation_48458415_48458415_115214.148.png", "Validation_48458415/Validation_48458415_48458415_115216.147.png", "Validation_48458415/Validation_48458415_48458415_115218.146.png" ], "turn_id": 5, "scan_id": "Validation_48458415", "origin_question": "Among these three objects(\"the large rectangular window with the white frame\",\"the cylindrical metallic silver bin with black accents\",\"the sleek rectangular stainless steel sink\" ), which one is the closest to you?", "option": [ "the large rectangular window with the white frame", "the cylindrical metallic silver bin with black accents", "the sleek rectangular stainless steel sink" ], "answer": "the large rectangular window with the white frame", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects(\"the large rectangular window with the white frame\",\"the cylindrical metallic silver bin with black accents\",\"the sleek rectangular stainless steel sink\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the large rectangular window with the white frame', 'the cylindrical metallic silver bin with black accents', 'the sleek rectangular stainless steel sink'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_48458415/Validation_48458415_48458415_115220.145.png", "Validation_48458415/Validation_48458415_48458415_115222.144.png", "Validation_48458415/Validation_48458415_48458415_115224.160.png", "Validation_48458415/Validation_48458415_48458415_115226.159.png" ], "turn_id": 6, "scan_id": "Validation_48458415", "origin_question": "At the end of which turn were both of \"the cylindrical metallic silver bin with black accents\" and \"the stainless steel kettle with soft blue or gray tones\" on your left side?", "option": [], "answer": "1", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the cylindrical metallic silver bin with black accents\" and \"the stainless steel kettle with soft blue or gray tones\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "Validation_48458415/Validation_48458415_48458415_115228.158.png", "Validation_48458415/Validation_48458415_48458415_115230.158.png", "Validation_48458415/Validation_48458415_48458415_115232.157.png", "Validation_48458415/Validation_48458415_48458415_115234.156.png" ], "turn_id": 7, "scan_id": "Validation_48458415", "origin_question": "Among these three objects(\"the cylindrical metallic silver bin with black accents\",\"the muted red bag with sturdy handles\",\"the sleek rectangular stainless steel sink\" ), which one is the farthest from you?", "option": [ "the cylindrical metallic silver bin with black accents", "the muted red bag with sturdy handles", "the sleek rectangular stainless steel sink" ], "answer": "the muted red bag with sturdy handles", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects(\"the cylindrical metallic silver bin with black accents\",\"the muted red bag with sturdy handles\",\"the sleek rectangular stainless steel sink\" ), which one is the farthest from you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the cylindrical metallic silver bin with black accents', 'the muted red bag with sturdy handles', 'the sleek rectangular stainless steel sink'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_48458415/Validation_48458415_48458415_115236.155.png", "Validation_48458415/Validation_48458415_48458415_115238.154.png", "Validation_48458415/Validation_48458415_48458415_115240.153.png", "Validation_48458415/Validation_48458415_48458415_115242.153.png" ], "turn_id": 8, "scan_id": "Validation_48458415", "origin_question": "Which two objects are on the same side of you now? \"the circular clock with the bold yellow frame\", \"the stainless steel kettle with soft blue or gray tones\", \"the large circular pan with the dark non-stick coating\". ", "option": [ "the circular clock with the bold yellow frame, the stainless steel kettle with soft blue or gray tones", "the circular clock with the bold yellow frame, the large circular pan with the dark non-stick coating", "the stainless steel kettle with soft blue or gray tones, the large circular pan with the dark non-stick coating" ], "answer": "the stainless steel kettle with soft blue or gray tones, the large circular pan with the dark non-stick coating", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the circular clock with the bold yellow frame\", \"the stainless steel kettle with soft blue or gray tones\", \"the large circular pan with the dark non-stick coating\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the circular clock with the bold yellow frame, the stainless steel kettle with soft blue or gray tones', 'the circular clock with the bold yellow frame, the large circular pan with the dark non-stick coating', 'the stainless steel kettle with soft blue or gray tones, the large circular pan with the dark non-stick coating'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_42899613/Validation_42899613_42899613_228273.033.png", "Validation_42899613/Validation_42899613_42899613_228275.016.png", "Validation_42899613/Validation_42899613_42899613_228277.015.png", "Validation_42899613/Validation_42899613_42899613_228279.014.png" ], "turn_id": 1, "scan_id": "Validation_42899613", "origin_question": "", "option": [], "answer": "", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42899613/Validation_42899613_42899613_228281.014.png", "Validation_42899613/Validation_42899613_42899613_228283.013.png", "Validation_42899613/Validation_42899613_42899613_228285.012.png", "Validation_42899613/Validation_42899613_42899613_228287.011.png" ], "turn_id": 2, "scan_id": "Validation_42899613", "origin_question": "Remember, have you seen any trophy(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any trophy(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_42899613/Validation_42899613_42899613_228289.027.png", "Validation_42899613/Validation_42899613_42899613_228291.026.png", "Validation_42899613/Validation_42899613_42899613_228293.025.png", "Validation_42899613/Validation_42899613_42899613_228295.025.png" ], "turn_id": 3, "scan_id": "Validation_42899613", "origin_question": "When did you first discover \"the glossy deep green ceramic vase\"? (index of the turn)", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the glossy deep green ceramic vase\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_42899613/Validation_42899613_42899613_228297.024.png", "Validation_42899613/Validation_42899613_42899613_228299.023.png", "Validation_42899613/Validation_42899613_42899613_228301.022.png", "Validation_42899613/Validation_42899613_42899613_228303.022.png" ], "turn_id": 4, "scan_id": "Validation_42899613", "origin_question": "Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "130.18", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_42899613/Validation_42899613_42899613_228305.021.png", "Validation_42899613/Validation_42899613_42899613_228307.020.png", "Validation_42899613/Validation_42899613_42899613_228309.019.png", "Validation_42899613/Validation_42899613_42899613_228311.018.png" ], "turn_id": 5, "scan_id": "Validation_42899613", "origin_question": "Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Right", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_42899613/Validation_42899613_42899613_228313.018.png", "Validation_42899613/Validation_42899613_42899613_228315.017.png", "Validation_42899613/Validation_42899613_42899613_228317.016.png", "Validation_42899613/Validation_42899613_42899613_228319.015.png" ], "turn_id": 6, "scan_id": "Validation_42899613", "origin_question": "Please recall \"the small bronze wall-mounted light with exposed bulbs\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "3.2", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the small bronze wall-mounted light with exposed bulbs\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_42899613/Validation_42899613_42899613_228321.014.png", "Validation_42899613/Validation_42899613_42899613_228323.014.png", "Validation_42899613/Validation_42899613_42899613_228325.013.png", "Validation_42899613/Validation_42899613_42899613_228327.012.png" ], "turn_id": 7, "scan_id": "Validation_42899613", "origin_question": "At the end of which turn were both of \"the glossy deep green ceramic vase\" and \"the large dark leather chair with wooden frame\" on your right side?", "option": [], "answer": "3", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the glossy deep green ceramic vase\" and \"the large dark leather chair with wooden frame\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47331068/Validation_47331068_47331068_131065.087.png", "Validation_47331068/Validation_47331068_47331068_131067.053.png", "Validation_47331068/Validation_47331068_47331068_131069.052.png", "Validation_47331068/Validation_47331068_47331068_131071.068.png" ], "turn_id": 1, "scan_id": "Validation_47331068", "origin_question": "Remember, have you seen any humidifier(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any humidifier(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47331068/Validation_47331068_47331068_131073.067.png", "Validation_47331068/Validation_47331068_47331068_131075.066.png", "Validation_47331068/Validation_47331068_47331068_131077.065.png", "Validation_47331068/Validation_47331068_47331068_131079.065.png" ], "turn_id": 2, "scan_id": "Validation_47331068", "origin_question": "In which turn were you closest to \"the round, bright yellow plastic basin\"? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the round, bright yellow plastic basin\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47331068/Validation_47331068_47331068_131081.064.png", "Validation_47331068/Validation_47331068_47331068_131083.063.png", "Validation_47331068/Validation_47331068_47331068_131085.062.png", "Validation_47331068/Validation_47331068_47331068_131087.061.png" ], "turn_id": 3, "scan_id": "Validation_47331068", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the round, bright yellow plastic basin\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the round, bright yellow plastic basin\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_47331068/Validation_47331068_47331068_131089.060.png", "Validation_47331068/Validation_47331068_47331068_131091.060.png", "Validation_47331068/Validation_47331068_47331068_131093.059.png", "Validation_47331068/Validation_47331068_47331068_131095.058.png" ], "turn_id": 4, "scan_id": "Validation_47331068", "origin_question": "In which turn did you see both \"the compact table made of waterproof wood\" and \"the round, bright yellow plastic basin\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the compact table made of waterproof wood\" and \"the round, bright yellow plastic basin\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47331068/Validation_47331068_47331068_131097.057.png", "Validation_47331068/Validation_47331068_47331068_131099.056.png", "Validation_47331068/Validation_47331068_47331068_131101.056.png", "Validation_47331068/Validation_47331068_47331068_131103.055.png" ], "turn_id": 5, "scan_id": "Validation_47331068", "origin_question": "Which direction is the round, bright yellow plastic basin in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the round, bright yellow plastic basin in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47331068/Validation_47331068_47331068_131105.054.png", "Validation_47331068/Validation_47331068_47331068_131107.053.png", "Validation_47331068/Validation_47331068_47331068_131109.052.png", "Validation_47331068/Validation_47331068_47331068_131111.068.png" ], "turn_id": 6, "scan_id": "Validation_47331068", "origin_question": "Which direction is the round, bright yellow plastic basin in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the round, bright yellow plastic basin in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47331068/Validation_47331068_47331068_131113.067.png", "Validation_47331068/Validation_47331068_47331068_131115.067.png", "Validation_47331068/Validation_47331068_47331068_131117.066.png", "Validation_47331068/Validation_47331068_47331068_131119.065.png" ], "turn_id": 7, "scan_id": "Validation_47331068", "origin_question": "Which direction is the round, bright yellow plastic basin in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the round, bright yellow plastic basin in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_47331068/Validation_47331068_47331068_131121.064.png", "Validation_47331068/Validation_47331068_47331068_131123.063.png", "Validation_47331068/Validation_47331068_47331068_131125.063.png", "Validation_47331068/Validation_47331068_47331068_131127.062.png" ], "turn_id": 8, "scan_id": "Validation_47331068", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the classic white ceramic toilet with the wooden seat\" from your position?", "option": [], "answer": "88.13", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the classic white ceramic toilet with the wooden seat\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_47331068/Validation_47331068_47331068_131129.061.png", "Validation_47331068/Validation_47331068_47331068_131131.060.png", "Validation_47331068/Validation_47331068_47331068_131133.059.png", "Validation_47331068/Validation_47331068_47331068_131135.059.png" ], "turn_id": 9, "scan_id": "Validation_47331068", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the round, bright yellow plastic basin\" from your position?", "option": [], "answer": "74.04", "user_message": "For the 9 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the round, bright yellow plastic basin\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_48458667/Validation_48458667_48458667_8355.676.png", "Validation_48458667/Validation_48458667_48458667_8357.659.png", "Validation_48458667/Validation_48458667_48458667_8359.658.png", "Validation_48458667/Validation_48458667_48458667_8361.657.png" ], "turn_id": 1, "scan_id": "Validation_48458667", "origin_question": "Remember, have you seen any clothes(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any clothes(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_48458667/Validation_48458667_48458667_8363.656.png", "Validation_48458667/Validation_48458667_48458667_8365.655.png", "Validation_48458667/Validation_48458667_48458667_8367.655.png", "Validation_48458667/Validation_48458667_48458667_8369.654.png" ], "turn_id": 2, "scan_id": "Validation_48458667", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the classic acoustic guitar in rich brown\" from your position?", "option": [], "answer": "178.41", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the classic acoustic guitar in rich brown\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_48458667/Validation_48458667_48458667_8371.653.png", "Validation_48458667/Validation_48458667_48458667_8373.652.png", "Validation_48458667/Validation_48458667_48458667_8375.651.png", "Validation_48458667/Validation_48458667_48458667_8377.650.png" ], "turn_id": 3, "scan_id": "Validation_48458667", "origin_question": "Which two objects are on the same side of you now? \"the modern white table with the smooth laminate finish\", \"the modern gray chair with the curved backrest\", \"the classic acoustic guitar in rich brown\". ", "option": [ "the modern white table with the smooth laminate finish, the modern gray chair with the curved backrest", "the modern white table with the smooth laminate finish, the classic acoustic guitar in rich brown", "the modern gray chair with the curved backrest, the classic acoustic guitar in rich brown" ], "answer": "the modern white table with the smooth laminate finish, the modern gray chair with the curved backrest", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the modern white table with the smooth laminate finish\", \"the modern gray chair with the curved backrest\", \"the classic acoustic guitar in rich brown\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the modern white table with the smooth laminate finish, the modern gray chair with the curved backrest', 'the modern white table with the smooth laminate finish, the classic acoustic guitar in rich brown', 'the modern gray chair with the curved backrest, the classic acoustic guitar in rich brown'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_48458667/Validation_48458667_48458667_8379.649.png", "Validation_48458667/Validation_48458667_48458667_8381.649.png", "Validation_48458667/Validation_48458667_48458667_8383.648.png", "Validation_48458667/Validation_48458667_48458667_8385.647.png" ], "turn_id": 4, "scan_id": "Validation_48458667", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the collection of clothes in various colors and textures\" from your position?", "option": [], "answer": "138.22", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the collection of clothes in various colors and textures\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_48458667/Validation_48458667_48458667_8387.646.png", "Validation_48458667/Validation_48458667_48458667_8389.662.png", "Validation_48458667/Validation_48458667_48458667_8391.661.png", "Validation_48458667/Validation_48458667_48458667_8393.660.png" ], "turn_id": 5, "scan_id": "Validation_48458667", "origin_question": "Which direction is the collection of clothes in various colors and textures in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the collection of clothes in various colors and textures in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_48458667/Validation_48458667_48458667_8395.659.png", "Validation_48458667/Validation_48458667_48458667_8397.659.png", "Validation_48458667/Validation_48458667_48458667_8399.658.png", "Validation_48458667/Validation_48458667_48458667_8401.657.png" ], "turn_id": 6, "scan_id": "Validation_48458667", "origin_question": "Which direction is the sleek rectangular monitor with the high-gloss black finish in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the sleek rectangular monitor with the high-gloss black finish in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_48458667/Validation_48458667_48458667_8403.656.png", "Validation_48458667/Validation_48458667_48458667_8405.655.png", "Validation_48458667/Validation_48458667_48458667_8407.655.png", "Validation_48458667/Validation_48458667_48458667_8409.654.png" ], "turn_id": 7, "scan_id": "Validation_48458667", "origin_question": "Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the vibrant colored poster printed on high-quality paper\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the vibrant colored poster printed on high-quality paper\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_48458667/Validation_48458667_48458667_8411.653.png", "Validation_48458667/Validation_48458667_48458667_8413.652.png", "Validation_48458667/Validation_48458667_48458667_8415.651.png", "Validation_48458667/Validation_48458667_48458667_8417.650.png" ], "turn_id": 8, "scan_id": "Validation_48458667", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the classic acoustic guitar in rich brown\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the classic acoustic guitar in rich brown\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "Validation_45663107/Validation_45663107_45663107_54806.464.png", "Validation_45663107/Validation_45663107_45663107_54808.447.png", "Validation_45663107/Validation_45663107_45663107_54810.446.png", "Validation_45663107/Validation_45663107_45663107_54812.445.png" ], "turn_id": 1, "scan_id": "Validation_45663107", "origin_question": "Remember, how many printer(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many printer(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_45663107/Validation_45663107_45663107_54814.444.png", "Validation_45663107/Validation_45663107_45663107_54816.444.png", "Validation_45663107/Validation_45663107_45663107_54818.459.png", "Validation_45663107/Validation_45663107_45663107_54820.459.png" ], "turn_id": 2, "scan_id": "Validation_45663107", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the rustic wooden cabinet with double doors and drawers\" from your position?", "option": [], "answer": "165.01", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the rustic wooden cabinet with double doors and drawers\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_45663107/Validation_45663107_45663107_54822.458.png", "Validation_45663107/Validation_45663107_45663107_54824.457.png", "Validation_45663107/Validation_45663107_45663107_54826.456.png", "Validation_45663107/Validation_45663107_45663107_54828.455.png" ], "turn_id": 3, "scan_id": "Validation_45663107", "origin_question": "Is the rustic wooden cabinet with double doors and drawers to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the rustic wooden cabinet with double doors and drawers to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_45663107/Validation_45663107_45663107_54830.454.png", "Validation_45663107/Validation_45663107_45663107_54832.454.png", "Validation_45663107/Validation_45663107_45663107_54834.453.png", "Validation_45663107/Validation_45663107_45663107_54836.452.png" ], "turn_id": 4, "scan_id": "Validation_45663107", "origin_question": "Which direction is the rectangular mirror with the warm wood-toned frame in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the rectangular mirror with the warm wood-toned frame in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_45663107/Validation_45663107_45663107_54838.451.png", "Validation_45663107/Validation_45663107_45663107_54840.450.png", "Validation_45663107/Validation_45663107_45663107_54842.450.png", "Validation_45663107/Validation_45663107_45663107_54844.449.png" ], "turn_id": 5, "scan_id": "Validation_45663107", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the rectangular mirror with the warm wood-toned frame\" from your position?", "option": [], "answer": "121.69", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the rectangular mirror with the warm wood-toned frame\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_45663107/Validation_45663107_45663107_54846.448.png", "Validation_45663107/Validation_45663107_45663107_54848.447.png", "Validation_45663107/Validation_45663107_45663107_54850.446.png", "Validation_45663107/Validation_45663107_45663107_54852.445.png" ], "turn_id": 6, "scan_id": "Validation_45663107", "origin_question": "Please recall \"the framed abstract picture in muted greens and whites\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.85", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the framed abstract picture in muted greens and whites\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_45663107/Validation_45663107_45663107_54854.445.png", "Validation_45663107/Validation_45663107_45663107_54856.444.png", "Validation_45663107/Validation_45663107_45663107_54858.460.png", "Validation_45663107/Validation_45663107_45663107_54860.459.png" ], "turn_id": 7, "scan_id": "Validation_45663107", "origin_question": "When did you first discover \"the rectangular mirror with the warm wood-toned frame\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the rectangular mirror with the warm wood-toned frame\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42445981/Validation_42445981_42445981_346892.760.png", "Validation_42445981/Validation_42445981_42445981_346894.743.png", "Validation_42445981/Validation_42445981_42445981_346896.742.png", "Validation_42445981/Validation_42445981_42445981_346898.741.png", "Validation_42445981/Validation_42445981_42445981_346900.740.png" ], "turn_id": 1, "scan_id": "Validation_42445981", "origin_question": "Remember, have you seen any table(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any table(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_42445981/Validation_42445981_42445981_346902.739.png", "Validation_42445981/Validation_42445981_42445981_346904.739.png", "Validation_42445981/Validation_42445981_42445981_346906.738.png", "Validation_42445981/Validation_42445981_42445981_346908.737.png", "Validation_42445981/Validation_42445981_42445981_346910.736.png" ], "turn_id": 2, "scan_id": "Validation_42445981", "origin_question": "Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Right", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_42445981/Validation_42445981_42445981_346912.735.png", "Validation_42445981/Validation_42445981_42445981_346914.735.png", "Validation_42445981/Validation_42445981_42445981_346916.734.png", "Validation_42445981/Validation_42445981_42445981_346918.733.png", "Validation_42445981/Validation_42445981_42445981_346920.732.png" ], "turn_id": 3, "scan_id": "Validation_42445981", "origin_question": "Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Right", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_42445981/Validation_42445981_42445981_346922.731.png", "Validation_42445981/Validation_42445981_42445981_346924.731.png", "Validation_42445981/Validation_42445981_42445981_346926.730.png", "Validation_42445981/Validation_42445981_42445981_346928.729.png", "Validation_42445981/Validation_42445981_42445981_346930.728.png" ], "turn_id": 4, "scan_id": "Validation_42445981", "origin_question": "What will be the first-time appearance order of the following categories: chair, pillow, plant?", "option": [ "plant, pillow, chair", "chair, plant, pillow", "pillow, chair, plant", "chair, pillow, plant" ], "answer": "chair, plant, pillow", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: chair, pillow, plant? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['plant, pillow, chair', 'chair, plant, pillow', 'pillow, chair, plant', 'chair, pillow, plant'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_42445981/Validation_42445981_42445981_346932.727.png", "Validation_42445981/Validation_42445981_42445981_346934.743.png", "Validation_42445981/Validation_42445981_42445981_346936.742.png", "Validation_42445981/Validation_42445981_42445981_346938.741.png", "Validation_42445981/Validation_42445981_42445981_346940.741.png" ], "turn_id": 5, "scan_id": "Validation_42445981", "origin_question": "In which turn were you closest to \"the cabinet\"? ", "option": [], "answer": "5", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the cabinet\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_42445981/Validation_42445981_42445981_346942.740.png", "Validation_42445981/Validation_42445981_42445981_346944.739.png", "Validation_42445981/Validation_42445981_42445981_346946.738.png", "Validation_42445981/Validation_42445981_42445981_346948.737.png", "Validation_42445981/Validation_42445981_42445981_346950.736.png" ], "turn_id": 6, "scan_id": "Validation_42445981", "origin_question": "Is the vibrant green plant to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the vibrant green plant to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_42445981/Validation_42445981_42445981_346952.736.png", "Validation_42445981/Validation_42445981_42445981_346954.735.png", "Validation_42445981/Validation_42445981_42445981_346956.734.png", "Validation_42445981/Validation_42445981_42445981_346958.733.png", "Validation_42445981/Validation_42445981_42445981_346960.732.png" ], "turn_id": 7, "scan_id": "Validation_42445981", "origin_question": "Compared to your position at the end of the 5 turn, are you now closer or farther away from \"the vibrant green plant\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 5 turn, are you now closer or farther away from \"the vibrant green plant\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_42445981/Validation_42445981_42445981_346962.732.png", "Validation_42445981/Validation_42445981_42445981_346964.731.png", "Validation_42445981/Validation_42445981_42445981_346966.730.png", "Validation_42445981/Validation_42445981_42445981_346968.729.png", "Validation_42445981/Validation_42445981_42445981_346970.728.png" ], "turn_id": 8, "scan_id": "Validation_42445981", "origin_question": "Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the modern black monitor with the glossy finish\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the modern black monitor with the glossy finish\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_42445981/Validation_42445981_42445981_346972.727.png", "Validation_42445981/Validation_42445981_42445981_346974.743.png", "Validation_42445981/Validation_42445981_42445981_346976.742.png", "Validation_42445981/Validation_42445981_42445981_346978.742.png", "Validation_42445981/Validation_42445981_42445981_346980.741.png" ], "turn_id": 9, "scan_id": "Validation_42445981", "origin_question": "By the end of the 4 turn, were you getting closer to or farther away from these two objects: X: \"the vibrant green plant\", Y: \"the modern black monitor with the glossy finish\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "farther away from both X and Y", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 4 turn, were you getting closer to or farther away from these two objects: X: \"the vibrant green plant\", Y: \"the modern black monitor with the glossy finish\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_42445981/Validation_42445981_42445981_346982.740.png", "Validation_42445981/Validation_42445981_42445981_346984.739.png", "Validation_42445981/Validation_42445981_42445981_346986.738.png", "Validation_42445981/Validation_42445981_42445981_346988.738.png", "Validation_42445981/Validation_42445981_42445981_346990.737.png" ], "turn_id": 10, "scan_id": "Validation_42445981", "origin_question": "Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the modern black monitor with the glossy finish\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the modern black monitor with the glossy finish\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_48018737/Validation_48018737_48018737_3469.827.png", "Validation_48018737/Validation_48018737_48018737_3471.327.png", "Validation_48018737/Validation_48018737_48018737_3472.826.png", "Validation_48018737/Validation_48018737_48018737_3474.325.png" ], "turn_id": 1, "scan_id": "Validation_48018737", "origin_question": "Remember, have you seen any toothbrush(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any toothbrush(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_48018737/Validation_48018737_48018737_3475.825.png", "Validation_48018737/Validation_48018737_48018737_3477.324.png", "Validation_48018737/Validation_48018737_48018737_3478.824.png", "Validation_48018737/Validation_48018737_48018737_3480.323.png" ], "turn_id": 2, "scan_id": "Validation_48018737", "origin_question": "In which turn were you farthest from \"the glossy white rectangular bathtub made of acrylic\"? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the glossy white rectangular bathtub made of acrylic\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_48018737/Validation_48018737_48018737_3481.822.png", "Validation_48018737/Validation_48018737_48018737_3483.322.png", "Validation_48018737/Validation_48018737_48018737_3484.821.png", "Validation_48018737/Validation_48018737_48018737_3486.320.png" ], "turn_id": 3, "scan_id": "Validation_48018737", "origin_question": "In which turn were you closest to \"the beige plastic box with brown trim and rounded corners\"? ", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the beige plastic box with brown trim and rounded corners\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_48018737/Validation_48018737_48018737_3487.820.png", "Validation_48018737/Validation_48018737_48018737_3489.319.png", "Validation_48018737/Validation_48018737_48018737_3490.819.png", "Validation_48018737/Validation_48018737_48018737_3492.318.png" ], "turn_id": 4, "scan_id": "Validation_48018737", "origin_question": "Please recall \"the medium-sized, deep red cylindrical plastic bottle\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.55", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the medium-sized, deep red cylindrical plastic bottle\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_48018737/Validation_48018737_48018737_3493.818.png", "Validation_48018737/Validation_48018737_48018737_3495.317.png", "Validation_48018737/Validation_48018737_48018737_3496.816.png", "Validation_48018737/Validation_48018737_48018737_3498.316.png" ], "turn_id": 5, "scan_id": "Validation_48018737", "origin_question": "In which turn did you see both \"the sleek, rectangular white ceramic sink\" and \"the beige plastic box with brown trim and rounded corners\" simultaneously? ", "option": [], "answer": "3", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the sleek, rectangular white ceramic sink\" and \"the beige plastic box with brown trim and rounded corners\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_45662921/Validation_45662921_45662921_21752.942.png", "Validation_45662921/Validation_45662921_45662921_21754.925.png", "Validation_45662921/Validation_45662921_45662921_21756.924.png", "Validation_45662921/Validation_45662921_45662921_21758.923.png", "Validation_45662921/Validation_45662921_45662921_21760.922.png" ], "turn_id": 1, "scan_id": "Validation_45662921", "origin_question": "Remember, have you seen any magazine(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any magazine(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_45662921/Validation_45662921_45662921_21762.921.png", "Validation_45662921/Validation_45662921_45662921_21764.921.png", "Validation_45662921/Validation_45662921_45662921_21766.920.png", "Validation_45662921/Validation_45662921_45662921_21768.919.png", "Validation_45662921/Validation_45662921_45662921_21770.918.png" ], "turn_id": 2, "scan_id": "Validation_45662921", "origin_question": "In which turn were you farthest from \"the glossy plant with deep green, oval leaves\"? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the glossy plant with deep green, oval leaves\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_45662921/Validation_45662921_45662921_21772.917.png", "Validation_45662921/Validation_45662921_45662921_21774.916.png", "Validation_45662921/Validation_45662921_45662921_21776.915.png", "Validation_45662921/Validation_45662921_45662921_21778.915.png", "Validation_45662921/Validation_45662921_45662921_21780.914.png" ], "turn_id": 3, "scan_id": "Validation_45662921", "origin_question": "Please recall \"the sleek, dark grey rectangular metal radiator\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "5.93", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the sleek, dark grey rectangular metal radiator\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_45662921/Validation_45662921_45662921_21782.913.png", "Validation_45662921/Validation_45662921_45662921_21784.929.png", "Validation_45662921/Validation_45662921_45662921_21786.928.png", "Validation_45662921/Validation_45662921_45662921_21788.927.png", "Validation_45662921/Validation_45662921_45662921_21790.926.png" ], "turn_id": 4, "scan_id": "Validation_45662921", "origin_question": "What will be the first-time appearance order of the following categories: chair, plant, stairs?", "option": [ "chair, stairs, plant", "plant, chair, stairs", "stairs, plant, chair", "stairs, chair, plant" ], "answer": "stairs, plant, chair", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: chair, plant, stairs? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['chair, stairs, plant', 'plant, chair, stairs', 'stairs, plant, chair', 'stairs, chair, plant'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_45662921/Validation_45662921_45662921_21792.926.png", "Validation_45662921/Validation_45662921_45662921_21794.925.png", "Validation_45662921/Validation_45662921_45662921_21796.924.png", "Validation_45662921/Validation_45662921_45662921_21798.923.png", "Validation_45662921/Validation_45662921_45662921_21800.922.png" ], "turn_id": 5, "scan_id": "Validation_45662921", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "3.62", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_45662921/Validation_45662921_45662921_21802.921.png", "Validation_45662921/Validation_45662921_45662921_21804.921.png", "Validation_45662921/Validation_45662921_45662921_21806.920.png", "Validation_45662921/Validation_45662921_45662921_21808.919.png", "Validation_45662921/Validation_45662921_45662921_21810.918.png" ], "turn_id": 6, "scan_id": "Validation_45662921", "origin_question": "Using your orientation at the end of turn 4 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "104.87", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 4 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_45662921/Validation_45662921_45662921_21812.917.png", "Validation_45662921/Validation_45662921_45662921_21814.916.png", "Validation_45662921/Validation_45662921_45662921_21816.916.png", "Validation_45662921/Validation_45662921_45662921_21818.915.png", "Validation_45662921/Validation_45662921_45662921_21820.914.png" ], "turn_id": 7, "scan_id": "Validation_45662921", "origin_question": "", "option": [], "answer": "", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_45662921/Validation_45662921_45662921_21822.913.png", "Validation_45662921/Validation_45662921_45662921_21824.929.png", "Validation_45662921/Validation_45662921_45662921_21826.928.png", "Validation_45662921/Validation_45662921_45662921_21828.927.png", "Validation_45662921/Validation_45662921_45662921_21830.926.png" ], "turn_id": 8, "scan_id": "Validation_45662921", "origin_question": "In which turn were you closest to \"the refined rectangular lightweight curtain in the neutral hue\"? ", "option": [], "answer": "8", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the refined rectangular lightweight curtain in the neutral hue\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "Validation_45662921/Validation_45662921_45662921_21832.926.png", "Validation_45662921/Validation_45662921_45662921_21834.925.png", "Validation_45662921/Validation_45662921_45662921_21836.924.png", "Validation_45662921/Validation_45662921_45662921_21838.923.png", "Validation_45662921/Validation_45662921_45662921_21840.922.png" ], "turn_id": 9, "scan_id": "Validation_45662921", "origin_question": "Remember, how many water cooler(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many water cooler(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_45662921/Validation_45662921_45662921_21842.921.png", "Validation_45662921/Validation_45662921_45662921_21844.921.png", "Validation_45662921/Validation_45662921_45662921_21846.920.png", "Validation_45662921/Validation_45662921_45662921_21848.919.png", "Validation_45662921/Validation_45662921_45662921_21850.918.png" ], "turn_id": 10, "scan_id": "Validation_45662921", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the sleek stainless steel stove with the silver-gray finish\" from your position?", "option": [], "answer": "117.82", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the sleek stainless steel stove with the silver-gray finish\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47430023/Validation_47430023_47430023_18859.957.png", "Validation_47430023/Validation_47430023_47430023_18861.939.png", "Validation_47430023/Validation_47430023_47430023_18863.938.png", "Validation_47430023/Validation_47430023_47430023_18865.937.png" ], "turn_id": 1, "scan_id": "Validation_47430023", "origin_question": "Remember, have you seen any cabinet(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any cabinet(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_47430023/Validation_47430023_47430023_18867.937.png", "Validation_47430023/Validation_47430023_47430023_18869.936.png", "Validation_47430023/Validation_47430023_47430023_18871.935.png", "Validation_47430023/Validation_47430023_47430023_18873.934.png" ], "turn_id": 2, "scan_id": "Validation_47430023", "origin_question": "When did you first discover \"the brown-tinted glass cylindrical bottle\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the brown-tinted glass cylindrical bottle\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47430023/Validation_47430023_47430023_18875.933.png", "Validation_47430023/Validation_47430023_47430023_18877.932.png", "Validation_47430023/Validation_47430023_47430023_18879.932.png", "Validation_47430023/Validation_47430023_47430023_18881.948.png" ], "turn_id": 3, "scan_id": "Validation_47430023", "origin_question": "Which direction is the brown-tinted glass cylindrical bottle in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the brown-tinted glass cylindrical bottle in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_47430023/Validation_47430023_47430023_18883.947.png", "Validation_47430023/Validation_47430023_47430023_18885.946.png", "Validation_47430023/Validation_47430023_47430023_18887.945.png", "Validation_47430023/Validation_47430023_47430023_18889.944.png" ], "turn_id": 4, "scan_id": "Validation_47430023", "origin_question": "Please recall \"the wooden statue with an elongated, dynamic form\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.64", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the wooden statue with an elongated, dynamic form\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_47430023/Validation_47430023_47430023_18891.944.png", "Validation_47430023/Validation_47430023_47430023_18893.943.png", "Validation_47430023/Validation_47430023_47430023_18895.942.png", "Validation_47430023/Validation_47430023_47430023_18897.941.png" ], "turn_id": 5, "scan_id": "Validation_47430023", "origin_question": "When did you first discover \"the large, glossy black monitor\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the large, glossy black monitor\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_47430023/Validation_47430023_47430023_18899.940.png", "Validation_47430023/Validation_47430023_47430023_18901.939.png", "Validation_47430023/Validation_47430023_47430023_18903.939.png", "Validation_47430023/Validation_47430023_47430023_18905.938.png" ], "turn_id": 6, "scan_id": "Validation_47430023", "origin_question": "What will be the first-time appearance order of the following categories: cabinet, door, monitor?", "option": [ "cabinet, monitor, door", "monitor, cabinet, door", "door, cabinet, monitor", "cabinet, door, monitor" ], "answer": "door, cabinet, monitor", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: cabinet, door, monitor? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['cabinet, monitor, door', 'monitor, cabinet, door', 'door, cabinet, monitor', 'cabinet, door, monitor'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_47430023/Validation_47430023_47430023_18907.937.png", "Validation_47430023/Validation_47430023_47430023_18909.936.png", "Validation_47430023/Validation_47430023_47430023_18911.935.png", "Validation_47430023/Validation_47430023_47430023_18913.934.png" ], "turn_id": 7, "scan_id": "Validation_47430023", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the large, glossy black monitor\" from your position?", "option": [], "answer": "172.56", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the large, glossy black monitor\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_47430023/Validation_47430023_47430023_18915.934.png", "Validation_47430023/Validation_47430023_47430023_18917.933.png", "Validation_47430023/Validation_47430023_47430023_18919.932.png", "Validation_47430023/Validation_47430023_47430023_18921.948.png" ], "turn_id": 8, "scan_id": "Validation_47430023", "origin_question": "Please recall \"the wooden statue with an elongated, dynamic form\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.48", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the wooden statue with an elongated, dynamic form\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42898508/Validation_42898508_42898508_749445.102.png", "Validation_42898508/Validation_42898508_42898508_749447.084.png", "Validation_42898508/Validation_42898508_42898508_749449.083.png", "Validation_42898508/Validation_42898508_42898508_749451.083.png", "Validation_42898508/Validation_42898508_42898508_749453.082.png" ], "turn_id": 1, "scan_id": "Validation_42898508", "origin_question": "Remember, have you seen any kettle(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any kettle(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_42898508/Validation_42898508_42898508_749455.081.png", "Validation_42898508/Validation_42898508_42898508_749457.080.png", "Validation_42898508/Validation_42898508_42898508_749459.079.png", "Validation_42898508/Validation_42898508_42898508_749461.078.png", "Validation_42898508/Validation_42898508_42898508_749463.078.png" ], "turn_id": 2, "scan_id": "Validation_42898508", "origin_question": "Which direction is the luxurious brown leather couch with broad armrests in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the luxurious brown leather couch with broad armrests in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_42898508/Validation_42898508_42898508_749465.077.png", "Validation_42898508/Validation_42898508_42898508_749467.076.png", "Validation_42898508/Validation_42898508_42898508_749469.092.png", "Validation_42898508/Validation_42898508_42898508_749471.091.png", "Validation_42898508/Validation_42898508_42898508_749473.090.png" ], "turn_id": 3, "scan_id": "Validation_42898508", "origin_question": "In which turn did you see both \"the luxurious brown leather couch with broad armrests\" and \"the medium-sized plant with vibrant green leaves\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the luxurious brown leather couch with broad armrests\" and \"the medium-sized plant with vibrant green leaves\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_42898508/Validation_42898508_42898508_749475.090.png", "Validation_42898508/Validation_42898508_42898508_749477.089.png", "Validation_42898508/Validation_42898508_42898508_749479.088.png", "Validation_42898508/Validation_42898508_42898508_749481.087.png", "Validation_42898508/Validation_42898508_42898508_749483.086.png" ], "turn_id": 4, "scan_id": "Validation_42898508", "origin_question": "Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "100.32", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_42898508/Validation_42898508_42898508_749485.085.png", "Validation_42898508/Validation_42898508_42898508_749487.084.png", "Validation_42898508/Validation_42898508_42898508_749489.084.png", "Validation_42898508/Validation_42898508_42898508_749491.083.png", "Validation_42898508/Validation_42898508_42898508_749493.082.png" ], "turn_id": 5, "scan_id": "Validation_42898508", "origin_question": "Using your orientation at the end of turn 4 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "157.43", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 4 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_42898508/Validation_42898508_42898508_749495.081.png", "Validation_42898508/Validation_42898508_42898508_749497.080.png", "Validation_42898508/Validation_42898508_42898508_749499.080.png", "Validation_42898508/Validation_42898508_42898508_749501.079.png", "Validation_42898508/Validation_42898508_42898508_749503.078.png" ], "turn_id": 6, "scan_id": "Validation_42898508", "origin_question": "Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_42898508/Validation_42898508_42898508_749505.077.png", "Validation_42898508/Validation_42898508_42898508_749507.076.png", "Validation_42898508/Validation_42898508_42898508_749509.092.png", "Validation_42898508/Validation_42898508_42898508_749511.091.png", "Validation_42898508/Validation_42898508_42898508_749513.091.png" ], "turn_id": 7, "scan_id": "Validation_42898508", "origin_question": "Please recall \"the medium-sized plant with vibrant green leaves\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.93", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the medium-sized plant with vibrant green leaves\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_42898508/Validation_42898508_42898508_749515.090.png", "Validation_42898508/Validation_42898508_42898508_749517.089.png", "Validation_42898508/Validation_42898508_42898508_749519.088.png", "Validation_42898508/Validation_42898508_42898508_749521.087.png", "Validation_42898508/Validation_42898508_42898508_749523.086.png" ], "turn_id": 8, "scan_id": "Validation_42898508", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "2.38", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_42898508/Validation_42898508_42898508_749525.086.png", "Validation_42898508/Validation_42898508_42898508_749527.085.png", "Validation_42898508/Validation_42898508_42898508_749529.084.png", "Validation_42898508/Validation_42898508_42898508_749531.083.png", "Validation_42898508/Validation_42898508_42898508_749533.082.png" ], "turn_id": 9, "scan_id": "Validation_42898508", "origin_question": "By the end of the 6 turn, were you getting closer to or farther away from these two objects: X: \"the medium-sized plant with vibrant green leaves\", Y: \"the large rectangular white window with sliding function\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 6 turn, were you getting closer to or farther away from these two objects: X: \"the medium-sized plant with vibrant green leaves\", Y: \"the large rectangular white window with sliding function\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_42898508/Validation_42898508_42898508_749535.081.png", "Validation_42898508/Validation_42898508_42898508_749537.081.png", "Validation_42898508/Validation_42898508_42898508_749539.080.png", "Validation_42898508/Validation_42898508_42898508_749541.079.png", "Validation_42898508/Validation_42898508_42898508_749543.078.png" ], "turn_id": 10, "scan_id": "Validation_42898508", "origin_question": "When did you first discover \"the medium-sized plant with vibrant green leaves\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the medium-sized plant with vibrant green leaves\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47333927/Validation_47333927_47333927_83990.218.png", "Validation_47333927/Validation_47333927_47333927_83992.201.png", "Validation_47333927/Validation_47333927_47333927_83994.200.png", "Validation_47333927/Validation_47333927_47333927_83996.199.png" ], "turn_id": 1, "scan_id": "Validation_47333927", "origin_question": "Remember, have you seen any detergent(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any detergent(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47333927/Validation_47333927_47333927_83998.198.png", "Validation_47333927/Validation_47333927_47333927_84000.197.png", "Validation_47333927/Validation_47333927_47333927_84002.213.png", "Validation_47333927/Validation_47333927_47333927_84004.212.png" ], "turn_id": 2, "scan_id": "Validation_47333927", "origin_question": "Remember, have you seen any soap(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any soap(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_47333927/Validation_47333927_47333927_84006.212.png", "Validation_47333927/Validation_47333927_47333927_84008.211.png", "Validation_47333927/Validation_47333927_47333927_84010.210.png", "Validation_47333927/Validation_47333927_47333927_84012.209.png" ], "turn_id": 3, "scan_id": "Validation_47333927", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the tall, glossy cylindrical glass vase\" from your position?", "option": [], "answer": "104.29", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the tall, glossy cylindrical glass vase\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_47333927/Validation_47333927_47333927_84014.208.png", "Validation_47333927/Validation_47333927_47333927_84016.207.png", "Validation_47333927/Validation_47333927_47333927_84018.207.png", "Validation_47333927/Validation_47333927_47333927_84020.206.png" ], "turn_id": 4, "scan_id": "Validation_47333927", "origin_question": "Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "72.24", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_47333927/Validation_47333927_47333927_84022.205.png", "Validation_47333927/Validation_47333927_47333927_84024.204.png", "Validation_47333927/Validation_47333927_47333927_84026.203.png", "Validation_47333927/Validation_47333927_47333927_84028.203.png" ], "turn_id": 5, "scan_id": "Validation_47333927", "origin_question": "Please recall \"an elegant ivory ceramic kettle with the traditional shape\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.67", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"an elegant ivory ceramic kettle with the traditional shape\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47333927/Validation_47333927_47333927_84030.202.png", "Validation_47333927/Validation_47333927_47333927_84032.201.png", "Validation_47333927/Validation_47333927_47333927_84034.200.png", "Validation_47333927/Validation_47333927_47333927_84036.199.png" ], "turn_id": 6, "scan_id": "Validation_47333927", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the compact gray cylindrical stool made of plastic\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the compact gray cylindrical stool made of plastic\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "Validation_47333927/Validation_47333927_47333927_84038.198.png", "Validation_47333927/Validation_47333927_47333927_84040.198.png", "Validation_47333927/Validation_47333927_47333927_84042.197.png", "Validation_47333927/Validation_47333927_47333927_84044.196.png" ], "turn_id": 7, "scan_id": "Validation_47333927", "origin_question": "Remember, how many candle(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many candle(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_47333927/Validation_47333927_47333927_84046.212.png", "Validation_47333927/Validation_47333927_47333927_84048.211.png", "Validation_47333927/Validation_47333927_47333927_84050.210.png", "Validation_47333927/Validation_47333927_47333927_84052.209.png" ], "turn_id": 8, "scan_id": "Validation_47333927", "origin_question": "What will be the first-time appearance order of the following categories: door, vase, window?", "option": [ "vase, door, window", "vase, window, door", "door, vase, window", "door, window, vase" ], "answer": "door, vase, window", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: door, vase, window? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['vase, door, window', 'vase, window, door', 'door, vase, window', 'door, window, vase'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42899726/Validation_42899726_42899726_799531.772.png", "Validation_42899726/Validation_42899726_42899726_799533.255.png", "Validation_42899726/Validation_42899726_42899726_799534.754.png", "Validation_42899726/Validation_42899726_42899726_799536.253.png" ], "turn_id": 1, "scan_id": "Validation_42899726", "origin_question": "Remember, have you seen any sink(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any sink(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_42899726/Validation_42899726_42899726_799537.753.png", "Validation_42899726/Validation_42899726_42899726_799539.252.png", "Validation_42899726/Validation_42899726_42899726_799540.752.png", "Validation_42899726/Validation_42899726_42899726_799542.251.png" ], "turn_id": 2, "scan_id": "Validation_42899726", "origin_question": "In which turn did you see both \"the sleek white ceramic rectangular sink\" and \"the bathtub\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the sleek white ceramic rectangular sink\" and \"the bathtub\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_42899726/Validation_42899726_42899726_799543.750.png", "Validation_42899726/Validation_42899726_42899726_799545.250.png", "Validation_42899726/Validation_42899726_42899726_799546.749.png", "Validation_42899726/Validation_42899726_42899726_799548.248.png" ], "turn_id": 3, "scan_id": "Validation_42899726", "origin_question": "In which turn did you see both \"the bathtub\" and \"the sleek, rounded, wall-mounted toilet in white ceramic\" simultaneously? ", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the bathtub\" and \"the sleek, rounded, wall-mounted toilet in white ceramic\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_42899726/Validation_42899726_42899726_799549.748.png", "Validation_42899726/Validation_42899726_42899726_799551.247.png", "Validation_42899726/Validation_42899726_42899726_799552.747.png", "Validation_42899726/Validation_42899726_42899726_799554.246.png" ], "turn_id": 4, "scan_id": "Validation_42899726", "origin_question": "In which turn did you see both \"the bathtub\" and \"the sleek metallic cylindrical bin\" simultaneously? ", "option": [], "answer": "3", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the bathtub\" and \"the sleek metallic cylindrical bin\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_42899726/Validation_42899726_42899726_799555.745.png", "Validation_42899726/Validation_42899726_42899726_799557.262.png", "Validation_42899726/Validation_42899726_42899726_799558.761.png", "Validation_42899726/Validation_42899726_42899726_799560.260.png" ], "turn_id": 5, "scan_id": "Validation_42899726", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the bathtub\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the bathtub\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_42899726/Validation_42899726_42899726_799561.760.png", "Validation_42899726/Validation_42899726_42899726_799563.259.png", "Validation_42899726/Validation_42899726_42899726_799564.758.png", "Validation_42899726/Validation_42899726_42899726_799566.258.png" ], "turn_id": 6, "scan_id": "Validation_42899726", "origin_question": "Please recall \"the sleek white ceramic rectangular sink\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.8", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the sleek white ceramic rectangular sink\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42899696/Validation_42899696_42899696_792367.199.png", "Validation_42899696/Validation_42899696_42899696_792369.181.png", "Validation_42899696/Validation_42899696_42899696_792371.181.png", "Validation_42899696/Validation_42899696_42899696_792373.180.png" ], "turn_id": 1, "scan_id": "Validation_42899696", "origin_question": "Remember, have you seen any stove(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any stove(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_42899696/Validation_42899696_42899696_792375.179.png", "Validation_42899696/Validation_42899696_42899696_792377.178.png", "Validation_42899696/Validation_42899696_42899696_792379.177.png", "Validation_42899696/Validation_42899696_42899696_792381.176.png" ], "turn_id": 2, "scan_id": "Validation_42899696", "origin_question": "When did you first discover \"the cylindrical metallic silver bin\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the cylindrical metallic silver bin\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_42899696/Validation_42899696_42899696_792383.176.png", "Validation_42899696/Validation_42899696_42899696_792385.175.png", "Validation_42899696/Validation_42899696_42899696_792387.174.png", "Validation_42899696/Validation_42899696_42899696_792389.173.png" ], "turn_id": 3, "scan_id": "Validation_42899696", "origin_question": "What will be the first-time appearance order of the following categories: bed, cabinet, table?", "option": [ "bed, cabinet, table", "table, bed, cabinet", "table, cabinet, bed", "cabinet, bed, table" ], "answer": "table, cabinet, bed", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bed, cabinet, table? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['bed, cabinet, table', 'table, bed, cabinet', 'table, cabinet, bed', 'cabinet, bed, table'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_42899696/Validation_42899696_42899696_792391.189.png", "Validation_42899696/Validation_42899696_42899696_792393.188.png", "Validation_42899696/Validation_42899696_42899696_792395.187.png", "Validation_42899696/Validation_42899696_42899696_792397.186.png" ], "turn_id": 4, "scan_id": "Validation_42899696", "origin_question": "Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "Validation_42899696/Validation_42899696_42899696_792399.186.png", "Validation_42899696/Validation_42899696_42899696_792401.185.png", "Validation_42899696/Validation_42899696_42899696_792403.184.png", "Validation_42899696/Validation_42899696_42899696_792405.183.png" ], "turn_id": 5, "scan_id": "Validation_42899696", "origin_question": "Among these three objects(\"the cylindrical metallic silver bin\",\"the sleek black rectangular monitor with slim bezels\",\"the large rectangular mirror with the sleek black frame\" ), which one is the closest to you?", "option": [ "the cylindrical metallic silver bin", "the sleek black rectangular monitor with slim bezels", "the large rectangular mirror with the sleek black frame" ], "answer": "the large rectangular mirror with the sleek black frame", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects(\"the cylindrical metallic silver bin\",\"the sleek black rectangular monitor with slim bezels\",\"the large rectangular mirror with the sleek black frame\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the cylindrical metallic silver bin', 'the sleek black rectangular monitor with slim bezels', 'the large rectangular mirror with the sleek black frame'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_42899696/Validation_42899696_42899696_792407.182.png", "Validation_42899696/Validation_42899696_42899696_792409.181.png", "Validation_42899696/Validation_42899696_42899696_792411.181.png", "Validation_42899696/Validation_42899696_42899696_792413.180.png" ], "turn_id": 6, "scan_id": "Validation_42899696", "origin_question": "At the end of which turn were both of \"the cylindrical metallic silver bin\" and \"the stylish, tufted gray bed with the striped duvet\" on your left side?", "option": [], "answer": "6", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the cylindrical metallic silver bin\" and \"the stylish, tufted gray bed with the striped duvet\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_42899696/Validation_42899696_42899696_792415.179.png", "Validation_42899696/Validation_42899696_42899696_792417.178.png", "Validation_42899696/Validation_42899696_42899696_792419.177.png", "Validation_42899696/Validation_42899696_42899696_792421.176.png" ], "turn_id": 7, "scan_id": "Validation_42899696", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_42899696/Validation_42899696_42899696_792423.176.png", "Validation_42899696/Validation_42899696_42899696_792425.175.png", "Validation_42899696/Validation_42899696_42899696_792427.174.png", "Validation_42899696/Validation_42899696_42899696_792429.173.png" ], "turn_id": 8, "scan_id": "Validation_42899696", "origin_question": "Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance forward or backward from that position?", "option": [ "Forward", "Backward" ], "answer": "Forward", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance forward or backward from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Forward', 'Backward'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_48018957/Validation_48018957_48018957_556503.860.png", "Validation_48018957/Validation_48018957_48018957_556505.343.png", "Validation_48018957/Validation_48018957_48018957_556506.842.png", "Validation_48018957/Validation_48018957_48018957_556508.358.png" ], "turn_id": 1, "scan_id": "Validation_48018957", "origin_question": "Remember, have you seen any fence(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any fence(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_48018957/Validation_48018957_48018957_556509.858.png", "Validation_48018957/Validation_48018957_48018957_556511.357.png", "Validation_48018957/Validation_48018957_48018957_556512.856.png", "Validation_48018957/Validation_48018957_48018957_556514.356.png" ], "turn_id": 2, "scan_id": "Validation_48018957", "origin_question": "In which turn were you closest to \"the plush, dark rectangular carpet with the shaggy texture\"? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the plush, dark rectangular carpet with the shaggy texture\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_48018957/Validation_48018957_48018957_556515.855.png", "Validation_48018957/Validation_48018957_48018957_556517.354.png", "Validation_48018957/Validation_48018957_48018957_556518.854.png", "Validation_48018957/Validation_48018957_48018957_556520.353.png" ], "turn_id": 3, "scan_id": "Validation_48018957", "origin_question": "", "option": [], "answer": "", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_48018957/Validation_48018957_48018957_556521.853.png", "Validation_48018957/Validation_48018957_48018957_556523.352.png", "Validation_48018957/Validation_48018957_48018957_556524.851.png", "Validation_48018957/Validation_48018957_48018957_556526.351.png" ], "turn_id": 4, "scan_id": "Validation_48018957", "origin_question": "In which turn were you farthest from \"the stylish front-loading washing machine with the polished white surface\"? ", "option": [], "answer": "1", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the stylish front-loading washing machine with the polished white surface\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_48018360/Validation_48018360_48018360_389156.371.png", "Validation_48018360/Validation_48018360_48018360_389157.854.png", "Validation_48018360/Validation_48018360_48018360_389159.353.png", "Validation_48018360/Validation_48018360_48018360_389160.853.png" ], "turn_id": 1, "scan_id": "Validation_48018360", "origin_question": "Remember, have you seen any poster(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any poster(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_48018360/Validation_48018360_48018360_389162.352.png", "Validation_48018360/Validation_48018360_48018360_389163.851.png", "Validation_48018360/Validation_48018360_48018360_389165.351.png", "Validation_48018360/Validation_48018360_48018360_389166.850.png" ], "turn_id": 2, "scan_id": "Validation_48018360", "origin_question": "At the end of which turn were both of \"the single-sized bed with checkered blue and red bedding\" and \"the window\" on your left side?", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the single-sized bed with checkered blue and red bedding\" and \"the window\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_48018360/Validation_48018360_48018360_389168.350.png", "Validation_48018360/Validation_48018360_48018360_389169.849.png", "Validation_48018360/Validation_48018360_48018360_389171.348.png", "Validation_48018360/Validation_48018360_48018360_389172.848.png" ], "turn_id": 3, "scan_id": "Validation_48018360", "origin_question": "When did you first discover \"the window\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the window\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_48018360/Validation_48018360_48018360_389174.347.png", "Validation_48018360/Validation_48018360_48018360_389175.846.png", "Validation_48018360/Validation_48018360_48018360_389177.346.png", "Validation_48018360/Validation_48018360_48018360_389178.845.png" ], "turn_id": 4, "scan_id": "Validation_48018360", "origin_question": "Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "154.89", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_48018360/Validation_48018360_48018360_389180.345.png", "Validation_48018360/Validation_48018360_48018360_389181.844.png", "Validation_48018360/Validation_48018360_48018360_389183.343.png", "Validation_48018360/Validation_48018360_48018360_389184.843.png" ], "turn_id": 5, "scan_id": "Validation_48018360", "origin_question": "Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_48018360/Validation_48018360_48018360_389186.342.png", "Validation_48018360/Validation_48018360_48018360_389187.842.png", "Validation_48018360/Validation_48018360_48018360_389189.341.png", "Validation_48018360/Validation_48018360_48018360_389190.840.png" ], "turn_id": 6, "scan_id": "Validation_48018360", "origin_question": "In which turn did you see both \"the metallic gold rectangular frame with the polished finish\" and \"the contemporary drum-shaped white metal and fabric light\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the metallic gold rectangular frame with the polished finish\" and \"the contemporary drum-shaped white metal and fabric light\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47334096/Validation_47334096_47334096_91955.452.png", "Validation_47334096/Validation_47334096_47334096_91957.435.png", "Validation_47334096/Validation_47334096_47334096_91959.434.png", "Validation_47334096/Validation_47334096_47334096_91961.433.png", "Validation_47334096/Validation_47334096_47334096_91963.432.png" ], "turn_id": 1, "scan_id": "Validation_47334096", "origin_question": "Remember, have you seen any chair(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any chair(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_47334096/Validation_47334096_47334096_91965.432.png", "Validation_47334096/Validation_47334096_47334096_91967.431.png", "Validation_47334096/Validation_47334096_47334096_91969.430.png", "Validation_47334096/Validation_47334096_47334096_91971.429.png", "Validation_47334096/Validation_47334096_47334096_91973.428.png" ], "turn_id": 2, "scan_id": "Validation_47334096", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the metallic silver rectangular microwave with the sleek design\", Y: \"the medium-sized round plastic flowerpot in the neutral color\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the metallic silver rectangular microwave with the sleek design\", Y: \"the medium-sized round plastic flowerpot in the neutral color\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47334096/Validation_47334096_47334096_91975.428.png", "Validation_47334096/Validation_47334096_47334096_91977.427.png", "Validation_47334096/Validation_47334096_47334096_91979.426.png", "Validation_47334096/Validation_47334096_47334096_91981.425.png", "Validation_47334096/Validation_47334096_47334096_91983.424.png" ], "turn_id": 3, "scan_id": "Validation_47334096", "origin_question": "In which turn were you closest to \"the metallic silver rectangular microwave with the sleek design\"? ", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the metallic silver rectangular microwave with the sleek design\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47334096/Validation_47334096_47334096_91985.424.png", "Validation_47334096/Validation_47334096_47334096_91987.439.png", "Validation_47334096/Validation_47334096_47334096_91989.439.png", "Validation_47334096/Validation_47334096_47334096_91991.438.png", "Validation_47334096/Validation_47334096_47334096_91993.437.png" ], "turn_id": 4, "scan_id": "Validation_47334096", "origin_question": "Remember, have you seen any boots(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any boots(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_47334096/Validation_47334096_47334096_91995.436.png", "Validation_47334096/Validation_47334096_47334096_91997.436.png", "Validation_47334096/Validation_47334096_47334096_91999.435.png", "Validation_47334096/Validation_47334096_47334096_92001.434.png", "Validation_47334096/Validation_47334096_47334096_92003.433.png" ], "turn_id": 5, "scan_id": "Validation_47334096", "origin_question": "What will be the first-time appearance order of the following categories: chair, sink, window?", "option": [ "chair, sink, window", "window, sink, chair", "sink, chair, window", "chair, window, sink" ], "answer": "sink, chair, window", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: chair, sink, window? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['chair, sink, window', 'window, sink, chair', 'sink, chair, window', 'chair, window, sink'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_47334096/Validation_47334096_47334096_92005.432.png", "Validation_47334096/Validation_47334096_47334096_92007.431.png", "Validation_47334096/Validation_47334096_47334096_92009.431.png", "Validation_47334096/Validation_47334096_47334096_92011.430.png", "Validation_47334096/Validation_47334096_47334096_92013.429.png" ], "turn_id": 6, "scan_id": "Validation_47334096", "origin_question": "In which turn did you see both \"the metallic silver rectangular microwave with the sleek design\" and \"the contemporary table with the glass top and dark wood frame\" simultaneously? ", "option": [], "answer": "6", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the metallic silver rectangular microwave with the sleek design\" and \"the contemporary table with the glass top and dark wood frame\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47334096/Validation_47334096_47334096_92015.428.png", "Validation_47334096/Validation_47334096_47334096_92017.427.png", "Validation_47334096/Validation_47334096_47334096_92019.427.png", "Validation_47334096/Validation_47334096_47334096_92021.426.png", "Validation_47334096/Validation_47334096_47334096_92023.425.png" ], "turn_id": 7, "scan_id": "Validation_47334096", "origin_question": "Which direction is the modern black stove with the glossy ceramic surface in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the modern black stove with the glossy ceramic surface in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_47334096/Validation_47334096_47334096_92025.424.png", "Validation_47334096/Validation_47334096_47334096_92027.440.png", "Validation_47334096/Validation_47334096_47334096_92029.439.png", "Validation_47334096/Validation_47334096_47334096_92031.438.png", "Validation_47334096/Validation_47334096_47334096_92033.438.png" ], "turn_id": 8, "scan_id": "Validation_47334096", "origin_question": "When did you first discover \"the modern black stove with the glossy ceramic surface\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the modern black stove with the glossy ceramic surface\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_47334096/Validation_47334096_47334096_92035.437.png", "Validation_47334096/Validation_47334096_47334096_92037.436.png", "Validation_47334096/Validation_47334096_47334096_92039.435.png", "Validation_47334096/Validation_47334096_47334096_92041.434.png", "Validation_47334096/Validation_47334096_47334096_92043.433.png" ], "turn_id": 9, "scan_id": "Validation_47334096", "origin_question": "Using your orientation at the end of turn 4 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "120.09", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 4 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47334096/Validation_47334096_47334096_92045.433.png", "Validation_47334096/Validation_47334096_47334096_92047.432.png", "Validation_47334096/Validation_47334096_47334096_92049.431.png", "Validation_47334096/Validation_47334096_47334096_92051.430.png", "Validation_47334096/Validation_47334096_47334096_92053.429.png" ], "turn_id": 10, "scan_id": "Validation_47334096", "origin_question": "", "option": [], "answer": "", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47332893/Validation_47332893_47332893_9187.706.png", "Validation_47332893/Validation_47332893_47332893_9189.688.png", "Validation_47332893/Validation_47332893_47332893_9191.687.png", "Validation_47332893/Validation_47332893_47332893_9193.686.png", "Validation_47332893/Validation_47332893_47332893_9195.686.png" ], "turn_id": 1, "scan_id": "Validation_47332893", "origin_question": "Remember, have you seen any curtain(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any curtain(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_47332893/Validation_47332893_47332893_9197.685.png", "Validation_47332893/Validation_47332893_47332893_9199.684.png", "Validation_47332893/Validation_47332893_47332893_9201.683.png", "Validation_47332893/Validation_47332893_47332893_9203.699.png", "Validation_47332893/Validation_47332893_47332893_9205.698.png" ], "turn_id": 2, "scan_id": "Validation_47332893", "origin_question": "Is the deep red smooth polyester curtain to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the deep red smooth polyester curtain to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47332893/Validation_47332893_47332893_9207.698.png", "Validation_47332893/Validation_47332893_47332893_9209.697.png", "Validation_47332893/Validation_47332893_47332893_9211.696.png", "Validation_47332893/Validation_47332893_47332893_9213.695.png", "Validation_47332893/Validation_47332893_47332893_9215.694.png" ], "turn_id": 3, "scan_id": "Validation_47332893", "origin_question": "Which direction is the deep red smooth polyester curtain in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the deep red smooth polyester curtain in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_47332893/Validation_47332893_47332893_9217.693.png", "Validation_47332893/Validation_47332893_47332893_9219.693.png", "Validation_47332893/Validation_47332893_47332893_9221.692.png", "Validation_47332893/Validation_47332893_47332893_9223.691.png", "Validation_47332893/Validation_47332893_47332893_9225.690.png" ], "turn_id": 4, "scan_id": "Validation_47332893", "origin_question": "Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_47332893/Validation_47332893_47332893_9227.689.png", "Validation_47332893/Validation_47332893_47332893_9229.689.png", "Validation_47332893/Validation_47332893_47332893_9231.688.png", "Validation_47332893/Validation_47332893_47332893_9233.687.png", "Validation_47332893/Validation_47332893_47332893_9235.686.png" ], "turn_id": 5, "scan_id": "Validation_47332893", "origin_question": "Please recall \"the deep red smooth polyester curtain\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.77", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the deep red smooth polyester curtain\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47332893/Validation_47332893_47332893_9237.685.png", "Validation_47332893/Validation_47332893_47332893_9239.685.png", "Validation_47332893/Validation_47332893_47332893_9241.684.png", "Validation_47332893/Validation_47332893_47332893_9243.683.png", "Validation_47332893/Validation_47332893_47332893_9245.699.png" ], "turn_id": 6, "scan_id": "Validation_47332893", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the vibrant green bed with playful design\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the vibrant green bed with playful design\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_47332893/Validation_47332893_47332893_9247.698.png", "Validation_47332893/Validation_47332893_47332893_9249.697.png", "Validation_47332893/Validation_47332893_47332893_9251.696.png", "Validation_47332893/Validation_47332893_47332893_9253.696.png", "Validation_47332893/Validation_47332893_47332893_9255.695.png" ], "turn_id": 7, "scan_id": "Validation_47332893", "origin_question": "Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance forward or backward from that position?", "option": [ "Forward", "Backward" ], "answer": "Backward", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance forward or backward from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Forward', 'Backward'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_47332893/Validation_47332893_47332893_9257.694.png", "Validation_47332893/Validation_47332893_47332893_9259.693.png", "Validation_47332893/Validation_47332893_47332893_9261.692.png", "Validation_47332893/Validation_47332893_47332893_9263.692.png", "Validation_47332893/Validation_47332893_47332893_9265.691.png" ], "turn_id": 8, "scan_id": "Validation_47332893", "origin_question": "Is the colorful cardboard box with vivid imagery to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the colorful cardboard box with vivid imagery to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_47332893/Validation_47332893_47332893_9267.690.png", "Validation_47332893/Validation_47332893_47332893_9269.689.png", "Validation_47332893/Validation_47332893_47332893_9271.688.png", "Validation_47332893/Validation_47332893_47332893_9273.687.png", "Validation_47332893/Validation_47332893_47332893_9275.687.png" ], "turn_id": 9, "scan_id": "Validation_47332893", "origin_question": "At the end of which turn were both of \"the sleek white steel radiator with the minimalist design\" and \"the modern, sleek monitor with the slim black bezel\" on your right side?", "option": [], "answer": "8", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the sleek white steel radiator with the minimalist design\" and \"the modern, sleek monitor with the slim black bezel\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_47332893/Validation_47332893_47332893_9277.686.png", "Validation_47332893/Validation_47332893_47332893_9279.685.png", "Validation_47332893/Validation_47332893_47332893_9281.684.png", "Validation_47332893/Validation_47332893_47332893_9283.683.png", "Validation_47332893/Validation_47332893_47332893_9285.699.png" ], "turn_id": 10, "scan_id": "Validation_47332893", "origin_question": "By the end of the 3 turn, were you getting closer to or farther away from these two objects: X: \"the vibrant green bed with playful design\", Y: \"the modern, sleek monitor with the slim black bezel\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 3 turn, were you getting closer to or farther away from these two objects: X: \"the vibrant green bed with playful design\", Y: \"the modern, sleek monitor with the slim black bezel\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47204605/Validation_47204605_47204605_2403.316.png", "Validation_47204605/Validation_47204605_47204605_2404.798.png", "Validation_47204605/Validation_47204605_47204605_2406.298.png", "Validation_47204605/Validation_47204605_47204605_2407.797.png" ], "turn_id": 1, "scan_id": "Validation_47204605", "origin_question": "Remember, have you seen any table(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any table(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47204605/Validation_47204605_47204605_2409.297.png", "Validation_47204605/Validation_47204605_47204605_2410.796.png", "Validation_47204605/Validation_47204605_47204605_2412.295.png", "Validation_47204605/Validation_47204605_47204605_2413.795.png" ], "turn_id": 2, "scan_id": "Validation_47204605", "origin_question": "In which turn were you farthest from \"the large rectangular black flat-screen monitor\"? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the large rectangular black flat-screen monitor\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47204605/Validation_47204605_47204605_2415.294.png", "Validation_47204605/Validation_47204605_47204605_2416.794.png", "Validation_47204605/Validation_47204605_47204605_2418.293.png", "Validation_47204605/Validation_47204605_47204605_2419.792.png" ], "turn_id": 3, "scan_id": "Validation_47204605", "origin_question": "In which turn were you farthest from \"the carpet\"? ", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the carpet\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_47204605/Validation_47204605_47204605_2421.292.png", "Validation_47204605/Validation_47204605_47204605_2422.791.png", "Validation_47204605/Validation_47204605_47204605_2424.291.png", "Validation_47204605/Validation_47204605_47204605_2425.790.png" ], "turn_id": 4, "scan_id": "Validation_47204605", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_47204605/Validation_47204605_47204605_2427.289.png", "Validation_47204605/Validation_47204605_47204605_2428.789.png", "Validation_47204605/Validation_47204605_47204605_2430.305.png", "Validation_47204605/Validation_47204605_47204605_2431.804.png" ], "turn_id": 5, "scan_id": "Validation_47204605", "origin_question": "At the end of which turn were both of \"the large rectangular black flat-screen monitor\" and \"the carpet\" on your right side?", "option": [], "answer": "5", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the large rectangular black flat-screen monitor\" and \"the carpet\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_47204605/Validation_47204605_47204605_2433.304.png", "Validation_47204605/Validation_47204605_47204605_2434.803.png", "Validation_47204605/Validation_47204605_47204605_2436.303.png", "Validation_47204605/Validation_47204605_47204605_2437.802.png" ], "turn_id": 6, "scan_id": "Validation_47204605", "origin_question": "At the end of which turn were both of \"the large rectangular black flat-screen monitor\" and \"the carpet\" on your left side?", "option": [], "answer": "6", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the large rectangular black flat-screen monitor\" and \"the carpet\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_48018356/Validation_48018356_48018356_388771.364.png", "Validation_48018356/Validation_48018356_48018356_388773.346.png", "Validation_48018356/Validation_48018356_48018356_388775.345.png", "Validation_48018356/Validation_48018356_48018356_388777.345.png" ], "turn_id": 1, "scan_id": "Validation_48018356", "origin_question": "Remember, have you seen any tablet(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any tablet(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_48018356/Validation_48018356_48018356_388779.344.png", "Validation_48018356/Validation_48018356_48018356_388781.343.png", "Validation_48018356/Validation_48018356_48018356_388783.342.png", "Validation_48018356/Validation_48018356_48018356_388785.341.png" ], "turn_id": 2, "scan_id": "Validation_48018356", "origin_question": "In which turn were you closest to \"the sleek black metal chair with the minimalist design\"? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the sleek black metal chair with the minimalist design\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_48018356/Validation_48018356_48018356_388787.341.png", "Validation_48018356/Validation_48018356_48018356_388789.340.png", "Validation_48018356/Validation_48018356_48018356_388791.339.png", "Validation_48018356/Validation_48018356_48018356_388793.338.png" ], "turn_id": 3, "scan_id": "Validation_48018356", "origin_question": "In which turn did you see both \"the window\" and \"the brownish-gray clothes with orange accents\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the window\" and \"the brownish-gray clothes with orange accents\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_48018356/Validation_48018356_48018356_388795.337.png", "Validation_48018356/Validation_48018356_48018356_388797.336.png", "Validation_48018356/Validation_48018356_48018356_388799.336.png", "Validation_48018356/Validation_48018356_48018356_388801.335.png" ], "turn_id": 4, "scan_id": "Validation_48018356", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the window\" from your position?", "option": [], "answer": "143.57", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the window\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_48018356/Validation_48018356_48018356_388803.334.png", "Validation_48018356/Validation_48018356_48018356_388805.333.png", "Validation_48018356/Validation_48018356_48018356_388807.332.png", "Validation_48018356/Validation_48018356_48018356_388809.348.png" ], "turn_id": 5, "scan_id": "Validation_48018356", "origin_question": "Is the window to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the window to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_48018356/Validation_48018356_48018356_388811.347.png", "Validation_48018356/Validation_48018356_48018356_388813.347.png", "Validation_48018356/Validation_48018356_48018356_388815.346.png", "Validation_48018356/Validation_48018356_48018356_388817.345.png" ], "turn_id": 6, "scan_id": "Validation_48018356", "origin_question": "What will be the first-time appearance order of the following categories: clothes, curtain, monitor?", "option": [ "curtain, clothes, monitor", "clothes, monitor, curtain", "monitor, clothes, curtain", "clothes, curtain, monitor" ], "answer": "curtain, clothes, monitor", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: clothes, curtain, monitor? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['curtain, clothes, monitor', 'clothes, monitor, curtain', 'monitor, clothes, curtain', 'clothes, curtain, monitor'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_48018356/Validation_48018356_48018356_388819.344.png", "Validation_48018356/Validation_48018356_48018356_388821.343.png", "Validation_48018356/Validation_48018356_48018356_388823.342.png", "Validation_48018356/Validation_48018356_48018356_388825.342.png" ], "turn_id": 7, "scan_id": "Validation_48018356", "origin_question": "When did you first discover \"the sleek black metal chair with the minimalist design\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the sleek black metal chair with the minimalist design\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47430485/Validation_47430485_47430485_313828.188.png", "Validation_47430485/Validation_47430485_47430485_313830.171.png", "Validation_47430485/Validation_47430485_47430485_313832.170.png", "Validation_47430485/Validation_47430485_47430485_313834.169.png", "Validation_47430485/Validation_47430485_47430485_313836.168.png" ], "turn_id": 1, "scan_id": "Validation_47430485", "origin_question": "Remember, have you seen any book(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any book(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_47430485/Validation_47430485_47430485_313838.168.png", "Validation_47430485/Validation_47430485_47430485_313840.167.png", "Validation_47430485/Validation_47430485_47430485_313842.183.png", "Validation_47430485/Validation_47430485_47430485_313844.182.png", "Validation_47430485/Validation_47430485_47430485_313846.181.png" ], "turn_id": 2, "scan_id": "Validation_47430485", "origin_question": "In which turn did you see both \"the modern gray couch with soft fabric and cushions\" and \"the modern white shelf with open cubed compartments\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the modern gray couch with soft fabric and cushions\" and \"the modern white shelf with open cubed compartments\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_47430485/Validation_47430485_47430485_313848.180.png", "Validation_47430485/Validation_47430485_47430485_313850.179.png", "Validation_47430485/Validation_47430485_47430485_313852.178.png", "Validation_47430485/Validation_47430485_47430485_313854.178.png", "Validation_47430485/Validation_47430485_47430485_313856.177.png" ], "turn_id": 3, "scan_id": "Validation_47430485", "origin_question": "At the end of which turn were both of \"the modern gray couch with soft fabric and cushions\" and \"the vibrant dinosaur toy with red and yellow colors\" on your left side?", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the modern gray couch with soft fabric and cushions\" and \"the vibrant dinosaur toy with red and yellow colors\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47430485/Validation_47430485_47430485_313858.176.png", "Validation_47430485/Validation_47430485_47430485_313860.175.png", "Validation_47430485/Validation_47430485_47430485_313862.175.png", "Validation_47430485/Validation_47430485_47430485_313864.174.png", "Validation_47430485/Validation_47430485_47430485_313866.173.png" ], "turn_id": 4, "scan_id": "Validation_47430485", "origin_question": "Which direction is the vibrant dinosaur toy with red and yellow colors in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the vibrant dinosaur toy with red and yellow colors in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_47430485/Validation_47430485_47430485_313868.172.png", "Validation_47430485/Validation_47430485_47430485_313870.171.png", "Validation_47430485/Validation_47430485_47430485_313872.170.png", "Validation_47430485/Validation_47430485_47430485_313874.170.png", "Validation_47430485/Validation_47430485_47430485_313876.169.png" ], "turn_id": 5, "scan_id": "Validation_47430485", "origin_question": "Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47430485/Validation_47430485_47430485_313878.168.png", "Validation_47430485/Validation_47430485_47430485_313880.167.png", "Validation_47430485/Validation_47430485_47430485_313882.166.png", "Validation_47430485/Validation_47430485_47430485_313884.182.png", "Validation_47430485/Validation_47430485_47430485_313886.181.png" ], "turn_id": 6, "scan_id": "Validation_47430485", "origin_question": "Which direction is the small white rectangular table with the smooth surface in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the small white rectangular table with the smooth surface in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47430485/Validation_47430485_47430485_313888.181.png", "Validation_47430485/Validation_47430485_47430485_313890.180.png", "Validation_47430485/Validation_47430485_47430485_313892.179.png", "Validation_47430485/Validation_47430485_47430485_313894.178.png", "Validation_47430485/Validation_47430485_47430485_313896.177.png" ], "turn_id": 7, "scan_id": "Validation_47430485", "origin_question": "", "option": [], "answer": "", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_47430485/Validation_47430485_47430485_313898.176.png", "Validation_47430485/Validation_47430485_47430485_313900.176.png", "Validation_47430485/Validation_47430485_47430485_313902.175.png", "Validation_47430485/Validation_47430485_47430485_313904.174.png", "Validation_47430485/Validation_47430485_47430485_313906.173.png" ], "turn_id": 8, "scan_id": "Validation_47430485", "origin_question": "Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "98.06", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_47430485/Validation_47430485_47430485_313908.172.png", "Validation_47430485/Validation_47430485_47430485_313910.171.png", "Validation_47430485/Validation_47430485_47430485_313912.171.png", "Validation_47430485/Validation_47430485_47430485_313914.170.png", "Validation_47430485/Validation_47430485_47430485_313916.169.png" ], "turn_id": 9, "scan_id": "Validation_47430485", "origin_question": "What will be the first-time appearance order of the following categories: couch, door, table?", "option": [ "table, door, couch", "couch, table, door", "door, couch, table", "door, table, couch" ], "answer": "couch, table, door", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: couch, door, table? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['table, door, couch', 'couch, table, door', 'door, couch, table', 'door, table, couch'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "Validation_47430485/Validation_47430485_47430485_313918.168.png", "Validation_47430485/Validation_47430485_47430485_313920.167.png", "Validation_47430485/Validation_47430485_47430485_313922.167.png", "Validation_47430485/Validation_47430485_47430485_313924.182.png", "Validation_47430485/Validation_47430485_47430485_313926.182.png" ], "turn_id": 10, "scan_id": "Validation_47430485", "origin_question": "Among these three objects(\"the modern gray couch with soft fabric and cushions\",\"the light gray relaxed clothes with wide sleeves\",\"the modern dark brown wooden door\" ), which one is the closest to you?", "option": [ "the modern gray couch with soft fabric and cushions", "the light gray relaxed clothes with wide sleeves", "the modern dark brown wooden door" ], "answer": "the modern gray couch with soft fabric and cushions", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the modern gray couch with soft fabric and cushions\",\"the light gray relaxed clothes with wide sleeves\",\"the modern dark brown wooden door\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the modern gray couch with soft fabric and cushions', 'the light gray relaxed clothes with wide sleeves', 'the modern dark brown wooden door'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_44358498/Validation_44358498_44358498_50463.415.png", "Validation_44358498/Validation_44358498_44358498_50465.398.png", "Validation_44358498/Validation_44358498_44358498_50467.397.png", "Validation_44358498/Validation_44358498_44358498_50469.396.png", "Validation_44358498/Validation_44358498_44358498_50471.395.png" ], "turn_id": 1, "scan_id": "Validation_44358498", "origin_question": "Remember, have you seen any table(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any table(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_44358498/Validation_44358498_44358498_50473.394.png", "Validation_44358498/Validation_44358498_44358498_50475.394.png", "Validation_44358498/Validation_44358498_44358498_50477.393.png", "Validation_44358498/Validation_44358498_44358498_50479.392.png", "Validation_44358498/Validation_44358498_44358498_50481.408.png" ], "turn_id": 2, "scan_id": "Validation_44358498", "origin_question": "In which turn were you farthest from \"the black cylindrical plastic bin with the closed lid\"? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the black cylindrical plastic bin with the closed lid\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_44358498/Validation_44358498_44358498_50483.407.png", "Validation_44358498/Validation_44358498_44358498_50485.406.png", "Validation_44358498/Validation_44358498_44358498_50487.405.png", "Validation_44358498/Validation_44358498_44358498_50489.404.png", "Validation_44358498/Validation_44358498_44358498_50491.404.png" ], "turn_id": 3, "scan_id": "Validation_44358498", "origin_question": "Please recall \"the black cylindrical plastic bin with the closed lid\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "3.52", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the black cylindrical plastic bin with the closed lid\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_44358498/Validation_44358498_44358498_50493.403.png", "Validation_44358498/Validation_44358498_44358498_50495.402.png", "Validation_44358498/Validation_44358498_44358498_50497.401.png", "Validation_44358498/Validation_44358498_44358498_50499.400.png", "Validation_44358498/Validation_44358498_44358498_50501.399.png" ], "turn_id": 4, "scan_id": "Validation_44358498", "origin_question": "Remember, have you seen any plate(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any plate(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_44358498/Validation_44358498_44358498_50503.399.png", "Validation_44358498/Validation_44358498_44358498_50505.398.png", "Validation_44358498/Validation_44358498_44358498_50507.397.png", "Validation_44358498/Validation_44358498_44358498_50509.396.png", "Validation_44358498/Validation_44358498_44358498_50511.395.png" ], "turn_id": 5, "scan_id": "Validation_44358498", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_44358498/Validation_44358498_44358498_50513.395.png", "Validation_44358498/Validation_44358498_44358498_50515.394.png", "Validation_44358498/Validation_44358498_44358498_50517.393.png", "Validation_44358498/Validation_44358498_44358498_50519.392.png", "Validation_44358498/Validation_44358498_44358498_50521.391.png" ], "turn_id": 6, "scan_id": "Validation_44358498", "origin_question": "In which turn did you see both \"the stainless steel dishwasher with the metallic silver finish\" and \"the black cylindrical plastic bin with the closed lid\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the stainless steel dishwasher with the metallic silver finish\" and \"the black cylindrical plastic bin with the closed lid\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_44358498/Validation_44358498_44358498_50523.390.png", "Validation_44358498/Validation_44358498_44358498_50525.406.png", "Validation_44358498/Validation_44358498_44358498_50527.405.png", "Validation_44358498/Validation_44358498_44358498_50529.405.png", "Validation_44358498/Validation_44358498_44358498_50531.404.png" ], "turn_id": 7, "scan_id": "Validation_44358498", "origin_question": "In which turn were you closest to \"the medium-sized rectangular stainless steel sink\"? ", "option": [], "answer": "7", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the medium-sized rectangular stainless steel sink\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_44358498/Validation_44358498_44358498_50533.403.png", "Validation_44358498/Validation_44358498_44358498_50535.402.png", "Validation_44358498/Validation_44358498_44358498_50537.401.png", "Validation_44358498/Validation_44358498_44358498_50539.400.png", "Validation_44358498/Validation_44358498_44358498_50541.400.png" ], "turn_id": 8, "scan_id": "Validation_44358498", "origin_question": "How far is your current position from where you were at the end of the turn 3? (in meters)", "option": [], "answer": "4.2", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 3? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "Validation_45261142/Validation_45261142_45261142_3886.482.png", "Validation_45261142/Validation_45261142_45261142_3887.965.png", "Validation_45261142/Validation_45261142_45261142_3889.464.png", "Validation_45261142/Validation_45261142_45261142_3890.963.png" ], "turn_id": 1, "scan_id": "Validation_45261142", "origin_question": "Remember, how many helmets(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many helmets(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_45261142/Validation_45261142_45261142_3892.463.png", "Validation_45261142/Validation_45261142_45261142_3893.962.png", "Validation_45261142/Validation_45261142_45261142_3895.461.png", "Validation_45261142/Validation_45261142_45261142_3896.961.png" ], "turn_id": 2, "scan_id": "Validation_45261142", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the sleek black rectangular monitor with the glass front\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the sleek black rectangular monitor with the glass front\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_45261142/Validation_45261142_45261142_3898.460.png", "Validation_45261142/Validation_45261142_45261142_3899.960.png", "Validation_45261142/Validation_45261142_45261142_3901.476.png", "Validation_45261142/Validation_45261142_45261142_3902.975.png" ], "turn_id": 3, "scan_id": "Validation_45261142", "origin_question": "When did you first discover \"the sleek white front-loading washing machine with the glass door\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the sleek white front-loading washing machine with the glass door\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_45261142/Validation_45261142_45261142_3904.474.png", "Validation_45261142/Validation_45261142_45261142_3905.974.png", "Validation_45261142/Validation_45261142_45261142_3907.473.png", "Validation_45261142/Validation_45261142_45261142_3908.972.png" ], "turn_id": 4, "scan_id": "Validation_45261142", "origin_question": "What will be the first-time appearance order of the following categories: pillow, table, washing machine?", "option": [ "washing machine, pillow, table", "washing machine, table, pillow", "pillow, washing machine, table", "table, pillow, washing machine" ], "answer": "table, pillow, washing machine", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: pillow, table, washing machine? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['washing machine, pillow, table', 'washing machine, table, pillow', 'pillow, washing machine, table', 'table, pillow, washing machine'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_45261142/Validation_45261142_45261142_3910.472.png", "Validation_45261142/Validation_45261142_45261142_3911.971.png", "Validation_45261142/Validation_45261142_45261142_3913.471.png", "Validation_45261142/Validation_45261142_45261142_3914.970.png" ], "turn_id": 5, "scan_id": "Validation_45261142", "origin_question": "What will be the first-time appearance order of the following categories: chair, picture, washing machine?", "option": [ "picture, washing machine, chair", "chair, picture, washing machine", "picture, chair, washing machine", "washing machine, picture, chair" ], "answer": "chair, picture, washing machine", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: chair, picture, washing machine? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['picture, washing machine, chair', 'chair, picture, washing machine', 'picture, chair, washing machine', 'washing machine, picture, chair'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_45261142/Validation_45261142_45261142_3916.469.png", "Validation_45261142/Validation_45261142_45261142_3917.969.png", "Validation_45261142/Validation_45261142_45261142_3919.468.png", "Validation_45261142/Validation_45261142_45261142_3920.968.png" ], "turn_id": 6, "scan_id": "Validation_45261142", "origin_question": "What will be the first-time appearance order of the following categories: monitor, pillow, plant?", "option": [ "plant, monitor, pillow", "plant, pillow, monitor", "monitor, pillow, plant", "pillow, plant, monitor" ], "answer": "monitor, pillow, plant", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: monitor, pillow, plant? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['plant, monitor, pillow', 'plant, pillow, monitor', 'monitor, pillow, plant', 'pillow, plant, monitor'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_45261142/Validation_45261142_45261142_3922.467.png", "Validation_45261142/Validation_45261142_45261142_3923.966.png", "Validation_45261142/Validation_45261142_45261142_3925.466.png", "Validation_45261142/Validation_45261142_45261142_3926.965.png" ], "turn_id": 7, "scan_id": "Validation_45261142", "origin_question": "Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Left", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_45261142/Validation_45261142_45261142_3928.464.png", "Validation_45261142/Validation_45261142_45261142_3929.964.png", "Validation_45261142/Validation_45261142_45261142_3931.463.png", "Validation_45261142/Validation_45261142_45261142_3932.963.png" ], "turn_id": 8, "scan_id": "Validation_45261142", "origin_question": "At the end of which turn were both of \"the sleek black rectangular monitor with the glass front\" and \"the sleek white front-loading washing machine with the glass door\" on your left side?", "option": [], "answer": "8", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the sleek black rectangular monitor with the glass front\" and \"the sleek white front-loading washing machine with the glass door\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47331668/Validation_47331668_47331668_23504.383.png", "Validation_47331668/Validation_47331668_47331668_23506.382.png", "Validation_47331668/Validation_47331668_47331668_23508.381.png", "Validation_47331668/Validation_47331668_47331668_23510.380.png", "Validation_47331668/Validation_47331668_47331668_23512.379.png" ], "turn_id": 1, "scan_id": "Validation_47331668", "origin_question": "Remember, have you seen any calendar(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any calendar(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_47331668/Validation_47331668_47331668_23514.379.png", "Validation_47331668/Validation_47331668_47331668_23516.378.png", "Validation_47331668/Validation_47331668_47331668_23518.394.png", "Validation_47331668/Validation_47331668_47331668_23520.393.png", "Validation_47331668/Validation_47331668_47331668_23522.392.png" ], "turn_id": 2, "scan_id": "Validation_47331668", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the wooden chair with floral-patterned green cushions\", Y: \"an octagonal clock with the sleek black frame\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the wooden chair with floral-patterned green cushions\", Y: \"an octagonal clock with the sleek black frame\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_47331668/Validation_47331668_47331668_23524.391.png", "Validation_47331668/Validation_47331668_47331668_23526.391.png", "Validation_47331668/Validation_47331668_47331668_23528.390.png", "Validation_47331668/Validation_47331668_47331668_23530.389.png", "Validation_47331668/Validation_47331668_47331668_23532.388.png" ], "turn_id": 3, "scan_id": "Validation_47331668", "origin_question": "Is an octagonal clock with the sleek black frame to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Is an octagonal clock with the sleek black frame to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_47331668/Validation_47331668_47331668_23534.387.png", "Validation_47331668/Validation_47331668_47331668_23536.386.png", "Validation_47331668/Validation_47331668_47331668_23538.386.png", "Validation_47331668/Validation_47331668_47331668_23540.385.png", "Validation_47331668/Validation_47331668_47331668_23542.384.png" ], "turn_id": 4, "scan_id": "Validation_47331668", "origin_question": "Which two objects are on the same side of you now? \"the wooden chair with floral-patterned green cushions\", \"an octagonal clock with the sleek black frame\", \"the bed\". ", "option": [ "the wooden chair with floral-patterned green cushions, an octagonal clock with the sleek black frame", "the wooden chair with floral-patterned green cushions, the bed", "an octagonal clock with the sleek black frame, the bed" ], "answer": "the wooden chair with floral-patterned green cushions, an octagonal clock with the sleek black frame", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the wooden chair with floral-patterned green cushions\", \"an octagonal clock with the sleek black frame\", \"the bed\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the wooden chair with floral-patterned green cushions, an octagonal clock with the sleek black frame', 'the wooden chair with floral-patterned green cushions, the bed', 'an octagonal clock with the sleek black frame, the bed'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_47331668/Validation_47331668_47331668_23544.383.png", "Validation_47331668/Validation_47331668_47331668_23546.382.png", "Validation_47331668/Validation_47331668_47331668_23548.381.png", "Validation_47331668/Validation_47331668_47331668_23550.381.png", "Validation_47331668/Validation_47331668_47331668_23552.380.png" ], "turn_id": 5, "scan_id": "Validation_47331668", "origin_question": "Is the bed to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the bed to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_47331668/Validation_47331668_47331668_23554.379.png", "Validation_47331668/Validation_47331668_47331668_23556.378.png", "Validation_47331668/Validation_47331668_47331668_23558.377.png", "Validation_47331668/Validation_47331668_47331668_23560.393.png", "Validation_47331668/Validation_47331668_47331668_23562.393.png" ], "turn_id": 6, "scan_id": "Validation_47331668", "origin_question": "Please recall \"the bed\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.51", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the bed\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47331668/Validation_47331668_47331668_23564.392.png", "Validation_47331668/Validation_47331668_47331668_23566.391.png", "Validation_47331668/Validation_47331668_47331668_23568.390.png", "Validation_47331668/Validation_47331668_47331668_23570.389.png", "Validation_47331668/Validation_47331668_47331668_23572.388.png" ], "turn_id": 7, "scan_id": "Validation_47331668", "origin_question": "Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the vintage crt monitor with the light gray casing\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the vintage crt monitor with the light gray casing\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_47331668/Validation_47331668_47331668_23574.388.png", "Validation_47331668/Validation_47331668_47331668_23576.387.png", "Validation_47331668/Validation_47331668_47331668_23578.386.png", "Validation_47331668/Validation_47331668_47331668_23580.385.png", "Validation_47331668/Validation_47331668_47331668_23582.384.png" ], "turn_id": 8, "scan_id": "Validation_47331668", "origin_question": "Using your orientation at the end of turn 7 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 7 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_48018560/Validation_48018560_48018560_3394.762.png", "Validation_48018560/Validation_48018560_48018560_3396.244.png", "Validation_48018560/Validation_48018560_48018560_3397.744.png", "Validation_48018560/Validation_48018560_48018560_3399.243.png" ], "turn_id": 1, "scan_id": "Validation_48018560", "origin_question": "Remember, have you seen any cabinet(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any cabinet(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_48018560/Validation_48018560_48018560_3400.742.png", "Validation_48018560/Validation_48018560_48018560_3402.242.png", "Validation_48018560/Validation_48018560_48018560_3403.741.png", "Validation_48018560/Validation_48018560_48018560_3405.241.png" ], "turn_id": 2, "scan_id": "Validation_48018560", "origin_question": "In which turn did you see both \"the oven\" and \"the medium-sized, rectangular green or grey towel\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the oven\" and \"the medium-sized, rectangular green or grey towel\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_48018560/Validation_48018560_48018560_3406.740.png", "Validation_48018560/Validation_48018560_48018560_3408.240.png", "Validation_48018560/Validation_48018560_48018560_3409.756.png", "Validation_48018560/Validation_48018560_48018560_3411.255.png" ], "turn_id": 3, "scan_id": "Validation_48018560", "origin_question": "What will be the first-time appearance order of the following categories: cabinet, kettle, towel?", "option": [ "towel, kettle, cabinet", "cabinet, kettle, towel", "cabinet, towel, kettle", "kettle, cabinet, towel" ], "answer": "cabinet, towel, kettle", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: cabinet, kettle, towel? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['towel, kettle, cabinet', 'cabinet, kettle, towel', 'cabinet, towel, kettle', 'kettle, cabinet, towel'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_48018560/Validation_48018560_48018560_3412.754.png", "Validation_48018560/Validation_48018560_48018560_3414.254.png", "Validation_48018560/Validation_48018560_48018560_3415.753.png", "Validation_48018560/Validation_48018560_48018560_3417.253.png" ], "turn_id": 4, "scan_id": "Validation_48018560", "origin_question": "Which two objects are on the same side of you now? \"the dome-shaped light made of translucent material\", \"the medium-sized, rectangular green or grey towel\", \"the stainless steel stove with four burners\". ", "option": [ "the dome-shaped light made of translucent material, the medium-sized, rectangular green or grey towel", "the dome-shaped light made of translucent material, the stainless steel stove with four burners", "the medium-sized, rectangular green or grey towel, the stainless steel stove with four burners" ], "answer": "the medium-sized, rectangular green or grey towel, the stainless steel stove with four burners", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the dome-shaped light made of translucent material\", \"the medium-sized, rectangular green or grey towel\", \"the stainless steel stove with four burners\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the dome-shaped light made of translucent material, the medium-sized, rectangular green or grey towel', 'the dome-shaped light made of translucent material, the stainless steel stove with four burners', 'the medium-sized, rectangular green or grey towel, the stainless steel stove with four burners'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_48018560/Validation_48018560_48018560_3418.752.png", "Validation_48018560/Validation_48018560_48018560_3420.251.png", "Validation_48018560/Validation_48018560_48018560_3421.751.png", "Validation_48018560/Validation_48018560_48018560_3423.250.png" ], "turn_id": 5, "scan_id": "Validation_48018560", "origin_question": "In which turn were you closest to \"the dome-shaped light made of translucent material\"? ", "option": [], "answer": "3", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the dome-shaped light made of translucent material\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "Validation_48018560/Validation_48018560_48018560_3424.750.png", "Validation_48018560/Validation_48018560_48018560_3426.249.png", "Validation_48018560/Validation_48018560_48018560_3427.748.png", "Validation_48018560/Validation_48018560_48018560_3429.248.png" ], "turn_id": 6, "scan_id": "Validation_48018560", "origin_question": "Among these three objects(\"the oven\",\"the sleek, silver round kettle\",\"the modern white refrigerator with ample storage space\" ), which one is the closest to you?", "option": [ "the oven", "the sleek, silver round kettle", "the modern white refrigerator with ample storage space" ], "answer": "the modern white refrigerator with ample storage space", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects(\"the oven\",\"the sleek, silver round kettle\",\"the modern white refrigerator with ample storage space\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the oven', 'the sleek, silver round kettle', 'the modern white refrigerator with ample storage space'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_48018560/Validation_48018560_48018560_3430.747.png", "Validation_48018560/Validation_48018560_48018560_3432.247.png", "Validation_48018560/Validation_48018560_48018560_3433.746.png", "Validation_48018560/Validation_48018560_48018560_3435.245.png" ], "turn_id": 7, "scan_id": "Validation_48018560", "origin_question": "When did you first discover \"the sleek stainless steel sink with rounded edges\"? (index of the turn)", "option": [], "answer": "3", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the sleek stainless steel sink with rounded edges\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_48018560/Validation_48018560_48018560_3436.745.png", "Validation_48018560/Validation_48018560_48018560_3438.244.png", "Validation_48018560/Validation_48018560_48018560_3439.744.png", "Validation_48018560/Validation_48018560_48018560_3441.243.png" ], "turn_id": 8, "scan_id": "Validation_48018560", "origin_question": "Remember, have you seen any light(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any light(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47332915/Validation_47332915_47332915_13466.990.png", "Validation_47332915/Validation_47332915_47332915_13468.972.png", "Validation_47332915/Validation_47332915_47332915_13470.972.png", "Validation_47332915/Validation_47332915_47332915_13472.971.png", "Validation_47332915/Validation_47332915_47332915_13474.970.png" ], "turn_id": 1, "scan_id": "Validation_47332915", "origin_question": "Remember, have you seen any door(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any door(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "Validation_47332915/Validation_47332915_47332915_13476.969.png", "Validation_47332915/Validation_47332915_47332915_13478.968.png", "Validation_47332915/Validation_47332915_47332915_13480.968.png", "Validation_47332915/Validation_47332915_47332915_13482.967.png", "Validation_47332915/Validation_47332915_47332915_13484.966.png" ], "turn_id": 2, "scan_id": "Validation_47332915", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the vibrant pink sheet on the rectangular bed frame\"; \"the rectangular window with the neutral-toned frame\"; \"the fresh green cylindrical flowerpot made of plastic\" ", "option": [ "the vibrant pink sheet on the rectangular bed frame", "the rectangular window with the neutral-toned frame", "the fresh green cylindrical flowerpot made of plastic" ], "answer": "the vibrant pink sheet on the rectangular bed frame", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the vibrant pink sheet on the rectangular bed frame\"; \"the rectangular window with the neutral-toned frame\"; \"the fresh green cylindrical flowerpot made of plastic\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the vibrant pink sheet on the rectangular bed frame', 'the rectangular window with the neutral-toned frame', 'the fresh green cylindrical flowerpot made of plastic'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_47332915/Validation_47332915_47332915_13486.965.png", "Validation_47332915/Validation_47332915_47332915_13488.964.png", "Validation_47332915/Validation_47332915_47332915_13490.963.png", "Validation_47332915/Validation_47332915_47332915_13492.963.png", "Validation_47332915/Validation_47332915_47332915_13494.978.png" ], "turn_id": 3, "scan_id": "Validation_47332915", "origin_question": "Please recall \"the rectangular window with the neutral-toned frame\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.76", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the rectangular window with the neutral-toned frame\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47332915/Validation_47332915_47332915_13496.978.png", "Validation_47332915/Validation_47332915_47332915_13498.977.png", "Validation_47332915/Validation_47332915_47332915_13500.976.png", "Validation_47332915/Validation_47332915_47332915_13502.975.png", "Validation_47332915/Validation_47332915_47332915_13504.974.png" ], "turn_id": 4, "scan_id": "Validation_47332915", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the fresh green cylindrical flowerpot made of plastic\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the fresh green cylindrical flowerpot made of plastic\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47332915/Validation_47332915_47332915_13506.974.png", "Validation_47332915/Validation_47332915_47332915_13508.973.png", "Validation_47332915/Validation_47332915_47332915_13510.972.png", "Validation_47332915/Validation_47332915_47332915_13512.971.png", "Validation_47332915/Validation_47332915_47332915_13514.970.png" ], "turn_id": 5, "scan_id": "Validation_47332915", "origin_question": "", "option": [], "answer": "", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_47332915/Validation_47332915_47332915_13516.969.png", "Validation_47332915/Validation_47332915_47332915_13518.969.png", "Validation_47332915/Validation_47332915_47332915_13520.968.png", "Validation_47332915/Validation_47332915_47332915_13522.967.png", "Validation_47332915/Validation_47332915_47332915_13524.966.png" ], "turn_id": 6, "scan_id": "Validation_47332915", "origin_question": "What will be the first-time appearance order of the following categories: monitor, pillow, window?", "option": [ "monitor, window, pillow", "monitor, pillow, window", "window, monitor, pillow", "pillow, monitor, window" ], "answer": "monitor, window, pillow", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: monitor, pillow, window? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['monitor, window, pillow', 'monitor, pillow, window', 'window, monitor, pillow', 'pillow, monitor, window'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_47332915/Validation_47332915_47332915_13526.965.png", "Validation_47332915/Validation_47332915_47332915_13528.965.png", "Validation_47332915/Validation_47332915_47332915_13530.964.png", "Validation_47332915/Validation_47332915_47332915_13532.963.png", "Validation_47332915/Validation_47332915_47332915_13534.962.png" ], "turn_id": 7, "scan_id": "Validation_47332915", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47332915/Validation_47332915_47332915_13536.978.png", "Validation_47332915/Validation_47332915_47332915_13538.977.png", "Validation_47332915/Validation_47332915_47332915_13540.976.png", "Validation_47332915/Validation_47332915_47332915_13542.976.png", "Validation_47332915/Validation_47332915_47332915_13544.975.png" ], "turn_id": 8, "scan_id": "Validation_47332915", "origin_question": "Which direction is the sleek white vertical metal radiator in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the sleek white vertical metal radiator in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_47332915/Validation_47332915_47332915_13546.974.png", "Validation_47332915/Validation_47332915_47332915_13548.973.png", "Validation_47332915/Validation_47332915_47332915_13550.972.png", "Validation_47332915/Validation_47332915_47332915_13552.971.png", "Validation_47332915/Validation_47332915_47332915_13554.971.png" ], "turn_id": 9, "scan_id": "Validation_47332915", "origin_question": "What will be the first-time appearance order of the following categories: bed, radiator, window?", "option": [ "radiator, bed, window", "window, bed, radiator", "bed, radiator, window", "radiator, window, bed" ], "answer": "window, bed, radiator", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bed, radiator, window? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['radiator, bed, window', 'window, bed, radiator', 'bed, radiator, window', 'radiator, window, bed'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47332915/Validation_47332915_47332915_13556.970.png", "Validation_47332915/Validation_47332915_47332915_13558.969.png", "Validation_47332915/Validation_47332915_47332915_13560.968.png", "Validation_47332915/Validation_47332915_47332915_13562.967.png", "Validation_47332915/Validation_47332915_47332915_13564.967.png" ], "turn_id": 10, "scan_id": "Validation_47332915", "origin_question": "Which direction is the rectangular window with the neutral-toned frame in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the rectangular window with the neutral-toned frame in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42897678/Validation_42897678_42897678_479553.578.png", "Validation_42897678/Validation_42897678_42897678_479555.578.png", "Validation_42897678/Validation_42897678_42897678_479557.577.png", "Validation_42897678/Validation_42897678_42897678_479559.576.png", "Validation_42897678/Validation_42897678_42897678_479561.575.png" ], "turn_id": 1, "scan_id": "Validation_42897678", "origin_question": "Remember, have you seen any lamp(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any lamp(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_42897678/Validation_42897678_42897678_479563.574.png", "Validation_42897678/Validation_42897678_42897678_479565.573.png", "Validation_42897678/Validation_42897678_42897678_479567.573.png", "Validation_42897678/Validation_42897678_42897678_479569.572.png", "Validation_42897678/Validation_42897678_42897678_479571.588.png" ], "turn_id": 2, "scan_id": "Validation_42897678", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_42897678/Validation_42897678_42897678_479573.587.png", "Validation_42897678/Validation_42897678_42897678_479575.586.png", "Validation_42897678/Validation_42897678_42897678_479577.585.png", "Validation_42897678/Validation_42897678_42897678_479579.584.png", "Validation_42897678/Validation_42897678_42897678_479581.583.png" ], "turn_id": 3, "scan_id": "Validation_42897678", "origin_question": "Is the medium-sized cream conical lamp with the gold base to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the medium-sized cream conical lamp with the gold base to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_42897678/Validation_42897678_42897678_479583.583.png", "Validation_42897678/Validation_42897678_42897678_479585.582.png", "Validation_42897678/Validation_42897678_42897678_479587.581.png", "Validation_42897678/Validation_42897678_42897678_479589.580.png", "Validation_42897678/Validation_42897678_42897678_479591.579.png" ], "turn_id": 4, "scan_id": "Validation_42897678", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_42897678/Validation_42897678_42897678_479593.578.png", "Validation_42897678/Validation_42897678_42897678_479595.578.png", "Validation_42897678/Validation_42897678_42897678_479597.577.png", "Validation_42897678/Validation_42897678_42897678_479599.576.png", "Validation_42897678/Validation_42897678_42897678_479601.575.png" ], "turn_id": 5, "scan_id": "Validation_42897678", "origin_question": "Is the vibrant rectangular picture of an urban scene to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the vibrant rectangular picture of an urban scene to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "Validation_42897678/Validation_42897678_42897678_479603.574.png", "Validation_42897678/Validation_42897678_42897678_479605.573.png", "Validation_42897678/Validation_42897678_42897678_479607.573.png", "Validation_42897678/Validation_42897678_42897678_479609.572.png", "Validation_42897678/Validation_42897678_42897678_479611.571.png" ], "turn_id": 6, "scan_id": "Validation_42897678", "origin_question": "Among these three objects(\"the luxurious, curved brown bed with light bedding\",\"the medium-sized cream conical lamp with the gold base\",\"the vibrant rectangular picture of an urban scene\" ), which one is the closest to you?", "option": [ "the luxurious, curved brown bed with light bedding", "the medium-sized cream conical lamp with the gold base", "the vibrant rectangular picture of an urban scene" ], "answer": "the luxurious, curved brown bed with light bedding", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the luxurious, curved brown bed with light bedding\",\"the medium-sized cream conical lamp with the gold base\",\"the vibrant rectangular picture of an urban scene\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the luxurious, curved brown bed with light bedding', 'the medium-sized cream conical lamp with the gold base', 'the vibrant rectangular picture of an urban scene'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_42897678/Validation_42897678_42897678_479613.570.png", "Validation_42897678/Validation_42897678_42897678_479615.586.png", "Validation_42897678/Validation_42897678_42897678_479617.585.png", "Validation_42897678/Validation_42897678_42897678_479619.584.png", "Validation_42897678/Validation_42897678_42897678_479621.584.png" ], "turn_id": 7, "scan_id": "Validation_42897678", "origin_question": "Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the medium-sized cream conical lamp with the gold base\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the medium-sized cream conical lamp with the gold base\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_42897678/Validation_42897678_42897678_479623.583.png", "Validation_42897678/Validation_42897678_42897678_479625.582.png", "Validation_42897678/Validation_42897678_42897678_479627.581.png", "Validation_42897678/Validation_42897678_42897678_479629.580.png", "Validation_42897678/Validation_42897678_42897678_479631.579.png" ], "turn_id": 8, "scan_id": "Validation_42897678", "origin_question": "Please recall \"the medium-sized cream conical lamp with the gold base\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "4.19", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the medium-sized cream conical lamp with the gold base\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_42897678/Validation_42897678_42897678_479633.579.png", "Validation_42897678/Validation_42897678_42897678_479635.578.png", "Validation_42897678/Validation_42897678_42897678_479637.577.png", "Validation_42897678/Validation_42897678_42897678_479639.576.png", "Validation_42897678/Validation_42897678_42897678_479641.575.png" ], "turn_id": 9, "scan_id": "Validation_42897678", "origin_question": "Using your orientation at the end of turn 7 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 7 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_42897678/Validation_42897678_42897678_479643.574.png", "Validation_42897678/Validation_42897678_42897678_479645.573.png", "Validation_42897678/Validation_42897678_42897678_479647.573.png", "Validation_42897678/Validation_42897678_42897678_479649.572.png", "Validation_42897678/Validation_42897678_42897678_479651.571.png" ], "turn_id": 10, "scan_id": "Validation_42897678", "origin_question": "Is the dark cylindrical plastic bin to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the dark cylindrical plastic bin to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47895350/Validation_47895350_47895350_96551.848.png", "Validation_47895350/Validation_47895350_47895350_96553.814.png", "Validation_47895350/Validation_47895350_47895350_96555.813.png", "Validation_47895350/Validation_47895350_47895350_96557.812.png", "Validation_47895350/Validation_47895350_47895350_96559.828.png" ], "turn_id": 1, "scan_id": "Validation_47895350", "origin_question": "Remember, have you seen any dustpan(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any dustpan(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47895350/Validation_47895350_47895350_96561.827.png", "Validation_47895350/Validation_47895350_47895350_96563.826.png", "Validation_47895350/Validation_47895350_47895350_96565.826.png", "Validation_47895350/Validation_47895350_47895350_96567.825.png", "Validation_47895350/Validation_47895350_47895350_96569.824.png" ], "turn_id": 2, "scan_id": "Validation_47895350", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the soft, off-white ergonomic chair\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the soft, off-white ergonomic chair\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47895350/Validation_47895350_47895350_96571.823.png", "Validation_47895350/Validation_47895350_47895350_96573.822.png", "Validation_47895350/Validation_47895350_47895350_96575.821.png", "Validation_47895350/Validation_47895350_47895350_96577.821.png", "Validation_47895350/Validation_47895350_47895350_96579.820.png" ], "turn_id": 3, "scan_id": "Validation_47895350", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"colorful books with uniform rectangular shapes\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"colorful books with uniform rectangular shapes\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47895350/Validation_47895350_47895350_96581.819.png", "Validation_47895350/Validation_47895350_47895350_96583.818.png", "Validation_47895350/Validation_47895350_47895350_96585.817.png", "Validation_47895350/Validation_47895350_47895350_96587.817.png", "Validation_47895350/Validation_47895350_47895350_96589.816.png" ], "turn_id": 4, "scan_id": "Validation_47895350", "origin_question": "Remember, have you seen any panel(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any panel(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47895350/Validation_47895350_47895350_96591.815.png", "Validation_47895350/Validation_47895350_47895350_96593.814.png", "Validation_47895350/Validation_47895350_47895350_96595.813.png", "Validation_47895350/Validation_47895350_47895350_96597.812.png", "Validation_47895350/Validation_47895350_47895350_96599.828.png" ], "turn_id": 5, "scan_id": "Validation_47895350", "origin_question": "Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the transparent rectangular case with the clear construction\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the transparent rectangular case with the clear construction\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_47895350/Validation_47895350_47895350_96601.828.png", "Validation_47895350/Validation_47895350_47895350_96603.827.png", "Validation_47895350/Validation_47895350_47895350_96605.826.png", "Validation_47895350/Validation_47895350_47895350_96607.825.png", "Validation_47895350/Validation_47895350_47895350_96609.824.png" ], "turn_id": 6, "scan_id": "Validation_47895350", "origin_question": "Is colorful books with uniform rectangular shapes to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Is colorful books with uniform rectangular shapes to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47895350/Validation_47895350_47895350_96611.823.png", "Validation_47895350/Validation_47895350_47895350_96613.823.png", "Validation_47895350/Validation_47895350_47895350_96615.822.png", "Validation_47895350/Validation_47895350_47895350_96617.821.png", "Validation_47895350/Validation_47895350_47895350_96619.820.png" ], "turn_id": 7, "scan_id": "Validation_47895350", "origin_question": "Compared to your position at the end of the 6 turn, are you now closer or farther away from \"the light wood rectangular table with the spacious surface\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 6 turn, are you now closer or farther away from \"the light wood rectangular table with the spacious surface\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47895350/Validation_47895350_47895350_96621.819.png", "Validation_47895350/Validation_47895350_47895350_96623.819.png", "Validation_47895350/Validation_47895350_47895350_96625.818.png", "Validation_47895350/Validation_47895350_47895350_96627.817.png", "Validation_47895350/Validation_47895350_47895350_96629.816.png" ], "turn_id": 8, "scan_id": "Validation_47895350", "origin_question": "Which direction is colorful books with uniform rectangular shapes in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is colorful books with uniform rectangular shapes in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_47895350/Validation_47895350_47895350_96631.815.png", "Validation_47895350/Validation_47895350_47895350_96633.815.png", "Validation_47895350/Validation_47895350_47895350_96635.814.png", "Validation_47895350/Validation_47895350_47895350_96637.813.png", "Validation_47895350/Validation_47895350_47895350_96639.812.png" ], "turn_id": 9, "scan_id": "Validation_47895350", "origin_question": "What will be the first-time appearance order of the following categories: book, cabinet, chair?", "option": [ "cabinet, chair, book", "chair, cabinet, book", "chair, book, cabinet", "book, cabinet, chair" ], "answer": "chair, book, cabinet", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: book, cabinet, chair? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['cabinet, chair, book', 'chair, cabinet, book', 'chair, book, cabinet', 'book, cabinet, chair'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47895350/Validation_47895350_47895350_96641.828.png", "Validation_47895350/Validation_47895350_47895350_96643.827.png", "Validation_47895350/Validation_47895350_47895350_96645.826.png", "Validation_47895350/Validation_47895350_47895350_96647.826.png", "Validation_47895350/Validation_47895350_47895350_96649.825.png" ], "turn_id": 10, "scan_id": "Validation_47895350", "origin_question": "", "option": [], "answer": "", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_48018972/Validation_48018972_48018972_556798.373.png", "Validation_48018972/Validation_48018972_48018972_556799.856.png", "Validation_48018972/Validation_48018972_48018972_556801.355.png", "Validation_48018972/Validation_48018972_48018972_556802.854.png" ], "turn_id": 1, "scan_id": "Validation_48018972", "origin_question": "Remember, have you seen any car(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any car(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "Validation_48018972/Validation_48018972_48018972_556804.354.png", "Validation_48018972/Validation_48018972_48018972_556805.853.png", "Validation_48018972/Validation_48018972_48018972_556807.353.png", "Validation_48018972/Validation_48018972_48018972_556808.852.png" ], "turn_id": 2, "scan_id": "Validation_48018972", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the round dark table with crisscrossed metal legs\"; \"the classic white wooden chair with the gray cushion\"; \"the rectangular gray pillow with multicolored patterns\" ", "option": [ "the round dark table with crisscrossed metal legs", "the classic white wooden chair with the gray cushion", "the rectangular gray pillow with multicolored patterns" ], "answer": "the round dark table with crisscrossed metal legs", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the round dark table with crisscrossed metal legs\"; \"the classic white wooden chair with the gray cushion\"; \"the rectangular gray pillow with multicolored patterns\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the round dark table with crisscrossed metal legs', 'the classic white wooden chair with the gray cushion', 'the rectangular gray pillow with multicolored patterns'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_48018972/Validation_48018972_48018972_556810.351.png", "Validation_48018972/Validation_48018972_48018972_556811.851.png", "Validation_48018972/Validation_48018972_48018972_556813.350.png", "Validation_48018972/Validation_48018972_48018972_556814.849.png" ], "turn_id": 3, "scan_id": "Validation_48018972", "origin_question": "By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the classic white wooden chair with the gray cushion\", Y: \"the bed\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to Y and farther away from X", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the classic white wooden chair with the gray cushion\", Y: \"the bed\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_48018972/Validation_48018972_48018972_556816.349.png", "Validation_48018972/Validation_48018972_48018972_556817.848.png", "Validation_48018972/Validation_48018972_48018972_556819.348.png", "Validation_48018972/Validation_48018972_48018972_556820.847.png" ], "turn_id": 4, "scan_id": "Validation_48018972", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the transparent rectangular box with the red lid\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the transparent rectangular box with the red lid\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_48018972/Validation_48018972_48018972_556822.346.png", "Validation_48018972/Validation_48018972_48018972_556823.862.png", "Validation_48018972/Validation_48018972_48018972_556825.362.png", "Validation_48018972/Validation_48018972_48018972_556826.861.png" ], "turn_id": 5, "scan_id": "Validation_48018972", "origin_question": "", "option": [], "answer": "", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_48018972/Validation_48018972_48018972_556828.360.png", "Validation_48018972/Validation_48018972_48018972_556829.860.png", "Validation_48018972/Validation_48018972_48018972_556831.359.png", "Validation_48018972/Validation_48018972_48018972_556832.859.png" ], "turn_id": 6, "scan_id": "Validation_48018972", "origin_question": "Which direction is the monitor in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the monitor in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42446079/Validation_42446079_42446079_348941.588.png", "Validation_42446079/Validation_42446079_42446079_348943.587.png", "Validation_42446079/Validation_42446079_42446079_348945.586.png", "Validation_42446079/Validation_42446079_42446079_348947.585.png", "Validation_42446079/Validation_42446079_42446079_348949.584.png" ], "turn_id": 1, "scan_id": "Validation_42446079", "origin_question": "Remember, have you seen any book(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any book(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_42446079/Validation_42446079_42446079_348951.583.png", "Validation_42446079/Validation_42446079_42446079_348953.583.png", "Validation_42446079/Validation_42446079_42446079_348955.582.png", "Validation_42446079/Validation_42446079_42446079_348957.581.png", "Validation_42446079/Validation_42446079_42446079_348959.580.png" ], "turn_id": 2, "scan_id": "Validation_42446079", "origin_question": "In which turn did you see both \"the sleek dark gray cylindrical bin made of plastic or metal\" and \"the white wooden panel door with the rectangular shape\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the sleek dark gray cylindrical bin made of plastic or metal\" and \"the white wooden panel door with the rectangular shape\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_42446079/Validation_42446079_42446079_348961.579.png", "Validation_42446079/Validation_42446079_42446079_348963.579.png", "Validation_42446079/Validation_42446079_42446079_348965.578.png", "Validation_42446079/Validation_42446079_42446079_348967.577.png", "Validation_42446079/Validation_42446079_42446079_348969.576.png" ], "turn_id": 3, "scan_id": "Validation_42446079", "origin_question": "Please recall \"the vibrant, glossy book with the captivating cover design\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.82", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the vibrant, glossy book with the captivating cover design\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_42446079/Validation_42446079_42446079_348971.575.png", "Validation_42446079/Validation_42446079_42446079_348973.574.png", "Validation_42446079/Validation_42446079_42446079_348975.590.png", "Validation_42446079/Validation_42446079_42446079_348977.589.png", "Validation_42446079/Validation_42446079_42446079_348979.589.png" ], "turn_id": 4, "scan_id": "Validation_42446079", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_42446079/Validation_42446079_42446079_348981.588.png", "Validation_42446079/Validation_42446079_42446079_348983.587.png", "Validation_42446079/Validation_42446079_42446079_348985.586.png", "Validation_42446079/Validation_42446079_42446079_348987.585.png", "Validation_42446079/Validation_42446079_42446079_348989.584.png" ], "turn_id": 5, "scan_id": "Validation_42446079", "origin_question": "Please recall \"the sleek dark gray cylindrical bin made of plastic or metal\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.95", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the sleek dark gray cylindrical bin made of plastic or metal\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_42446079/Validation_42446079_42446079_348991.584.png", "Validation_42446079/Validation_42446079_42446079_348993.583.png", "Validation_42446079/Validation_42446079_42446079_348995.582.png", "Validation_42446079/Validation_42446079_42446079_348997.581.png", "Validation_42446079/Validation_42446079_42446079_348999.580.png" ], "turn_id": 6, "scan_id": "Validation_42446079", "origin_question": "Which two objects are on the same side of you now? \"the sleek dark gray cylindrical bin made of plastic or metal\", \"the white wooden panel door with the rectangular shape\", \"the vibrant, glossy book with the captivating cover design\". ", "option": [ "the sleek dark gray cylindrical bin made of plastic or metal, the white wooden panel door with the rectangular shape", "the sleek dark gray cylindrical bin made of plastic or metal, the vibrant, glossy book with the captivating cover design", "the white wooden panel door with the rectangular shape, the vibrant, glossy book with the captivating cover design" ], "answer": "the sleek dark gray cylindrical bin made of plastic or metal, the white wooden panel door with the rectangular shape", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the sleek dark gray cylindrical bin made of plastic or metal\", \"the white wooden panel door with the rectangular shape\", \"the vibrant, glossy book with the captivating cover design\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the sleek dark gray cylindrical bin made of plastic or metal, the white wooden panel door with the rectangular shape', 'the sleek dark gray cylindrical bin made of plastic or metal, the vibrant, glossy book with the captivating cover design', 'the white wooden panel door with the rectangular shape, the vibrant, glossy book with the captivating cover design'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_42446079/Validation_42446079_42446079_349001.580.png", "Validation_42446079/Validation_42446079_42446079_349003.579.png", "Validation_42446079/Validation_42446079_42446079_349005.578.png", "Validation_42446079/Validation_42446079_42446079_349007.577.png", "Validation_42446079/Validation_42446079_42446079_349009.576.png" ], "turn_id": 7, "scan_id": "Validation_42446079", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the white wooden panel door with the rectangular shape\" from your position?", "option": [], "answer": "155.68", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the white wooden panel door with the rectangular shape\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_42446079/Validation_42446079_42446079_349011.575.png", "Validation_42446079/Validation_42446079_42446079_349013.575.png", "Validation_42446079/Validation_42446079_42446079_349015.574.png", "Validation_42446079/Validation_42446079_42446079_349017.590.png", "Validation_42446079/Validation_42446079_42446079_349019.589.png" ], "turn_id": 8, "scan_id": "Validation_42446079", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_42446079/Validation_42446079_42446079_349021.588.png", "Validation_42446079/Validation_42446079_42446079_349023.587.png", "Validation_42446079/Validation_42446079_42446079_349025.586.png", "Validation_42446079/Validation_42446079_42446079_349027.586.png", "Validation_42446079/Validation_42446079_42446079_349029.585.png" ], "turn_id": 9, "scan_id": "Validation_42446079", "origin_question": "At the end of which turn were both of \"the rectangular off-white wooden table with the sturdy design\" and \"the sleek dark gray cylindrical bin made of plastic or metal\" on your left side?", "option": [], "answer": "3", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the rectangular off-white wooden table with the sturdy design\" and \"the sleek dark gray cylindrical bin made of plastic or metal\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_42446079/Validation_42446079_42446079_349031.584.png", "Validation_42446079/Validation_42446079_42446079_349033.583.png", "Validation_42446079/Validation_42446079_42446079_349035.582.png", "Validation_42446079/Validation_42446079_42446079_349037.581.png", "Validation_42446079/Validation_42446079_42446079_349039.581.png" ], "turn_id": 10, "scan_id": "Validation_42446079", "origin_question": "Is the sleek dark gray cylindrical bin made of plastic or metal to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the sleek dark gray cylindrical bin made of plastic or metal to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47334110/Validation_47334110_47334110_93524.492.png", "Validation_47334110/Validation_47334110_47334110_93526.491.png", "Validation_47334110/Validation_47334110_47334110_93528.490.png", "Validation_47334110/Validation_47334110_47334110_93530.489.png", "Validation_47334110/Validation_47334110_47334110_93532.489.png" ], "turn_id": 1, "scan_id": "Validation_47334110", "origin_question": "Remember, have you seen any door(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any door(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "Validation_47334110/Validation_47334110_47334110_93534.488.png", "Validation_47334110/Validation_47334110_47334110_93536.487.png", "Validation_47334110/Validation_47334110_47334110_93538.486.png", "Validation_47334110/Validation_47334110_47334110_93540.502.png", "Validation_47334110/Validation_47334110_47334110_93542.501.png" ], "turn_id": 2, "scan_id": "Validation_47334110", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the sleek oval white porcelain toilet\"; \"the medium-sized, sleek stainless steel box\"; \"the rich brown wooden door with frosted glass panels\" ", "option": [ "the sleek oval white porcelain toilet", "the medium-sized, sleek stainless steel box", "the rich brown wooden door with frosted glass panels" ], "answer": "the medium-sized, sleek stainless steel box", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the sleek oval white porcelain toilet\"; \"the medium-sized, sleek stainless steel box\"; \"the rich brown wooden door with frosted glass panels\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the sleek oval white porcelain toilet', 'the medium-sized, sleek stainless steel box', 'the rich brown wooden door with frosted glass panels'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_47334110/Validation_47334110_47334110_93544.500.png", "Validation_47334110/Validation_47334110_47334110_93546.500.png", "Validation_47334110/Validation_47334110_47334110_93548.499.png", "Validation_47334110/Validation_47334110_47334110_93550.498.png", "Validation_47334110/Validation_47334110_47334110_93552.497.png" ], "turn_id": 3, "scan_id": "Validation_47334110", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the sleek oval white porcelain toilet\" from your position?", "option": [], "answer": "107.25", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the sleek oval white porcelain toilet\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47334110/Validation_47334110_47334110_93554.496.png", "Validation_47334110/Validation_47334110_47334110_93556.495.png", "Validation_47334110/Validation_47334110_47334110_93558.495.png", "Validation_47334110/Validation_47334110_47334110_93560.494.png", "Validation_47334110/Validation_47334110_47334110_93562.493.png" ], "turn_id": 4, "scan_id": "Validation_47334110", "origin_question": "", "option": [], "answer": "", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_47334110/Validation_47334110_47334110_93564.492.png", "Validation_47334110/Validation_47334110_47334110_93566.491.png", "Validation_47334110/Validation_47334110_47334110_93568.491.png", "Validation_47334110/Validation_47334110_47334110_93570.490.png", "Validation_47334110/Validation_47334110_47334110_93572.489.png" ], "turn_id": 5, "scan_id": "Validation_47334110", "origin_question": "At the end of which turn were both of \"the sleek white front-loading washing machine\" and \"the medium-sized, sleek stainless steel box\" on your right side?", "option": [], "answer": "5", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the sleek white front-loading washing machine\" and \"the medium-sized, sleek stainless steel box\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_47334110/Validation_47334110_47334110_93574.488.png", "Validation_47334110/Validation_47334110_47334110_93576.487.png", "Validation_47334110/Validation_47334110_47334110_93578.486.png", "Validation_47334110/Validation_47334110_47334110_93580.502.png", "Validation_47334110/Validation_47334110_47334110_93582.502.png" ], "turn_id": 6, "scan_id": "Validation_47334110", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_47334110/Validation_47334110_47334110_93584.501.png", "Validation_47334110/Validation_47334110_47334110_93586.500.png", "Validation_47334110/Validation_47334110_47334110_93588.499.png", "Validation_47334110/Validation_47334110_47334110_93590.498.png", "Validation_47334110/Validation_47334110_47334110_93592.497.png" ], "turn_id": 7, "scan_id": "Validation_47334110", "origin_question": "Please recall \"the rich brown wooden door with frosted glass panels\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.91", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the rich brown wooden door with frosted glass panels\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_47334110/Validation_47334110_47334110_93594.497.png", "Validation_47334110/Validation_47334110_47334110_93596.496.png", "Validation_47334110/Validation_47334110_47334110_93598.495.png", "Validation_47334110/Validation_47334110_47334110_93600.494.png", "Validation_47334110/Validation_47334110_47334110_93602.493.png" ], "turn_id": 8, "scan_id": "Validation_47334110", "origin_question": "At the end of which turn were both of \"the sleek oval white porcelain toilet\" and \"the medium-sized, sleek stainless steel box\" on your right side?", "option": [], "answer": "6", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the sleek oval white porcelain toilet\" and \"the medium-sized, sleek stainless steel box\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47334110/Validation_47334110_47334110_93604.493.png", "Validation_47334110/Validation_47334110_47334110_93606.492.png", "Validation_47334110/Validation_47334110_47334110_93608.491.png", "Validation_47334110/Validation_47334110_47334110_93610.490.png", "Validation_47334110/Validation_47334110_47334110_93612.489.png" ], "turn_id": 9, "scan_id": "Validation_47334110", "origin_question": "", "option": [], "answer": "", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47334110/Validation_47334110_47334110_93614.489.png", "Validation_47334110/Validation_47334110_47334110_93616.488.png", "Validation_47334110/Validation_47334110_47334110_93618.487.png", "Validation_47334110/Validation_47334110_47334110_93620.486.png", "Validation_47334110/Validation_47334110_47334110_93622.502.png" ], "turn_id": 10, "scan_id": "Validation_47334110", "origin_question": "Which direction is the sleek oval white porcelain toilet in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the sleek oval white porcelain toilet in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_45261185/Validation_45261185_45261185_3547.503.png", "Validation_45261185/Validation_45261185_45261185_3549.485.png", "Validation_45261185/Validation_45261185_45261185_3551.484.png", "Validation_45261185/Validation_45261185_45261185_3553.484.png" ], "turn_id": 1, "scan_id": "Validation_45261185", "origin_question": "Remember, have you seen any pillow(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any pillow(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_45261185/Validation_45261185_45261185_3555.483.png", "Validation_45261185/Validation_45261185_45261185_3557.482.png", "Validation_45261185/Validation_45261185_45261185_3559.481.png", "Validation_45261185/Validation_45261185_45261185_3561.480.png" ], "turn_id": 2, "scan_id": "Validation_45261185", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_45261185/Validation_45261185_45261185_3563.480.png", "Validation_45261185/Validation_45261185_45261185_3565.479.png", "Validation_45261185/Validation_45261185_45261185_3567.478.png", "Validation_45261185/Validation_45261185_45261185_3569.494.png" ], "turn_id": 3, "scan_id": "Validation_45261185", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the sleek, light-colored bed with no headboard\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the sleek, light-colored bed with no headboard\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_45261185/Validation_45261185_45261185_3571.493.png", "Validation_45261185/Validation_45261185_45261185_3573.492.png", "Validation_45261185/Validation_45261185_45261185_3575.492.png", "Validation_45261185/Validation_45261185_45261185_3577.491.png" ], "turn_id": 4, "scan_id": "Validation_45261185", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the sleek rectangular monitor with the dark frame\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the sleek rectangular monitor with the dark frame\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_45261185/Validation_45261185_45261185_3579.490.png", "Validation_45261185/Validation_45261185_45261185_3581.489.png", "Validation_45261185/Validation_45261185_45261185_3583.488.png", "Validation_45261185/Validation_45261185_45261185_3585.488.png" ], "turn_id": 5, "scan_id": "Validation_45261185", "origin_question": "What will be the first-time appearance order of the following categories: bed, clothes, monitor?", "option": [ "monitor, bed, clothes", "bed, monitor, clothes", "clothes, bed, monitor", "clothes, monitor, bed" ], "answer": "bed, monitor, clothes", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bed, clothes, monitor? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['monitor, bed, clothes', 'bed, monitor, clothes', 'clothes, bed, monitor', 'clothes, monitor, bed'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_45261185/Validation_45261185_45261185_3587.487.png", "Validation_45261185/Validation_45261185_45261185_3589.486.png", "Validation_45261185/Validation_45261185_45261185_3591.485.png", "Validation_45261185/Validation_45261185_45261185_3593.485.png" ], "turn_id": 6, "scan_id": "Validation_45261185", "origin_question": "Assuming the direction you are facing at the end of the turn 5 is forward, did you move a certain distance forward or backward from that position?", "option": [ "Forward", "Backward" ], "answer": "Forward", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 5 is forward, did you move a certain distance forward or backward from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Forward', 'Backward'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_45261185/Validation_45261185_45261185_3595.484.png", "Validation_45261185/Validation_45261185_45261185_3597.483.png", "Validation_45261185/Validation_45261185_45261185_3599.482.png", "Validation_45261185/Validation_45261185_45261185_3601.481.png" ], "turn_id": 7, "scan_id": "Validation_45261185", "origin_question": "Which direction is the sleek, light-colored bed with no headboard in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the sleek, light-colored bed with no headboard in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_45261185/Validation_45261185_45261185_3603.481.png", "Validation_45261185/Validation_45261185_45261185_3605.480.png", "Validation_45261185/Validation_45261185_45261185_3607.479.png", "Validation_45261185/Validation_45261185_45261185_3609.478.png" ], "turn_id": 8, "scan_id": "Validation_45261185", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the stack of neatly arranged, colorful clothes\" from your position?", "option": [], "answer": "115.05", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the stack of neatly arranged, colorful clothes\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_45662979/Validation_45662979_45662979_26029.749.png", "Validation_45662979/Validation_45662979_45662979_26031.232.png", "Validation_45662979/Validation_45662979_45662979_26032.731.png", "Validation_45662979/Validation_45662979_45662979_26034.231.png" ], "turn_id": 1, "scan_id": "Validation_45662979", "origin_question": "Remember, have you seen any sink(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any sink(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_45662979/Validation_45662979_45662979_26035.730.png", "Validation_45662979/Validation_45662979_45662979_26037.229.png", "Validation_45662979/Validation_45662979_45662979_26038.729.png", "Validation_45662979/Validation_45662979_45662979_26040.228.png" ], "turn_id": 2, "scan_id": "Validation_45662979", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the tall vertical shelf with light wood shelves\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the tall vertical shelf with light wood shelves\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_45662979/Validation_45662979_45662979_26041.728.png", "Validation_45662979/Validation_45662979_45662979_26043.227.png", "Validation_45662979/Validation_45662979_45662979_26044.726.png", "Validation_45662979/Validation_45662979_45662979_26046.226.png" ], "turn_id": 3, "scan_id": "Validation_45662979", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the sink\", Y: \"the large rectangular woven basket in natural brown\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the sink\", Y: \"the large rectangular woven basket in natural brown\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_45662979/Validation_45662979_45662979_26047.725.png", "Validation_45662979/Validation_45662979_45662979_26049.225.png", "Validation_45662979/Validation_45662979_45662979_26050.724.png", "Validation_45662979/Validation_45662979_45662979_26052.223.png" ], "turn_id": 4, "scan_id": "Validation_45662979", "origin_question": "Which two objects are on the same side of you now? \"the lush green plant in the medium-sized pot\", \"the rectangular window with the white frame\", \"the tall vertical shelf with light wood shelves\". ", "option": [ "the lush green plant in the medium-sized pot, the rectangular window with the white frame", "the lush green plant in the medium-sized pot, the tall vertical shelf with light wood shelves", "the rectangular window with the white frame, the tall vertical shelf with light wood shelves" ], "answer": "the lush green plant in the medium-sized pot, the rectangular window with the white frame", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the lush green plant in the medium-sized pot\", \"the rectangular window with the white frame\", \"the tall vertical shelf with light wood shelves\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the lush green plant in the medium-sized pot, the rectangular window with the white frame', 'the lush green plant in the medium-sized pot, the tall vertical shelf with light wood shelves', 'the rectangular window with the white frame, the tall vertical shelf with light wood shelves'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47331319/Validation_47331319_47331319_179039.062.png", "Validation_47331319/Validation_47331319_47331319_179041.028.png", "Validation_47331319/Validation_47331319_47331319_179043.027.png", "Validation_47331319/Validation_47331319_47331319_179045.026.png" ], "turn_id": 1, "scan_id": "Validation_47331319", "origin_question": "Remember, have you seen any chair(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any chair(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_47331319/Validation_47331319_47331319_179047.025.png", "Validation_47331319/Validation_47331319_47331319_179049.041.png", "Validation_47331319/Validation_47331319_47331319_179051.040.png", "Validation_47331319/Validation_47331319_47331319_179053.040.png" ], "turn_id": 2, "scan_id": "Validation_47331319", "origin_question": "In which turn did you see both \"the sleek angular stainless steel range hood\" and \"the medium-sized, muted brown ceramic flowerpot\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the sleek angular stainless steel range hood\" and \"the medium-sized, muted brown ceramic flowerpot\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_47331319/Validation_47331319_47331319_179055.039.png", "Validation_47331319/Validation_47331319_47331319_179057.038.png", "Validation_47331319/Validation_47331319_47331319_179059.037.png", "Validation_47331319/Validation_47331319_47331319_179061.036.png" ], "turn_id": 3, "scan_id": "Validation_47331319", "origin_question": "In which turn did you see both \"the refrigerator\" and \"the glossy black glass-ceramic stove with the sleek design\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the refrigerator\" and \"the glossy black glass-ceramic stove with the sleek design\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_47331319/Validation_47331319_47331319_179063.035.png", "Validation_47331319/Validation_47331319_47331319_179065.035.png", "Validation_47331319/Validation_47331319_47331319_179067.034.png", "Validation_47331319/Validation_47331319_47331319_179069.033.png" ], "turn_id": 4, "scan_id": "Validation_47331319", "origin_question": "Please recall \"the medium-sized, muted brown ceramic flowerpot\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.27", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the medium-sized, muted brown ceramic flowerpot\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47331319/Validation_47331319_47331319_179071.032.png", "Validation_47331319/Validation_47331319_47331319_179073.031.png", "Validation_47331319/Validation_47331319_47331319_179075.031.png", "Validation_47331319/Validation_47331319_47331319_179077.030.png" ], "turn_id": 5, "scan_id": "Validation_47331319", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the sleek silver microwave with dark accents\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the sleek silver microwave with dark accents\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47331319/Validation_47331319_47331319_179079.029.png", "Validation_47331319/Validation_47331319_47331319_179081.028.png", "Validation_47331319/Validation_47331319_47331319_179083.027.png", "Validation_47331319/Validation_47331319_47331319_179085.027.png" ], "turn_id": 6, "scan_id": "Validation_47331319", "origin_question": "Remember, have you seen any range hood(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any range hood(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "Validation_47331319/Validation_47331319_47331319_179087.026.png", "Validation_47331319/Validation_47331319_47331319_179089.025.png", "Validation_47331319/Validation_47331319_47331319_179091.041.png", "Validation_47331319/Validation_47331319_47331319_179093.040.png" ], "turn_id": 7, "scan_id": "Validation_47331319", "origin_question": "Among these three objects(\"the glossy black glass-ceramic stove with the sleek design\",\"the oven\",\"the circular stainless steel pan with the metallic finish\" ), which one is the closest to you?", "option": [ "the glossy black glass-ceramic stove with the sleek design", "the oven", "the circular stainless steel pan with the metallic finish" ], "answer": "the glossy black glass-ceramic stove with the sleek design", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects(\"the glossy black glass-ceramic stove with the sleek design\",\"the oven\",\"the circular stainless steel pan with the metallic finish\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the glossy black glass-ceramic stove with the sleek design', 'the oven', 'the circular stainless steel pan with the metallic finish'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47334239/Validation_47334239_47334239_111104.040.png", "Validation_47334239/Validation_47334239_47334239_111106.023.png", "Validation_47334239/Validation_47334239_47334239_111108.022.png", "Validation_47334239/Validation_47334239_47334239_111110.021.png" ], "turn_id": 1, "scan_id": "Validation_47334239", "origin_question": "Remember, have you seen any cabinet(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any cabinet(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_47334239/Validation_47334239_47334239_111112.020.png", "Validation_47334239/Validation_47334239_47334239_111114.019.png", "Validation_47334239/Validation_47334239_47334239_111116.019.png", "Validation_47334239/Validation_47334239_47334239_111118.018.png" ], "turn_id": 2, "scan_id": "Validation_47334239", "origin_question": "When did you first discover \"the sleek white glossy cabinet with the rectangular shape\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the sleek white glossy cabinet with the rectangular shape\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47334239/Validation_47334239_47334239_111120.017.png", "Validation_47334239/Validation_47334239_47334239_111122.016.png", "Validation_47334239/Validation_47334239_47334239_111124.015.png", "Validation_47334239/Validation_47334239_47334239_111126.015.png" ], "turn_id": 3, "scan_id": "Validation_47334239", "origin_question": "In which turn were you closest to \"the medium-sized white oval porcelain sink\"? ", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the medium-sized white oval porcelain sink\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47334239/Validation_47334239_47334239_111128.030.png", "Validation_47334239/Validation_47334239_47334239_111130.030.png", "Validation_47334239/Validation_47334239_47334239_111132.029.png", "Validation_47334239/Validation_47334239_47334239_111134.028.png" ], "turn_id": 4, "scan_id": "Validation_47334239", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the sleek white front-loading washing machine\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the sleek white front-loading washing machine\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_47334239/Validation_47334239_47334239_111136.027.png", "Validation_47334239/Validation_47334239_47334239_111138.026.png", "Validation_47334239/Validation_47334239_47334239_111140.026.png", "Validation_47334239/Validation_47334239_47334239_111142.025.png" ], "turn_id": 5, "scan_id": "Validation_47334239", "origin_question": "Which two objects are on the same side of you now? \"the medium-sized white oval porcelain sink\", \"the sleek, rounded white porcelain toilet\", \"the sleek white front-loading washing machine\". ", "option": [ "the medium-sized white oval porcelain sink, the sleek, rounded white porcelain toilet", "the medium-sized white oval porcelain sink, the sleek white front-loading washing machine", "the sleek, rounded white porcelain toilet, the sleek white front-loading washing machine" ], "answer": "the medium-sized white oval porcelain sink, the sleek, rounded white porcelain toilet", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the medium-sized white oval porcelain sink\", \"the sleek, rounded white porcelain toilet\", \"the sleek white front-loading washing machine\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the medium-sized white oval porcelain sink, the sleek, rounded white porcelain toilet', 'the medium-sized white oval porcelain sink, the sleek white front-loading washing machine', 'the sleek, rounded white porcelain toilet, the sleek white front-loading washing machine'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_47334239/Validation_47334239_47334239_111144.024.png", "Validation_47334239/Validation_47334239_47334239_111146.023.png", "Validation_47334239/Validation_47334239_47334239_111148.022.png", "Validation_47334239/Validation_47334239_47334239_111150.022.png" ], "turn_id": 6, "scan_id": "Validation_47334239", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "62.33", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_45260903/Validation_45260903_45260903_1519.114.png", "Validation_45260903/Validation_45260903_45260903_1521.096.png", "Validation_45260903/Validation_45260903_45260903_1523.095.png", "Validation_45260903/Validation_45260903_45260903_1525.095.png", "Validation_45260903/Validation_45260903_45260903_1527.094.png" ], "turn_id": 1, "scan_id": "Validation_45260903", "origin_question": "Remember, have you seen any bicycle(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any bicycle(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_45260903/Validation_45260903_45260903_1529.093.png", "Validation_45260903/Validation_45260903_45260903_1531.092.png", "Validation_45260903/Validation_45260903_45260903_1533.091.png", "Validation_45260903/Validation_45260903_45260903_1535.091.png", "Validation_45260903/Validation_45260903_45260903_1537.090.png" ], "turn_id": 2, "scan_id": "Validation_45260903", "origin_question": "When did you first discover \"the medium-sized indoor plant with broad, glossy leaves\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the medium-sized indoor plant with broad, glossy leaves\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_45260903/Validation_45260903_45260903_1539.089.png", "Validation_45260903/Validation_45260903_45260903_1541.088.png", "Validation_45260903/Validation_45260903_45260903_1543.088.png", "Validation_45260903/Validation_45260903_45260903_1545.087.png", "Validation_45260903/Validation_45260903_45260903_1547.103.png" ], "turn_id": 3, "scan_id": "Validation_45260903", "origin_question": "What will be the first-time appearance order of the following categories: cabinet, fireplace, pillow?", "option": [ "fireplace, cabinet, pillow", "pillow, fireplace, cabinet", "cabinet, pillow, fireplace", "pillow, cabinet, fireplace" ], "answer": "pillow, cabinet, fireplace", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: cabinet, fireplace, pillow? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['fireplace, cabinet, pillow', 'pillow, fireplace, cabinet', 'cabinet, pillow, fireplace', 'pillow, cabinet, fireplace'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_45260903/Validation_45260903_45260903_1549.102.png", "Validation_45260903/Validation_45260903_45260903_1551.101.png", "Validation_45260903/Validation_45260903_45260903_1553.100.png", "Validation_45260903/Validation_45260903_45260903_1555.100.png", "Validation_45260903/Validation_45260903_45260903_1557.099.png" ], "turn_id": 4, "scan_id": "Validation_45260903", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_45260903/Validation_45260903_45260903_1559.098.png", "Validation_45260903/Validation_45260903_45260903_1561.097.png", "Validation_45260903/Validation_45260903_45260903_1563.096.png", "Validation_45260903/Validation_45260903_45260903_1565.095.png", "Validation_45260903/Validation_45260903_45260903_1567.095.png" ], "turn_id": 5, "scan_id": "Validation_45260903", "origin_question": "Which direction is the medium-sized monitor with the slim matte black frame in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the medium-sized monitor with the slim matte black frame in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_45260903/Validation_45260903_45260903_1569.094.png", "Validation_45260903/Validation_45260903_45260903_1571.093.png", "Validation_45260903/Validation_45260903_45260903_1573.092.png", "Validation_45260903/Validation_45260903_45260903_1575.091.png", "Validation_45260903/Validation_45260903_45260903_1577.091.png" ], "turn_id": 6, "scan_id": "Validation_45260903", "origin_question": "Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the fireplace\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the fireplace\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_45260903/Validation_45260903_45260903_1579.090.png", "Validation_45260903/Validation_45260903_45260903_1581.089.png", "Validation_45260903/Validation_45260903_45260903_1583.088.png", "Validation_45260903/Validation_45260903_45260903_1585.088.png", "Validation_45260903/Validation_45260903_45260903_1587.087.png" ], "turn_id": 7, "scan_id": "Validation_45260903", "origin_question": "Please recall \"the medium-sized monitor with the slim matte black frame\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "4.11", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the medium-sized monitor with the slim matte black frame\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_45260903/Validation_45260903_45260903_1589.103.png", "Validation_45260903/Validation_45260903_45260903_1591.102.png", "Validation_45260903/Validation_45260903_45260903_1593.101.png", "Validation_45260903/Validation_45260903_45260903_1595.100.png", "Validation_45260903/Validation_45260903_45260903_1597.100.png" ], "turn_id": 8, "scan_id": "Validation_45260903", "origin_question": "Assuming the direction you are facing at the end of the turn 3 is forward, did you move a certain distance forward or backward from that position?", "option": [ "Forward", "Backward" ], "answer": "Forward", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 3 is forward, did you move a certain distance forward or backward from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Forward', 'Backward'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "Validation_42446114/Validation_42446114_42446114_78188.985.png", "Validation_42446114/Validation_42446114_42446114_78190.985.png", "Validation_42446114/Validation_42446114_42446114_78192.984.png", "Validation_42446114/Validation_42446114_42446114_78194.983.png", "Validation_42446114/Validation_42446114_42446114_78196.982.png" ], "turn_id": 1, "scan_id": "Validation_42446114", "origin_question": "Remember, how many chair(s) have you seen so far? ", "option": [], "answer": "3", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many chair(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "Validation_42446114/Validation_42446114_42446114_78198.981.png", "Validation_42446114/Validation_42446114_42446114_78200.981.png", "Validation_42446114/Validation_42446114_42446114_78202.997.png", "Validation_42446114/Validation_42446114_42446114_78204.996.png", "Validation_42446114/Validation_42446114_42446114_78206.995.png" ], "turn_id": 2, "scan_id": "Validation_42446114", "origin_question": "Remember, how many shower(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many shower(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_42446114/Validation_42446114_42446114_78208.994.png", "Validation_42446114/Validation_42446114_42446114_78210.993.png", "Validation_42446114/Validation_42446114_42446114_78212.993.png", "Validation_42446114/Validation_42446114_42446114_78214.992.png", "Validation_42446114/Validation_42446114_42446114_78216.991.png" ], "turn_id": 3, "scan_id": "Validation_42446114", "origin_question": "In which turn were you closest to \"the sleek stainless steel built-in oven\"? ", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the sleek stainless steel built-in oven\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_42446114/Validation_42446114_42446114_78218.990.png", "Validation_42446114/Validation_42446114_42446114_78220.989.png", "Validation_42446114/Validation_42446114_42446114_78222.989.png", "Validation_42446114/Validation_42446114_42446114_78224.988.png", "Validation_42446114/Validation_42446114_42446114_78226.987.png" ], "turn_id": 4, "scan_id": "Validation_42446114", "origin_question": "When did you first discover \"the stylish cylindrical white bin with the black lid\"? (index of the turn)", "option": [], "answer": "4", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the stylish cylindrical white bin with the black lid\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_42446114/Validation_42446114_42446114_78228.986.png", "Validation_42446114/Validation_42446114_42446114_78230.985.png", "Validation_42446114/Validation_42446114_42446114_78232.985.png", "Validation_42446114/Validation_42446114_42446114_78234.984.png", "Validation_42446114/Validation_42446114_42446114_78236.983.png" ], "turn_id": 5, "scan_id": "Validation_42446114", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the three-seater gray fabric couch with contrasting cushions\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the three-seater gray fabric couch with contrasting cushions\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_42446114/Validation_42446114_42446114_78238.982.png", "Validation_42446114/Validation_42446114_42446114_78240.981.png", "Validation_42446114/Validation_42446114_42446114_78242.981.png", "Validation_42446114/Validation_42446114_42446114_78244.997.png", "Validation_42446114/Validation_42446114_42446114_78246.996.png" ], "turn_id": 6, "scan_id": "Validation_42446114", "origin_question": "In which turn were you closest to \"the three-seater gray fabric couch with contrasting cushions\"? ", "option": [], "answer": "3", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the three-seater gray fabric couch with contrasting cushions\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_42446114/Validation_42446114_42446114_78248.995.png", "Validation_42446114/Validation_42446114_42446114_78250.994.png", "Validation_42446114/Validation_42446114_42446114_78252.993.png", "Validation_42446114/Validation_42446114_42446114_78254.993.png", "Validation_42446114/Validation_42446114_42446114_78256.992.png" ], "turn_id": 7, "scan_id": "Validation_42446114", "origin_question": "By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the slender, soft blue glossy vase\", Y: \"the medium-sized rectangular wooden shelf in soft yellow\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to X and farther away from Y", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the slender, soft blue glossy vase\", Y: \"the medium-sized rectangular wooden shelf in soft yellow\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_42446114/Validation_42446114_42446114_78258.991.png", "Validation_42446114/Validation_42446114_42446114_78260.990.png", "Validation_42446114/Validation_42446114_42446114_78262.989.png", "Validation_42446114/Validation_42446114_42446114_78264.989.png", "Validation_42446114/Validation_42446114_42446114_78266.988.png" ], "turn_id": 8, "scan_id": "Validation_42446114", "origin_question": "Is the slender, soft blue glossy vase to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the slender, soft blue glossy vase to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_42446114/Validation_42446114_42446114_78268.987.png", "Validation_42446114/Validation_42446114_42446114_78270.986.png", "Validation_42446114/Validation_42446114_42446114_78272.985.png", "Validation_42446114/Validation_42446114_42446114_78274.985.png", "Validation_42446114/Validation_42446114_42446114_78276.984.png" ], "turn_id": 9, "scan_id": "Validation_42446114", "origin_question": "Assuming the direction you are facing at the end of the turn 6 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Right", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 6 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_42446114/Validation_42446114_42446114_78278.983.png", "Validation_42446114/Validation_42446114_42446114_78280.982.png", "Validation_42446114/Validation_42446114_42446114_78282.981.png", "Validation_42446114/Validation_42446114_42446114_78284.997.png", "Validation_42446114/Validation_42446114_42446114_78286.997.png" ], "turn_id": 10, "scan_id": "Validation_42446114", "origin_question": "Compared to your position at the end of the 8 turn, are you now closer or farther away from \"the slender, soft blue glossy vase\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 8 turn, are you now closer or farther away from \"the slender, soft blue glossy vase\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "Validation_48458472/Validation_48458472_48458472_110525.154.png", "Validation_48458472/Validation_48458472_48458472_110527.154.png", "Validation_48458472/Validation_48458472_48458472_110529.153.png", "Validation_48458472/Validation_48458472_48458472_110531.152.png", "Validation_48458472/Validation_48458472_48458472_110533.151.png" ], "turn_id": 1, "scan_id": "Validation_48458472", "origin_question": "Remember, how many pillow(s) have you seen so far? ", "option": [], "answer": "3", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many pillow(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_48458472/Validation_48458472_48458472_110535.150.png", "Validation_48458472/Validation_48458472_48458472_110537.150.png", "Validation_48458472/Validation_48458472_48458472_110539.149.png", "Validation_48458472/Validation_48458472_48458472_110541.148.png", "Validation_48458472/Validation_48458472_48458472_110543.147.png" ], "turn_id": 2, "scan_id": "Validation_48458472", "origin_question": "Which two objects are on the same side of you now? \"the rich, deep brown rectangular wooden table\", \"the medium-sized lamp with the spherical metal base\", \"the monitor\". ", "option": [ "the rich, deep brown rectangular wooden table, the medium-sized lamp with the spherical metal base", "the rich, deep brown rectangular wooden table, the monitor", "the medium-sized lamp with the spherical metal base, the monitor" ], "answer": "the rich, deep brown rectangular wooden table, the medium-sized lamp with the spherical metal base", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the rich, deep brown rectangular wooden table\", \"the medium-sized lamp with the spherical metal base\", \"the monitor\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rich, deep brown rectangular wooden table, the medium-sized lamp with the spherical metal base', 'the rich, deep brown rectangular wooden table, the monitor', 'the medium-sized lamp with the spherical metal base, the monitor'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_48458472/Validation_48458472_48458472_110545.146.png", "Validation_48458472/Validation_48458472_48458472_110547.146.png", "Validation_48458472/Validation_48458472_48458472_110549.145.png", "Validation_48458472/Validation_48458472_48458472_110551.144.png", "Validation_48458472/Validation_48458472_48458472_110553.143.png" ], "turn_id": 3, "scan_id": "Validation_48458472", "origin_question": "Which direction is the large round glossy white plate in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the large round glossy white plate in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_48458472/Validation_48458472_48458472_110555.143.png", "Validation_48458472/Validation_48458472_48458472_110557.142.png", "Validation_48458472/Validation_48458472_48458472_110559.141.png", "Validation_48458472/Validation_48458472_48458472_110561.140.png", "Validation_48458472/Validation_48458472_48458472_110563.139.png" ], "turn_id": 4, "scan_id": "Validation_48458472", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the large round glossy white plate\" from your position?", "option": [], "answer": "91.61", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the large round glossy white plate\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_48458472/Validation_48458472_48458472_110565.155.png", "Validation_48458472/Validation_48458472_48458472_110567.154.png", "Validation_48458472/Validation_48458472_48458472_110569.154.png", "Validation_48458472/Validation_48458472_48458472_110571.153.png", "Validation_48458472/Validation_48458472_48458472_110573.152.png" ], "turn_id": 5, "scan_id": "Validation_48458472", "origin_question": "By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the large round glossy white plate\", Y: \"the monitor\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "farther away from both X and Y", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the large round glossy white plate\", Y: \"the monitor\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_48458472/Validation_48458472_48458472_110575.151.png", "Validation_48458472/Validation_48458472_48458472_110577.151.png", "Validation_48458472/Validation_48458472_48458472_110579.150.png", "Validation_48458472/Validation_48458472_48458472_110581.149.png", "Validation_48458472/Validation_48458472_48458472_110583.148.png" ], "turn_id": 6, "scan_id": "Validation_48458472", "origin_question": "When did you first discover \"the medium-sized lamp with the spherical metal base\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the medium-sized lamp with the spherical metal base\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_48458472/Validation_48458472_48458472_110585.247.png", "Validation_48458472/Validation_48458472_48458472_110587.246.png", "Validation_48458472/Validation_48458472_48458472_110589.246.png", "Validation_48458472/Validation_48458472_48458472_110591.245.png", "Validation_48458472/Validation_48458472_48458472_110593.244.png" ], "turn_id": 7, "scan_id": "Validation_48458472", "origin_question": "Assuming the direction you are facing at the end of the turn 5 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Right", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 5 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_48458472/Validation_48458472_48458472_110595.243.png", "Validation_48458472/Validation_48458472_48458472_110597.242.png", "Validation_48458472/Validation_48458472_48458472_110599.242.png", "Validation_48458472/Validation_48458472_48458472_110601.241.png", "Validation_48458472/Validation_48458472_48458472_110603.240.png" ], "turn_id": 8, "scan_id": "Validation_48458472", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the vibrant, cube-shaped toy in red, yellow, green, and blue\" from your position?", "option": [], "answer": "148.62", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the vibrant, cube-shaped toy in red, yellow, green, and blue\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_48458472/Validation_48458472_48458472_110605.239.png", "Validation_48458472/Validation_48458472_48458472_110607.255.png", "Validation_48458472/Validation_48458472_48458472_110609.254.png", "Validation_48458472/Validation_48458472_48458472_110611.254.png", "Validation_48458472/Validation_48458472_48458472_110613.253.png" ], "turn_id": 9, "scan_id": "Validation_48458472", "origin_question": "At the end of which turn were both of \"the sleek, polished metal teapot with the classic design\" and \"the vibrant, cube-shaped toy in red, yellow, green, and blue\" on your left side?", "option": [], "answer": "9", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the sleek, polished metal teapot with the classic design\" and \"the vibrant, cube-shaped toy in red, yellow, green, and blue\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42899461/Validation_42899461_42899461_195785.452.png", "Validation_42899461/Validation_42899461_42899461_195787.435.png", "Validation_42899461/Validation_42899461_42899461_195789.434.png", "Validation_42899461/Validation_42899461_42899461_195791.433.png", "Validation_42899461/Validation_42899461_42899461_195793.432.png" ], "turn_id": 1, "scan_id": "Validation_42899461", "origin_question": "Remember, have you seen any chair(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any chair(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_42899461/Validation_42899461_42899461_195795.431.png", "Validation_42899461/Validation_42899461_42899461_195797.430.png", "Validation_42899461/Validation_42899461_42899461_195799.430.png", "Validation_42899461/Validation_42899461_42899461_195801.429.png", "Validation_42899461/Validation_42899461_42899461_195803.428.png" ], "turn_id": 2, "scan_id": "Validation_42899461", "origin_question": "When did you first discover \"the compact, healthy green plant in the glass container\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the compact, healthy green plant in the glass container\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_42899461/Validation_42899461_42899461_195805.427.png", "Validation_42899461/Validation_42899461_42899461_195807.426.png", "Validation_42899461/Validation_42899461_42899461_195809.425.png", "Validation_42899461/Validation_42899461_42899461_195811.425.png", "Validation_42899461/Validation_42899461_42899461_195813.424.png" ], "turn_id": 3, "scan_id": "Validation_42899461", "origin_question": "Which two objects are on the same side of you now? \"an elegant pendant lamp with antique metallic finish\", \"the light rectangular cotton towel\", \"the spiral ladder made of warm, golden-brown wood\". ", "option": [ "an elegant pendant lamp with antique metallic finish, the light rectangular cotton towel", "an elegant pendant lamp with antique metallic finish, the spiral ladder made of warm, golden-brown wood", "the light rectangular cotton towel, the spiral ladder made of warm, golden-brown wood" ], "answer": "an elegant pendant lamp with antique metallic finish, the light rectangular cotton towel", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"an elegant pendant lamp with antique metallic finish\", \"the light rectangular cotton towel\", \"the spiral ladder made of warm, golden-brown wood\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['an elegant pendant lamp with antique metallic finish, the light rectangular cotton towel', 'an elegant pendant lamp with antique metallic finish, the spiral ladder made of warm, golden-brown wood', 'the light rectangular cotton towel, the spiral ladder made of warm, golden-brown wood'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_42899461/Validation_42899461_42899461_195815.423.png", "Validation_42899461/Validation_42899461_42899461_195817.422.png", "Validation_42899461/Validation_42899461_42899461_195819.421.png", "Validation_42899461/Validation_42899461_42899461_195821.421.png", "Validation_42899461/Validation_42899461_42899461_195823.420.png" ], "turn_id": 4, "scan_id": "Validation_42899461", "origin_question": "Please recall \"an elegant pendant lamp with antique metallic finish\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "5.4", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"an elegant pendant lamp with antique metallic finish\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_42899461/Validation_42899461_42899461_195825.435.png", "Validation_42899461/Validation_42899461_42899461_195827.435.png", "Validation_42899461/Validation_42899461_42899461_195829.434.png", "Validation_42899461/Validation_42899461_42899461_195831.433.png", "Validation_42899461/Validation_42899461_42899461_195833.432.png" ], "turn_id": 5, "scan_id": "Validation_42899461", "origin_question": "", "option": [], "answer": "", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_42899461/Validation_42899461_42899461_195835.431.png", "Validation_42899461/Validation_42899461_42899461_195837.430.png", "Validation_42899461/Validation_42899461_42899461_195839.430.png", "Validation_42899461/Validation_42899461_42899461_195841.429.png", "Validation_42899461/Validation_42899461_42899461_195843.428.png" ], "turn_id": 6, "scan_id": "Validation_42899461", "origin_question": "What will be the first-time appearance order of the following categories: fireplace, microwave, plant?", "option": [ "fireplace, microwave, plant", "fireplace, plant, microwave", "microwave, fireplace, plant", "plant, microwave, fireplace" ], "answer": "fireplace, plant, microwave", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: fireplace, microwave, plant? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['fireplace, microwave, plant', 'fireplace, plant, microwave', 'microwave, fireplace, plant', 'plant, microwave, fireplace'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_42899461/Validation_42899461_42899461_195845.427.png", "Validation_42899461/Validation_42899461_42899461_195847.426.png", "Validation_42899461/Validation_42899461_42899461_195849.426.png", "Validation_42899461/Validation_42899461_42899461_195851.425.png", "Validation_42899461/Validation_42899461_42899461_195853.424.png" ], "turn_id": 7, "scan_id": "Validation_42899461", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "1.34", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "Validation_42899461/Validation_42899461_42899461_195855.423.png", "Validation_42899461/Validation_42899461_42899461_195857.422.png", "Validation_42899461/Validation_42899461_42899461_195859.421.png", "Validation_42899461/Validation_42899461_42899461_195861.421.png", "Validation_42899461/Validation_42899461_42899461_195863.420.png" ], "turn_id": 8, "scan_id": "Validation_42899461", "origin_question": "Among these three objects(\"the deep brown rectangular couch with cushions\",\"the light rectangular cotton towel\",\"the compact metallic gray machine with the sleek design\" ), which one is the farthest from you?", "option": [ "the deep brown rectangular couch with cushions", "the light rectangular cotton towel", "the compact metallic gray machine with the sleek design" ], "answer": "the compact metallic gray machine with the sleek design", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the deep brown rectangular couch with cushions\",\"the light rectangular cotton towel\",\"the compact metallic gray machine with the sleek design\" ), which one is the farthest from you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the deep brown rectangular couch with cushions', 'the light rectangular cotton towel', 'the compact metallic gray machine with the sleek design'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_42899461/Validation_42899461_42899461_195865.419.png", "Validation_42899461/Validation_42899461_42899461_195867.418.png", "Validation_42899461/Validation_42899461_42899461_195869.434.png", "Validation_42899461/Validation_42899461_42899461_195871.433.png", "Validation_42899461/Validation_42899461_42899461_195873.432.png" ], "turn_id": 9, "scan_id": "Validation_42899461", "origin_question": "When did you first discover \"the sleek teal stainless steel kettle\"? (index of the turn)", "option": [], "answer": "7", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the sleek teal stainless steel kettle\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_42899461/Validation_42899461_42899461_195875.431.png", "Validation_42899461/Validation_42899461_42899461_195877.431.png", "Validation_42899461/Validation_42899461_42899461_195879.430.png", "Validation_42899461/Validation_42899461_42899461_195881.429.png", "Validation_42899461/Validation_42899461_42899461_195883.428.png" ], "turn_id": 10, "scan_id": "Validation_42899461", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the light rectangular cotton towel\" from your position?", "option": [], "answer": "83.3", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the light rectangular cotton towel\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47334103/Validation_47334103_47334103_92683.567.png", "Validation_47334103/Validation_47334103_47334103_92685.549.png", "Validation_47334103/Validation_47334103_47334103_92687.549.png", "Validation_47334103/Validation_47334103_47334103_92689.548.png", "Validation_47334103/Validation_47334103_47334103_92691.547.png" ], "turn_id": 1, "scan_id": "Validation_47334103", "origin_question": "Remember, have you seen any monitor(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any monitor(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_47334103/Validation_47334103_47334103_92693.546.png", "Validation_47334103/Validation_47334103_47334103_92695.545.png", "Validation_47334103/Validation_47334103_47334103_92697.545.png", "Validation_47334103/Validation_47334103_47334103_92699.560.png", "Validation_47334103/Validation_47334103_47334103_92701.560.png" ], "turn_id": 2, "scan_id": "Validation_47334103", "origin_question": "In which turn did you see both \"the tall blue wooden cabinet with closed doors\" and \"the glossy white rectangular sink\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the tall blue wooden cabinet with closed doors\" and \"the glossy white rectangular sink\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47334103/Validation_47334103_47334103_92703.559.png", "Validation_47334103/Validation_47334103_47334103_92705.558.png", "Validation_47334103/Validation_47334103_47334103_92707.557.png", "Validation_47334103/Validation_47334103_47334103_92709.556.png", "Validation_47334103/Validation_47334103_47334103_92711.556.png" ], "turn_id": 3, "scan_id": "Validation_47334103", "origin_question": "", "option": [], "answer": "", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_47334103/Validation_47334103_47334103_92713.555.png", "Validation_47334103/Validation_47334103_47334103_92715.554.png", "Validation_47334103/Validation_47334103_47334103_92717.553.png", "Validation_47334103/Validation_47334103_47334103_92719.552.png", "Validation_47334103/Validation_47334103_47334103_92721.552.png" ], "turn_id": 4, "scan_id": "Validation_47334103", "origin_question": "What will be the first-time appearance order of the following categories: monitor, sink, stool?", "option": [ "sink, monitor, stool", "stool, monitor, sink", "stool, sink, monitor", "monitor, stool, sink" ], "answer": "stool, sink, monitor", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: monitor, sink, stool? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['sink, monitor, stool', 'stool, monitor, sink', 'stool, sink, monitor', 'monitor, stool, sink'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47334103/Validation_47334103_47334103_92723.551.png", "Validation_47334103/Validation_47334103_47334103_92725.550.png", "Validation_47334103/Validation_47334103_47334103_92727.549.png", "Validation_47334103/Validation_47334103_47334103_92729.548.png", "Validation_47334103/Validation_47334103_47334103_92731.548.png" ], "turn_id": 5, "scan_id": "Validation_47334103", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the glossy white rectangular sink\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the glossy white rectangular sink\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_47334103/Validation_47334103_47334103_92733.547.png", "Validation_47334103/Validation_47334103_47334103_92735.546.png", "Validation_47334103/Validation_47334103_47334103_92737.545.png", "Validation_47334103/Validation_47334103_47334103_92739.544.png", "Validation_47334103/Validation_47334103_47334103_92741.560.png" ], "turn_id": 6, "scan_id": "Validation_47334103", "origin_question": "Please recall \"the stool\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "4.09", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the stool\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_47334103/Validation_47334103_47334103_92743.559.png", "Validation_47334103/Validation_47334103_47334103_92745.558.png", "Validation_47334103/Validation_47334103_47334103_92747.558.png", "Validation_47334103/Validation_47334103_47334103_92749.557.png", "Validation_47334103/Validation_47334103_47334103_92751.556.png" ], "turn_id": 7, "scan_id": "Validation_47334103", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "66.41", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_47334103/Validation_47334103_47334103_92753.555.png", "Validation_47334103/Validation_47334103_47334103_92755.554.png", "Validation_47334103/Validation_47334103_47334103_92757.554.png", "Validation_47334103/Validation_47334103_47334103_92759.553.png", "Validation_47334103/Validation_47334103_47334103_92761.552.png" ], "turn_id": 8, "scan_id": "Validation_47334103", "origin_question": "Is the stool to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the stool to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "Validation_47333940/Validation_47333940_47333940_85410.981.png", "Validation_47333940/Validation_47333940_47333940_85412.963.png", "Validation_47333940/Validation_47333940_47333940_85414.962.png", "Validation_47333940/Validation_47333940_47333940_85416.978.png" ], "turn_id": 1, "scan_id": "Validation_47333940", "origin_question": "Remember, how many album(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many album(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47333940/Validation_47333940_47333940_85418.977.png", "Validation_47333940/Validation_47333940_47333940_85420.977.png", "Validation_47333940/Validation_47333940_47333940_85422.976.png", "Validation_47333940/Validation_47333940_47333940_85424.975.png" ], "turn_id": 2, "scan_id": "Validation_47333940", "origin_question": "In which turn were you closest to \"the white door with the frosted glass panel\"? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the white door with the frosted glass panel\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_47333940/Validation_47333940_47333940_85426.974.png", "Validation_47333940/Validation_47333940_47333940_85428.973.png", "Validation_47333940/Validation_47333940_47333940_85430.973.png", "Validation_47333940/Validation_47333940_47333940_85432.972.png" ], "turn_id": 3, "scan_id": "Validation_47333940", "origin_question": "When did you first discover \"the radiator\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the radiator\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_47333940/Validation_47333940_47333940_85434.971.png", "Validation_47333940/Validation_47333940_47333940_85436.970.png", "Validation_47333940/Validation_47333940_47333940_85438.970.png", "Validation_47333940/Validation_47333940_47333940_85440.969.png" ], "turn_id": 4, "scan_id": "Validation_47333940", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "1.41", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47333940/Validation_47333940_47333940_85442.968.png", "Validation_47333940/Validation_47333940_47333940_85444.967.png", "Validation_47333940/Validation_47333940_47333940_85446.966.png", "Validation_47333940/Validation_47333940_47333940_85448.965.png" ], "turn_id": 5, "scan_id": "Validation_47333940", "origin_question": "Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the vibrant red rectangular plush pillow\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the vibrant red rectangular plush pillow\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47333940/Validation_47333940_47333940_85450.965.png", "Validation_47333940/Validation_47333940_47333940_85452.964.png", "Validation_47333940/Validation_47333940_47333940_85454.963.png", "Validation_47333940/Validation_47333940_47333940_85456.962.png" ], "turn_id": 6, "scan_id": "Validation_47333940", "origin_question": "", "option": [], "answer": "", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_47333940/Validation_47333940_47333940_85458.978.png", "Validation_47333940/Validation_47333940_47333940_85460.977.png", "Validation_47333940/Validation_47333940_47333940_85462.976.png", "Validation_47333940/Validation_47333940_47333940_85464.976.png" ], "turn_id": 7, "scan_id": "Validation_47333940", "origin_question": "Is the vibrant red rectangular plush pillow to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the vibrant red rectangular plush pillow to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47333940/Validation_47333940_47333940_85466.975.png", "Validation_47333940/Validation_47333940_47333940_85468.974.png", "Validation_47333940/Validation_47333940_47333940_85470.973.png", "Validation_47333940/Validation_47333940_47333940_85472.972.png" ], "turn_id": 8, "scan_id": "Validation_47333940", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the white door with the frosted glass panel\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the white door with the frosted glass panel\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_47333940/Validation_47333940_47333940_85474.972.png", "Validation_47333940/Validation_47333940_47333940_85476.971.png", "Validation_47333940/Validation_47333940_47333940_85478.970.png", "Validation_47333940/Validation_47333940_47333940_85480.969.png" ], "turn_id": 9, "scan_id": "Validation_47333940", "origin_question": "Using your orientation at the end of turn 5 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 9 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 5 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_45261619/Validation_45261619_45261619_55065.702.png", "Validation_45261619/Validation_45261619_45261619_55067.685.png", "Validation_45261619/Validation_45261619_45261619_55069.684.png", "Validation_45261619/Validation_45261619_45261619_55071.683.png" ], "turn_id": 1, "scan_id": "Validation_45261619", "origin_question": "Remember, have you seen any couch(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any couch(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "Validation_45261619/Validation_45261619_45261619_55073.683.png", "Validation_45261619/Validation_45261619_45261619_55075.682.png", "Validation_45261619/Validation_45261619_45261619_55077.681.png", "Validation_45261619/Validation_45261619_45261619_55079.680.png" ], "turn_id": 2, "scan_id": "Validation_45261619", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the metallic-framed neutral-colored window with blinds\"; \"the light-colored rectangular door\"; \"the rectangular gray and white picture with yellow notes\" ", "option": [ "the metallic-framed neutral-colored window with blinds", "the light-colored rectangular door", "the rectangular gray and white picture with yellow notes" ], "answer": "the metallic-framed neutral-colored window with blinds", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the metallic-framed neutral-colored window with blinds\"; \"the light-colored rectangular door\"; \"the rectangular gray and white picture with yellow notes\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the metallic-framed neutral-colored window with blinds', 'the light-colored rectangular door', 'the rectangular gray and white picture with yellow notes'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "Validation_45261619/Validation_45261619_45261619_55081.679.png", "Validation_45261619/Validation_45261619_45261619_55083.679.png", "Validation_45261619/Validation_45261619_45261619_55085.678.png", "Validation_45261619/Validation_45261619_45261619_55087.694.png" ], "turn_id": 3, "scan_id": "Validation_45261619", "origin_question": "Among these three objects(\"the sleek, glossy black rectangular monitor\",\"the rectangular gray and white picture with yellow notes\",\"the square, dark-colored pillow with smooth fabric\" ), which one is the closest to you?", "option": [ "the sleek, glossy black rectangular monitor", "the rectangular gray and white picture with yellow notes", "the square, dark-colored pillow with smooth fabric" ], "answer": "the sleek, glossy black rectangular monitor", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects(\"the sleek, glossy black rectangular monitor\",\"the rectangular gray and white picture with yellow notes\",\"the square, dark-colored pillow with smooth fabric\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the sleek, glossy black rectangular monitor', 'the rectangular gray and white picture with yellow notes', 'the square, dark-colored pillow with smooth fabric'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_45261619/Validation_45261619_45261619_55089.693.png", "Validation_45261619/Validation_45261619_45261619_55091.692.png", "Validation_45261619/Validation_45261619_45261619_55093.691.png", "Validation_45261619/Validation_45261619_45261619_55095.691.png" ], "turn_id": 4, "scan_id": "Validation_45261619", "origin_question": "Which two objects are on the same side of you now? \"the sleek, glossy black rectangular monitor\", \"the white rectangular wall-mounted radiator\", \"the metallic-framed neutral-colored window with blinds\". ", "option": [ "the sleek, glossy black rectangular monitor, the white rectangular wall-mounted radiator", "the sleek, glossy black rectangular monitor, the metallic-framed neutral-colored window with blinds", "the white rectangular wall-mounted radiator, the metallic-framed neutral-colored window with blinds" ], "answer": "the white rectangular wall-mounted radiator, the metallic-framed neutral-colored window with blinds", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the sleek, glossy black rectangular monitor\", \"the white rectangular wall-mounted radiator\", \"the metallic-framed neutral-colored window with blinds\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the sleek, glossy black rectangular monitor, the white rectangular wall-mounted radiator', 'the sleek, glossy black rectangular monitor, the metallic-framed neutral-colored window with blinds', 'the white rectangular wall-mounted radiator, the metallic-framed neutral-colored window with blinds'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_45261619/Validation_45261619_45261619_55097.690.png", "Validation_45261619/Validation_45261619_45261619_55099.689.png", "Validation_45261619/Validation_45261619_45261619_55101.688.png", "Validation_45261619/Validation_45261619_45261619_55103.687.png" ], "turn_id": 5, "scan_id": "Validation_45261619", "origin_question": "By the end of the 4 turn, were you getting closer to or farther away from these two objects: X: \"the light-colored rectangular door\", Y: \"the metallic-framed neutral-colored window with blinds\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "farther away from both X and Y", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 4 turn, were you getting closer to or farther away from these two objects: X: \"the light-colored rectangular door\", Y: \"the metallic-framed neutral-colored window with blinds\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_45261619/Validation_45261619_45261619_55105.687.png", "Validation_45261619/Validation_45261619_45261619_55107.686.png", "Validation_45261619/Validation_45261619_45261619_55109.685.png", "Validation_45261619/Validation_45261619_45261619_55111.684.png" ], "turn_id": 6, "scan_id": "Validation_45261619", "origin_question": "Which two objects are on the same side of you now? \"the vibrant plant with green leaves\", \"the rectangular gray and white picture with yellow notes\", \"the light-colored rectangular door\". ", "option": [ "the vibrant plant with green leaves, the rectangular gray and white picture with yellow notes", "the vibrant plant with green leaves, the light-colored rectangular door", "the rectangular gray and white picture with yellow notes, the light-colored rectangular door" ], "answer": "the vibrant plant with green leaves, the rectangular gray and white picture with yellow notes", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the vibrant plant with green leaves\", \"the rectangular gray and white picture with yellow notes\", \"the light-colored rectangular door\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the vibrant plant with green leaves, the rectangular gray and white picture with yellow notes', 'the vibrant plant with green leaves, the light-colored rectangular door', 'the rectangular gray and white picture with yellow notes, the light-colored rectangular door'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_45261619/Validation_45261619_45261619_55113.683.png", "Validation_45261619/Validation_45261619_45261619_55115.683.png", "Validation_45261619/Validation_45261619_45261619_55117.682.png", "Validation_45261619/Validation_45261619_45261619_55119.681.png" ], "turn_id": 7, "scan_id": "Validation_45261619", "origin_question": "Please recall \"the white rectangular wall-mounted radiator\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.13", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the white rectangular wall-mounted radiator\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_45261619/Validation_45261619_45261619_55121.680.png", "Validation_45261619/Validation_45261619_45261619_55123.679.png", "Validation_45261619/Validation_45261619_45261619_55125.679.png", "Validation_45261619/Validation_45261619_45261619_55127.678.png" ], "turn_id": 8, "scan_id": "Validation_45261619", "origin_question": "When did you first discover \"the rectangular gray and white picture with yellow notes\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the rectangular gray and white picture with yellow notes\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "Validation_47333918/Validation_47333918_47333918_82456.471.png", "Validation_47333918/Validation_47333918_47333918_82458.454.png", "Validation_47333918/Validation_47333918_47333918_82460.453.png", "Validation_47333918/Validation_47333918_47333918_82462.452.png", "Validation_47333918/Validation_47333918_47333918_82464.451.png" ], "turn_id": 1, "scan_id": "Validation_47333918", "origin_question": "Remember, how many dishwasher(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many dishwasher(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "Validation_47333918/Validation_47333918_47333918_82466.451.png", "Validation_47333918/Validation_47333918_47333918_82468.450.png", "Validation_47333918/Validation_47333918_47333918_82470.466.png", "Validation_47333918/Validation_47333918_47333918_82472.465.png", "Validation_47333918/Validation_47333918_47333918_82474.464.png" ], "turn_id": 2, "scan_id": "Validation_47333918", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"an off-white dome-shaped light\"; \"the large rectangular window with the white frame\"; \"the slim black rectangular monitor with the matte finish\" ", "option": [ "an off-white dome-shaped light", "the large rectangular window with the white frame", "the slim black rectangular monitor with the matte finish" ], "answer": "the slim black rectangular monitor with the matte finish", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"an off-white dome-shaped light\"; \"the large rectangular window with the white frame\"; \"the slim black rectangular monitor with the matte finish\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['an off-white dome-shaped light', 'the large rectangular window with the white frame', 'the slim black rectangular monitor with the matte finish'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_47333918/Validation_47333918_47333918_82476.463.png", "Validation_47333918/Validation_47333918_47333918_82478.462.png", "Validation_47333918/Validation_47333918_47333918_82480.462.png", "Validation_47333918/Validation_47333918_47333918_82482.461.png", "Validation_47333918/Validation_47333918_47333918_82484.460.png" ], "turn_id": 3, "scan_id": "Validation_47333918", "origin_question": "Is the large rectangular window with the white frame to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the large rectangular window with the white frame to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_47333918/Validation_47333918_47333918_82486.459.png", "Validation_47333918/Validation_47333918_47333918_82488.458.png", "Validation_47333918/Validation_47333918_47333918_82490.458.png", "Validation_47333918/Validation_47333918_47333918_82492.457.png", "Validation_47333918/Validation_47333918_47333918_82494.456.png" ], "turn_id": 4, "scan_id": "Validation_47333918", "origin_question": "Is an off-white dome-shaped light to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Is an off-white dome-shaped light to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_47333918/Validation_47333918_47333918_82496.455.png", "Validation_47333918/Validation_47333918_47333918_82498.454.png", "Validation_47333918/Validation_47333918_47333918_82500.454.png", "Validation_47333918/Validation_47333918_47333918_82502.453.png", "Validation_47333918/Validation_47333918_47333918_82504.452.png" ], "turn_id": 5, "scan_id": "Validation_47333918", "origin_question": "In which turn did you see both \"an off-white dome-shaped light\" and \"the slim black rectangular monitor with the matte finish\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"an off-white dome-shaped light\" and \"the slim black rectangular monitor with the matte finish\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_47333918/Validation_47333918_47333918_82506.451.png", "Validation_47333918/Validation_47333918_47333918_82508.450.png", "Validation_47333918/Validation_47333918_47333918_82510.466.png", "Validation_47333918/Validation_47333918_47333918_82512.465.png", "Validation_47333918/Validation_47333918_47333918_82514.465.png" ], "turn_id": 6, "scan_id": "Validation_47333918", "origin_question": "Please recall \"an off-white dome-shaped light\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.66", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"an off-white dome-shaped light\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_47333918/Validation_47333918_47333918_82516.464.png", "Validation_47333918/Validation_47333918_47333918_82518.463.png", "Validation_47333918/Validation_47333918_47333918_82520.462.png", "Validation_47333918/Validation_47333918_47333918_82522.461.png", "Validation_47333918/Validation_47333918_47333918_82524.460.png" ], "turn_id": 7, "scan_id": "Validation_47333918", "origin_question": "What will be the first-time appearance order of the following categories: curtain, monitor, stool?", "option": [ "monitor, curtain, stool", "curtain, monitor, stool", "stool, monitor, curtain", "stool, curtain, monitor" ], "answer": "curtain, monitor, stool", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: curtain, monitor, stool? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['monitor, curtain, stool', 'curtain, monitor, stool', 'stool, monitor, curtain', 'stool, curtain, monitor'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_47333918/Validation_47333918_47333918_82526.460.png", "Validation_47333918/Validation_47333918_47333918_82528.459.png", "Validation_47333918/Validation_47333918_47333918_82530.458.png", "Validation_47333918/Validation_47333918_47333918_82532.457.png", "Validation_47333918/Validation_47333918_47333918_82534.456.png" ], "turn_id": 8, "scan_id": "Validation_47333918", "origin_question": "In which turn did you see both \"the contemporary rectangular door with the light frame\" and \"the compact, square stool upholstered in dark fabric\" simultaneously? ", "option": [], "answer": "7", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the contemporary rectangular door with the light frame\" and \"the compact, square stool upholstered in dark fabric\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_48018345/Validation_48018345_48018345_388665.057.png", "Validation_48018345/Validation_48018345_48018345_388666.540.png", "Validation_48018345/Validation_48018345_48018345_388668.040.png", "Validation_48018345/Validation_48018345_48018345_388669.539.png" ], "turn_id": 1, "scan_id": "Validation_48018345", "origin_question": "", "option": [], "answer": "", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_48018345/Validation_48018345_48018345_388671.038.png", "Validation_48018345/Validation_48018345_48018345_388672.538.png", "Validation_48018345/Validation_48018345_48018345_388674.037.png", "Validation_48018345/Validation_48018345_48018345_388675.536.png" ], "turn_id": 2, "scan_id": "Validation_48018345", "origin_question": "At the end of which turn were both of \"the modern rectangular light wood cabinet\" and \"the single bed with vibrant striped bedding\" on your right side?", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the modern rectangular light wood cabinet\" and \"the single bed with vibrant striped bedding\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_48018345/Validation_48018345_48018345_388677.036.png", "Validation_48018345/Validation_48018345_48018345_388678.535.png", "Validation_48018345/Validation_48018345_48018345_388680.035.png", "Validation_48018345/Validation_48018345_48018345_388681.534.png" ], "turn_id": 3, "scan_id": "Validation_48018345", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_48018345/Validation_48018345_48018345_388683.034.png", "Validation_48018345/Validation_48018345_48018345_388684.550.png", "Validation_48018345/Validation_48018345_48018345_388686.049.png", "Validation_48018345/Validation_48018345_48018345_388687.548.png" ], "turn_id": 4, "scan_id": "Validation_48018345", "origin_question": "Assuming the direction you are facing at the end of the turn 3 is forward, did you move a certain distance forward or backward from that position?", "option": [ "Forward", "Backward" ], "answer": "Forward", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 3 is forward, did you move a certain distance forward or backward from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Forward', 'Backward'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_48018345/Validation_48018345_48018345_388689.048.png", "Validation_48018345/Validation_48018345_48018345_388690.547.png", "Validation_48018345/Validation_48018345_48018345_388692.046.png", "Validation_48018345/Validation_48018345_48018345_388693.546.png" ], "turn_id": 5, "scan_id": "Validation_48018345", "origin_question": "Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_48018345/Validation_48018345_48018345_388695.045.png", "Validation_48018345/Validation_48018345_48018345_388696.545.png", "Validation_48018345/Validation_48018345_48018345_388698.044.png", "Validation_48018345/Validation_48018345_48018345_388699.543.png" ], "turn_id": 6, "scan_id": "Validation_48018345", "origin_question": "Is the small, square-shaped chair with pink and purple stripes to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the small, square-shaped chair with pink and purple stripes to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_48018345/Validation_48018345_48018345_388701.043.png", "Validation_48018345/Validation_48018345_48018345_388702.542.png", "Validation_48018345/Validation_48018345_48018345_388704.041.png", "Validation_48018345/Validation_48018345_48018345_388705.541.png" ], "turn_id": 7, "scan_id": "Validation_48018345", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the single bed with vibrant striped bedding\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the single bed with vibrant striped bedding\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_48018345/Validation_48018345_48018345_388707.040.png", "Validation_48018345/Validation_48018345_48018345_388708.540.png", "Validation_48018345/Validation_48018345_48018345_388710.039.png", "Validation_48018345/Validation_48018345_48018345_388711.538.png" ], "turn_id": 8, "scan_id": "Validation_48018345", "origin_question": "In which turn were you closest to \"the small, square-shaped chair with pink and purple stripes\"? ", "option": [], "answer": "5", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the small, square-shaped chair with pink and purple stripes\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47895782/Validation_47895782_47895782_386935.462.png", "Validation_47895782/Validation_47895782_47895782_386936.961.png", "Validation_47895782/Validation_47895782_47895782_386938.460.png", "Validation_47895782/Validation_47895782_47895782_386939.960.png" ], "turn_id": 1, "scan_id": "Validation_47895782", "origin_question": "Remember, have you seen any cabinet(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any cabinet(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47895782/Validation_47895782_47895782_386941.459.png", "Validation_47895782/Validation_47895782_47895782_386942.959.png", "Validation_47895782/Validation_47895782_47895782_386944.458.png", "Validation_47895782/Validation_47895782_47895782_386945.957.png" ], "turn_id": 2, "scan_id": "Validation_47895782", "origin_question": "Remember, have you seen any teapot(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any teapot(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_47895782/Validation_47895782_47895782_386947.457.png", "Validation_47895782/Validation_47895782_47895782_386948.956.png", "Validation_47895782/Validation_47895782_47895782_386950.455.png", "Validation_47895782/Validation_47895782_47895782_386951.955.png" ], "turn_id": 3, "scan_id": "Validation_47895782", "origin_question": "At the end of which turn were both of \"the white vitreous chinthe toilet with the streamlined bowl\" and \"the sleek circular ceiling light with the frosted finish\" on your right side?", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the white vitreous chinthe toilet with the streamlined bowl\" and \"the sleek circular ceiling light with the frosted finish\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47895782/Validation_47895782_47895782_386953.454.png", "Validation_47895782/Validation_47895782_47895782_386954.954.png", "Validation_47895782/Validation_47895782_47895782_386956.453.png", "Validation_47895782/Validation_47895782_47895782_386957.953.png" ], "turn_id": 4, "scan_id": "Validation_47895782", "origin_question": "Which direction is the sleek circular ceiling light with the frosted finish in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the sleek circular ceiling light with the frosted finish in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0084_region10/3393cfe75d8e4593bf4b05d00b7eb2c2_i1_3.jpg", "1mp3d_0084_region10/3393cfe75d8e4593bf4b05d00b7eb2c2_i2_3.jpg", "1mp3d_0084_region10/44762c24817a4a4fab97961ebaaee07f_i1_1.jpg", "1mp3d_0084_region10/44762c24817a4a4fab97961ebaaee07f_i2_1.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0084_region10", "origin_question": "Remember, have you seen any tray(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any tray(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "1mp3d_0084_region10/6333e669c4174c5dbcbc70e6da1c08f6_i1_1.jpg", "1mp3d_0084_region10/6333e669c4174c5dbcbc70e6da1c08f6_i2_1.jpg", "1mp3d_0084_region10/6333e669c4174c5dbcbc70e6da1c08f6_i1_0.jpg", "1mp3d_0084_region10/6333e669c4174c5dbcbc70e6da1c08f6_i2_0.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0084_region10", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the painting in a golden frame depicting a horse\"; \"the toilet\"; \"the bathtub\" ", "option": [ "the painting in a golden frame depicting a horse", "the toilet", "the bathtub" ], "answer": "the bathtub", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the painting in a golden frame depicting a horse\"; \"the toilet\"; \"the bathtub\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the painting in a golden frame depicting a horse', 'the toilet', 'the bathtub'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "1mp3d_0084_region10/6333e669c4174c5dbcbc70e6da1c08f6_i2_5.jpg", "1mp3d_0084_region10/44762c24817a4a4fab97961ebaaee07f_i1_5.jpg", "1mp3d_0084_region10/44762c24817a4a4fab97961ebaaee07f_i2_5.jpg", "1mp3d_0084_region10/3393cfe75d8e4593bf4b05d00b7eb2c2_i1_4.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0084_region10", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the moderate-sized plant in a white flower pot\" from your position?", "option": [], "answer": "66.03", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the moderate-sized plant in a white flower pot\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "1mp3d_0084_region10/3393cfe75d8e4593bf4b05d00b7eb2c2_i2_4.jpg", "1mp3d_0084_region10/1eeba340f7204705bda0b0016a98e59a_i1_1.jpg", "1mp3d_0084_region10/1eeba340f7204705bda0b0016a98e59a_i2_1.jpg", "1mp3d_0084_region10/1eeba340f7204705bda0b0016a98e59a_i1_2.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0084_region10", "origin_question": "What will be the first-time appearance order of the following categories: cabinet, curtain, window?", "option": [ "window, curtain, cabinet", "curtain, cabinet, window", "curtain, window, cabinet", "cabinet, curtain, window" ], "answer": "cabinet, curtain, window", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: cabinet, curtain, window? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['window, curtain, cabinet', 'curtain, cabinet, window', 'curtain, window, cabinet', 'cabinet, curtain, window'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "1mp3d_0084_region10/1eeba340f7204705bda0b0016a98e59a_i2_2.jpg", "1mp3d_0084_region10/1eeba340f7204705bda0b0016a98e59a_i1_0.jpg", "1mp3d_0084_region10/1eeba340f7204705bda0b0016a98e59a_i2_0.jpg", "1mp3d_0084_region10/ac87cf2f5d1f4c6e82be217339e7f0a6_i1_2.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0084_region10", "origin_question": "At the end of which turn were both of \"the toilet\" and \"the light brown granite counter with a warm atmosphere\" on your left side?", "option": [], "answer": "3", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the toilet\" and \"the light brown granite counter with a warm atmosphere\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0006_01/00000.jpg", "scene0006_01/00040.jpg", "scene0006_01/00080.jpg", "scene0006_01/00120.jpg", "scene0006_01/00160.jpg" ], "turn_id": 1, "scan_id": "scene0006_01", "origin_question": "Remember, how many clothes dryer(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many clothes dryer(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0006_01/00200.jpg", "scene0006_01/00240.jpg", "scene0006_01/00280.jpg", "scene0006_01/00320.jpg", "scene0006_01/00360.jpg" ], "turn_id": 2, "scan_id": "scene0006_01", "origin_question": "When did you first discover \"the dark wooden round stand for placing items\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the dark wooden round stand for placing items\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0006_01/00400.jpg", "scene0006_01/00440.jpg", "scene0006_01/00480.jpg", "scene0006_01/00520.jpg", "scene0006_01/00560.jpg" ], "turn_id": 3, "scan_id": "scene0006_01", "origin_question": "Please recall \"the dark wooden round stand for placing items\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.95", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the dark wooden round stand for placing items\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0006_01/00600.jpg", "scene0006_01/00640.jpg", "scene0006_01/00680.jpg", "scene0006_01/00720.jpg", "scene0006_01/00760.jpg" ], "turn_id": 4, "scan_id": "scene0006_01", "origin_question": "Is the small rectangular white plastic switch to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the small rectangular white plastic switch to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0006_01/00800.jpg", "scene0006_01/00840.jpg", "scene0006_01/00880.jpg", "scene0006_01/00920.jpg", "scene0006_01/00960.jpg" ], "turn_id": 5, "scan_id": "scene0006_01", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the dark wooden round stand for placing items\" from your position?", "option": [], "answer": "160.36", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the dark wooden round stand for placing items\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0006_01/01000.jpg", "scene0006_01/01040.jpg", "scene0006_01/01080.jpg", "scene0006_01/01120.jpg", "scene0006_01/01160.jpg" ], "turn_id": 6, "scan_id": "scene0006_01", "origin_question": "Which direction is the small rectangular white plastic switch in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the small rectangular white plastic switch in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0006_01/01200.jpg", "scene0006_01/01240.jpg", "scene0006_01/01280.jpg", "scene0006_01/01320.jpg", "scene0006_01/01360.jpg" ], "turn_id": 7, "scan_id": "scene0006_01", "origin_question": "Is the rectangular black purse made of durable fabric to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the rectangular black purse made of durable fabric to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "scene0006_01/01400.jpg", "scene0006_01/01440.jpg", "scene0006_01/01480.jpg", "scene0006_01/01520.jpg", "scene0006_01/01560.jpg" ], "turn_id": 8, "scan_id": "scene0006_01", "origin_question": "Among these three objects(\"the small rectangular white plastic switch\",\"the medium-sized rectangular brown wooden desk\",\"the cylindrical green metal bottle suitable for drinking\" ), which one is the closest to you?", "option": [ "the small rectangular white plastic switch", "the medium-sized rectangular brown wooden desk", "the cylindrical green metal bottle suitable for drinking" ], "answer": "the medium-sized rectangular brown wooden desk", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the small rectangular white plastic switch\",\"the medium-sized rectangular brown wooden desk\",\"the cylindrical green metal bottle suitable for drinking\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the small rectangular white plastic switch', 'the medium-sized rectangular brown wooden desk', 'the cylindrical green metal bottle suitable for drinking'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "scene0006_01/01600.jpg", "scene0006_01/01640.jpg", "scene0006_01/01680.jpg", "scene0006_01/01720.jpg", "scene0006_01/01760.jpg" ], "turn_id": 9, "scan_id": "scene0006_01", "origin_question": "Among these three objects(\"the small rectangular white plastic switch\",\"the medium-sized rectangular brown wooden desk\",\"the sleek silver and black coffee maker with a round design\" ), which one is the closest to you?", "option": [ "the small rectangular white plastic switch", "the medium-sized rectangular brown wooden desk", "the sleek silver and black coffee maker with a round design" ], "answer": "the medium-sized rectangular brown wooden desk", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the small rectangular white plastic switch\",\"the medium-sized rectangular brown wooden desk\",\"the sleek silver and black coffee maker with a round design\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the small rectangular white plastic switch', 'the medium-sized rectangular brown wooden desk', 'the sleek silver and black coffee maker with a round design'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0006_01/01800.jpg", "scene0006_01/01840.jpg", "scene0006_01/01880.jpg", "scene0006_01/01920.jpg", "scene0006_01/01960.jpg" ], "turn_id": 10, "scan_id": "scene0006_01", "origin_question": "In which turn did you see both \"the cylindrical green metal bottle suitable for drinking\" and \"the flat, dark cloth shoe with a rectangular shape\" simultaneously? ", "option": [], "answer": "6", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the cylindrical green metal bottle suitable for drinking\" and \"the flat, dark cloth shoe with a rectangular shape\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0626_02/00000.jpg", "scene0626_02/00030.jpg", "scene0626_02/00060.jpg", "scene0626_02/00090.jpg" ], "turn_id": 1, "scan_id": "scene0626_02", "origin_question": "Remember, have you seen any umbrella(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any umbrella(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0626_02/00120.jpg", "scene0626_02/00150.jpg", "scene0626_02/00180.jpg", "scene0626_02/00210.jpg" ], "turn_id": 2, "scan_id": "scene0626_02", "origin_question": "Which direction is the black rectangular plastic keyboard with a flat surface in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the black rectangular plastic keyboard with a flat surface in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0626_02/00240.jpg", "scene0626_02/00270.jpg", "scene0626_02/00300.jpg", "scene0626_02/00330.jpg" ], "turn_id": 3, "scan_id": "scene0626_02", "origin_question": "Please recall \"the black rectangular plastic keyboard with a flat surface\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.22", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the black rectangular plastic keyboard with a flat surface\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0626_02/00360.jpg", "scene0626_02/00390.jpg", "scene0626_02/00420.jpg", "scene0626_02/00450.jpg" ], "turn_id": 4, "scan_id": "scene0626_02", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the black rectangular plastic keyboard with a flat surface\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the black rectangular plastic keyboard with a flat surface\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0626_02/00480.jpg", "scene0626_02/00510.jpg", "scene0626_02/00540.jpg", "scene0626_02/00570.jpg" ], "turn_id": 5, "scan_id": "scene0626_02", "origin_question": "By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the black rectangular plastic keyboard with a flat surface\", Y: \"the purple medium-sized umbrella with a black handle\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the black rectangular plastic keyboard with a flat surface\", Y: \"the purple medium-sized umbrella with a black handle\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0626_02/00600.jpg", "scene0626_02/00630.jpg", "scene0626_02/00660.jpg", "scene0626_02/00690.jpg" ], "turn_id": 6, "scan_id": "scene0626_02", "origin_question": "Which direction is the purple medium-sized umbrella with a black handle in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the purple medium-sized umbrella with a black handle in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0040_region43/0454f54d1be8466ba4760bc4ca275f43_i1_3.jpg", "1mp3d_0040_region43/0454f54d1be8466ba4760bc4ca275f43_i2_3.jpg", "1mp3d_0040_region43/0454f54d1be8466ba4760bc4ca275f43_i1_4.jpg", "1mp3d_0040_region43/0454f54d1be8466ba4760bc4ca275f43_i2_4.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0040_region43", "origin_question": "Remember, have you seen any door(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any door(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "1mp3d_0040_region43/545da9c1d46243758c2f904ff4812ab0_i1_2.jpg", "1mp3d_0040_region43/545da9c1d46243758c2f904ff4812ab0_i2_2.jpg", "1mp3d_0040_region43/a93a97e5e78946438c199367964bd080_i1_0.jpg", "1mp3d_0040_region43/a93a97e5e78946438c199367964bd080_i2_0.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0040_region43", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the deep brown rectangular wooden stand\"; \"the fruit\"; \"the flat-screen TV with a black frame\" ", "option": [ "the deep brown rectangular wooden stand", "the fruit", "the flat-screen TV with a black frame" ], "answer": "the flat-screen TV with a black frame", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the deep brown rectangular wooden stand\"; \"the fruit\"; \"the flat-screen TV with a black frame\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the deep brown rectangular wooden stand', 'the fruit', 'the flat-screen TV with a black frame'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "1mp3d_0040_region43/a93a97e5e78946438c199367964bd080_i1_5.jpg", "1mp3d_0040_region43/a93a97e5e78946438c199367964bd080_i2_5.jpg", "1mp3d_0040_region43/a93a97e5e78946438c199367964bd080_i1_4.jpg", "1mp3d_0040_region43/a93a97e5e78946438c199367964bd080_i2_4.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0040_region43", "origin_question": "Which direction is the flat-screen TV with a black frame in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the flat-screen TV with a black frame in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0040_region43/545da9c1d46243758c2f904ff4812ab0_i1_3.jpg", "1mp3d_0040_region43/545da9c1d46243758c2f904ff4812ab0_i2_3.jpg", "1mp3d_0040_region43/c8a5472a5ef243319ffa4f88d3ddb4bd_i1_2.jpg", "1mp3d_0040_region43/c8a5472a5ef243319ffa4f88d3ddb4bd_i2_2.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0040_region43", "origin_question": "Remember, have you seen any plant(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any plant(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "1mp3d_0040_region43/c8a5472a5ef243319ffa4f88d3ddb4bd_i1_3.jpg", "1mp3d_0040_region43/c8a5472a5ef243319ffa4f88d3ddb4bd_i2_3.jpg", "1mp3d_0040_region43/c8a5472a5ef243319ffa4f88d3ddb4bd_i1_4.jpg", "1mp3d_0040_region43/c8a5472a5ef243319ffa4f88d3ddb4bd_i2_4.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0040_region43", "origin_question": "Please recall \"the deep brown rectangular wooden stand\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.86", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the deep brown rectangular wooden stand\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "1mp3d_0040_region43/545da9c1d46243758c2f904ff4812ab0_i1_4.jpg", "1mp3d_0040_region43/545da9c1d46243758c2f904ff4812ab0_i2_4.jpg", "1mp3d_0040_region43/0454f54d1be8466ba4760bc4ca275f43_i1_2.jpg", "1mp3d_0040_region43/0454f54d1be8466ba4760bc4ca275f43_i2_2.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0040_region43", "origin_question": "In which turn did you see both \"the fruit\" and \"the flat-screen TV with a black frame\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the fruit\" and \"the flat-screen TV with a black frame\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0300_00/00000.jpg", "scene0300_00/00040.jpg", "scene0300_00/00080.jpg", "scene0300_00/00120.jpg", "scene0300_00/00160.jpg" ], "turn_id": 1, "scan_id": "scene0300_00", "origin_question": "Remember, have you seen any table(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any table(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0300_00/00200.jpg", "scene0300_00/00240.jpg", "scene0300_00/00280.jpg", "scene0300_00/00320.jpg", "scene0300_00/00360.jpg" ], "turn_id": 2, "scan_id": "scene0300_00", "origin_question": "In which turn were you closest to \"the light gray rectangular fabric curtain\"? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the light gray rectangular fabric curtain\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0300_00/00400.jpg", "scene0300_00/00440.jpg", "scene0300_00/00480.jpg", "scene0300_00/00520.jpg", "scene0300_00/00560.jpg" ], "turn_id": 3, "scan_id": "scene0300_00", "origin_question": "In which turn did you see both \"the rectangular glossy plastic bin\" and \"the divider\" simultaneously? ", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the rectangular glossy plastic bin\" and \"the divider\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0300_00/00600.jpg", "scene0300_00/00640.jpg", "scene0300_00/00680.jpg", "scene0300_00/00720.jpg", "scene0300_00/00760.jpg" ], "turn_id": 4, "scan_id": "scene0300_00", "origin_question": "Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "61.24", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0300_00/00800.jpg", "scene0300_00/00840.jpg", "scene0300_00/00880.jpg", "scene0300_00/00920.jpg", "scene0300_00/00960.jpg" ], "turn_id": 5, "scan_id": "scene0300_00", "origin_question": "What will be the first-time appearance order of the following categories: bin, chair, picture?", "option": [ "chair, bin, picture", "chair, picture, bin", "bin, picture, chair", "picture, chair, bin" ], "answer": "chair, bin, picture", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bin, chair, picture? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['chair, bin, picture', 'chair, picture, bin', 'bin, picture, chair', 'picture, chair, bin'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0300_00/01000.jpg", "scene0300_00/01040.jpg", "scene0300_00/01080.jpg", "scene0300_00/01120.jpg", "scene0300_00/01160.jpg" ], "turn_id": 6, "scan_id": "scene0300_00", "origin_question": "When did you first discover \"the large rectangular gray window with a white frame\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the large rectangular gray window with a white frame\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0300_00/01200.jpg", "scene0300_00/01240.jpg", "scene0300_00/01280.jpg", "scene0300_00/01320.jpg", "scene0300_00/01360.jpg" ], "turn_id": 7, "scan_id": "scene0300_00", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the rectangular glossy plastic bin\" from your position?", "option": [], "answer": "39.98", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the rectangular glossy plastic bin\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0300_00/01400.jpg", "scene0300_00/01440.jpg", "scene0300_00/01480.jpg", "scene0300_00/01520.jpg", "scene0300_00/01560.jpg" ], "turn_id": 8, "scan_id": "scene0300_00", "origin_question": "Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0084_01/00000.jpg", "scene0084_01/00040.jpg", "scene0084_01/00080.jpg", "scene0084_01/00120.jpg", "scene0084_01/00160.jpg" ], "turn_id": 1, "scan_id": "scene0084_01", "origin_question": "Remember, have you seen any jacket(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any jacket(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0084_01/00200.jpg", "scene0084_01/00240.jpg", "scene0084_01/00280.jpg", "scene0084_01/00320.jpg", "scene0084_01/00360.jpg" ], "turn_id": 2, "scan_id": "scene0084_01", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the rectangular black and white dispenser\"; \"the sink\"; \"the modern rectangular white plastic soap dispenser\" ", "option": [ "the rectangular black and white dispenser", "the sink", "the modern rectangular white plastic soap dispenser" ], "answer": "the modern rectangular white plastic soap dispenser", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the rectangular black and white dispenser\"; \"the sink\"; \"the modern rectangular white plastic soap dispenser\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular black and white dispenser', 'the sink', 'the modern rectangular white plastic soap dispenser'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0084_01/00400.jpg", "scene0084_01/00440.jpg", "scene0084_01/00480.jpg", "scene0084_01/00520.jpg", "scene0084_01/00560.jpg" ], "turn_id": 3, "scan_id": "scene0084_01", "origin_question": "Please recall \"the rectangular black and white dispenser\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.67", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the rectangular black and white dispenser\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0084_01/00600.jpg", "scene0084_01/00640.jpg", "scene0084_01/00680.jpg", "scene0084_01/00720.jpg", "scene0084_01/00760.jpg" ], "turn_id": 4, "scan_id": "scene0084_01", "origin_question": "How far is your current position from where you were at the end of the turn 1? (in meters)", "option": [], "answer": "1.08", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 1? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0084_01/00800.jpg", "scene0084_01/00840.jpg", "scene0084_01/00880.jpg", "scene0084_01/00920.jpg", "scene0084_01/00960.jpg" ], "turn_id": 5, "scan_id": "scene0084_01", "origin_question": "Remember, have you seen any ladder(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any ladder(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0084_01/01000.jpg", "scene0084_01/01040.jpg", "scene0084_01/01080.jpg", "scene0084_01/01120.jpg", "scene0084_01/01160.jpg" ], "turn_id": 6, "scan_id": "scene0084_01", "origin_question": "Which direction is the rectangular yellow warning sign with \"Caution\" text in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the rectangular yellow warning sign with \"Caution\" text in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0084_01/01200.jpg", "scene0084_01/01240.jpg", "scene0084_01/01280.jpg", "scene0084_01/01320.jpg", "scene0084_01/01360.jpg" ], "turn_id": 7, "scan_id": "scene0084_01", "origin_question": "Please recall \"the gray slim-fit windproof jacket made of synthetic material\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "3.88", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the gray slim-fit windproof jacket made of synthetic material\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0084_01/01400.jpg", "scene0084_01/01440.jpg", "scene0084_01/01480.jpg", "scene0084_01/01520.jpg", "scene0084_01/01560.jpg" ], "turn_id": 8, "scan_id": "scene0084_01", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the white rectangular metal radiator with a stylish design\" from your position?", "option": [], "answer": "113.48", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the white rectangular metal radiator with a stylish design\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0697_02/00000.jpg", "scene0697_02/00040.jpg", "scene0697_02/00080.jpg", "scene0697_02/00120.jpg", "scene0697_02/00160.jpg" ], "turn_id": 1, "scan_id": "scene0697_02", "origin_question": "Remember, have you seen any lamp(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any lamp(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0697_02/00200.jpg", "scene0697_02/00240.jpg", "scene0697_02/00280.jpg", "scene0697_02/00320.jpg", "scene0697_02/00360.jpg" ], "turn_id": 2, "scan_id": "scene0697_02", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the book\"; \"the round silver clock with a black dial and yellow hands\"; \"the rectangular tissue box with a zigzag pattern\" ", "option": [ "the book", "the round silver clock with a black dial and yellow hands", "the rectangular tissue box with a zigzag pattern" ], "answer": "the book", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the book\"; \"the round silver clock with a black dial and yellow hands\"; \"the rectangular tissue box with a zigzag pattern\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the book', 'the round silver clock with a black dial and yellow hands', 'the rectangular tissue box with a zigzag pattern'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0697_02/00400.jpg", "scene0697_02/00440.jpg", "scene0697_02/00480.jpg", "scene0697_02/00520.jpg", "scene0697_02/00560.jpg" ], "turn_id": 3, "scan_id": "scene0697_02", "origin_question": "Which direction is the rectangular tissue box with a zigzag pattern in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the rectangular tissue box with a zigzag pattern in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0697_02/00600.jpg", "scene0697_02/00640.jpg", "scene0697_02/00680.jpg", "scene0697_02/00720.jpg", "scene0697_02/00760.jpg" ], "turn_id": 4, "scan_id": "scene0697_02", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the small white rectangular plastic shelf\" from your position?", "option": [], "answer": "122.38", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the small white rectangular plastic shelf\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0697_02/00800.jpg", "scene0697_02/00840.jpg", "scene0697_02/00880.jpg", "scene0697_02/00920.jpg", "scene0697_02/00960.jpg" ], "turn_id": 5, "scan_id": "scene0697_02", "origin_question": "What will be the first-time appearance order of the following categories: book, light, picture?", "option": [ "book, light, picture", "light, book, picture", "picture, light, book", "book, picture, light" ], "answer": "light, book, picture", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: book, light, picture? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['book, light, picture', 'light, book, picture', 'picture, light, book', 'book, picture, light'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0697_02/01000.jpg", "scene0697_02/01040.jpg", "scene0697_02/01080.jpg", "scene0697_02/01120.jpg", "scene0697_02/01160.jpg" ], "turn_id": 6, "scan_id": "scene0697_02", "origin_question": "At the end of which turn were both of \"the toy\" and \"the framed rectangular picture with a black border\" on your right side?", "option": [], "answer": "4", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the toy\" and \"the framed rectangular picture with a black border\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0697_02/01200.jpg", "scene0697_02/01240.jpg", "scene0697_02/01280.jpg", "scene0697_02/01320.jpg", "scene0697_02/01360.jpg" ], "turn_id": 7, "scan_id": "scene0697_02", "origin_question": "Using your orientation at the end of turn 5 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 5 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "scene0697_02/01400.jpg", "scene0697_02/01440.jpg", "scene0697_02/01480.jpg", "scene0697_02/01520.jpg", "scene0697_02/01560.jpg" ], "turn_id": 8, "scan_id": "scene0697_02", "origin_question": "Among these three objects(\"the rectangular tissue box with a zigzag pattern\",\"the toy\",\"the medium-sized black wall-mounted tv\" ), which one is the closest to you?", "option": [ "the rectangular tissue box with a zigzag pattern", "the toy", "the medium-sized black wall-mounted tv" ], "answer": "the medium-sized black wall-mounted tv", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the rectangular tissue box with a zigzag pattern\",\"the toy\",\"the medium-sized black wall-mounted tv\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular tissue box with a zigzag pattern', 'the toy', 'the medium-sized black wall-mounted tv'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0697_02/01600.jpg", "scene0697_02/01640.jpg", "scene0697_02/01680.jpg", "scene0697_02/01720.jpg", "scene0697_02/01760.jpg" ], "turn_id": 9, "scan_id": "scene0697_02", "origin_question": "Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the rectangular tissue box with a zigzag pattern\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the rectangular tissue box with a zigzag pattern\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0697_02/01800.jpg", "scene0697_02/01840.jpg", "scene0697_02/01880.jpg", "scene0697_02/01920.jpg", "scene0697_02/01960.jpg" ], "turn_id": 10, "scan_id": "scene0697_02", "origin_question": "Is the framed rectangular picture with a black border to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the framed rectangular picture with a black border to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0044_region29/da17b34f8eba45ab9947ffd3e150d2ff_i1_3.jpg", "1mp3d_0044_region29/da17b34f8eba45ab9947ffd3e150d2ff_i2_3.jpg", "1mp3d_0044_region29/55b05c4aa35a45b386e4169472aeb5a9_i1_1.jpg", "1mp3d_0044_region29/da17b34f8eba45ab9947ffd3e150d2ff_i1_2.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0044_region29", "origin_question": "Remember, have you seen any tv(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any tv(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0044_region29/da17b34f8eba45ab9947ffd3e150d2ff_i2_2.jpg", "1mp3d_0044_region29/a26f1480cb934ad2ba3fa36542dffaf0_i1_2.jpg", "1mp3d_0044_region29/a26f1480cb934ad2ba3fa36542dffaf0_i2_2.jpg", "1mp3d_0044_region29/561ddcda6a2d4a74b9dd2d42e6c8e863_i1_1.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0044_region29", "origin_question": "Remember, have you seen any window(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any window(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "1mp3d_0044_region29/561ddcda6a2d4a74b9dd2d42e6c8e863_i2_1.jpg", "1mp3d_0044_region29/7ac84f4049164948b4349cec7fab2675_i1_2.jpg", "1mp3d_0044_region29/561ddcda6a2d4a74b9dd2d42e6c8e863_i1_2.jpg", "1mp3d_0044_region29/561ddcda6a2d4a74b9dd2d42e6c8e863_i2_2.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0044_region29", "origin_question": "When did you first discover \"the large rectangular white metal garage door\"? (index of the turn)", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the large rectangular white metal garage door\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "1mp3d_0044_region29/7ac84f4049164948b4349cec7fab2675_i1_3.jpg", "1mp3d_0044_region29/f40ae796663b46ba8fe48392eada6053_i1_2.jpg", "1mp3d_0044_region29/f40ae796663b46ba8fe48392eada6053_i2_2.jpg", "1mp3d_0044_region29/8c6a1212bf9f41f190e6c0b43be35b16_i1_3.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0044_region29", "origin_question": "What will be the first-time appearance order of the following categories: door, garage door, window?", "option": [ "garage door, door, window", "door, garage door, window", "window, door, garage door", "window, garage door, door" ], "answer": "window, garage door, door", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: door, garage door, window? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['garage door, door, window', 'door, garage door, window', 'window, door, garage door', 'window, garage door, door'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "1mp3d_0044_region29/8c6a1212bf9f41f190e6c0b43be35b16_i2_3.jpg", "1mp3d_0044_region29/de3b8549d5784404af13fc30f4a5ddf2_i1_1.jpg", "1mp3d_0044_region29/de3b8549d5784404af13fc30f4a5ddf2_i2_1.jpg", "1mp3d_0044_region29/e669d53f0aae49b3b17a20fcd51e54fe_i1_2.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0044_region29", "origin_question": "How far is your current position from where you were at the end of the turn 1? (in meters)", "option": [], "answer": "8.58", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 1? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "1mp3d_0044_region29/e669d53f0aae49b3b17a20fcd51e54fe_i2_2.jpg", "1mp3d_0044_region29/e669d53f0aae49b3b17a20fcd51e54fe_i1_3.jpg", "1mp3d_0044_region29/e669d53f0aae49b3b17a20fcd51e54fe_i2_3.jpg", "1mp3d_0044_region29/e669d53f0aae49b3b17a20fcd51e54fe_i1_4.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0044_region29", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "7.18", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "1mp3d_0044_region29/e669d53f0aae49b3b17a20fcd51e54fe_i2_4.jpg", "1mp3d_0044_region29/75cc2ac383b54733b202fd94cf2a6cf5_i1_5.jpg", "1mp3d_0044_region29/75cc2ac383b54733b202fd94cf2a6cf5_i2_5.jpg", "1mp3d_0044_region29/75cc2ac383b54733b202fd94cf2a6cf5_i1_4.jpg" ], "turn_id": 7, "scan_id": "1mp3d_0044_region29", "origin_question": "What will be the first-time appearance order of the following categories: door, garage door, light?", "option": [ "garage door, light, door", "light, garage door, door", "door, light, garage door", "door, garage door, light" ], "answer": "light, garage door, door", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: door, garage door, light? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['garage door, light, door', 'light, garage door, door', 'door, light, garage door', 'door, garage door, light'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "1mp3d_0044_region29/75cc2ac383b54733b202fd94cf2a6cf5_i2_4.jpg", "1mp3d_0044_region29/75cc2ac383b54733b202fd94cf2a6cf5_i1_3.jpg", "1mp3d_0044_region29/75cc2ac383b54733b202fd94cf2a6cf5_i2_3.jpg", "1mp3d_0044_region29/3978074e349c44548a76c346bde34b51_i1_5.jpg" ], "turn_id": 8, "scan_id": "1mp3d_0044_region29", "origin_question": "Assuming the direction you are facing at the end of the turn 6 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Left", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 6 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0316_00/00000.jpg", "scene0316_00/00030.jpg", "scene0316_00/00060.jpg", "scene0316_00/00090.jpg" ], "turn_id": 1, "scan_id": "scene0316_00", "origin_question": "Remember, have you seen any washing machine(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any washing machine(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0316_00/00120.jpg", "scene0316_00/00150.jpg", "scene0316_00/00180.jpg", "scene0316_00/00210.jpg" ], "turn_id": 2, "scan_id": "scene0316_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0316_00/00240.jpg", "scene0316_00/00270.jpg", "scene0316_00/00300.jpg", "scene0316_00/00330.jpg" ], "turn_id": 3, "scan_id": "scene0316_00", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "70.54", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0316_00/00360.jpg", "scene0316_00/00390.jpg", "scene0316_00/00420.jpg", "scene0316_00/00450.jpg" ], "turn_id": 4, "scan_id": "scene0316_00", "origin_question": "What will be the first-time appearance order of the following categories: chair, door, ottoman?", "option": [ "ottoman, chair, door", "door, chair, ottoman", "chair, ottoman, door", "chair, door, ottoman" ], "answer": "chair, door, ottoman", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: chair, door, ottoman? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['ottoman, chair, door', 'door, chair, ottoman', 'chair, ottoman, door', 'chair, door, ottoman'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0316_00/00480.jpg", "scene0316_00/00510.jpg", "scene0316_00/00540.jpg", "scene0316_00/00570.jpg" ], "turn_id": 5, "scan_id": "scene0316_00", "origin_question": "In which turn did you see both \"the door\" and \"the round white glass table with a smooth surface\" simultaneously? ", "option": [], "answer": "3", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the door\" and \"the round white glass table with a smooth surface\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0316_00/00600.jpg", "scene0316_00/00630.jpg", "scene0316_00/00660.jpg", "scene0316_00/00690.jpg" ], "turn_id": 6, "scan_id": "scene0316_00", "origin_question": "Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the round white glass table with a smooth surface\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the round white glass table with a smooth surface\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0331_00/00000.jpg", "scene0331_00/00040.jpg", "scene0331_00/00080.jpg", "scene0331_00/00120.jpg", "scene0331_00/00160.jpg" ], "turn_id": 1, "scan_id": "scene0331_00", "origin_question": "Remember, have you seen any statue(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any statue(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0331_00/00200.jpg", "scene0331_00/00240.jpg", "scene0331_00/00280.jpg", "scene0331_00/00320.jpg", "scene0331_00/00360.jpg" ], "turn_id": 2, "scan_id": "scene0331_00", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the chair\"; \"the medium rectangular light beige bin\"; \"the rectangular brown wooden door\" ", "option": [ "the chair", "the medium rectangular light beige bin", "the rectangular brown wooden door" ], "answer": "the chair", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the chair\"; \"the medium rectangular light beige bin\"; \"the rectangular brown wooden door\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the chair', 'the medium rectangular light beige bin', 'the rectangular brown wooden door'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0331_00/00400.jpg", "scene0331_00/00440.jpg", "scene0331_00/00480.jpg", "scene0331_00/00520.jpg", "scene0331_00/00560.jpg" ], "turn_id": 3, "scan_id": "scene0331_00", "origin_question": "Is the bright red and black medium-sized backpack to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the bright red and black medium-sized backpack to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0331_00/00600.jpg", "scene0331_00/00640.jpg", "scene0331_00/00680.jpg", "scene0331_00/00720.jpg", "scene0331_00/00760.jpg" ], "turn_id": 4, "scan_id": "scene0331_00", "origin_question": "Please recall \"the vibrant bunch of purple and pink plants\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.51", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the vibrant bunch of purple and pink plants\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0331_00/00800.jpg", "scene0331_00/00840.jpg", "scene0331_00/00880.jpg", "scene0331_00/00920.jpg", "scene0331_00/00960.jpg" ], "turn_id": 5, "scan_id": "scene0331_00", "origin_question": "Is the medium-sized yellow and white fabric bag to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the medium-sized yellow and white fabric bag to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0331_00/01000.jpg", "scene0331_00/01040.jpg", "scene0331_00/01080.jpg", "scene0331_00/01120.jpg", "scene0331_00/01160.jpg" ], "turn_id": 6, "scan_id": "scene0331_00", "origin_question": "In which turn were you closest to \"the bright red and black medium-sized backpack\"? ", "option": [], "answer": "6", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the bright red and black medium-sized backpack\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0331_00/01200.jpg", "scene0331_00/01240.jpg", "scene0331_00/01280.jpg", "scene0331_00/01320.jpg", "scene0331_00/01360.jpg" ], "turn_id": 7, "scan_id": "scene0331_00", "origin_question": "By the end of the 6 turn, were you getting closer to or farther away from these two objects: X: \"the medium-sized yellow and white fabric bag\", Y: \"the small white oscillating fan with a circular shape\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "farther away from both X and Y", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 6 turn, were you getting closer to or farther away from these two objects: X: \"the medium-sized yellow and white fabric bag\", Y: \"the small white oscillating fan with a circular shape\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0331_00/01400.jpg", "scene0331_00/01440.jpg", "scene0331_00/01480.jpg", "scene0331_00/01520.jpg", "scene0331_00/01560.jpg" ], "turn_id": 8, "scan_id": "scene0331_00", "origin_question": "Please recall \"the large round white plastic hamper\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.51", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the large round white plastic hamper\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0331_00/01600.jpg", "scene0331_00/01640.jpg", "scene0331_00/01680.jpg", "scene0331_00/01720.jpg", "scene0331_00/01760.jpg" ], "turn_id": 9, "scan_id": "scene0331_00", "origin_question": "Which two objects are on the same side of you now? \"the large round white plastic hamper\", \"the small white oscillating fan with a circular shape\", \"the small gray and black woven purse made of cloth\". ", "option": [ "the large round white plastic hamper, the small white oscillating fan with a circular shape", "the large round white plastic hamper, the small gray and black woven purse made of cloth", "the small white oscillating fan with a circular shape, the small gray and black woven purse made of cloth" ], "answer": "the small white oscillating fan with a circular shape, the small gray and black woven purse made of cloth", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the large round white plastic hamper\", \"the small white oscillating fan with a circular shape\", \"the small gray and black woven purse made of cloth\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the large round white plastic hamper, the small white oscillating fan with a circular shape', 'the large round white plastic hamper, the small gray and black woven purse made of cloth', 'the small white oscillating fan with a circular shape, the small gray and black woven purse made of cloth'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0331_00/01800.jpg", "scene0331_00/01840.jpg", "scene0331_00/01880.jpg", "scene0331_00/01920.jpg", "scene0331_00/01960.jpg" ], "turn_id": 10, "scan_id": "scene0331_00", "origin_question": "Which two objects are on the same side of you now? \"the chair\", \"the medium-sized yellow and white fabric bag\", \"the slim white plastic or metal rod\". ", "option": [ "the chair, the medium-sized yellow and white fabric bag", "the chair, the slim white plastic or metal rod", "the medium-sized yellow and white fabric bag, the slim white plastic or metal rod" ], "answer": "the medium-sized yellow and white fabric bag, the slim white plastic or metal rod", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the chair\", \"the medium-sized yellow and white fabric bag\", \"the slim white plastic or metal rod\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the chair, the medium-sized yellow and white fabric bag', 'the chair, the slim white plastic or metal rod', 'the medium-sized yellow and white fabric bag, the slim white plastic or metal rod'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0047_region23/37cb64e0cddc4f20b309abc4ff5344c2_i1_2.jpg", "1mp3d_0047_region23/37cb64e0cddc4f20b309abc4ff5344c2_i2_2.jpg", "1mp3d_0047_region23/3ba76deab6be444c8b21652763d82795_i1_2.jpg", "1mp3d_0047_region23/3ba76deab6be444c8b21652763d82795_i2_2.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0047_region23", "origin_question": "Remember, have you seen any couch(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any couch(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "1mp3d_0047_region23/38bb31f87b6d4f19afedea1ad2d232f9_i1_4.jpg", "1mp3d_0047_region23/38bb31f87b6d4f19afedea1ad2d232f9_i2_4.jpg", "1mp3d_0047_region23/626c0f49c8cb4d9b93a19215f540d8c7_i1_5.jpg", "1mp3d_0047_region23/626c0f49c8cb4d9b93a19215f540d8c7_i2_5.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0047_region23", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the wooden single bed topped with a red blanket\", Y: \"the large purple couch with curved backrest and armrests\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "farther away from both X and Y", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the wooden single bed topped with a red blanket\", Y: \"the large purple couch with curved backrest and armrests\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "1mp3d_0047_region23/38bb31f87b6d4f19afedea1ad2d232f9_i1_5.jpg", "1mp3d_0047_region23/38bb31f87b6d4f19afedea1ad2d232f9_i2_5.jpg", "1mp3d_0047_region23/3ba76deab6be444c8b21652763d82795_i1_3.jpg", "1mp3d_0047_region23/3ba76deab6be444c8b21652763d82795_i2_3.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0047_region23", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "43.27", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "1mp3d_0047_region23/043526f4441e4a4b8bf7e686047a0e96_i1_4.jpg", "1mp3d_0047_region23/043526f4441e4a4b8bf7e686047a0e96_i2_4.jpg", "1mp3d_0047_region23/043526f4441e4a4b8bf7e686047a0e96_i1_3.jpg", "1mp3d_0047_region23/043526f4441e4a4b8bf7e686047a0e96_i2_3.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0047_region23", "origin_question": "At the end of which turn were both of \"the large purple couch with curved backrest and armrests\" and \"the bin\" on your left side?", "option": [], "answer": "3", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the large purple couch with curved backrest and armrests\" and \"the bin\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0047_region23/c0d5fe43f0b143df80de8b5f78858474_i1_4.jpg", "1mp3d_0047_region23/c0d5fe43f0b143df80de8b5f78858474_i2_4.jpg", "1mp3d_0047_region23/7ae91b7a25d44157a89f31cc230d6200_i1_5.jpg", "1mp3d_0047_region23/7ae91b7a25d44157a89f31cc230d6200_i2_5.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0047_region23", "origin_question": "Remember, have you seen any pillow(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any pillow(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "1mp3d_0047_region23/1782ab3735854bc0baa81b292b059f06_i1_0.jpg", "1mp3d_0047_region23/1782ab3735854bc0baa81b292b059f06_i2_0.jpg", "1mp3d_0047_region23/1782ab3735854bc0baa81b292b059f06_i1_1.jpg", "1mp3d_0047_region23/1782ab3735854bc0baa81b292b059f06_i2_1.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0047_region23", "origin_question": "When did you first discover \"the bin\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the bin\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0231_01/00000.jpg", "scene0231_01/00040.jpg", "scene0231_01/00080.jpg", "scene0231_01/00120.jpg", "scene0231_01/00160.jpg" ], "turn_id": 1, "scan_id": "scene0231_01", "origin_question": "Remember, have you seen any desk(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any desk(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0231_01/00200.jpg", "scene0231_01/00240.jpg", "scene0231_01/00280.jpg", "scene0231_01/00320.jpg", "scene0231_01/00360.jpg" ], "turn_id": 2, "scan_id": "scene0231_01", "origin_question": "Is the rectangular light-colored paper calendar to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the rectangular light-colored paper calendar to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0231_01/00400.jpg", "scene0231_01/00440.jpg", "scene0231_01/00480.jpg", "scene0231_01/00520.jpg", "scene0231_01/00560.jpg" ], "turn_id": 3, "scan_id": "scene0231_01", "origin_question": "Which direction is the fireplace in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the fireplace in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0231_01/00600.jpg", "scene0231_01/00640.jpg", "scene0231_01/00680.jpg", "scene0231_01/00720.jpg", "scene0231_01/00760.jpg" ], "turn_id": 4, "scan_id": "scene0231_01", "origin_question": "Which direction is the poster in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the poster in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0231_01/00800.jpg", "scene0231_01/00840.jpg", "scene0231_01/00880.jpg", "scene0231_01/00920.jpg", "scene0231_01/00960.jpg" ], "turn_id": 5, "scan_id": "scene0231_01", "origin_question": "Which two objects are on the same side of you now? \"the large rectangular dark wooden desk\", \"the white modern machine with a digital display\", \"the medium-sized rectangular white wooden door\". ", "option": [ "the large rectangular dark wooden desk, the white modern machine with a digital display", "the large rectangular dark wooden desk, the medium-sized rectangular white wooden door", "the white modern machine with a digital display, the medium-sized rectangular white wooden door" ], "answer": "the large rectangular dark wooden desk, the white modern machine with a digital display", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the large rectangular dark wooden desk\", \"the white modern machine with a digital display\", \"the medium-sized rectangular white wooden door\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the large rectangular dark wooden desk, the white modern machine with a digital display', 'the large rectangular dark wooden desk, the medium-sized rectangular white wooden door', 'the white modern machine with a digital display, the medium-sized rectangular white wooden door'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0231_01/01000.jpg", "scene0231_01/01040.jpg", "scene0231_01/01080.jpg", "scene0231_01/01120.jpg", "scene0231_01/01160.jpg" ], "turn_id": 6, "scan_id": "scene0231_01", "origin_question": "Which direction is the rectangular black and silver stove in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the rectangular black and silver stove in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0231_01/01200.jpg", "scene0231_01/01240.jpg", "scene0231_01/01280.jpg", "scene0231_01/01320.jpg", "scene0231_01/01360.jpg" ], "turn_id": 7, "scan_id": "scene0231_01", "origin_question": "Which direction is the rectangular light-colored paper calendar in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the rectangular light-colored paper calendar in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0231_01/01400.jpg", "scene0231_01/01440.jpg", "scene0231_01/01480.jpg", "scene0231_01/01520.jpg", "scene0231_01/01560.jpg" ], "turn_id": 8, "scan_id": "scene0231_01", "origin_question": "What will be the first-time appearance order of the following categories: calendar, dishwasher, poster?", "option": [ "dishwasher, calendar, poster", "calendar, dishwasher, poster", "dishwasher, poster, calendar", "calendar, poster, dishwasher" ], "answer": "calendar, poster, dishwasher", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: calendar, dishwasher, poster? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['dishwasher, calendar, poster', 'calendar, dishwasher, poster', 'dishwasher, poster, calendar', 'calendar, poster, dishwasher'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "scene0231_01/01600.jpg", "scene0231_01/01640.jpg", "scene0231_01/01680.jpg", "scene0231_01/01720.jpg", "scene0231_01/01760.jpg" ], "turn_id": 9, "scan_id": "scene0231_01", "origin_question": "Among these three objects(\"the rectangular light-colored paper calendar\",\"the poster\",\"the modern rectangular stainless steel dishwasher\" ), which one is the closest to you?", "option": [ "the rectangular light-colored paper calendar", "the poster", "the modern rectangular stainless steel dishwasher" ], "answer": "the modern rectangular stainless steel dishwasher", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the rectangular light-colored paper calendar\",\"the poster\",\"the modern rectangular stainless steel dishwasher\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular light-colored paper calendar', 'the poster', 'the modern rectangular stainless steel dishwasher'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0231_01/01800.jpg", "scene0231_01/01840.jpg", "scene0231_01/01880.jpg", "scene0231_01/01920.jpg", "scene0231_01/01960.jpg" ], "turn_id": 10, "scan_id": "scene0231_01", "origin_question": "Please recall \"the white modern machine with a digital display\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.5", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the white modern machine with a digital display\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0044_region16/9b17f6bce0cf41d9ad7eacc6dfeb73e1_i1_0.jpg", "1mp3d_0044_region16/9b17f6bce0cf41d9ad7eacc6dfeb73e1_i2_0.jpg", "1mp3d_0044_region16/9b17f6bce0cf41d9ad7eacc6dfeb73e1_i2_1.jpg", "1mp3d_0044_region16/9b17f6bce0cf41d9ad7eacc6dfeb73e1_i1_4.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0044_region16", "origin_question": "Remember, have you seen any scale(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any scale(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0044_region16/9b17f6bce0cf41d9ad7eacc6dfeb73e1_i2_4.jpg", "1mp3d_0044_region16/d9bfec8420e341d1911b9de6ba824f0c_i1_4.jpg", "1mp3d_0044_region16/d9bfec8420e341d1911b9de6ba824f0c_i2_4.jpg", "1mp3d_0044_region16/7f61452466834ca9b795a540df1a3c83_i1_3.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0044_region16", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the rectangular framed picture of a landscape\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the rectangular framed picture of a landscape\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "1mp3d_0044_region16/7f61452466834ca9b795a540df1a3c83_i2_3.jpg", "1mp3d_0044_region16/d9bfec8420e341d1911b9de6ba824f0c_i1_0.jpg", "1mp3d_0044_region16/d9bfec8420e341d1911b9de6ba824f0c_i2_0.jpg", "1mp3d_0044_region16/82060e76b8e743c39c7eba0fb02cb5dc_i1_4.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0044_region16", "origin_question": "How far is your current position from where you were at the end of the turn 1? (in meters)", "option": [], "answer": "1.75", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 1? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0044_region16/82060e76b8e743c39c7eba0fb02cb5dc_i2_4.jpg", "1mp3d_0044_region16/de2c508124c34fb79faa373c8da2ffbe_i1_2.jpg", "1mp3d_0044_region16/de2c508124c34fb79faa373c8da2ffbe_i2_2.jpg", "1mp3d_0044_region16/222996ba85c34c3f99b4a25c15adbf4f_i1_2.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0044_region16", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the long, slender black metal rod\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the long, slender black metal rod\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "1mp3d_0044_region16/222996ba85c34c3f99b4a25c15adbf4f_i2_2.jpg", "1mp3d_0044_region16/de2c508124c34fb79faa373c8da2ffbe_i1_3.jpg", "1mp3d_0044_region16/de2c508124c34fb79faa373c8da2ffbe_i2_3.jpg", "1mp3d_0044_region16/de2c508124c34fb79faa373c8da2ffbe_i1_4.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0044_region16", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "1mp3d_0044_region16/de2c508124c34fb79faa373c8da2ffbe_i2_4.jpg", "1mp3d_0044_region16/222996ba85c34c3f99b4a25c15adbf4f_i1_0.jpg", "1mp3d_0044_region16/222996ba85c34c3f99b4a25c15adbf4f_i2_0.jpg", "1mp3d_0044_region16/222996ba85c34c3f99b4a25c15adbf4f_i1_1.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0044_region16", "origin_question": "How far is your current position from where you were at the end of the turn 5? (in meters)", "option": [], "answer": "1.59", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 5? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "1mp3d_0044_region16/222996ba85c34c3f99b4a25c15adbf4f_i2_1.jpg", "1mp3d_0044_region16/27613d374ac34b5cb912e45c1581b9f0_i1_2.jpg", "1mp3d_0044_region16/27613d374ac34b5cb912e45c1581b9f0_i2_2.jpg", "1mp3d_0044_region16/27613d374ac34b5cb912e45c1581b9f0_i1_3.jpg" ], "turn_id": 7, "scan_id": "1mp3d_0044_region16", "origin_question": "Which two objects are on the same side of you now? \"the rectangular dark brown wooden shelf\", \"the large round black metal chandelier with candle lights\", \"the long, slender black metal rod\". ", "option": [ "the rectangular dark brown wooden shelf, the large round black metal chandelier with candle lights", "the rectangular dark brown wooden shelf, the long, slender black metal rod", "the large round black metal chandelier with candle lights, the long, slender black metal rod" ], "answer": "the rectangular dark brown wooden shelf, the large round black metal chandelier with candle lights", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the rectangular dark brown wooden shelf\", \"the large round black metal chandelier with candle lights\", \"the long, slender black metal rod\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular dark brown wooden shelf, the large round black metal chandelier with candle lights', 'the rectangular dark brown wooden shelf, the long, slender black metal rod', 'the large round black metal chandelier with candle lights, the long, slender black metal rod'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0687_00/00000.jpg", "scene0687_00/00040.jpg", "scene0687_00/00080.jpg", "scene0687_00/00120.jpg" ], "turn_id": 1, "scan_id": "scene0687_00", "origin_question": "Remember, how many kettle(s) have you seen so far? ", "option": [], "answer": "3", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many kettle(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0687_00/00160.jpg", "scene0687_00/00200.jpg", "scene0687_00/00240.jpg", "scene0687_00/00280.jpg" ], "turn_id": 2, "scan_id": "scene0687_00", "origin_question": "Which direction is the deep gray automatic hand soap dispenser made of plastic in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the deep gray automatic hand soap dispenser made of plastic in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0687_00/00320.jpg", "scene0687_00/00360.jpg", "scene0687_00/00400.jpg", "scene0687_00/00440.jpg" ], "turn_id": 3, "scan_id": "scene0687_00", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular plastic bin with gray lining\", Y: \"the sleek black microwave with a digital display screen\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to Y and farther away from X", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular plastic bin with gray lining\", Y: \"the sleek black microwave with a digital display screen\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0687_00/00480.jpg", "scene0687_00/00520.jpg", "scene0687_00/00560.jpg", "scene0687_00/00600.jpg" ], "turn_id": 4, "scan_id": "scene0687_00", "origin_question": "Is the two-tier rectangular wooden dish rack with slats to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the two-tier rectangular wooden dish rack with slats to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0687_00/00640.jpg", "scene0687_00/00680.jpg", "scene0687_00/00720.jpg", "scene0687_00/00760.jpg" ], "turn_id": 5, "scan_id": "scene0687_00", "origin_question": "Please recall \"the two-tier rectangular wooden dish rack with slats\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.45", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the two-tier rectangular wooden dish rack with slats\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0687_00/00800.jpg", "scene0687_00/00840.jpg", "scene0687_00/00880.jpg", "scene0687_00/00920.jpg" ], "turn_id": 6, "scan_id": "scene0687_00", "origin_question": "Which direction is the deep gray automatic hand soap dispenser made of plastic in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the deep gray automatic hand soap dispenser made of plastic in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "1mp3d_0044_region1/d2c0c06ce09249f9a9105fca6ba197b9_i1_2.jpg", "1mp3d_0044_region1/d2c0c06ce09249f9a9105fca6ba197b9_i2_2.jpg", "1mp3d_0044_region1/b2f83db410a044e0aa9ee7c46fad25e3_i1_2.jpg", "1mp3d_0044_region1/b2f83db410a044e0aa9ee7c46fad25e3_i2_2.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0044_region1", "origin_question": "Remember, how many footstool(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many footstool(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "1mp3d_0044_region1/d2c0c06ce09249f9a9105fca6ba197b9_i1_1.jpg", "1mp3d_0044_region1/d2c0c06ce09249f9a9105fca6ba197b9_i2_1.jpg", "1mp3d_0044_region1/b2f83db410a044e0aa9ee7c46fad25e3_i1_1.jpg", "1mp3d_0044_region1/b2f83db410a044e0aa9ee7c46fad25e3_i2_1.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0044_region1", "origin_question": "In which turn did you see both \"the dark wooden single bed with intricate carvings\" and \"the large rectangular mirror with a dark wooden frame\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the dark wooden single bed with intricate carvings\" and \"the large rectangular mirror with a dark wooden frame\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "1mp3d_0044_region1/585178d4731e40b1a1ad71cf27e00af5_i1_3.jpg", "1mp3d_0044_region1/585178d4731e40b1a1ad71cf27e00af5_i2_3.jpg", "1mp3d_0044_region1/19464a09a14b460fb627a2edbf56bee1_i1_3.jpg", "1mp3d_0044_region1/19464a09a14b460fb627a2edbf56bee1_i2_3.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0044_region1", "origin_question": "Among these three objects(\"the dark wooden single bed with intricate carvings\",\"the fan\",\"the medium-sized rectangular flat-panel TV with a shiny black coating\" ), which one is the farthest from you?", "option": [ "the dark wooden single bed with intricate carvings", "the fan", "the medium-sized rectangular flat-panel TV with a shiny black coating" ], "answer": "the dark wooden single bed with intricate carvings", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects(\"the dark wooden single bed with intricate carvings\",\"the fan\",\"the medium-sized rectangular flat-panel TV with a shiny black coating\" ), which one is the farthest from you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the dark wooden single bed with intricate carvings', 'the fan', 'the medium-sized rectangular flat-panel TV with a shiny black coating'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "1mp3d_0044_region1/d9f270cf602b4deba2ff90fae3aed483_i1_3.jpg", "1mp3d_0044_region1/d9f270cf602b4deba2ff90fae3aed483_i2_3.jpg", "1mp3d_0044_region1/d9f270cf602b4deba2ff90fae3aed483_i1_2.jpg", "1mp3d_0044_region1/d9f270cf602b4deba2ff90fae3aed483_i2_2.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0044_region1", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the fan\" from your position?", "option": [], "answer": "41.14", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the fan\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "1mp3d_0044_region1/d9f270cf602b4deba2ff90fae3aed483_i1_1.jpg", "1mp3d_0044_region1/d9f270cf602b4deba2ff90fae3aed483_i2_1.jpg", "1mp3d_0044_region1/d9f270cf602b4deba2ff90fae3aed483_i1_0.jpg", "1mp3d_0044_region1/d9f270cf602b4deba2ff90fae3aed483_i2_0.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0044_region1", "origin_question": "Which two objects are on the same side of you now? \"the dark wooden single bed with intricate carvings\", \"the large black plastic pot\", \"the large rectangular mirror with a dark wooden frame\". ", "option": [ "the dark wooden single bed with intricate carvings, the large black plastic pot", "the dark wooden single bed with intricate carvings, the large rectangular mirror with a dark wooden frame", "the large black plastic pot, the large rectangular mirror with a dark wooden frame" ], "answer": "the dark wooden single bed with intricate carvings, the large black plastic pot", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the dark wooden single bed with intricate carvings\", \"the large black plastic pot\", \"the large rectangular mirror with a dark wooden frame\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the dark wooden single bed with intricate carvings, the large black plastic pot', 'the dark wooden single bed with intricate carvings, the large rectangular mirror with a dark wooden frame', 'the large black plastic pot, the large rectangular mirror with a dark wooden frame'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "1mp3d_0044_region1/d9f270cf602b4deba2ff90fae3aed483_i1_5.jpg", "1mp3d_0044_region1/d9f270cf602b4deba2ff90fae3aed483_i2_5.jpg", "1mp3d_0044_region1/19464a09a14b460fb627a2edbf56bee1_i1_0.jpg", "1mp3d_0044_region1/19464a09a14b460fb627a2edbf56bee1_i2_0.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0044_region1", "origin_question": "Which two objects are on the same side of you now? \"the dark wooden single bed with intricate carvings\", \"the classic brown chair with decorative patterns\", \"the medium-sized rectangular flat-panel TV with a shiny black coating\". ", "option": [ "the dark wooden single bed with intricate carvings, the classic brown chair with decorative patterns", "the dark wooden single bed with intricate carvings, the medium-sized rectangular flat-panel TV with a shiny black coating", "the classic brown chair with decorative patterns, the medium-sized rectangular flat-panel TV with a shiny black coating" ], "answer": "the classic brown chair with decorative patterns, the medium-sized rectangular flat-panel TV with a shiny black coating", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the dark wooden single bed with intricate carvings\", \"the classic brown chair with decorative patterns\", \"the medium-sized rectangular flat-panel TV with a shiny black coating\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the dark wooden single bed with intricate carvings, the classic brown chair with decorative patterns', 'the dark wooden single bed with intricate carvings, the medium-sized rectangular flat-panel TV with a shiny black coating', 'the classic brown chair with decorative patterns, the medium-sized rectangular flat-panel TV with a shiny black coating'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "1mp3d_0044_region1/585178d4731e40b1a1ad71cf27e00af5_i1_5.jpg", "1mp3d_0044_region1/585178d4731e40b1a1ad71cf27e00af5_i2_5.jpg", "1mp3d_0044_region1/585178d4731e40b1a1ad71cf27e00af5_i1_4.jpg", "1mp3d_0044_region1/585178d4731e40b1a1ad71cf27e00af5_i2_4.jpg" ], "turn_id": 7, "scan_id": "1mp3d_0044_region1", "origin_question": "In which turn did you see both \"the fan\" and \"the large rectangular mirror with a dark wooden frame\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the fan\" and \"the large rectangular mirror with a dark wooden frame\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "1mp3d_0044_region1/17f8aac81c134b039374f07333e76b2e_i1_3.jpg", "1mp3d_0044_region1/17f8aac81c134b039374f07333e76b2e_i2_3.jpg", "1mp3d_0044_region1/b4ddd79b58f3485990e2d304b7db988d_i1_3.jpg", "1mp3d_0044_region1/b4ddd79b58f3485990e2d304b7db988d_i2_3.jpg" ], "turn_id": 8, "scan_id": "1mp3d_0044_region1", "origin_question": "Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Right", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0114_02/00000.jpg", "scene0114_02/00040.jpg", "scene0114_02/00080.jpg", "scene0114_02/00120.jpg", "scene0114_02/00160.jpg" ], "turn_id": 1, "scan_id": "scene0114_02", "origin_question": "Remember, have you seen any plant(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any plant(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0114_02/00200.jpg", "scene0114_02/00240.jpg", "scene0114_02/00280.jpg", "scene0114_02/00320.jpg", "scene0114_02/00360.jpg" ], "turn_id": 2, "scan_id": "scene0114_02", "origin_question": "Please recall \"the white corded landline telephone with a rectangular shape\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.97", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the white corded landline telephone with a rectangular shape\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0114_02/00400.jpg", "scene0114_02/00440.jpg", "scene0114_02/00480.jpg", "scene0114_02/00520.jpg", "scene0114_02/00560.jpg" ], "turn_id": 3, "scan_id": "scene0114_02", "origin_question": "Which two objects are on the same side of you now? \"the white rectangular metal radiator with fins\", \"the standard-sized brown ceramic cup\", \"the small round wooden table with a brown color\". ", "option": [ "the white rectangular metal radiator with fins, the standard-sized brown ceramic cup", "the white rectangular metal radiator with fins, the small round wooden table with a brown color", "the standard-sized brown ceramic cup, the small round wooden table with a brown color" ], "answer": "the white rectangular metal radiator with fins, the standard-sized brown ceramic cup", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the white rectangular metal radiator with fins\", \"the standard-sized brown ceramic cup\", \"the small round wooden table with a brown color\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the white rectangular metal radiator with fins, the standard-sized brown ceramic cup', 'the white rectangular metal radiator with fins, the small round wooden table with a brown color', 'the standard-sized brown ceramic cup, the small round wooden table with a brown color'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0114_02/00600.jpg", "scene0114_02/00640.jpg", "scene0114_02/00680.jpg", "scene0114_02/00720.jpg", "scene0114_02/00760.jpg" ], "turn_id": 4, "scan_id": "scene0114_02", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the white corded landline telephone with a rectangular shape\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the white corded landline telephone with a rectangular shape\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0114_02/00800.jpg", "scene0114_02/00840.jpg", "scene0114_02/00880.jpg", "scene0114_02/00920.jpg", "scene0114_02/00960.jpg" ], "turn_id": 5, "scan_id": "scene0114_02", "origin_question": "What will be the first-time appearance order of the following categories: cabinet, door, window?", "option": [ "cabinet, window, door", "cabinet, door, window", "door, window, cabinet", "window, cabinet, door" ], "answer": "window, cabinet, door", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: cabinet, door, window? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['cabinet, window, door', 'cabinet, door, window', 'door, window, cabinet', 'window, cabinet, door'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0114_02/01000.jpg", "scene0114_02/01040.jpg", "scene0114_02/01080.jpg", "scene0114_02/01120.jpg", "scene0114_02/01160.jpg" ], "turn_id": 6, "scan_id": "scene0114_02", "origin_question": "In which turn did you see both \"the lush indoor plant with bright green leaves\" and \"the rectangular, light brown wooden door\" simultaneously? ", "option": [], "answer": "5", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the lush indoor plant with bright green leaves\" and \"the rectangular, light brown wooden door\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0114_02/01200.jpg", "scene0114_02/01240.jpg", "scene0114_02/01280.jpg", "scene0114_02/01320.jpg", "scene0114_02/01360.jpg" ], "turn_id": 7, "scan_id": "scene0114_02", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the white rectangular metal radiator with fins\" from your position?", "option": [], "answer": "41.07", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the white rectangular metal radiator with fins\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0114_02/01400.jpg", "scene0114_02/01440.jpg", "scene0114_02/01480.jpg", "scene0114_02/01520.jpg", "scene0114_02/01560.jpg" ], "turn_id": 8, "scan_id": "scene0114_02", "origin_question": "In which turn did you see both \"the rectangular, light brown wooden door\" and \"the bin\" simultaneously? ", "option": [], "answer": "5", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the rectangular, light brown wooden door\" and \"the bin\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0004_region5/1981c29bb2b049138b7ada0f2b35c526_i1_5.jpg", "1mp3d_0004_region5/1981c29bb2b049138b7ada0f2b35c526_i2_5.jpg", "1mp3d_0004_region5/7619c3a3428f4448b33615b460606508_i1_3.jpg", "1mp3d_0004_region5/7619c3a3428f4448b33615b460606508_i2_3.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0004_region5", "origin_question": "Remember, have you seen any picture(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any picture(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0004_region5/e20ca2d896bc41bb9c03a679bf94fd68_i1_0.jpg", "1mp3d_0004_region5/e20ca2d896bc41bb9c03a679bf94fd68_i2_0.jpg", "1mp3d_0004_region5/da07d639cbd44e46abf3c23d9f7fa2c5_i1_2.jpg", "1mp3d_0004_region5/da07d639cbd44e46abf3c23d9f7fa2c5_i2_2.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0004_region5", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the white cylindrical bottle with a glossy surface\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the white cylindrical bottle with a glossy surface\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "1mp3d_0004_region5/89a5978345854bd486450d81c204a79d_i1_5.jpg", "1mp3d_0004_region5/89a5978345854bd486450d81c204a79d_i2_5.jpg", "1mp3d_0004_region5/78d6e1d2e0d042af9f1c96b180d449d6_i1_2.jpg", "1mp3d_0004_region5/78d6e1d2e0d042af9f1c96b180d449d6_i2_2.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0004_region5", "origin_question": "What will be the first-time appearance order of the following categories: bottle, door, picture?", "option": [ "door, bottle, picture", "picture, bottle, door", "picture, door, bottle", "bottle, picture, door" ], "answer": "picture, door, bottle", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bottle, door, picture? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['door, bottle, picture', 'picture, bottle, door', 'picture, door, bottle', 'bottle, picture, door'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0004_region5/78d6e1d2e0d042af9f1c96b180d449d6_i1_1.jpg", "1mp3d_0004_region5/78d6e1d2e0d042af9f1c96b180d449d6_i2_1.jpg", "1mp3d_0004_region5/78d6e1d2e0d042af9f1c96b180d449d6_i1_0.jpg", "1mp3d_0004_region5/78d6e1d2e0d042af9f1c96b180d449d6_i2_0.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0004_region5", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the white cylindrical lamp with a circular base\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the white cylindrical lamp with a circular base\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "1mp3d_0004_region5/89a5978345854bd486450d81c204a79d_i1_4.jpg", "1mp3d_0004_region5/89a5978345854bd486450d81c204a79d_i2_4.jpg", "1mp3d_0004_region5/da07d639cbd44e46abf3c23d9f7fa2c5_i1_1.jpg", "1mp3d_0004_region5/da07d639cbd44e46abf3c23d9f7fa2c5_i2_1.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0004_region5", "origin_question": "Assuming the direction you are facing at the end of the turn 4 is forward, did you move a certain distance forward or backward from that position?", "option": [ "Forward", "Backward" ], "answer": "Backward", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 4 is forward, did you move a certain distance forward or backward from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Forward', 'Backward'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "1mp3d_0004_region5/e20ca2d896bc41bb9c03a679bf94fd68_i1_4.jpg", "1mp3d_0004_region5/e20ca2d896bc41bb9c03a679bf94fd68_i2_4.jpg", "1mp3d_0004_region5/e20ca2d896bc41bb9c03a679bf94fd68_i1_5.jpg", "1mp3d_0004_region5/e20ca2d896bc41bb9c03a679bf94fd68_i2_5.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0004_region5", "origin_question": "At the end of which turn were both of \"the rectangular white window with a white frame\" and \"the rectangular mirror with a luxurious golden frame\" on your right side?", "option": [], "answer": "4", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the rectangular white window with a white frame\" and \"the rectangular mirror with a luxurious golden frame\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0004_region12/34b8aac0033445599e8932a355b34695_i1_5.jpg", "1mp3d_0004_region12/34b8aac0033445599e8932a355b34695_i2_5.jpg", "1mp3d_0004_region12/550a4ce414334860aa94a9f9e1a3833c_i1_5.jpg", "1mp3d_0004_region12/550a4ce414334860aa94a9f9e1a3833c_i2_5.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0004_region12", "origin_question": "Remember, have you seen any coat hanger(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any coat hanger(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "1mp3d_0004_region12/550a4ce414334860aa94a9f9e1a3833c_i1_2.jpg", "1mp3d_0004_region12/550a4ce414334860aa94a9f9e1a3833c_i2_2.jpg", "1mp3d_0004_region12/550a4ce414334860aa94a9f9e1a3833c_i1_0.jpg", "1mp3d_0004_region12/550a4ce414334860aa94a9f9e1a3833c_i2_0.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0004_region12", "origin_question": "In which turn were you closest to \"the roof\"? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the roof\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0004_region12/550a4ce414334860aa94a9f9e1a3833c_i1_1.jpg", "1mp3d_0004_region12/550a4ce414334860aa94a9f9e1a3833c_i2_1.jpg", "1mp3d_0004_region12/550a4ce414334860aa94a9f9e1a3833c_i1_4.jpg", "1mp3d_0004_region12/550a4ce414334860aa94a9f9e1a3833c_i2_4.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0004_region12", "origin_question": "Remember, have you seen any water cooler(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any water cooler(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "1mp3d_0004_region12/550a4ce414334860aa94a9f9e1a3833c_i1_3.jpg", "1mp3d_0004_region12/550a4ce414334860aa94a9f9e1a3833c_i2_3.jpg", "1mp3d_0004_region12/34b8aac0033445599e8932a355b34695_i1_1.jpg", "1mp3d_0004_region12/34b8aac0033445599e8932a355b34695_i2_1.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0004_region12", "origin_question": "When did you first discover \"the large round gray ceramic flowerpot with a handle\"? (index of the turn)", "option": [], "answer": "4", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the large round gray ceramic flowerpot with a handle\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "1mp3d_0004_region12/4dbe1da290ac444cb724401272429465_i1_2.jpg", "1mp3d_0004_region12/4dbe1da290ac444cb724401272429465_i2_2.jpg", "1mp3d_0004_region12/fa1d02186f8e4eba917dad99cf856b35_i1_1.jpg", "1mp3d_0004_region12/fa1d02186f8e4eba917dad99cf856b35_i2_1.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0004_region12", "origin_question": "By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the dark, elaborate framed picture\", Y: \"the large round gray ceramic flowerpot with a handle\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the dark, elaborate framed picture\", Y: \"the large round gray ceramic flowerpot with a handle\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "1mp3d_0004_region12/fa1d02186f8e4eba917dad99cf856b35_i1_2.jpg", "1mp3d_0004_region12/fa1d02186f8e4eba917dad99cf856b35_i2_2.jpg", "1mp3d_0004_region12/813a76fa308b44f292ab40a25d1565ab_i1_0.jpg", "1mp3d_0004_region12/813a76fa308b44f292ab40a25d1565ab_i2_0.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0004_region12", "origin_question": "What will be the first-time appearance order of the following categories: door, table, washbasin?", "option": [ "door, table, washbasin", "washbasin, door, table", "table, door, washbasin", "door, washbasin, table" ], "answer": "door, washbasin, table", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: door, table, washbasin? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['door, table, washbasin', 'washbasin, door, table', 'table, door, washbasin', 'door, washbasin, table'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0004_region12/813a76fa308b44f292ab40a25d1565ab_i1_5.jpg", "1mp3d_0004_region12/813a76fa308b44f292ab40a25d1565ab_i2_5.jpg", "1mp3d_0004_region12/d2de17fa68294749a01641894b6ab12b_i1_1.jpg", "1mp3d_0004_region12/d2de17fa68294749a01641894b6ab12b_i2_1.jpg" ], "turn_id": 7, "scan_id": "1mp3d_0004_region12", "origin_question": "", "option": [], "answer": "", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "1mp3d_0004_region12/5ca4f5d44ce34f8f91aad68587563d32_i1_5.jpg", "1mp3d_0004_region12/5ca4f5d44ce34f8f91aad68587563d32_i2_5.jpg", "1mp3d_0004_region12/5ca4f5d44ce34f8f91aad68587563d32_i1_4.jpg", "1mp3d_0004_region12/5ca4f5d44ce34f8f91aad68587563d32_i2_4.jpg" ], "turn_id": 8, "scan_id": "1mp3d_0004_region12", "origin_question": "What will be the first-time appearance order of the following categories: flowerpot, light, table?", "option": [ "light, flowerpot, table", "table, flowerpot, light", "light, table, flowerpot", "flowerpot, light, table" ], "answer": "light, table, flowerpot", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: flowerpot, light, table? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['light, flowerpot, table', 'table, flowerpot, light', 'light, table, flowerpot', 'flowerpot, light, table'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "1mp3d_0004_region12/5ca4f5d44ce34f8f91aad68587563d32_i1_3.jpg", "1mp3d_0004_region12/5ca4f5d44ce34f8f91aad68587563d32_i2_3.jpg", "1mp3d_0004_region12/4d059bb6efaf411985e6ee5eba2b0c64_i1_5.jpg", "1mp3d_0004_region12/4d059bb6efaf411985e6ee5eba2b0c64_i2_5.jpg" ], "turn_id": 9, "scan_id": "1mp3d_0004_region12", "origin_question": "What will be the first-time appearance order of the following categories: door, light, table?", "option": [ "door, table, light", "light, door, table", "light, table, door", "table, door, light" ], "answer": "door, table, light", "user_message": "For the 9 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: door, light, table? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['door, table, light', 'light, door, table', 'light, table, door', 'table, door, light'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0084_region6/c2dcba0e507748e0b4eb18690058666e_i1_1.jpg", "1mp3d_0084_region6/c2dcba0e507748e0b4eb18690058666e_i2_1.jpg", "1mp3d_0084_region6/ffb7f7310ae84246a7184818cbdbba38_i1_1.jpg", "1mp3d_0084_region6/ffb7f7310ae84246a7184818cbdbba38_i2_1.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0084_region6", "origin_question": "Remember, have you seen any washing machine(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any washing machine(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "1mp3d_0084_region6/3ed3115dbc9d46fdbd694e3e636d50d3_i1_4.jpg", "1mp3d_0084_region6/3ed3115dbc9d46fdbd694e3e636d50d3_i2_4.jpg", "1mp3d_0084_region6/3ed3115dbc9d46fdbd694e3e636d50d3_i1_3.jpg", "1mp3d_0084_region6/3ed3115dbc9d46fdbd694e3e636d50d3_i2_3.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0084_region6", "origin_question": "Is the white oval-shaped porcelain sink with a chrome faucet to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the white oval-shaped porcelain sink with a chrome faucet to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "1mp3d_0084_region6/3ed3115dbc9d46fdbd694e3e636d50d3_i1_2.jpg", "1mp3d_0084_region6/3ed3115dbc9d46fdbd694e3e636d50d3_i2_2.jpg", "1mp3d_0084_region6/3e51eeaac8404b31ad8a950bb2bb953d_i1_4.jpg", "1mp3d_0084_region6/3e51eeaac8404b31ad8a950bb2bb953d_i2_4.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0084_region6", "origin_question": "Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "90.84", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "1mp3d_0084_region6/1767bad8f63a4a8497c8122da77ee7d9_i1_3.jpg", "1mp3d_0084_region6/1767bad8f63a4a8497c8122da77ee7d9_i2_3.jpg", "1mp3d_0084_region6/1767bad8f63a4a8497c8122da77ee7d9_i1_2.jpg", "1mp3d_0084_region6/1767bad8f63a4a8497c8122da77ee7d9_i2_2.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0084_region6", "origin_question": "What will be the first-time appearance order of the following categories: bowl, toilet, toilet paper?", "option": [ "bowl, toilet, toilet paper", "toilet, bowl, toilet paper", "toilet, toilet paper, bowl", "toilet paper, toilet, bowl" ], "answer": "bowl, toilet, toilet paper", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bowl, toilet, toilet paper? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['bowl, toilet, toilet paper', 'toilet, bowl, toilet paper', 'toilet, toilet paper, bowl', 'toilet paper, toilet, bowl'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "1mp3d_0084_region6/3e51eeaac8404b31ad8a950bb2bb953d_i1_5.jpg", "1mp3d_0084_region6/3e51eeaac8404b31ad8a950bb2bb953d_i2_5.jpg", "1mp3d_0084_region6/1767bad8f63a4a8497c8122da77ee7d9_i1_1.jpg", "1mp3d_0084_region6/1767bad8f63a4a8497c8122da77ee7d9_i2_1.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0084_region6", "origin_question": "When did you first discover \"the rectangular white book with a black cover\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the rectangular white book with a black cover\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0040_region50/c640736df26241529c55198137868ecd_i1_3.jpg", "1mp3d_0040_region50/c640736df26241529c55198137868ecd_i2_3.jpg", "1mp3d_0040_region50/80a5f01f4277498385b71fff85b85bba_i1_2.jpg", "1mp3d_0040_region50/80a5f01f4277498385b71fff85b85bba_i2_2.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0040_region50", "origin_question": "Remember, have you seen any table(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any table(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0040_region50/3193cd5f6d7a4b2bbf85d5043cbc6352_i1_2.jpg", "1mp3d_0040_region50/3193cd5f6d7a4b2bbf85d5043cbc6352_i2_2.jpg", "1mp3d_0040_region50/421a3c211c2c4abdb53c55ab1c2e21eb_i1_4.jpg", "1mp3d_0040_region50/421a3c211c2c4abdb53c55ab1c2e21eb_i2_4.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0040_region50", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "1mp3d_0040_region50/421a3c211c2c4abdb53c55ab1c2e21eb_i1_3.jpg", "1mp3d_0040_region50/421a3c211c2c4abdb53c55ab1c2e21eb_i2_3.jpg", "1mp3d_0040_region50/421a3c211c2c4abdb53c55ab1c2e21eb_i1_2.jpg", "1mp3d_0040_region50/421a3c211c2c4abdb53c55ab1c2e21eb_i2_2.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0040_region50", "origin_question": "What will be the first-time appearance order of the following categories: light, pillar, table?", "option": [ "light, pillar, table", "light, table, pillar", "table, pillar, light", "table, light, pillar" ], "answer": "table, light, pillar", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: light, pillar, table? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['light, pillar, table', 'light, table, pillar', 'table, pillar, light', 'table, light, pillar'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0040_region50/3193cd5f6d7a4b2bbf85d5043cbc6352_i1_1.jpg", "1mp3d_0040_region50/3193cd5f6d7a4b2bbf85d5043cbc6352_i2_1.jpg", "1mp3d_0040_region50/3193cd5f6d7a4b2bbf85d5043cbc6352_i1_3.jpg", "1mp3d_0040_region50/3193cd5f6d7a4b2bbf85d5043cbc6352_i2_3.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0040_region50", "origin_question": "Remember, have you seen any mirror(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any mirror(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "1mp3d_0040_region50/80a5f01f4277498385b71fff85b85bba_i1_1.jpg", "1mp3d_0040_region50/80a5f01f4277498385b71fff85b85bba_i2_1.jpg", "1mp3d_0040_region50/c640736df26241529c55198137868ecd_i1_5.jpg", "1mp3d_0040_region50/c640736df26241529c55198137868ecd_i2_5.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0040_region50", "origin_question": "By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the classic, intricate brown desk with marble-like top\", Y: \"the large oval-shaped mirror with a dark brown frame\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "farther away from both X and Y", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the classic, intricate brown desk with marble-like top\", Y: \"the large oval-shaped mirror with a dark brown frame\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "1mp3d_0040_region50/c640736df26241529c55198137868ecd_i1_0.jpg", "1mp3d_0040_region50/c640736df26241529c55198137868ecd_i2_0.jpg", "1mp3d_0040_region50/ef592c66cb3e474c9a814c9973712487_i1_0.jpg", "1mp3d_0040_region50/ef592c66cb3e474c9a814c9973712487_i2_0.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0040_region50", "origin_question": "When did you first discover \"the large oval-shaped mirror with a dark brown frame\"? (index of the turn)", "option": [], "answer": "3", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the large oval-shaped mirror with a dark brown frame\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0040_region7/3d0386c86a4a4ada8dc4a3ac54e7315c_i1_4.jpg", "1mp3d_0040_region7/3d0386c86a4a4ada8dc4a3ac54e7315c_i2_4.jpg", "1mp3d_0040_region7/de4d37507ef94c05b71156cdb6af3067_i1_4.jpg", "1mp3d_0040_region7/de4d37507ef94c05b71156cdb6af3067_i2_4.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0040_region7", "origin_question": "Remember, have you seen any picture(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any picture(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0040_region7/3d0386c86a4a4ada8dc4a3ac54e7315c_i1_0.jpg", "1mp3d_0040_region7/3d0386c86a4a4ada8dc4a3ac54e7315c_i2_0.jpg", "1mp3d_0040_region7/6862217358884d56a90b1012aad23d36_i1_2.jpg", "1mp3d_0040_region7/6862217358884d56a90b1012aad23d36_i2_2.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0040_region7", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "1mp3d_0040_region7/f08ba349cbf2424ca4500e05f7c4c337_i1_2.jpg", "1mp3d_0040_region7/f08ba349cbf2424ca4500e05f7c4c337_i2_2.jpg", "1mp3d_0040_region7/f08ba349cbf2424ca4500e05f7c4c337_i1_1.jpg", "1mp3d_0040_region7/f08ba349cbf2424ca4500e05f7c4c337_i2_1.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0040_region7", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "1.48", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "1mp3d_0040_region7/6862217358884d56a90b1012aad23d36_i1_0.jpg", "1mp3d_0040_region7/6862217358884d56a90b1012aad23d36_i2_0.jpg", "1mp3d_0040_region7/f08ba349cbf2424ca4500e05f7c4c337_i1_3.jpg", "1mp3d_0040_region7/f08ba349cbf2424ca4500e05f7c4c337_i2_3.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0040_region7", "origin_question": "Which direction is the small, white wooden square table in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the small, white wooden square table in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "1mp3d_0040_region7/6862217358884d56a90b1012aad23d36_i1_1.jpg", "1mp3d_0040_region7/6862217358884d56a90b1012aad23d36_i2_1.jpg", "1mp3d_0040_region7/3d0386c86a4a4ada8dc4a3ac54e7315c_i1_3.jpg", "1mp3d_0040_region7/3d0386c86a4a4ada8dc4a3ac54e7315c_i2_3.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0040_region7", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "33.56", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "1mp3d_0040_region7/3d0386c86a4a4ada8dc4a3ac54e7315c_i1_2.jpg", "1mp3d_0040_region7/3d0386c86a4a4ada8dc4a3ac54e7315c_i2_2.jpg", "1mp3d_0040_region7/3d0386c86a4a4ada8dc4a3ac54e7315c_i1_1.jpg", "1mp3d_0040_region7/3d0386c86a4a4ada8dc4a3ac54e7315c_i2_1.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0040_region7", "origin_question": "Which direction is the rectangular flat-screen TV with a sleek black coating in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the rectangular flat-screen TV with a sleek black coating in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0264_00/00000.jpg", "scene0264_00/00040.jpg", "scene0264_00/00080.jpg", "scene0264_00/00120.jpg", "scene0264_00/00160.jpg" ], "turn_id": 1, "scan_id": "scene0264_00", "origin_question": "Remember, have you seen any chair(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any chair(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0264_00/00200.jpg", "scene0264_00/00240.jpg", "scene0264_00/00280.jpg", "scene0264_00/00320.jpg", "scene0264_00/00360.jpg" ], "turn_id": 2, "scan_id": "scene0264_00", "origin_question": "In which turn were you closest to \"the light brown rectangular wooden tray\"? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the light brown rectangular wooden tray\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0264_00/00400.jpg", "scene0264_00/00440.jpg", "scene0264_00/00480.jpg", "scene0264_00/00520.jpg", "scene0264_00/00560.jpg" ], "turn_id": 3, "scan_id": "scene0264_00", "origin_question": "In which turn were you closest to \"the door\"? ", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the door\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0264_00/00600.jpg", "scene0264_00/00640.jpg", "scene0264_00/00680.jpg", "scene0264_00/00720.jpg", "scene0264_00/00760.jpg" ], "turn_id": 4, "scan_id": "scene0264_00", "origin_question": "At the end of which turn were both of \"the light brown rectangular wooden tray\" and \"the rectangular black plastic keyboard\" on your right side?", "option": [], "answer": "3", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the light brown rectangular wooden tray\" and \"the rectangular black plastic keyboard\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0264_00/00800.jpg", "scene0264_00/00840.jpg", "scene0264_00/00880.jpg", "scene0264_00/00920.jpg", "scene0264_00/00960.jpg" ], "turn_id": 5, "scan_id": "scene0264_00", "origin_question": "When did you first discover \"the door\"? (index of the turn)", "option": [], "answer": "3", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the door\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0264_00/01000.jpg", "scene0264_00/01040.jpg", "scene0264_00/01080.jpg", "scene0264_00/01120.jpg", "scene0264_00/01160.jpg" ], "turn_id": 6, "scan_id": "scene0264_00", "origin_question": "When did you first discover \"the bottle\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the bottle\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0264_00/01200.jpg", "scene0264_00/01240.jpg", "scene0264_00/01280.jpg", "scene0264_00/01320.jpg", "scene0264_00/01360.jpg" ], "turn_id": 7, "scan_id": "scene0264_00", "origin_question": "In which turn did you see both \"the rectangular white plastic switch\" and \"the door\" simultaneously? ", "option": [], "answer": "3", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the rectangular white plastic switch\" and \"the door\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0264_00/01400.jpg", "scene0264_00/01440.jpg", "scene0264_00/01480.jpg", "scene0264_00/01520.jpg", "scene0264_00/01560.jpg" ], "turn_id": 8, "scan_id": "scene0264_00", "origin_question": "Please recall \"the rectangular black plastic keyboard\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.94", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the rectangular black plastic keyboard\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0357_01/00000.jpg", "scene0357_01/00030.jpg", "scene0357_01/00060.jpg", "scene0357_01/00090.jpg" ], "turn_id": 1, "scan_id": "scene0357_01", "origin_question": "Remember, have you seen any door(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any door(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0357_01/00120.jpg", "scene0357_01/00150.jpg", "scene0357_01/00180.jpg", "scene0357_01/00210.jpg" ], "turn_id": 2, "scan_id": "scene0357_01", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the towel\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the towel\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0357_01/00240.jpg", "scene0357_01/00270.jpg", "scene0357_01/00300.jpg", "scene0357_01/00330.jpg" ], "turn_id": 3, "scan_id": "scene0357_01", "origin_question": "Which direction is the window in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the window in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0357_01/00360.jpg", "scene0357_01/00390.jpg", "scene0357_01/00420.jpg", "scene0357_01/00450.jpg" ], "turn_id": 4, "scan_id": "scene0357_01", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0357_01/00480.jpg", "scene0357_01/00510.jpg", "scene0357_01/00540.jpg", "scene0357_01/00570.jpg" ], "turn_id": 5, "scan_id": "scene0357_01", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the rectangular smooth white sheet of paper\" from your position?", "option": [], "answer": "162.31", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the rectangular smooth white sheet of paper\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0357_01/00600.jpg", "scene0357_01/00630.jpg", "scene0357_01/00660.jpg", "scene0357_01/00690.jpg" ], "turn_id": 6, "scan_id": "scene0357_01", "origin_question": "Which direction is the green and yellow rectangular box in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the green and yellow rectangular box in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0357_01/00720.jpg", "scene0357_01/00750.jpg", "scene0357_01/00780.jpg", "scene0357_01/00810.jpg" ], "turn_id": 7, "scan_id": "scene0357_01", "origin_question": "Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "119.28", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0608_01/00000.jpg", "scene0608_01/00040.jpg", "scene0608_01/00080.jpg", "scene0608_01/00120.jpg", "scene0608_01/00160.jpg" ], "turn_id": 1, "scan_id": "scene0608_01", "origin_question": "Remember, have you seen any table(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any table(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0608_01/00200.jpg", "scene0608_01/00240.jpg", "scene0608_01/00280.jpg", "scene0608_01/00320.jpg", "scene0608_01/00360.jpg" ], "turn_id": 2, "scan_id": "scene0608_01", "origin_question": "At the end of which turn were both of \"the table\" and \"the sleek black flat-panel television on a shiny stand\" on your left side?", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the table\" and \"the sleek black flat-panel television on a shiny stand\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0608_01/00400.jpg", "scene0608_01/00440.jpg", "scene0608_01/00480.jpg", "scene0608_01/00520.jpg", "scene0608_01/00560.jpg" ], "turn_id": 3, "scan_id": "scene0608_01", "origin_question": "Please recall \"the natural wood-colored acoustic guitar\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.17", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the natural wood-colored acoustic guitar\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0608_01/00600.jpg", "scene0608_01/00640.jpg", "scene0608_01/00680.jpg", "scene0608_01/00720.jpg", "scene0608_01/00760.jpg" ], "turn_id": 4, "scan_id": "scene0608_01", "origin_question": "What will be the first-time appearance order of the following categories: paper, piano, pillow?", "option": [ "paper, piano, pillow", "piano, pillow, paper", "paper, pillow, piano", "pillow, paper, piano" ], "answer": "piano, pillow, paper", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: paper, piano, pillow? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['paper, piano, pillow', 'piano, pillow, paper', 'paper, pillow, piano', 'pillow, paper, piano'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0608_01/00800.jpg", "scene0608_01/00840.jpg", "scene0608_01/00880.jpg", "scene0608_01/00920.jpg", "scene0608_01/00960.jpg" ], "turn_id": 5, "scan_id": "scene0608_01", "origin_question": "Which direction is the sleek black flat-panel television on a shiny stand in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the sleek black flat-panel television on a shiny stand in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0608_01/01000.jpg", "scene0608_01/01040.jpg", "scene0608_01/01080.jpg", "scene0608_01/01120.jpg", "scene0608_01/01160.jpg" ], "turn_id": 6, "scan_id": "scene0608_01", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the rectangular, light-colored wooden door\"; \"the medium rectangular cardboard box with green and white designs\"; \"the tall, cylindrical plastic bottle with a red lid\" ", "option": [ "the rectangular, light-colored wooden door", "the medium rectangular cardboard box with green and white designs", "the tall, cylindrical plastic bottle with a red lid" ], "answer": "the tall, cylindrical plastic bottle with a red lid", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the rectangular, light-colored wooden door\"; \"the medium rectangular cardboard box with green and white designs\"; \"the tall, cylindrical plastic bottle with a red lid\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular, light-colored wooden door', 'the medium rectangular cardboard box with green and white designs', 'the tall, cylindrical plastic bottle with a red lid'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0608_01/01200.jpg", "scene0608_01/01240.jpg", "scene0608_01/01280.jpg", "scene0608_01/01320.jpg", "scene0608_01/01360.jpg" ], "turn_id": 7, "scan_id": "scene0608_01", "origin_question": "Please recall \"the sleek black flat-panel television on a shiny stand\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.32", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the sleek black flat-panel television on a shiny stand\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0608_01/01400.jpg", "scene0608_01/01440.jpg", "scene0608_01/01480.jpg", "scene0608_01/01520.jpg", "scene0608_01/01560.jpg" ], "turn_id": 8, "scan_id": "scene0608_01", "origin_question": "Which two objects are on the same side of you now? \"the natural wood-colored acoustic guitar\", \"the sleek black flat-panel television on a shiny stand\", \"the rectangular dark-colored ottoman with a smooth surface\". ", "option": [ "the natural wood-colored acoustic guitar, the sleek black flat-panel television on a shiny stand", "the natural wood-colored acoustic guitar, the rectangular dark-colored ottoman with a smooth surface", "the sleek black flat-panel television on a shiny stand, the rectangular dark-colored ottoman with a smooth surface" ], "answer": "the natural wood-colored acoustic guitar, the sleek black flat-panel television on a shiny stand", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the natural wood-colored acoustic guitar\", \"the sleek black flat-panel television on a shiny stand\", \"the rectangular dark-colored ottoman with a smooth surface\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the natural wood-colored acoustic guitar, the sleek black flat-panel television on a shiny stand', 'the natural wood-colored acoustic guitar, the rectangular dark-colored ottoman with a smooth surface', 'the sleek black flat-panel television on a shiny stand, the rectangular dark-colored ottoman with a smooth surface'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0608_01/01600.jpg", "scene0608_01/01640.jpg", "scene0608_01/01680.jpg", "scene0608_01/01720.jpg", "scene0608_01/01760.jpg" ], "turn_id": 9, "scan_id": "scene0608_01", "origin_question": "Which two objects are on the same side of you now? \"the window\", \"the medium-sized, white pleated lamp\", \"the rectangular dark-colored ottoman with a smooth surface\". ", "option": [ "the window, the medium-sized, white pleated lamp", "the window, the rectangular dark-colored ottoman with a smooth surface", "the medium-sized, white pleated lamp, the rectangular dark-colored ottoman with a smooth surface" ], "answer": "the medium-sized, white pleated lamp, the rectangular dark-colored ottoman with a smooth surface", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the window\", \"the medium-sized, white pleated lamp\", \"the rectangular dark-colored ottoman with a smooth surface\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the window, the medium-sized, white pleated lamp', 'the window, the rectangular dark-colored ottoman with a smooth surface', 'the medium-sized, white pleated lamp, the rectangular dark-colored ottoman with a smooth surface'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0608_01/01800.jpg", "scene0608_01/01840.jpg", "scene0608_01/01880.jpg", "scene0608_01/01920.jpg", "scene0608_01/01960.jpg" ], "turn_id": 10, "scan_id": "scene0608_01", "origin_question": "", "option": [], "answer": "", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0076_00/00000.jpg", "scene0076_00/00030.jpg", "scene0076_00/00060.jpg", "scene0076_00/00090.jpg" ], "turn_id": 1, "scan_id": "scene0076_00", "origin_question": "Remember, have you seen any tv(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any tv(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0076_00/00120.jpg", "scene0076_00/00150.jpg", "scene0076_00/00180.jpg", "scene0076_00/00210.jpg" ], "turn_id": 2, "scan_id": "scene0076_00", "origin_question": "In which turn were you closest to \"the table\"? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the table\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0076_00/00240.jpg", "scene0076_00/00270.jpg", "scene0076_00/00300.jpg", "scene0076_00/00330.jpg" ], "turn_id": 3, "scan_id": "scene0076_00", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the dispenser\"; \"the modern silver stainless steel teapot with a rounded shape\"; \"the counter\" ", "option": [ "the dispenser", "the modern silver stainless steel teapot with a rounded shape", "the counter" ], "answer": "the dispenser", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the dispenser\"; \"the modern silver stainless steel teapot with a rounded shape\"; \"the counter\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the dispenser', 'the modern silver stainless steel teapot with a rounded shape', 'the counter'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0076_00/00360.jpg", "scene0076_00/00390.jpg", "scene0076_00/00420.jpg", "scene0076_00/00450.jpg" ], "turn_id": 4, "scan_id": "scene0076_00", "origin_question": "Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "73.33", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "scene0076_00/00480.jpg", "scene0076_00/00510.jpg", "scene0076_00/00540.jpg", "scene0076_00/00570.jpg" ], "turn_id": 5, "scan_id": "scene0076_00", "origin_question": "Among these three objects(\"the sleek rectangular stainless steel microwave\",\"the medium upright refrigerator with a stainless steel surface\",\"the dispenser\" ), which one is the closest to you?", "option": [ "the sleek rectangular stainless steel microwave", "the medium upright refrigerator with a stainless steel surface", "the dispenser" ], "answer": "the medium upright refrigerator with a stainless steel surface", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects(\"the sleek rectangular stainless steel microwave\",\"the medium upright refrigerator with a stainless steel surface\",\"the dispenser\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the sleek rectangular stainless steel microwave', 'the medium upright refrigerator with a stainless steel surface', 'the dispenser'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0076_00/00600.jpg", "scene0076_00/00630.jpg", "scene0076_00/00660.jpg", "scene0076_00/00690.jpg" ], "turn_id": 6, "scan_id": "scene0076_00", "origin_question": "Is the rectangular stainless steel sink to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the rectangular stainless steel sink to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0076_00/00720.jpg", "scene0076_00/00750.jpg", "scene0076_00/00780.jpg", "scene0076_00/00990.jpg" ], "turn_id": 7, "scan_id": "scene0076_00", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the tv\" from your position?", "option": [], "answer": "75.22", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the tv\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0076_00/01020.jpg", "scene0076_00/01050.jpg", "scene0076_00/01080.jpg", "scene0076_00/01110.jpg" ], "turn_id": 8, "scan_id": "scene0076_00", "origin_question": "At the end of which turn were both of \"the modern silver stainless steel teapot with a rounded shape\" and \"the white cylindrical wall-mounted soap dispenser\" on your left side?", "option": [], "answer": "3", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the modern silver stainless steel teapot with a rounded shape\" and \"the white cylindrical wall-mounted soap dispenser\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0695_02/00000.jpg", "scene0695_02/00040.jpg", "scene0695_02/00080.jpg", "scene0695_02/00120.jpg", "scene0695_02/00160.jpg" ], "turn_id": 1, "scan_id": "scene0695_02", "origin_question": "Remember, have you seen any clothes(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any clothes(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0695_02/00200.jpg", "scene0695_02/00240.jpg", "scene0695_02/00280.jpg", "scene0695_02/00320.jpg", "scene0695_02/00360.jpg" ], "turn_id": 2, "scan_id": "scene0695_02", "origin_question": "At the end of which turn were both of \"the slim silver laptop with teal edges\" and \"the rectangular brown paper bag\" on your right side?", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the slim silver laptop with teal edges\" and \"the rectangular brown paper bag\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0695_02/00400.jpg", "scene0695_02/00440.jpg", "scene0695_02/00480.jpg", "scene0695_02/00520.jpg", "scene0695_02/00560.jpg" ], "turn_id": 3, "scan_id": "scene0695_02", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "scene0695_02/00690.jpg", "scene0695_02/00730.jpg", "scene0695_02/00770.jpg", "scene0695_02/00810.jpg", "scene0695_02/00850.jpg" ], "turn_id": 4, "scan_id": "scene0695_02", "origin_question": "Among these three objects(\"the bed\",\"the charming brown toy with a soft plush material\",\"the medium-sized, deep red nylon backpack\" ), which one is the closest to you?", "option": [ "the bed", "the charming brown toy with a soft plush material", "the medium-sized, deep red nylon backpack" ], "answer": "the bed", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the bed\",\"the charming brown toy with a soft plush material\",\"the medium-sized, deep red nylon backpack\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the bed', 'the charming brown toy with a soft plush material', 'the medium-sized, deep red nylon backpack'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0695_02/00890.jpg", "scene0695_02/00930.jpg", "scene0695_02/00970.jpg", "scene0695_02/01010.jpg", "scene0695_02/01050.jpg" ], "turn_id": 5, "scan_id": "scene0695_02", "origin_question": "At the end of which turn were both of \"the large, rectangular soft, plush blanket with patterns\" and \"the medium-sized, deep red nylon backpack\" on your right side?", "option": [], "answer": "3", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the large, rectangular soft, plush blanket with patterns\" and \"the medium-sized, deep red nylon backpack\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0695_02/01090.jpg", "scene0695_02/01130.jpg", "scene0695_02/01170.jpg", "scene0695_02/01210.jpg", "scene0695_02/01250.jpg" ], "turn_id": 6, "scan_id": "scene0695_02", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular black computer monitor with a slim profile\", Y: \"the cylindrical, transparent hamper with a handle\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to X and farther away from Y", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular black computer monitor with a slim profile\", Y: \"the cylindrical, transparent hamper with a handle\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "scene0695_02/01290.jpg", "scene0695_02/01330.jpg", "scene0695_02/01370.jpg", "scene0695_02/01410.jpg", "scene0695_02/01450.jpg" ], "turn_id": 7, "scan_id": "scene0695_02", "origin_question": "Among these three objects(\"the rectangular black computer monitor with a slim profile\",\"the rectangular wooden shelf with a simple design\",\"the cylindrical, transparent hamper with a handle\" ), which one is the farthest from you?", "option": [ "the rectangular black computer monitor with a slim profile", "the rectangular wooden shelf with a simple design", "the cylindrical, transparent hamper with a handle" ], "answer": "the cylindrical, transparent hamper with a handle", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the rectangular black computer monitor with a slim profile\",\"the rectangular wooden shelf with a simple design\",\"the cylindrical, transparent hamper with a handle\" ), which one is the farthest from you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular black computer monitor with a slim profile', 'the rectangular wooden shelf with a simple design', 'the cylindrical, transparent hamper with a handle'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0695_02/01490.jpg", "scene0695_02/01530.jpg", "scene0695_02/01570.jpg", "scene0695_02/01610.jpg", "scene0695_02/01650.jpg" ], "turn_id": 8, "scan_id": "scene0695_02", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the cylindrical, transparent hamper with a handle\" from your position?", "option": [], "answer": "36.16", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the cylindrical, transparent hamper with a handle\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0695_02/01690.jpg", "scene0695_02/01730.jpg", "scene0695_02/01770.jpg", "scene0695_02/01810.jpg", "scene0695_02/01850.jpg" ], "turn_id": 9, "scan_id": "scene0695_02", "origin_question": "Which two objects are on the same side of you now? \"the mat\", \"the cylindrical, transparent hamper with a handle\", \"the charming brown toy with a soft plush material\". ", "option": [ "the mat, the cylindrical, transparent hamper with a handle", "the mat, the charming brown toy with a soft plush material", "the cylindrical, transparent hamper with a handle, the charming brown toy with a soft plush material" ], "answer": "the mat, the cylindrical, transparent hamper with a handle", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the mat\", \"the cylindrical, transparent hamper with a handle\", \"the charming brown toy with a soft plush material\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the mat, the cylindrical, transparent hamper with a handle', 'the mat, the charming brown toy with a soft plush material', 'the cylindrical, transparent hamper with a handle, the charming brown toy with a soft plush material'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0695_02/01890.jpg", "scene0695_02/01930.jpg", "scene0695_02/01970.jpg", "scene0695_02/02010.jpg", "scene0695_02/02050.jpg" ], "turn_id": 10, "scan_id": "scene0695_02", "origin_question": "By the end of the 3 turn, were you getting closer to or farther away from these two objects: X: \"the pillow\", Y: \"the rectangular wooden shelf with a simple design\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to X and farther away from Y", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 3 turn, were you getting closer to or farther away from these two objects: X: \"the pillow\", Y: \"the rectangular wooden shelf with a simple design\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0415_00/00000.jpg", "scene0415_00/00040.jpg", "scene0415_00/00080.jpg", "scene0415_00/00120.jpg" ], "turn_id": 1, "scan_id": "scene0415_00", "origin_question": "Remember, have you seen any furniture(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any furniture(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0415_00/00160.jpg", "scene0415_00/00200.jpg", "scene0415_00/00240.jpg", "scene0415_00/00280.jpg" ], "turn_id": 2, "scan_id": "scene0415_00", "origin_question": "When did you first discover \"the framed picture of colorful tree leaves\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the framed picture of colorful tree leaves\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0415_00/00320.jpg", "scene0415_00/00360.jpg", "scene0415_00/00400.jpg", "scene0415_00/00440.jpg" ], "turn_id": 3, "scan_id": "scene0415_00", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the framed picture of colorful tree leaves\" from your position?", "option": [], "answer": "58.96", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the framed picture of colorful tree leaves\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0415_00/00480.jpg", "scene0415_00/00520.jpg", "scene0415_00/00560.jpg", "scene0415_00/00600.jpg" ], "turn_id": 4, "scan_id": "scene0415_00", "origin_question": "Is the framed picture of colorful tree leaves to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the framed picture of colorful tree leaves to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0415_00/00640.jpg", "scene0415_00/00680.jpg", "scene0415_00/00720.jpg", "scene0415_00/00760.jpg" ], "turn_id": 5, "scan_id": "scene0415_00", "origin_question": "By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the medium-sized white oval porcelain toilet\", Y: \"the framed picture of colorful tree leaves\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the medium-sized white oval porcelain toilet\", Y: \"the framed picture of colorful tree leaves\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0415_00/00800.jpg", "scene0415_00/00840.jpg", "scene0415_00/00880.jpg", "scene0415_00/00920.jpg" ], "turn_id": 6, "scan_id": "scene0415_00", "origin_question": "Which direction is the single roll of white toilet paper in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the single roll of white toilet paper in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0415_00/00960.jpg", "scene0415_00/01000.jpg", "scene0415_00/01040.jpg", "scene0415_00/01080.jpg" ], "turn_id": 7, "scan_id": "scene0415_00", "origin_question": "What will be the first-time appearance order of the following categories: bin, mirror, picture?", "option": [ "picture, mirror, bin", "mirror, picture, bin", "mirror, bin, picture", "bin, picture, mirror" ], "answer": "bin, picture, mirror", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bin, mirror, picture? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['picture, mirror, bin', 'mirror, picture, bin', 'mirror, bin, picture', 'bin, picture, mirror'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0583_01/00000.jpg", "scene0583_01/00040.jpg", "scene0583_01/00080.jpg", "scene0583_01/00120.jpg" ], "turn_id": 1, "scan_id": "scene0583_01", "origin_question": "Remember, have you seen any picture(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any picture(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0583_01/00160.jpg", "scene0583_01/00200.jpg", "scene0583_01/00240.jpg", "scene0583_01/00280.jpg" ], "turn_id": 2, "scan_id": "scene0583_01", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the creamy white rectangular vanity with a smooth surface\", Y: \"the stool\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the creamy white rectangular vanity with a smooth surface\", Y: \"the stool\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0583_01/00320.jpg", "scene0583_01/00360.jpg", "scene0583_01/00400.jpg", "scene0583_01/00440.jpg" ], "turn_id": 3, "scan_id": "scene0583_01", "origin_question": "Is the stool to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the stool to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0583_01/00480.jpg", "scene0583_01/00520.jpg", "scene0583_01/00560.jpg", "scene0583_01/00600.jpg" ], "turn_id": 4, "scan_id": "scene0583_01", "origin_question": "Is the creamy white rectangular vanity with a smooth surface to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the creamy white rectangular vanity with a smooth surface to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0583_01/00640.jpg", "scene0583_01/00680.jpg", "scene0583_01/00720.jpg", "scene0583_01/00760.jpg" ], "turn_id": 5, "scan_id": "scene0583_01", "origin_question": "At the end of which turn were both of \"the rectangular mirror with a white wooden border\" and \"the white rectangular plastic bottle with a pump top\" on your left side?", "option": [], "answer": "2", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the rectangular mirror with a white wooden border\" and \"the white rectangular plastic bottle with a pump top\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0583_01/00800.jpg", "scene0583_01/00840.jpg", "scene0583_01/00880.jpg", "scene0583_01/00920.jpg" ], "turn_id": 6, "scan_id": "scene0583_01", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the regular cylindrical roll of white toilet paper\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the regular cylindrical roll of white toilet paper\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "scene0583_01/00960.jpg", "scene0583_01/01000.jpg", "scene0583_01/01040.jpg", "scene0583_01/01080.jpg" ], "turn_id": 7, "scan_id": "scene0583_01", "origin_question": "Among these three objects(\"the double-hung window with a white frame and glass sashes\",\"the white ceramic toilet with a round bowl\",\"the shiny stainless steel stopcock with a curved shape\" ), which one is the closest to you?", "option": [ "the double-hung window with a white frame and glass sashes", "the white ceramic toilet with a round bowl", "the shiny stainless steel stopcock with a curved shape" ], "answer": "the double-hung window with a white frame and glass sashes", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects(\"the double-hung window with a white frame and glass sashes\",\"the white ceramic toilet with a round bowl\",\"the shiny stainless steel stopcock with a curved shape\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the double-hung window with a white frame and glass sashes', 'the white ceramic toilet with a round bowl', 'the shiny stainless steel stopcock with a curved shape'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0583_01/01120.jpg", "scene0583_01/01160.jpg", "scene0583_01/01200.jpg", "scene0583_01/01240.jpg" ], "turn_id": 8, "scan_id": "scene0583_01", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the shiny stainless steel stopcock with a curved shape\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the shiny stainless steel stopcock with a curved shape\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0583_01/01290.jpg", "scene0583_01/01330.jpg", "scene0583_01/01370.jpg", "scene0583_01/01410.jpg" ], "turn_id": 9, "scan_id": "scene0583_01", "origin_question": "How far is your current position from where you were at the end of the turn 1? (in meters)", "option": [], "answer": "1.07", "user_message": "For the 9 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 1? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0029_region16/cd200cef3d824df19eadb99c5224e733_i2_4.jpg", "1mp3d_0029_region16/7f2b92531a51488daa9fa851f8c19a52_i1_4.jpg", "1mp3d_0029_region16/7f2b92531a51488daa9fa851f8c19a52_i2_4.jpg", "1mp3d_0029_region16/cd200cef3d824df19eadb99c5224e733_i1_3.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0029_region16", "origin_question": "Remember, have you seen any stairs(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any stairs(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 3 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "1mp3d_0029_region16/cd200cef3d824df19eadb99c5224e733_i2_3.jpg", "1mp3d_0029_region16/7f2b92531a51488daa9fa851f8c19a52_i2_0.jpg", "1mp3d_0029_region16/b6dea13e3e414d718e5977ae827651ca_i2_0.jpg", "1mp3d_0029_region16/7f2b92531a51488daa9fa851f8c19a52_i1_3.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0029_region16", "origin_question": "When did you first discover \"the white rectangular wooden rail\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the white rectangular wooden rail\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 3 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0029_region16/7f2b92531a51488daa9fa851f8c19a52_i2_3.jpg", "1mp3d_0029_region16/0662f8da68b94d83a47af1e38f28f4e5_i1_2.jpg", "1mp3d_0029_region16/0662f8da68b94d83a47af1e38f28f4e5_i2_2.jpg", "1mp3d_0029_region16/9a503b8a42274ecc9c1fde72a893b75f_i1_3.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0029_region16", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the white rectangular wooden rail\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the white rectangular wooden rail\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 3 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0559_01/00000.jpg", "scene0559_01/00030.jpg", "scene0559_01/00060.jpg", "scene0559_01/00090.jpg" ], "turn_id": 1, "scan_id": "scene0559_01", "origin_question": "Remember, have you seen any bench(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any bench(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 3 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0559_01/00120.jpg", "scene0559_01/00150.jpg", "scene0559_01/00180.jpg", "scene0559_01/00210.jpg" ], "turn_id": 2, "scan_id": "scene0559_01", "origin_question": "Is the light brown rectangular cardboard box to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the light brown rectangular cardboard box to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 3 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0559_01/00240.jpg", "scene0559_01/00270.jpg", "scene0559_01/00300.jpg", "scene0559_01/00330.jpg" ], "turn_id": 3, "scan_id": "scene0559_01", "origin_question": "Please recall \"the pair of black and white leather clothes\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.75", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the pair of black and white leather clothes\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 3 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0692_04/00000.jpg", "scene0692_04/00030.jpg", "scene0692_04/00060.jpg", "scene0692_04/00090.jpg" ], "turn_id": 1, "scan_id": "scene0692_04", "origin_question": "Remember, have you seen any mirror(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any mirror(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0692_04/00120.jpg", "scene0692_04/00150.jpg", "scene0692_04/00210.jpg", "scene0692_04/00280.jpg" ], "turn_id": 2, "scan_id": "scene0692_04", "origin_question": "Which direction is the backpack in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the backpack in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0692_04/00620.jpg", "scene0692_04/00650.jpg", "scene0692_04/00680.jpg", "scene0692_04/00710.jpg" ], "turn_id": 3, "scan_id": "scene0692_04", "origin_question": "In which turn were you closest to \"the mat\"? ", "option": [], "answer": "2", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the mat\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0692_04/00740.jpg", "scene0692_04/00770.jpg", "scene0692_04/00800.jpg", "scene0692_04/00830.jpg" ], "turn_id": 4, "scan_id": "scene0692_04", "origin_question": "What will be the first-time appearance order of the following categories: door, mirror, stairs?", "option": [ "door, mirror, stairs", "stairs, door, mirror", "mirror, stairs, door", "stairs, mirror, door" ], "answer": "mirror, stairs, door", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: door, mirror, stairs? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['door, mirror, stairs', 'stairs, door, mirror', 'mirror, stairs, door', 'stairs, mirror, door'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0692_04/00860.jpg", "scene0692_04/00890.jpg", "scene0692_04/00920.jpg", "scene0692_04/00950.jpg" ], "turn_id": 5, "scan_id": "scene0692_04", "origin_question": "What will be the first-time appearance order of the following categories: backpack, door, stairs?", "option": [ "stairs, backpack, door", "door, stairs, backpack", "stairs, door, backpack", "backpack, stairs, door" ], "answer": "backpack, stairs, door", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: backpack, door, stairs? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['stairs, backpack, door', 'door, stairs, backpack', 'stairs, door, backpack', 'backpack, stairs, door'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0692_04/00980.jpg", "scene0692_04/01010.jpg", "scene0692_04/01040.jpg", "scene0692_04/01070.jpg" ], "turn_id": 6, "scan_id": "scene0692_04", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the mirror\" from your position?", "option": [], "answer": "145.21", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the mirror\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0692_04/01100.jpg", "scene0692_04/01130.jpg", "scene0692_04/01160.jpg", "scene0692_04/01190.jpg" ], "turn_id": 7, "scan_id": "scene0692_04", "origin_question": "When did you first discover \"the rail\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the rail\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0692_04/01220.jpg", "scene0692_04/01280.jpg", "scene0692_04/01370.jpg", "scene0692_04/01400.jpg" ], "turn_id": 8, "scan_id": "scene0692_04", "origin_question": "Is the mat to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the mat to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "1mp3d_0057_region6/76dd8d8a97814d16b3f5401888e7c1e1_i1_1.jpg", "1mp3d_0057_region6/76dd8d8a97814d16b3f5401888e7c1e1_i2_1.jpg", "1mp3d_0057_region6/f2bd01837bb04271a6ec99c31ea850ba_i2_2.jpg", "1mp3d_0057_region6/f2bd01837bb04271a6ec99c31ea850ba_i2_1.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0057_region6", "origin_question": "Remember, how many clock(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many clock(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0057_region6/76dd8d8a97814d16b3f5401888e7c1e1_i1_2.jpg", "1mp3d_0057_region6/76dd8d8a97814d16b3f5401888e7c1e1_i2_2.jpg", "1mp3d_0057_region6/63a2a4cb2cba477bba0ad596573efbf9_i1_2.jpg", "1mp3d_0057_region6/63a2a4cb2cba477bba0ad596573efbf9_i2_2.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0057_region6", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "1mp3d_0057_region6/6e7445ceeb244238b3162f7fa478bb6a_i2_2.jpg", "1mp3d_0057_region6/8ca6a0d87d1d4741aac838c39171e4a8_i1_0.jpg", "1mp3d_0057_region6/8ca6a0d87d1d4741aac838c39171e4a8_i2_0.jpg", "1mp3d_0057_region6/c8aed1383418416eaa977b5da1adac11_i1_1.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0057_region6", "origin_question": "In which turn were you closest to \"the silver stainless steel pan with a black lid\"? ", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the silver stainless steel pan with a black lid\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0057_region6/c8aed1383418416eaa977b5da1adac11_i2_1.jpg", "1mp3d_0057_region6/c8aed1383418416eaa977b5da1adac11_i1_2.jpg", "1mp3d_0057_region6/c8aed1383418416eaa977b5da1adac11_i2_2.jpg", "1mp3d_0057_region6/1b208b02fa6f46c999355dd1f59f29fb_i1_1.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0057_region6", "origin_question": "", "option": [], "answer": "", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "1mp3d_0057_region6/1b208b02fa6f46c999355dd1f59f29fb_i2_1.jpg", "1mp3d_0057_region6/1b208b02fa6f46c999355dd1f59f29fb_i1_2.jpg", "1mp3d_0057_region6/1b208b02fa6f46c999355dd1f59f29fb_i2_2.jpg", "1mp3d_0057_region6/4914bb5db5db465dbba150bcb124bcf9_i1_1.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0057_region6", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "5.56", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0057_region6/4914bb5db5db465dbba150bcb124bcf9_i2_1.jpg", "1mp3d_0057_region6/a70ce41dd9094145bb4e73eeb6fe67b5_i1_5.jpg", "1mp3d_0057_region6/a70ce41dd9094145bb4e73eeb6fe67b5_i2_5.jpg", "1mp3d_0057_region6/d8704782e15f407d94c06a2b89ea8411_i1_3.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0057_region6", "origin_question": "Compared to your position at the end of the 5 turn, are you now closer or farther away from \"the rectangular brown wooden door with horizontal blinds\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 5 turn, are you now closer or farther away from \"the rectangular brown wooden door with horizontal blinds\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "1mp3d_0057_region6/d8704782e15f407d94c06a2b89ea8411_i2_3.jpg", "1mp3d_0057_region6/466bd9b20f7940fd96f766b51641e3fb_i1_5.jpg", "1mp3d_0057_region6/466bd9b20f7940fd96f766b51641e3fb_i2_5.jpg", "1mp3d_0057_region6/249227316e5c491c955577587c5c518b_i1_3.jpg" ], "turn_id": 7, "scan_id": "1mp3d_0057_region6", "origin_question": "Using your orientation at the end of turn 4 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "116.27", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 4 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "1mp3d_0057_region6/249227316e5c491c955577587c5c518b_i2_3.jpg", "1mp3d_0057_region6/466bd9b20f7940fd96f766b51641e3fb_i1_4.jpg", "1mp3d_0057_region6/466bd9b20f7940fd96f766b51641e3fb_i2_4.jpg", "1mp3d_0057_region6/d8704782e15f407d94c06a2b89ea8411_i1_4.jpg" ], "turn_id": 8, "scan_id": "1mp3d_0057_region6", "origin_question": "In which turn did you see both \"the rectangular brown wooden door with horizontal blinds\" and \"the rectangular window with a light-colored wooden frame\" simultaneously? ", "option": [], "answer": "5", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the rectangular brown wooden door with horizontal blinds\" and \"the rectangular window with a light-colored wooden frame\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0067_region20/91f4d6ee66934fe39f4b1cb95d4bf2b9_i1_1.jpg", "1mp3d_0067_region20/91f4d6ee66934fe39f4b1cb95d4bf2b9_i2_1.jpg", "1mp3d_0067_region20/72f5856a420c45a9a790ed3f9ae17a2e_i1_1.jpg", "1mp3d_0067_region20/72f5856a420c45a9a790ed3f9ae17a2e_i2_1.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0067_region20", "origin_question": "Remember, have you seen any dish rack(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any dish rack(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0067_region20/91f4d6ee66934fe39f4b1cb95d4bf2b9_i1_0.jpg", "1mp3d_0067_region20/91f4d6ee66934fe39f4b1cb95d4bf2b9_i2_0.jpg", "1mp3d_0067_region20/b711654d9789435da19337b03e468de1_i1_0.jpg", "1mp3d_0067_region20/b711654d9789435da19337b03e468de1_i2_0.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0067_region20", "origin_question": "Remember, have you seen any switch(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any switch(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0067_region20/b711654d9789435da19337b03e468de1_i1_1.jpg", "1mp3d_0067_region20/b711654d9789435da19337b03e468de1_i2_1.jpg", "1mp3d_0067_region20/78ba509a09f84cb5a0ae52168825c95e_i1_4.jpg", "1mp3d_0067_region20/78ba509a09f84cb5a0ae52168825c95e_i2_4.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0067_region20", "origin_question": "", "option": [], "answer": "", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "1mp3d_0067_region20/5dea2611951e4a908ab308cb165eb241_i1_3.jpg", "1mp3d_0067_region20/5dea2611951e4a908ab308cb165eb241_i2_3.jpg", "1mp3d_0067_region20/78ba509a09f84cb5a0ae52168825c95e_i1_3.jpg", "1mp3d_0067_region20/78ba509a09f84cb5a0ae52168825c95e_i2_3.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0067_region20", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "1mp3d_0067_region20/b711654d9789435da19337b03e468de1_i1_5.jpg", "1mp3d_0067_region20/b711654d9789435da19337b03e468de1_i2_5.jpg", "1mp3d_0067_region20/78ba509a09f84cb5a0ae52168825c95e_i1_1.jpg", "1mp3d_0067_region20/78ba509a09f84cb5a0ae52168825c95e_i2_1.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0067_region20", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "1.6", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "1mp3d_0067_region20/78ba509a09f84cb5a0ae52168825c95e_i1_2.jpg", "1mp3d_0067_region20/78ba509a09f84cb5a0ae52168825c95e_i2_2.jpg", "1mp3d_0067_region20/5dea2611951e4a908ab308cb165eb241_i1_2.jpg", "1mp3d_0067_region20/5dea2611951e4a908ab308cb165eb241_i2_2.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0067_region20", "origin_question": "Please recall \"the door\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.89", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the door\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0153_01/00000.jpg", "scene0153_01/00030.jpg", "scene0153_01/00060.jpg", "scene0153_01/00090.jpg" ], "turn_id": 1, "scan_id": "scene0153_01", "origin_question": "Remember, have you seen any bed(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any bed(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0153_01/00120.jpg", "scene0153_01/00150.jpg", "scene0153_01/00180.jpg", "scene0153_01/00210.jpg" ], "turn_id": 2, "scan_id": "scene0153_01", "origin_question": "Which two objects are on the same side of you now? \"the medium roll of white toilet paper\", \"the plunger\", \"the green soft bath towel of standard size\". ", "option": [ "the medium roll of white toilet paper, the plunger", "the medium roll of white toilet paper, the green soft bath towel of standard size", "the plunger, the green soft bath towel of standard size" ], "answer": "the medium roll of white toilet paper, the plunger", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the medium roll of white toilet paper\", \"the plunger\", \"the green soft bath towel of standard size\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the medium roll of white toilet paper, the plunger', 'the medium roll of white toilet paper, the green soft bath towel of standard size', 'the plunger, the green soft bath towel of standard size'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0153_01/00240.jpg", "scene0153_01/00270.jpg", "scene0153_01/00300.jpg", "scene0153_01/00330.jpg" ], "turn_id": 3, "scan_id": "scene0153_01", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the small round silver stopcock made of metal\", Y: \"the standard size, plain ceramic or plastic cup\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the small round silver stopcock made of metal\", Y: \"the standard size, plain ceramic or plastic cup\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0153_01/00360.jpg", "scene0153_01/00390.jpg", "scene0153_01/00420.jpg", "scene0153_01/00450.jpg" ], "turn_id": 4, "scan_id": "scene0153_01", "origin_question": "Which direction is the standard white oval porcelain sink in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the standard white oval porcelain sink in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0153_01/00480.jpg", "scene0153_01/00510.jpg", "scene0153_01/00540.jpg", "scene0153_01/00570.jpg" ], "turn_id": 5, "scan_id": "scene0153_01", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "116.29", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0153_01/00600.jpg", "scene0153_01/00630.jpg", "scene0153_01/00660.jpg", "scene0153_01/00690.jpg" ], "turn_id": 6, "scan_id": "scene0153_01", "origin_question": "Which direction is the modern rectangular scale with a white exterior in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the modern rectangular scale with a white exterior in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0079_01/00000.jpg", "scene0079_01/00030.jpg", "scene0079_01/00060.jpg", "scene0079_01/00090.jpg" ], "turn_id": 1, "scan_id": "scene0079_01", "origin_question": "Remember, have you seen any shirt(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any shirt(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 3 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0079_01/00120.jpg", "scene0079_01/00150.jpg", "scene0079_01/00180.jpg", "scene0079_01/00210.jpg" ], "turn_id": 2, "scan_id": "scene0079_01", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the tall, white rectangular machine with a digital display\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the tall, white rectangular machine with a digital display\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 3 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0079_01/00240.jpg", "scene0079_01/00270.jpg", "scene0079_01/00300.jpg", "scene0079_01/00330.jpg" ], "turn_id": 3, "scan_id": "scene0079_01", "origin_question": "Is the round stainless steel plate with a smooth surface to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the round stainless steel plate with a smooth surface to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 3 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0260_02/00000.jpg", "scene0260_02/00040.jpg", "scene0260_02/00080.jpg", "scene0260_02/00120.jpg" ], "turn_id": 1, "scan_id": "scene0260_02", "origin_question": "Remember, have you seen any umbrella(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any umbrella(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0260_02/00160.jpg", "scene0260_02/00200.jpg", "scene0260_02/00240.jpg", "scene0260_02/00280.jpg" ], "turn_id": 2, "scan_id": "scene0260_02", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the vibrant canvas picture framed in black\"; \"the shiny metallic water stopcock with a curved shape\"; \"the medium roll of white toilet paper\" ", "option": [ "the vibrant canvas picture framed in black", "the shiny metallic water stopcock with a curved shape", "the medium roll of white toilet paper" ], "answer": "the vibrant canvas picture framed in black", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the vibrant canvas picture framed in black\"; \"the shiny metallic water stopcock with a curved shape\"; \"the medium roll of white toilet paper\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the vibrant canvas picture framed in black', 'the shiny metallic water stopcock with a curved shape', 'the medium roll of white toilet paper'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0260_02/00320.jpg", "scene0260_02/00360.jpg", "scene0260_02/00400.jpg", "scene0260_02/00440.jpg" ], "turn_id": 3, "scan_id": "scene0260_02", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the small, round silver metal bin\"; \"the white ceramic toilet with a standard design\"; \"the green cylindrical plastic bottle\" ", "option": [ "the small, round silver metal bin", "the white ceramic toilet with a standard design", "the green cylindrical plastic bottle" ], "answer": "the green cylindrical plastic bottle", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the small, round silver metal bin\"; \"the white ceramic toilet with a standard design\"; \"the green cylindrical plastic bottle\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the small, round silver metal bin', 'the white ceramic toilet with a standard design', 'the green cylindrical plastic bottle'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0260_02/00480.jpg", "scene0260_02/00520.jpg", "scene0260_02/00560.jpg", "scene0260_02/00600.jpg" ], "turn_id": 4, "scan_id": "scene0260_02", "origin_question": "Please recall \"the shiny metallic water stopcock with a curved shape\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.44", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the shiny metallic water stopcock with a curved shape\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0260_02/00640.jpg", "scene0260_02/00680.jpg", "scene0260_02/00720.jpg", "scene0260_02/00760.jpg" ], "turn_id": 5, "scan_id": "scene0260_02", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the sleek rounded black hair dryer made of plastic\" from your position?", "option": [], "answer": "97.92", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the sleek rounded black hair dryer made of plastic\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0260_02/00800.jpg", "scene0260_02/00840.jpg", "scene0260_02/00880.jpg", "scene0260_02/00920.jpg" ], "turn_id": 6, "scan_id": "scene0260_02", "origin_question": "Which direction is the sleek rounded black hair dryer made of plastic in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the sleek rounded black hair dryer made of plastic in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0260_02/00960.jpg", "scene0260_02/01000.jpg", "scene0260_02/01040.jpg", "scene0260_02/01080.jpg" ], "turn_id": 7, "scan_id": "scene0260_02", "origin_question": "Which two objects are on the same side of you now? \"the white ceramic toilet with a standard design\", \"the vibrant canvas picture framed in black\", \"the small rectangular plastic switch on a wall\". ", "option": [ "the white ceramic toilet with a standard design, the vibrant canvas picture framed in black", "the white ceramic toilet with a standard design, the small rectangular plastic switch on a wall", "the vibrant canvas picture framed in black, the small rectangular plastic switch on a wall" ], "answer": "the white ceramic toilet with a standard design, the vibrant canvas picture framed in black", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the white ceramic toilet with a standard design\", \"the vibrant canvas picture framed in black\", \"the small rectangular plastic switch on a wall\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the white ceramic toilet with a standard design, the vibrant canvas picture framed in black', 'the white ceramic toilet with a standard design, the small rectangular plastic switch on a wall', 'the vibrant canvas picture framed in black, the small rectangular plastic switch on a wall'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0260_02/01120.jpg", "scene0260_02/01160.jpg", "scene0260_02/01200.jpg", "scene0260_02/01240.jpg" ], "turn_id": 8, "scan_id": "scene0260_02", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the white tube of toothpaste with blue and red labels\" from your position?", "option": [], "answer": "113.17", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the white tube of toothpaste with blue and red labels\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0260_02/01280.jpg", "scene0260_02/01320.jpg", "scene0260_02/01360.jpg", "scene0260_02/01400.jpg" ], "turn_id": 9, "scan_id": "scene0260_02", "origin_question": "Is the sink to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 9 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the sink to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0067_region16/364f2ca5c2dd45cd95dc8d28b0d53903_i1_3.jpg", "1mp3d_0067_region16/364f2ca5c2dd45cd95dc8d28b0d53903_i2_3.jpg", "1mp3d_0067_region16/364f2ca5c2dd45cd95dc8d28b0d53903_i1_2.jpg", "1mp3d_0067_region16/364f2ca5c2dd45cd95dc8d28b0d53903_i2_2.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0067_region16", "origin_question": "Remember, have you seen any door(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any door(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "1mp3d_0067_region16/e1d8c24154834577ad8029b907e98de8_i1_3.jpg", "1mp3d_0067_region16/e1d8c24154834577ad8029b907e98de8_i2_3.jpg", "1mp3d_0067_region16/90023a1989ac4787bd7c18e8651e3d3a_i1_3.jpg", "1mp3d_0067_region16/90023a1989ac4787bd7c18e8651e3d3a_i2_3.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0067_region16", "origin_question": "When did you first discover \"the medium oval-shaped white porcelain bidet\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the medium oval-shaped white porcelain bidet\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "1mp3d_0067_region16/331098855d30477582b2ad4d3fcdda62_i1_0.jpg", "1mp3d_0067_region16/331098855d30477582b2ad4d3fcdda62_i2_0.jpg", "1mp3d_0067_region16/90023a1989ac4787bd7c18e8651e3d3a_i1_4.jpg", "1mp3d_0067_region16/90023a1989ac4787bd7c18e8651e3d3a_i2_4.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0067_region16", "origin_question": "When did you first discover \"the white rectangular ceramic toilet with a modern design\"? (index of the turn)", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the white rectangular ceramic toilet with a modern design\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "1mp3d_0067_region16/90023a1989ac4787bd7c18e8651e3d3a_i2_5.jpg", "1mp3d_0067_region16/e1d8c24154834577ad8029b907e98de8_i2_5.jpg", "1mp3d_0067_region16/e1d8c24154834577ad8029b907e98de8_i1_4.jpg", "1mp3d_0067_region16/e1d8c24154834577ad8029b907e98de8_i2_4.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0067_region16", "origin_question": "What will be the first-time appearance order of the following categories: door, shelf, toilet paper?", "option": [ "toilet paper, door, shelf", "toilet paper, shelf, door", "door, shelf, toilet paper", "door, toilet paper, shelf" ], "answer": "door, shelf, toilet paper", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: door, shelf, toilet paper? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['toilet paper, door, shelf', 'toilet paper, shelf, door', 'door, shelf, toilet paper', 'door, toilet paper, shelf'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0693_00/00000.jpg", "scene0693_00/00040.jpg", "scene0693_00/00080.jpg", "scene0693_00/00120.jpg", "scene0693_00/00160.jpg" ], "turn_id": 1, "scan_id": "scene0693_00", "origin_question": "Remember, have you seen any towel(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any towel(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0693_00/00200.jpg", "scene0693_00/00240.jpg", "scene0693_00/00280.jpg", "scene0693_00/00320.jpg", "scene0693_00/00360.jpg" ], "turn_id": 2, "scan_id": "scene0693_00", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the vanity\"; \"the green rectangular plastic toothpaste tube\"; \"the large rectangular mirror with a silver border\" ", "option": [ "the vanity", "the green rectangular plastic toothpaste tube", "the large rectangular mirror with a silver border" ], "answer": "the large rectangular mirror with a silver border", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the vanity\"; \"the green rectangular plastic toothpaste tube\"; \"the large rectangular mirror with a silver border\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the vanity', 'the green rectangular plastic toothpaste tube', 'the large rectangular mirror with a silver border'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0693_00/00400.jpg", "scene0693_00/00440.jpg", "scene0693_00/00480.jpg", "scene0693_00/00520.jpg", "scene0693_00/00560.jpg" ], "turn_id": 3, "scan_id": "scene0693_00", "origin_question": "In which turn did you see both \"the green rectangular plastic toothpaste tube\" and \"the small rectangular window with a white frame\" simultaneously? ", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the green rectangular plastic toothpaste tube\" and \"the small rectangular window with a white frame\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0693_00/00600.jpg", "scene0693_00/00640.jpg", "scene0693_00/00680.jpg", "scene0693_00/00720.jpg", "scene0693_00/00760.jpg" ], "turn_id": 4, "scan_id": "scene0693_00", "origin_question": "Which direction is the green rectangular plastic toothpaste tube in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the green rectangular plastic toothpaste tube in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0693_00/00800.jpg", "scene0693_00/00840.jpg", "scene0693_00/00880.jpg", "scene0693_00/00920.jpg", "scene0693_00/00960.jpg" ], "turn_id": 5, "scan_id": "scene0693_00", "origin_question": "Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0693_00/01000.jpg", "scene0693_00/01040.jpg", "scene0693_00/01080.jpg", "scene0693_00/01120.jpg", "scene0693_00/01160.jpg" ], "turn_id": 6, "scan_id": "scene0693_00", "origin_question": "Please recall \"the green rectangular plastic toothpaste tube\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.55", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the green rectangular plastic toothpaste tube\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0693_00/01200.jpg", "scene0693_00/01240.jpg", "scene0693_00/01280.jpg", "scene0693_00/01320.jpg", "scene0693_00/01360.jpg" ], "turn_id": 7, "scan_id": "scene0693_00", "origin_question": "Which direction is the vanity in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the vanity in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0693_00/01400.jpg", "scene0693_00/01440.jpg", "scene0693_00/01480.jpg", "scene0693_00/01520.jpg", "scene0693_00/01560.jpg" ], "turn_id": 8, "scan_id": "scene0693_00", "origin_question": "Which direction is the medium-sized white rectangular porcelain bathtub in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the medium-sized white rectangular porcelain bathtub in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0598_00/00000.jpg", "scene0598_00/00040.jpg", "scene0598_00/00080.jpg", "scene0598_00/00120.jpg" ], "turn_id": 1, "scan_id": "scene0598_00", "origin_question": "Remember, have you seen any backpack(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any backpack(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0598_00/00160.jpg", "scene0598_00/00200.jpg", "scene0598_00/00240.jpg", "scene0598_00/00280.jpg" ], "turn_id": 2, "scan_id": "scene0598_00", "origin_question": "In which turn were you farthest from \"the medium-sized deep blue and black backpack\"? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the medium-sized deep blue and black backpack\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0598_00/00320.jpg", "scene0598_00/00360.jpg", "scene0598_00/00400.jpg", "scene0598_00/00440.jpg" ], "turn_id": 3, "scan_id": "scene0598_00", "origin_question": "Which direction is the medium-sized deep blue and black backpack in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the medium-sized deep blue and black backpack in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0598_00/00480.jpg", "scene0598_00/00520.jpg", "scene0598_00/00560.jpg", "scene0598_00/00600.jpg" ], "turn_id": 4, "scan_id": "scene0598_00", "origin_question": "Is the white rectangular wooden shelf with a minimalist design to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the white rectangular wooden shelf with a minimalist design to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0598_00/00640.jpg", "scene0598_00/00680.jpg", "scene0598_00/00720.jpg", "scene0598_00/00760.jpg" ], "turn_id": 5, "scan_id": "scene0598_00", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the medium-sized deep blue and black backpack\" from your position?", "option": [], "answer": "115.16", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the medium-sized deep blue and black backpack\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0598_00/00800.jpg", "scene0598_00/00840.jpg", "scene0598_00/00880.jpg", "scene0598_00/00920.jpg" ], "turn_id": 6, "scan_id": "scene0598_00", "origin_question": "Remember, how many mouse(s) have you seen so far? ", "option": [], "answer": "2", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many mouse(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0691_00/00000.jpg", "scene0691_00/00040.jpg", "scene0691_00/00080.jpg", "scene0691_00/00120.jpg" ], "turn_id": 1, "scan_id": "scene0691_00", "origin_question": "Remember, have you seen any cabinet(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any cabinet(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0691_00/00160.jpg", "scene0691_00/00200.jpg", "scene0691_00/00240.jpg", "scene0691_00/00280.jpg" ], "turn_id": 2, "scan_id": "scene0691_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0691_00/00320.jpg", "scene0691_00/00360.jpg", "scene0691_00/00400.jpg", "scene0691_00/00440.jpg" ], "turn_id": 3, "scan_id": "scene0691_00", "origin_question": "Which direction is the red and silver cylindrical metal fire extinguisher in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the red and silver cylindrical metal fire extinguisher in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0691_00/00480.jpg", "scene0691_00/00520.jpg", "scene0691_00/00560.jpg", "scene0691_00/00600.jpg" ], "turn_id": 4, "scan_id": "scene0691_00", "origin_question": "At the end of which turn were both of \"the red and silver cylindrical metal fire extinguisher\" and \"the standard-sized white and black metal printer\" on your right side?", "option": [], "answer": "1", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the red and silver cylindrical metal fire extinguisher\" and \"the standard-sized white and black metal printer\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0691_00/00640.jpg", "scene0691_00/00680.jpg", "scene0691_00/00720.jpg", "scene0691_00/00760.jpg" ], "turn_id": 5, "scan_id": "scene0691_00", "origin_question": "What will be the first-time appearance order of the following categories: fire extinguisher, printer, table?", "option": [ "printer, fire extinguisher, table", "table, fire extinguisher, printer", "printer, table, fire extinguisher", "table, printer, fire extinguisher" ], "answer": "table, fire extinguisher, printer", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: fire extinguisher, printer, table? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['printer, fire extinguisher, table', 'table, fire extinguisher, printer', 'printer, table, fire extinguisher', 'table, printer, fire extinguisher'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "scene0691_00/00800.jpg", "scene0691_00/00840.jpg", "scene0691_00/00880.jpg", "scene0691_00/00920.jpg" ], "turn_id": 6, "scan_id": "scene0691_00", "origin_question": "Assuming the direction you are facing at the end of the turn 5 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Left", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 5 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0691_00/00960.jpg", "scene0691_00/01000.jpg", "scene0691_00/01040.jpg", "scene0691_00/01080.jpg" ], "turn_id": 7, "scan_id": "scene0691_00", "origin_question": "Using your orientation at the end of turn 5 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "119.86", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 5 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0394_00/00000.jpg", "scene0394_00/00040.jpg", "scene0394_00/00080.jpg", "scene0394_00/00120.jpg", "scene0394_00/00160.jpg" ], "turn_id": 1, "scan_id": "scene0394_00", "origin_question": "Remember, have you seen any stool(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any stool(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0394_00/00200.jpg", "scene0394_00/00240.jpg", "scene0394_00/00280.jpg", "scene0394_00/00320.jpg", "scene0394_00/00360.jpg" ], "turn_id": 2, "scan_id": "scene0394_00", "origin_question": "Remember, how many stand(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many stand(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0394_00/00400.jpg", "scene0394_00/00440.jpg", "scene0394_00/00480.jpg", "scene0394_00/00520.jpg", "scene0394_00/00560.jpg" ], "turn_id": 3, "scan_id": "scene0394_00", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the rectangular white porcelain sink with a chrome faucet\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the rectangular white porcelain sink with a chrome faucet\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0394_00/00600.jpg", "scene0394_00/00640.jpg", "scene0394_00/00680.jpg", "scene0394_00/00720.jpg", "scene0394_00/00760.jpg" ], "turn_id": 4, "scan_id": "scene0394_00", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "1.1", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0394_00/00800.jpg", "scene0394_00/00840.jpg", "scene0394_00/00880.jpg", "scene0394_00/00920.jpg", "scene0394_00/00960.jpg" ], "turn_id": 5, "scan_id": "scene0394_00", "origin_question": "Please recall \"the white rectangular table with a smooth surface\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.76", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the white rectangular table with a smooth surface\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0394_00/01000.jpg", "scene0394_00/01040.jpg", "scene0394_00/01080.jpg", "scene0394_00/01120.jpg", "scene0394_00/01160.jpg" ], "turn_id": 6, "scan_id": "scene0394_00", "origin_question": "What will be the first-time appearance order of the following categories: carpet, plant, towel?", "option": [ "carpet, towel, plant", "plant, carpet, towel", "towel, carpet, plant", "carpet, plant, towel" ], "answer": "plant, carpet, towel", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: carpet, plant, towel? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['carpet, towel, plant', 'plant, carpet, towel', 'towel, carpet, plant', 'carpet, plant, towel'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0394_00/01200.jpg", "scene0394_00/01240.jpg", "scene0394_00/01280.jpg", "scene0394_00/01320.jpg", "scene0394_00/01360.jpg" ], "turn_id": 7, "scan_id": "scene0394_00", "origin_question": "Which direction is the round white mirror with decorative patterns in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the round white mirror with decorative patterns in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0394_00/01400.jpg", "scene0394_00/01440.jpg", "scene0394_00/01480.jpg", "scene0394_00/01520.jpg", "scene0394_00/01560.jpg" ], "turn_id": 8, "scan_id": "scene0394_00", "origin_question": "Compared to your position at the end of the 6 turn, are you now closer or farther away from \"the round white mirror with decorative patterns\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 6 turn, are you now closer or farther away from \"the round white mirror with decorative patterns\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0394_00/01600.jpg", "scene0394_00/01640.jpg", "scene0394_00/01680.jpg", "scene0394_00/01720.jpg", "scene0394_00/01760.jpg" ], "turn_id": 9, "scan_id": "scene0394_00", "origin_question": "Compared to your position at the end of the 6 turn, are you now closer or farther away from \"the simple white square stool made of sturdy material\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 6 turn, are you now closer or farther away from \"the simple white square stool made of sturdy material\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0394_00/01800.jpg", "scene0394_00/01840.jpg", "scene0394_00/01880.jpg", "scene0394_00/01920.jpg", "scene0394_00/01960.jpg" ], "turn_id": 10, "scan_id": "scene0394_00", "origin_question": "Please recall \"the round white mirror with decorative patterns\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.05", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the round white mirror with decorative patterns\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0044_region33/353e251a1e6d4b089653a705a4f4ddc3_i1_5.jpg", "1mp3d_0044_region33/353e251a1e6d4b089653a705a4f4ddc3_i2_5.jpg", "1mp3d_0044_region33/6f8bd4c7663a4fa7a5c288b7f74a7915_i1_5.jpg", "1mp3d_0044_region33/353e251a1e6d4b089653a705a4f4ddc3_i1_4.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0044_region33", "origin_question": "", "option": [], "answer": "", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "1mp3d_0044_region33/353e251a1e6d4b089653a705a4f4ddc3_i2_4.jpg", "1mp3d_0044_region33/6f8bd4c7663a4fa7a5c288b7f74a7915_i1_4.jpg", "1mp3d_0044_region33/6f8bd4c7663a4fa7a5c288b7f74a7915_i2_4.jpg", "1mp3d_0044_region33/353e251a1e6d4b089653a705a4f4ddc3_i1_2.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0044_region33", "origin_question": "When did you first discover \"the rectangular black-framed glass window\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the rectangular black-framed glass window\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0044_region33/353e251a1e6d4b089653a705a4f4ddc3_i2_2.jpg", "1mp3d_0044_region33/6f8bd4c7663a4fa7a5c288b7f74a7915_i1_3.jpg", "1mp3d_0044_region33/6f8bd4c7663a4fa7a5c288b7f74a7915_i2_3.jpg", "1mp3d_0044_region33/b41525c3c3594e60a6601783c4914e27_i1_3.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0044_region33", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the rectangular dark door made of sturdy materials\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the rectangular dark door made of sturdy materials\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "1mp3d_0044_region33/b41525c3c3594e60a6601783c4914e27_i2_3.jpg", "1mp3d_0044_region33/511c857f1f3d499cb0e944e9f2edf664_i2_3.jpg", "1mp3d_0044_region33/b41525c3c3594e60a6601783c4914e27_i1_2.jpg", "1mp3d_0044_region33/b41525c3c3594e60a6601783c4914e27_i2_2.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0044_region33", "origin_question": "Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "60.2", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0004_region24/7fc36de72634404d8b06b7779328e791_i1_2.jpg", "1mp3d_0004_region24/7fc36de72634404d8b06b7779328e791_i2_2.jpg", "1mp3d_0004_region24/d2de17fa68294749a01641894b6ab12b_i1_5.jpg", "1mp3d_0004_region24/d2de17fa68294749a01641894b6ab12b_i2_5.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0004_region24", "origin_question": "Remember, have you seen any stairs(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any stairs(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0004_region24/7fc36de72634404d8b06b7779328e791_i1_4.jpg", "1mp3d_0004_region24/7fc36de72634404d8b06b7779328e791_i2_4.jpg", "1mp3d_0004_region24/7fc36de72634404d8b06b7779328e791_i1_3.jpg", "1mp3d_0004_region24/7fc36de72634404d8b06b7779328e791_i2_3.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0004_region24", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "1mp3d_0004_region24/2697bd35e68844d09efc27d0712be5c7_i1_2.jpg", "1mp3d_0004_region24/2697bd35e68844d09efc27d0712be5c7_i2_2.jpg", "1mp3d_0004_region24/e58c9b6bf156474299c139009061c1ed_i1_3.jpg", "1mp3d_0004_region24/e58c9b6bf156474299c139009061c1ed_i2_3.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0004_region24", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "1mp3d_0004_region24/e58c9b6bf156474299c139009061c1ed_i1_4.jpg", "1mp3d_0004_region24/e58c9b6bf156474299c139009061c1ed_i2_4.jpg", "1mp3d_0004_region24/2697bd35e68844d09efc27d0712be5c7_i1_5.jpg", "1mp3d_0004_region24/2697bd35e68844d09efc27d0712be5c7_i2_5.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0004_region24", "origin_question": "How far is your current position from where you were at the end of the turn 1? (in meters)", "option": [], "answer": "1.52", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 1? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "1mp3d_0004_region24/2697bd35e68844d09efc27d0712be5c7_i1_1.jpg", "1mp3d_0004_region24/2697bd35e68844d09efc27d0712be5c7_i2_1.jpg", "1mp3d_0004_region24/e58c9b6bf156474299c139009061c1ed_i1_5.jpg", "1mp3d_0004_region24/e58c9b6bf156474299c139009061c1ed_i2_5.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0004_region24", "origin_question": "When did you first discover \"the curved black metal rail\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the curved black metal rail\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "1mp3d_0004_region24/5951cc1ffb5a47789fff75051dee56fb_i1_2.jpg", "1mp3d_0004_region24/5951cc1ffb5a47789fff75051dee56fb_i2_2.jpg", "1mp3d_0004_region24/5951cc1ffb5a47789fff75051dee56fb_i1_1.jpg", "1mp3d_0004_region24/5951cc1ffb5a47789fff75051dee56fb_i2_1.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0004_region24", "origin_question": "In which turn did you see both \"the curved black metal rail\" and \"the window frame\" simultaneously? ", "option": [], "answer": "4", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the curved black metal rail\" and \"the window frame\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0575_02/00000.jpg", "scene0575_02/00040.jpg", "scene0575_02/00080.jpg", "scene0575_02/00120.jpg", "scene0575_02/00160.jpg" ], "turn_id": 1, "scan_id": "scene0575_02", "origin_question": "Remember, have you seen any bin(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any bin(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0575_02/00200.jpg", "scene0575_02/00240.jpg", "scene0575_02/00280.jpg", "scene0575_02/00320.jpg", "scene0575_02/00360.jpg" ], "turn_id": 2, "scan_id": "scene0575_02", "origin_question": "In which turn were you farthest from \"the rectangular, light-colored wooden table\"? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the rectangular, light-colored wooden table\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0575_02/00400.jpg", "scene0575_02/00440.jpg", "scene0575_02/00480.jpg", "scene0575_02/00520.jpg", "scene0575_02/00560.jpg" ], "turn_id": 3, "scan_id": "scene0575_02", "origin_question": "", "option": [], "answer": "", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0575_02/00600.jpg", "scene0575_02/00640.jpg", "scene0575_02/00680.jpg", "scene0575_02/00720.jpg", "scene0575_02/00760.jpg" ], "turn_id": 4, "scan_id": "scene0575_02", "origin_question": "In which turn did you see both \"the rectangular metal door with glass panels\" and \"the window\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the rectangular metal door with glass panels\" and \"the window\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "scene0575_02/00800.jpg", "scene0575_02/00840.jpg", "scene0575_02/00880.jpg", "scene0575_02/00920.jpg", "scene0575_02/00960.jpg" ], "turn_id": 5, "scan_id": "scene0575_02", "origin_question": "Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Left", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0575_02/01000.jpg", "scene0575_02/01040.jpg", "scene0575_02/01080.jpg", "scene0575_02/01120.jpg", "scene0575_02/01160.jpg" ], "turn_id": 6, "scan_id": "scene0575_02", "origin_question": "Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "144.0", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0575_02/01200.jpg", "scene0575_02/01240.jpg", "scene0575_02/01280.jpg", "scene0575_02/01320.jpg", "scene0575_02/01360.jpg" ], "turn_id": 7, "scan_id": "scene0575_02", "origin_question": "Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the black cylindrical plastic bin\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the black cylindrical plastic bin\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "scene0575_02/01400.jpg", "scene0575_02/01440.jpg", "scene0575_02/01480.jpg", "scene0575_02/01520.jpg", "scene0575_02/01560.jpg" ], "turn_id": 8, "scan_id": "scene0575_02", "origin_question": "Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance forward or backward from that position?", "option": [ "Forward", "Backward" ], "answer": "Forward", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance forward or backward from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Forward', 'Backward'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0575_02/01600.jpg", "scene0575_02/01640.jpg", "scene0575_02/01680.jpg", "scene0575_02/01720.jpg", "scene0575_02/01760.jpg" ], "turn_id": 9, "scan_id": "scene0575_02", "origin_question": "When did you first discover \"the rectangular metal door with glass panels\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the rectangular metal door with glass panels\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0575_02/01800.jpg", "scene0575_02/01840.jpg", "scene0575_02/01880.jpg", "scene0575_02/01920.jpg", "scene0575_02/01960.jpg" ], "turn_id": 10, "scan_id": "scene0575_02", "origin_question": "", "option": [], "answer": "", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0518_00/00000.jpg", "scene0518_00/00040.jpg", "scene0518_00/00080.jpg", "scene0518_00/00120.jpg" ], "turn_id": 1, "scan_id": "scene0518_00", "origin_question": "Remember, how many pillow(s) have you seen so far? ", "option": [], "answer": "3", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many pillow(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0518_00/00160.jpg", "scene0518_00/00200.jpg", "scene0518_00/00240.jpg", "scene0518_00/00280.jpg" ], "turn_id": 2, "scan_id": "scene0518_00", "origin_question": "Which direction is the rectangular dark wooden cabinet with two doors in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the rectangular dark wooden cabinet with two doors in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0518_00/00320.jpg", "scene0518_00/00360.jpg", "scene0518_00/00400.jpg", "scene0518_00/00440.jpg" ], "turn_id": 3, "scan_id": "scene0518_00", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0518_00/00480.jpg", "scene0518_00/00520.jpg", "scene0518_00/00560.jpg", "scene0518_00/00600.jpg" ], "turn_id": 4, "scan_id": "scene0518_00", "origin_question": "Which direction is the flat-panel LED or OLED TV with a deep black color in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the flat-panel LED or OLED TV with a deep black color in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0518_00/00640.jpg", "scene0518_00/00680.jpg", "scene0518_00/00720.jpg", "scene0518_00/00760.jpg" ], "turn_id": 5, "scan_id": "scene0518_00", "origin_question": "Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the flat-panel LED or OLED TV with a deep black color\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the flat-panel LED or OLED TV with a deep black color\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0518_00/00800.jpg", "scene0518_00/00840.jpg", "scene0518_00/00880.jpg", "scene0518_00/00920.jpg" ], "turn_id": 6, "scan_id": "scene0518_00", "origin_question": "At the end of which turn were both of \"the medium-sized rectangular orange couch made of soft fabric\" and \"the brightly colored, medium-sized rectangular backpack\" on your right side?", "option": [], "answer": "2", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the medium-sized rectangular orange couch made of soft fabric\" and \"the brightly colored, medium-sized rectangular backpack\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0518_00/00960.jpg", "scene0518_00/01000.jpg", "scene0518_00/01040.jpg", "scene0518_00/01080.jpg" ], "turn_id": 7, "scan_id": "scene0518_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0518_00/01120.jpg", "scene0518_00/01160.jpg", "scene0518_00/01200.jpg", "scene0518_00/01240.jpg" ], "turn_id": 8, "scan_id": "scene0518_00", "origin_question": "Is the rectangular brown wall-mounted shelf with open compartments to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the rectangular brown wall-mounted shelf with open compartments to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0518_00/01280.jpg", "scene0518_00/01320.jpg", "scene0518_00/01360.jpg", "scene0518_00/01400.jpg" ], "turn_id": 9, "scan_id": "scene0518_00", "origin_question": "Remember, how many chair(s) have you seen so far? ", "option": [], "answer": "2", "user_message": "For the 9 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many chair(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0449_02/00000.jpg", "scene0449_02/00040.jpg", "scene0449_02/00080.jpg", "scene0449_02/00120.jpg", "scene0449_02/00160.jpg" ], "turn_id": 1, "scan_id": "scene0449_02", "origin_question": "Remember, have you seen any toilet(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any toilet(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0449_02/00200.jpg", "scene0449_02/00240.jpg", "scene0449_02/00280.jpg", "scene0449_02/00320.jpg", "scene0449_02/00360.jpg" ], "turn_id": 2, "scan_id": "scene0449_02", "origin_question": "In which turn did you see both \"the bathtub\" and \"the white oval marble sink with a chrome faucet\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the bathtub\" and \"the white oval marble sink with a chrome faucet\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0449_02/00400.jpg", "scene0449_02/00440.jpg", "scene0449_02/00480.jpg", "scene0449_02/00520.jpg", "scene0449_02/00560.jpg" ], "turn_id": 3, "scan_id": "scene0449_02", "origin_question": "Which direction is the toilet in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the toilet in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0449_02/00600.jpg", "scene0449_02/00640.jpg", "scene0449_02/00680.jpg", "scene0449_02/00720.jpg", "scene0449_02/00760.jpg" ], "turn_id": 4, "scan_id": "scene0449_02", "origin_question": "Is the vanity to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the vanity to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0449_02/00800.jpg", "scene0449_02/00840.jpg", "scene0449_02/00880.jpg", "scene0449_02/00920.jpg", "scene0449_02/00960.jpg" ], "turn_id": 5, "scan_id": "scene0449_02", "origin_question": "Which direction is the white oval marble sink with a chrome faucet in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the white oval marble sink with a chrome faucet in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0449_02/01000.jpg", "scene0449_02/01040.jpg", "scene0449_02/01080.jpg", "scene0449_02/01120.jpg", "scene0449_02/01160.jpg" ], "turn_id": 6, "scan_id": "scene0449_02", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the medium-sized white drawstring bag made of woven fabric\" from your position?", "option": [], "answer": "88.62", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the medium-sized white drawstring bag made of woven fabric\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0449_02/01200.jpg", "scene0449_02/01240.jpg", "scene0449_02/01280.jpg", "scene0449_02/01320.jpg", "scene0449_02/01360.jpg" ], "turn_id": 7, "scan_id": "scene0449_02", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the simple, white vertical curtain made of light material\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the simple, white vertical curtain made of light material\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0449_02/01400.jpg", "scene0449_02/01440.jpg", "scene0449_02/01480.jpg", "scene0449_02/01520.jpg", "scene0449_02/01560.jpg" ], "turn_id": 8, "scan_id": "scene0449_02", "origin_question": "Which direction is the slender stainless steel bar with a polished finish in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the slender stainless steel bar with a polished finish in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0495_00/00000.jpg", "scene0495_00/00040.jpg", "scene0495_00/00080.jpg", "scene0495_00/00120.jpg" ], "turn_id": 1, "scan_id": "scene0495_00", "origin_question": "Remember, have you seen any pillow(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any pillow(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0495_00/00160.jpg", "scene0495_00/00200.jpg", "scene0495_00/00240.jpg", "scene0495_00/00280.jpg" ], "turn_id": 2, "scan_id": "scene0495_00", "origin_question": "Remember, how many bin(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many bin(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0495_00/00320.jpg", "scene0495_00/00360.jpg", "scene0495_00/00400.jpg", "scene0495_00/00440.jpg" ], "turn_id": 3, "scan_id": "scene0495_00", "origin_question": "Remember, how many package(s) have you seen so far? ", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many package(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0495_00/00480.jpg", "scene0495_00/00520.jpg", "scene0495_00/00560.jpg", "scene0495_00/00600.jpg" ], "turn_id": 4, "scan_id": "scene0495_00", "origin_question": "Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0495_00/00640.jpg", "scene0495_00/00680.jpg", "scene0495_00/00720.jpg", "scene0495_00/00760.jpg" ], "turn_id": 5, "scan_id": "scene0495_00", "origin_question": "Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "142.35", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0495_00/00800.jpg", "scene0495_00/00840.jpg", "scene0495_00/00880.jpg", "scene0495_00/00920.jpg" ], "turn_id": 6, "scan_id": "scene0495_00", "origin_question": "Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "66.15", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0067_region5/bd4a599f645342adafc1b09a953431e8_i1_0.jpg", "1mp3d_0067_region5/bd4a599f645342adafc1b09a953431e8_i2_0.jpg", "1mp3d_0067_region5/a7c9d5a193dd450fb2d2f04f71935ca4_i1_0.jpg", "1mp3d_0067_region5/a7c9d5a193dd450fb2d2f04f71935ca4_i2_0.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0067_region5", "origin_question": "Remember, have you seen any chair(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any chair(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0067_region5/bd4a599f645342adafc1b09a953431e8_i1_1.jpg", "1mp3d_0067_region5/bd4a599f645342adafc1b09a953431e8_i2_1.jpg", "1mp3d_0067_region5/a7c9d5a193dd450fb2d2f04f71935ca4_i1_1.jpg", "1mp3d_0067_region5/a7c9d5a193dd450fb2d2f04f71935ca4_i2_1.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0067_region5", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "1mp3d_0067_region5/bd4a599f645342adafc1b09a953431e8_i1_2.jpg", "1mp3d_0067_region5/bd4a599f645342adafc1b09a953431e8_i2_2.jpg", "1mp3d_0067_region5/c370bbcbfbf649e38524754b668bb1b2_i1_3.jpg", "1mp3d_0067_region5/c370bbcbfbf649e38524754b668bb1b2_i2_3.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0067_region5", "origin_question": "How far is your current position from where you were at the end of the turn 1? (in meters)", "option": [], "answer": "3.51", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 1? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "1mp3d_0067_region5/7fb0cc6c0754461ca29084c1d793b600_i1_4.jpg", "1mp3d_0067_region5/7fb0cc6c0754461ca29084c1d793b600_i2_4.jpg", "1mp3d_0067_region5/c370bbcbfbf649e38524754b668bb1b2_i1_5.jpg", "1mp3d_0067_region5/c370bbcbfbf649e38524754b668bb1b2_i2_5.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0067_region5", "origin_question": "When did you first discover \"the rectangular blue picture in perfect condition\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the rectangular blue picture in perfect condition\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0067_region5/c370bbcbfbf649e38524754b668bb1b2_i1_0.jpg", "1mp3d_0067_region5/c370bbcbfbf649e38524754b668bb1b2_i2_0.jpg", "1mp3d_0067_region5/c370bbcbfbf649e38524754b668bb1b2_i1_1.jpg", "1mp3d_0067_region5/c370bbcbfbf649e38524754b668bb1b2_i2_1.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0067_region5", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the rectangular blue picture in perfect condition\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the rectangular blue picture in perfect condition\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0067_region5/7fb0cc6c0754461ca29084c1d793b600_i1_0.jpg", "1mp3d_0067_region5/7fb0cc6c0754461ca29084c1d793b600_i2_0.jpg", "1mp3d_0067_region5/7fb0cc6c0754461ca29084c1d793b600_i1_5.jpg", "1mp3d_0067_region5/7fb0cc6c0754461ca29084c1d793b600_i2_5.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0067_region5", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the rectangular blue picture in perfect condition\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the rectangular blue picture in perfect condition\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0362_03/00000.jpg", "scene0362_03/00040.jpg", "scene0362_03/00080.jpg", "scene0362_03/00120.jpg" ], "turn_id": 1, "scan_id": "scene0362_03", "origin_question": "Remember, have you seen any table(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any table(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0362_03/00160.jpg", "scene0362_03/00200.jpg", "scene0362_03/00240.jpg", "scene0362_03/00280.jpg" ], "turn_id": 2, "scan_id": "scene0362_03", "origin_question": "In which turn did you see both \"the full-length rectangular beige curtain made of fabric\" and \"the horizontal light beige metal radiator with protrusions\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the full-length rectangular beige curtain made of fabric\" and \"the horizontal light beige metal radiator with protrusions\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0362_03/00320.jpg", "scene0362_03/00360.jpg", "scene0362_03/00400.jpg", "scene0362_03/00440.jpg" ], "turn_id": 3, "scan_id": "scene0362_03", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0362_03/00480.jpg", "scene0362_03/00520.jpg", "scene0362_03/00560.jpg", "scene0362_03/00600.jpg" ], "turn_id": 4, "scan_id": "scene0362_03", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the modern abstract picture with geometric shapes in red, yellow, and beige\" from your position?", "option": [], "answer": "138.75", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the modern abstract picture with geometric shapes in red, yellow, and beige\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "scene0362_03/00640.jpg", "scene0362_03/00680.jpg", "scene0362_03/00720.jpg", "scene0362_03/00760.jpg" ], "turn_id": 5, "scan_id": "scene0362_03", "origin_question": "Among these three objects(\"the rectangular two-seater covered in rich brown fabric\",\"the modern abstract picture with geometric shapes in red, yellow, and beige\",\"the white standard-sized paper with black text\" ), which one is the closest to you?", "option": [ "the rectangular two-seater covered in rich brown fabric", "the modern abstract picture with geometric shapes in red, yellow, and beige", "the white standard-sized paper with black text" ], "answer": "the rectangular two-seater covered in rich brown fabric", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects(\"the rectangular two-seater covered in rich brown fabric\",\"the modern abstract picture with geometric shapes in red, yellow, and beige\",\"the white standard-sized paper with black text\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular two-seater covered in rich brown fabric', 'the modern abstract picture with geometric shapes in red, yellow, and beige', 'the white standard-sized paper with black text'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0362_03/00800.jpg", "scene0362_03/00840.jpg", "scene0362_03/00880.jpg", "scene0362_03/00920.jpg" ], "turn_id": 6, "scan_id": "scene0362_03", "origin_question": "Which direction is the white standard-sized paper with black text in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the white standard-sized paper with black text in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0362_03/00960.jpg", "scene0362_03/01000.jpg", "scene0362_03/01040.jpg", "scene0362_03/01080.jpg" ], "turn_id": 7, "scan_id": "scene0362_03", "origin_question": "Is the long, dark-colored wool jacket to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the long, dark-colored wool jacket to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0382_00/00000.jpg", "scene0382_00/00030.jpg", "scene0382_00/00060.jpg", "scene0382_00/00090.jpg" ], "turn_id": 1, "scan_id": "scene0382_00", "origin_question": "Remember, have you seen any table(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any table(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0382_00/00120.jpg", "scene0382_00/00150.jpg", "scene0382_00/00180.jpg", "scene0382_00/00210.jpg" ], "turn_id": 2, "scan_id": "scene0382_00", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the white rectangular metal radiator with holes on top\"; \"the black ergonomic plastic chair with a curved backrest\"; \"the rectangular light-colored wooden shelf\" ", "option": [ "the white rectangular metal radiator with holes on top", "the black ergonomic plastic chair with a curved backrest", "the rectangular light-colored wooden shelf" ], "answer": "the white rectangular metal radiator with holes on top", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the white rectangular metal radiator with holes on top\"; \"the black ergonomic plastic chair with a curved backrest\"; \"the rectangular light-colored wooden shelf\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the white rectangular metal radiator with holes on top', 'the black ergonomic plastic chair with a curved backrest', 'the rectangular light-colored wooden shelf'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0382_00/00240.jpg", "scene0382_00/00270.jpg", "scene0382_00/00300.jpg", "scene0382_00/00330.jpg" ], "turn_id": 3, "scan_id": "scene0382_00", "origin_question": "Which direction is the rectangular light-colored wooden table with cylindrical legs in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the rectangular light-colored wooden table with cylindrical legs in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0382_00/00360.jpg", "scene0382_00/00390.jpg", "scene0382_00/00420.jpg", "scene0382_00/00450.jpg" ], "turn_id": 4, "scan_id": "scene0382_00", "origin_question": "Please recall \"the black ergonomic plastic chair with a curved backrest\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.55", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the black ergonomic plastic chair with a curved backrest\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0382_00/00480.jpg", "scene0382_00/00510.jpg", "scene0382_00/00540.jpg", "scene0382_00/00570.jpg" ], "turn_id": 5, "scan_id": "scene0382_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0382_00/00600.jpg", "scene0382_00/00870.jpg", "scene0382_00/00900.jpg", "scene0382_00/00950.jpg" ], "turn_id": 6, "scan_id": "scene0382_00", "origin_question": "Is the rectangular light-colored wooden shelf to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the rectangular light-colored wooden shelf to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0382_00/00980.jpg", "scene0382_00/01010.jpg", "scene0382_00/01040.jpg", "scene0382_00/01070.jpg" ], "turn_id": 7, "scan_id": "scene0382_00", "origin_question": "By the end of the 5 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular light-colored wooden shelf\", Y: \"the black ergonomic plastic chair with a curved backrest\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to Y and farther away from X", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 5 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular light-colored wooden shelf\", Y: \"the black ergonomic plastic chair with a curved backrest\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0047_region6/0eb44473678c4fa292d991f7e8b1f85f_i1_4.jpg", "1mp3d_0047_region6/0eb44473678c4fa292d991f7e8b1f85f_i2_4.jpg", "1mp3d_0047_region6/60eecd51a84c423a851efcaa74df2ac9_i1_4.jpg", "1mp3d_0047_region6/60eecd51a84c423a851efcaa74df2ac9_i2_4.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0047_region6", "origin_question": "Remember, have you seen any chair(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any chair(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0047_region6/de07efde6cf54bc28fd6f33b7cffe069_i1_5.jpg", "1mp3d_0047_region6/de07efde6cf54bc28fd6f33b7cffe069_i2_5.jpg", "1mp3d_0047_region6/3a59799301704bfbbcc96922e94c306a_i1_1.jpg", "1mp3d_0047_region6/3a59799301704bfbbcc96922e94c306a_i2_1.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0047_region6", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "1mp3d_0047_region6/de07efde6cf54bc28fd6f33b7cffe069_i1_0.jpg", "1mp3d_0047_region6/de07efde6cf54bc28fd6f33b7cffe069_i2_0.jpg", "1mp3d_0047_region6/0cef156ab53041da97dd6a70d3d5af0b_i1_1.jpg", "1mp3d_0047_region6/0cef156ab53041da97dd6a70d3d5af0b_i2_1.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0047_region6", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "2.4", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "1mp3d_0047_region6/de07efde6cf54bc28fd6f33b7cffe069_i1_1.jpg", "1mp3d_0047_region6/de07efde6cf54bc28fd6f33b7cffe069_i2_1.jpg", "1mp3d_0047_region6/60eecd51a84c423a851efcaa74df2ac9_i1_5.jpg", "1mp3d_0047_region6/60eecd51a84c423a851efcaa74df2ac9_i2_5.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0047_region6", "origin_question": "How far is your current position from where you were at the end of the turn 3? (in meters)", "option": [], "answer": "3.2", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 3? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "1mp3d_0047_region6/0eb44473678c4fa292d991f7e8b1f85f_i1_5.jpg", "1mp3d_0047_region6/0eb44473678c4fa292d991f7e8b1f85f_i2_5.jpg", "1mp3d_0047_region6/8b675c426b5f4d7ca4ef9000da426789_i1_4.jpg", "1mp3d_0047_region6/8b675c426b5f4d7ca4ef9000da426789_i2_4.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0047_region6", "origin_question": "How far is your current position from where you were at the end of the turn 3? (in meters)", "option": [], "answer": "5.17", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 3? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "1mp3d_0047_region6/78f4fd1a913946059be779b7507870b3_i1_3.jpg", "1mp3d_0047_region6/78f4fd1a913946059be779b7507870b3_i2_3.jpg", "1mp3d_0047_region6/8b675c426b5f4d7ca4ef9000da426789_i1_3.jpg", "1mp3d_0047_region6/8b675c426b5f4d7ca4ef9000da426789_i2_3.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0047_region6", "origin_question": "Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Right", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0269_01/00000.jpg", "scene0269_01/00030.jpg", "scene0269_01/00060.jpg", "scene0269_01/00090.jpg" ], "turn_id": 1, "scan_id": "scene0269_01", "origin_question": "Remember, have you seen any crib(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any crib(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "scene0269_01/00120.jpg", "scene0269_01/00150.jpg", "scene0269_01/00180.jpg", "scene0269_01/00210.jpg" ], "turn_id": 2, "scan_id": "scene0269_01", "origin_question": "Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Right", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "scene0269_01/00240.jpg", "scene0269_01/00270.jpg", "scene0269_01/00300.jpg", "scene0269_01/00330.jpg" ], "turn_id": 3, "scan_id": "scene0269_01", "origin_question": "Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Right", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "scene0269_01/00360.jpg", "scene0269_01/00390.jpg", "scene0269_01/00420.jpg", "scene0269_01/00450.jpg" ], "turn_id": 4, "scan_id": "scene0269_01", "origin_question": "Assuming the direction you are facing at the end of the turn 3 is forward, did you move a certain distance forward or backward from that position?", "option": [ "Forward", "Backward" ], "answer": "Forward", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 3 is forward, did you move a certain distance forward or backward from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Forward', 'Backward'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0269_01/00480.jpg", "scene0269_01/00510.jpg", "scene0269_01/00540.jpg", "scene0269_01/00570.jpg" ], "turn_id": 5, "scan_id": "scene0269_01", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "3.15", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "scene0269_01/00600.jpg", "scene0269_01/00650.jpg", "scene0269_01/00680.jpg", "scene0269_01/00710.jpg" ], "turn_id": 6, "scan_id": "scene0269_01", "origin_question": "Assuming the direction you are facing at the end of the turn 3 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Left", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 3 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0269_01/00740.jpg", "scene0269_01/00770.jpg", "scene0269_01/00800.jpg", "scene0269_01/00830.jpg" ], "turn_id": 7, "scan_id": "scene0269_01", "origin_question": "How far is your current position from where you were at the end of the turn 1? (in meters)", "option": [], "answer": "1.99", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 1? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0078_region12/a860e5bf7afe44179bcdc838db421802_i2_2.jpg", "1mp3d_0078_region12/ab0f99b210bb4556b7fc18190126480b_i1_5.jpg", "1mp3d_0078_region12/ab0f99b210bb4556b7fc18190126480b_i2_5.jpg", "1mp3d_0078_region12/85849923dc9e46eda0168f62c0fb43f3_i2_5.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0078_region12", "origin_question": "Remember, have you seen any stairs(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any stairs(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "1mp3d_0078_region12/ab0f99b210bb4556b7fc18190126480b_i2_0.jpg", "1mp3d_0078_region12/0836acc78091475dbba186b98e05792a_i1_2.jpg", "1mp3d_0078_region12/0836acc78091475dbba186b98e05792a_i2_2.jpg", "1mp3d_0078_region12/0836acc78091475dbba186b98e05792a_i2_3.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0078_region12", "origin_question": "At the end of which turn were both of \"the rectangular window with a light brown wooden frame\" and \"the dark, hand-carved wooden statue with intricate carvings\" on your left side?", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the rectangular window with a light brown wooden frame\" and \"the dark, hand-carved wooden statue with intricate carvings\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0078_region12/9ddd0d8326b44a038d3985d96ffbb2a1_i1_2.jpg", "1mp3d_0078_region12/9ddd0d8326b44a038d3985d96ffbb2a1_i2_2.jpg", "1mp3d_0078_region12/9ddd0d8326b44a038d3985d96ffbb2a1_i1_3.jpg", "1mp3d_0078_region12/9ddd0d8326b44a038d3985d96ffbb2a1_i2_3.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0078_region12", "origin_question": "", "option": [], "answer": "", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "1mp3d_0078_region12/959878d8792948b5888684622f4eca36_i1_1.jpg", "1mp3d_0078_region12/959878d8792948b5888684622f4eca36_i2_1.jpg", "1mp3d_0078_region12/0896e14268a54c1faa6a5648eb8eb63b_i2_0.jpg", "1mp3d_0078_region12/0896e14268a54c1faa6a5648eb8eb63b_i2_5.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0078_region12", "origin_question": "At the end of which turn were both of \"Light-colored wooden stairs with a unique design\" and \"the long, light brown wooden rail\" on your left side?", "option": [], "answer": "4", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"Light-colored wooden stairs with a unique design\" and \"the long, light brown wooden rail\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0086_01/00000.jpg", "scene0086_01/00040.jpg", "scene0086_01/00080.jpg", "scene0086_01/00120.jpg" ], "turn_id": 1, "scan_id": "scene0086_01", "origin_question": "Remember, have you seen any keyboard(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any keyboard(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0086_01/00160.jpg", "scene0086_01/00200.jpg", "scene0086_01/00240.jpg", "scene0086_01/00280.jpg" ], "turn_id": 2, "scan_id": "scene0086_01", "origin_question": "When did you first discover \"the sleek black plastic soap dispenser with gray interior\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the sleek black plastic soap dispenser with gray interior\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0086_01/00320.jpg", "scene0086_01/00360.jpg", "scene0086_01/00400.jpg", "scene0086_01/00440.jpg" ], "turn_id": 3, "scan_id": "scene0086_01", "origin_question": "What will be the first-time appearance order of the following categories: grab bar, sink, toilet?", "option": [ "grab bar, sink, toilet", "grab bar, toilet, sink", "sink, toilet, grab bar", "sink, grab bar, toilet" ], "answer": "sink, grab bar, toilet", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: grab bar, sink, toilet? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['grab bar, sink, toilet', 'grab bar, toilet, sink', 'sink, toilet, grab bar', 'sink, grab bar, toilet'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0086_01/00480.jpg", "scene0086_01/00520.jpg", "scene0086_01/00560.jpg", "scene0086_01/00600.jpg" ], "turn_id": 4, "scan_id": "scene0086_01", "origin_question": "Remember, how many mailbox(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many mailbox(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0086_01/00640.jpg", "scene0086_01/00680.jpg", "scene0086_01/00720.jpg", "scene0086_01/00760.jpg" ], "turn_id": 5, "scan_id": "scene0086_01", "origin_question": "Which direction is the sleek black plastic soap dispenser with gray interior in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the sleek black plastic soap dispenser with gray interior in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0086_01/00800.jpg", "scene0086_01/00840.jpg", "scene0086_01/00880.jpg", "scene0086_01/00920.jpg" ], "turn_id": 6, "scan_id": "scene0086_01", "origin_question": "Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the sleek black plastic soap dispenser with gray interior\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the sleek black plastic soap dispenser with gray interior\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0086_01/00960.jpg", "scene0086_01/01000.jpg", "scene0086_01/01040.jpg", "scene0086_01/01080.jpg" ], "turn_id": 7, "scan_id": "scene0086_01", "origin_question": "What will be the first-time appearance order of the following categories: bin, door, mirror?", "option": [ "bin, door, mirror", "bin, mirror, door", "door, mirror, bin", "door, bin, mirror" ], "answer": "bin, door, mirror", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bin, door, mirror? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['bin, door, mirror', 'bin, mirror, door', 'door, mirror, bin', 'door, bin, mirror'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0086_01/01120.jpg", "scene0086_01/01160.jpg", "scene0086_01/01200.jpg", "scene0086_01/01240.jpg" ], "turn_id": 8, "scan_id": "scene0086_01", "origin_question": "What will be the first-time appearance order of the following categories: bin, bottle, door?", "option": [ "door, bottle, bin", "door, bin, bottle", "bin, door, bottle", "bottle, door, bin" ], "answer": "bin, door, bottle", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bin, bottle, door? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['door, bottle, bin', 'door, bin, bottle', 'bin, door, bottle', 'bottle, door, bin'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0685_00/00000.jpg", "scene0685_00/00040.jpg", "scene0685_00/00080.jpg", "scene0685_00/00120.jpg", "scene0685_00/00160.jpg" ], "turn_id": 1, "scan_id": "scene0685_00", "origin_question": "Remember, have you seen any bin(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any bin(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0685_00/00200.jpg", "scene0685_00/00240.jpg", "scene0685_00/00280.jpg", "scene0685_00/00320.jpg", "scene0685_00/00360.jpg" ], "turn_id": 2, "scan_id": "scene0685_00", "origin_question": "Remember, have you seen any door(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any door(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0685_00/00400.jpg", "scene0685_00/00440.jpg", "scene0685_00/00480.jpg", "scene0685_00/00520.jpg", "scene0685_00/00560.jpg" ], "turn_id": 3, "scan_id": "scene0685_00", "origin_question": "In which turn did you see both \"the green rectangular plastic bin with a lid\" and \"the blue roll of plastic adhesive tape\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the green rectangular plastic bin with a lid\" and \"the blue roll of plastic adhesive tape\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0685_00/00600.jpg", "scene0685_00/00640.jpg", "scene0685_00/00680.jpg", "scene0685_00/00720.jpg", "scene0685_00/00760.jpg" ], "turn_id": 4, "scan_id": "scene0685_00", "origin_question": "When did you first discover \"the green rectangular plastic bin with a lid\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the green rectangular plastic bin with a lid\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0685_00/00800.jpg", "scene0685_00/00840.jpg", "scene0685_00/00880.jpg", "scene0685_00/00920.jpg", "scene0685_00/00960.jpg" ], "turn_id": 5, "scan_id": "scene0685_00", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0685_00/01000.jpg", "scene0685_00/01040.jpg", "scene0685_00/01080.jpg", "scene0685_00/01120.jpg", "scene0685_00/01160.jpg" ], "turn_id": 6, "scan_id": "scene0685_00", "origin_question": "Using your orientation at the end of turn 5 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 5 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0685_00/01200.jpg", "scene0685_00/01240.jpg", "scene0685_00/01280.jpg", "scene0685_00/01320.jpg", "scene0685_00/01360.jpg" ], "turn_id": 7, "scan_id": "scene0685_00", "origin_question": "At the end of which turn were both of \"the green rectangular plastic bin with a lid\" and \"the blue roll of plastic adhesive tape\" on your right side?", "option": [], "answer": "4", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the green rectangular plastic bin with a lid\" and \"the blue roll of plastic adhesive tape\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0685_00/01400.jpg", "scene0685_00/01440.jpg", "scene0685_00/01480.jpg", "scene0685_00/01520.jpg", "scene0685_00/01560.jpg" ], "turn_id": 8, "scan_id": "scene0685_00", "origin_question": "By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the large rectangular glossy paper with markings\", Y: \"the rectangular light-colored wooden cabinet with two drawers\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "farther away from both X and Y", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the large rectangular glossy paper with markings\", Y: \"the rectangular light-colored wooden cabinet with two drawers\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0685_00/01600.jpg", "scene0685_00/01640.jpg", "scene0685_00/01680.jpg", "scene0685_00/01720.jpg", "scene0685_00/01760.jpg" ], "turn_id": 9, "scan_id": "scene0685_00", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the large rectangular glossy paper with markings\" from your position?", "option": [], "answer": "150.29", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the large rectangular glossy paper with markings\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "1mp3d_0044_region8/530469cd7a1949f98fc2ff465b67a201_i1_4.jpg", "1mp3d_0044_region8/530469cd7a1949f98fc2ff465b67a201_i2_4.jpg", "1mp3d_0044_region8/8c5462fa7d924d5b8ae9877ea7d3e089_i1_1.jpg", "1mp3d_0044_region8/8c5462fa7d924d5b8ae9877ea7d3e089_i2_1.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0044_region8", "origin_question": "Remember, how many printer(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many printer(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "1mp3d_0044_region8/530469cd7a1949f98fc2ff465b67a201_i1_0.jpg", "1mp3d_0044_region8/530469cd7a1949f98fc2ff465b67a201_i2_0.jpg", "1mp3d_0044_region8/8e05a52536d6480c8b3ccea3332e8d39_i1_2.jpg", "1mp3d_0044_region8/8e05a52536d6480c8b3ccea3332e8d39_i2_2.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0044_region8", "origin_question": "Please recall \"the granite counter with a speckled brown and yellow pattern\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "3.25", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the granite counter with a speckled brown and yellow pattern\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "1mp3d_0044_region8/8e05a52536d6480c8b3ccea3332e8d39_i1_1.jpg", "1mp3d_0044_region8/8e05a52536d6480c8b3ccea3332e8d39_i2_1.jpg", "1mp3d_0044_region8/02fc112c72f54dad9c127dc99ad242be_i1_1.jpg", "1mp3d_0044_region8/02fc112c72f54dad9c127dc99ad242be_i2_1.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0044_region8", "origin_question": "In which turn were you farthest from \"the rectangular stainless steel microwave oven with a black door\"? ", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the rectangular stainless steel microwave oven with a black door\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "1mp3d_0044_region8/dd6fb65d75ca41d2aa13e81410291aa1_i1_3.jpg", "1mp3d_0044_region8/dd6fb65d75ca41d2aa13e81410291aa1_i2_3.jpg", "1mp3d_0044_region8/ef2b67b567304849bc8724e63f2f2721_i1_1.jpg", "1mp3d_0044_region8/ef2b67b567304849bc8724e63f2f2721_i2_1.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0044_region8", "origin_question": "When did you first discover \"the rectangular modern gas stove with stainless steel exterior\"? (index of the turn)", "option": [], "answer": "3", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the rectangular modern gas stove with stainless steel exterior\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "1mp3d_0044_region8/c7c12a431655490198f83edfa2ba64ae_i1_1.jpg", "1mp3d_0044_region8/c7c12a431655490198f83edfa2ba64ae_i2_1.jpg", "1mp3d_0044_region8/ece144083db74fd4b97d109d6c920934_i1_4.jpg", "1mp3d_0044_region8/ece144083db74fd4b97d109d6c920934_i2_4.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0044_region8", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the faucet\" from your position?", "option": [], "answer": "60.55", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the faucet\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0044_region8/ece144083db74fd4b97d109d6c920934_i1_3.jpg", "1mp3d_0044_region8/ece144083db74fd4b97d109d6c920934_i2_3.jpg", "1mp3d_0044_region8/c7c12a431655490198f83edfa2ba64ae_i1_0.jpg", "1mp3d_0044_region8/c7c12a431655490198f83edfa2ba64ae_i2_0.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0044_region8", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the rectangular, brown stainless steel range hood with carvings\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the rectangular, brown stainless steel range hood with carvings\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "1mp3d_0044_region8/ef2b67b567304849bc8724e63f2f2721_i1_2.jpg", "1mp3d_0044_region8/ef2b67b567304849bc8724e63f2f2721_i2_2.jpg", "1mp3d_0044_region8/dd6fb65d75ca41d2aa13e81410291aa1_i1_2.jpg", "1mp3d_0044_region8/dd6fb65d75ca41d2aa13e81410291aa1_i2_2.jpg" ], "turn_id": 7, "scan_id": "1mp3d_0044_region8", "origin_question": "Assuming the direction you are facing at the end of the turn 6 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Right", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 6 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "1mp3d_0044_region8/02fc112c72f54dad9c127dc99ad242be_i1_5.jpg", "1mp3d_0044_region8/02fc112c72f54dad9c127dc99ad242be_i2_5.jpg", "1mp3d_0044_region8/02fc112c72f54dad9c127dc99ad242be_i1_2.jpg", "1mp3d_0044_region8/02fc112c72f54dad9c127dc99ad242be_i2_2.jpg" ], "turn_id": 8, "scan_id": "1mp3d_0044_region8", "origin_question": "What will be the first-time appearance order of the following categories: faucet, plant, vase?", "option": [ "plant, faucet, vase", "faucet, vase, plant", "vase, faucet, plant", "vase, plant, faucet" ], "answer": "plant, faucet, vase", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: faucet, plant, vase? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['plant, faucet, vase', 'faucet, vase, plant', 'vase, faucet, plant', 'vase, plant, faucet'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "1mp3d_0044_region8/02fc112c72f54dad9c127dc99ad242be_i1_0.jpg", "1mp3d_0044_region8/02fc112c72f54dad9c127dc99ad242be_i2_0.jpg", "1mp3d_0044_region8/bf5ac8319e8d48568b82d9f36a064969_i1_4.jpg", "1mp3d_0044_region8/bf5ac8319e8d48568b82d9f36a064969_i2_4.jpg" ], "turn_id": 9, "scan_id": "1mp3d_0044_region8", "origin_question": "Remember, how many bathtub(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 9 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many bathtub(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0656_03/00000.jpg", "scene0656_03/00040.jpg", "scene0656_03/00080.jpg", "scene0656_03/00120.jpg" ], "turn_id": 1, "scan_id": "scene0656_03", "origin_question": "Remember, have you seen any chair(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any chair(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0656_03/00160.jpg", "scene0656_03/00200.jpg", "scene0656_03/00240.jpg", "scene0656_03/00280.jpg" ], "turn_id": 2, "scan_id": "scene0656_03", "origin_question": "Please recall \"the rectangular white cardboard tissue box\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.52", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the rectangular white cardboard tissue box\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0656_03/00320.jpg", "scene0656_03/00360.jpg", "scene0656_03/00400.jpg", "scene0656_03/00440.jpg" ], "turn_id": 3, "scan_id": "scene0656_03", "origin_question": "What will be the first-time appearance order of the following categories: bin, picture, shelf?", "option": [ "picture, shelf, bin", "picture, bin, shelf", "bin, picture, shelf", "shelf, bin, picture" ], "answer": "bin, picture, shelf", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bin, picture, shelf? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['picture, shelf, bin', 'picture, bin, shelf', 'bin, picture, shelf', 'shelf, bin, picture'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0656_03/00480.jpg", "scene0656_03/00520.jpg", "scene0656_03/00560.jpg", "scene0656_03/00600.jpg" ], "turn_id": 4, "scan_id": "scene0656_03", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the rectangular white cardboard tissue box\" from your position?", "option": [], "answer": "91.21", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the rectangular white cardboard tissue box\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0656_03/00640.jpg", "scene0656_03/00680.jpg", "scene0656_03/00720.jpg", "scene0656_03/00760.jpg" ], "turn_id": 5, "scan_id": "scene0656_03", "origin_question": "In which turn were you closest to \"the rectangular white cardboard tissue box\"? ", "option": [], "answer": "5", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the rectangular white cardboard tissue box\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0656_03/00800.jpg", "scene0656_03/00840.jpg", "scene0656_03/00880.jpg", "scene0656_03/00920.jpg" ], "turn_id": 6, "scan_id": "scene0656_03", "origin_question": "What will be the first-time appearance order of the following categories: curtain, door, vanity?", "option": [ "door, vanity, curtain", "curtain, door, vanity", "curtain, vanity, door", "door, curtain, vanity" ], "answer": "curtain, door, vanity", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: curtain, door, vanity? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['door, vanity, curtain', 'curtain, door, vanity', 'curtain, vanity, door', 'door, curtain, vanity'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0656_03/00960.jpg", "scene0656_03/01000.jpg", "scene0656_03/01040.jpg", "scene0656_03/01080.jpg" ], "turn_id": 7, "scan_id": "scene0656_03", "origin_question": "Which two objects are on the same side of you now? \"Creamy white double-layered curtains with subtle patterns\", \"the deep brown wooden table with brass handles\", \"the medium-sized rectangular dark brown wooden dresser\". ", "option": [ "Creamy white double-layered curtains with subtle patterns, the deep brown wooden table with brass handles", "Creamy white double-layered curtains with subtle patterns, the medium-sized rectangular dark brown wooden dresser", "the deep brown wooden table with brass handles, the medium-sized rectangular dark brown wooden dresser" ], "answer": "Creamy white double-layered curtains with subtle patterns, the deep brown wooden table with brass handles", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"Creamy white double-layered curtains with subtle patterns\", \"the deep brown wooden table with brass handles\", \"the medium-sized rectangular dark brown wooden dresser\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Creamy white double-layered curtains with subtle patterns, the deep brown wooden table with brass handles', 'Creamy white double-layered curtains with subtle patterns, the medium-sized rectangular dark brown wooden dresser', 'the deep brown wooden table with brass handles, the medium-sized rectangular dark brown wooden dresser'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0656_03/01120.jpg", "scene0656_03/01160.jpg", "scene0656_03/01200.jpg", "scene0656_03/01240.jpg" ], "turn_id": 8, "scan_id": "scene0656_03", "origin_question": "Which two objects are on the same side of you now? \"the bed\", \"the clean, rectangular white sheet of paper\", \"the three-tiered wooden shelf for storage\". ", "option": [ "the bed, the clean, rectangular white sheet of paper", "the bed, the three-tiered wooden shelf for storage", "the clean, rectangular white sheet of paper, the three-tiered wooden shelf for storage" ], "answer": "the clean, rectangular white sheet of paper, the three-tiered wooden shelf for storage", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the bed\", \"the clean, rectangular white sheet of paper\", \"the three-tiered wooden shelf for storage\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the bed, the clean, rectangular white sheet of paper', 'the bed, the three-tiered wooden shelf for storage', 'the clean, rectangular white sheet of paper, the three-tiered wooden shelf for storage'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0656_03/01280.jpg", "scene0656_03/01320.jpg", "scene0656_03/01360.jpg", "scene0656_03/01400.jpg" ], "turn_id": 9, "scan_id": "scene0656_03", "origin_question": "When did you first discover \"the pink cylindrical plastic bin\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 9 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the pink cylindrical plastic bin\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0228_00/00000.jpg", "scene0228_00/00040.jpg", "scene0228_00/00080.jpg", "scene0228_00/00120.jpg", "scene0228_00/00160.jpg" ], "turn_id": 1, "scan_id": "scene0228_00", "origin_question": "Remember, how many laptop(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many laptop(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0228_00/00200.jpg", "scene0228_00/00240.jpg", "scene0228_00/00280.jpg", "scene0228_00/00320.jpg", "scene0228_00/00360.jpg" ], "turn_id": 2, "scan_id": "scene0228_00", "origin_question": "Remember, have you seen any bin(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any bin(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0228_00/00400.jpg", "scene0228_00/00440.jpg", "scene0228_00/00480.jpg", "scene0228_00/00520.jpg", "scene0228_00/00560.jpg" ], "turn_id": 3, "scan_id": "scene0228_00", "origin_question": "How far is your current position from where you were at the end of the turn 1? (in meters)", "option": [], "answer": "6.41", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 1? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0228_00/00600.jpg", "scene0228_00/00640.jpg", "scene0228_00/00680.jpg", "scene0228_00/00720.jpg", "scene0228_00/00760.jpg" ], "turn_id": 4, "scan_id": "scene0228_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0228_00/00800.jpg", "scene0228_00/00840.jpg", "scene0228_00/00880.jpg", "scene0228_00/00920.jpg", "scene0228_00/00960.jpg" ], "turn_id": 5, "scan_id": "scene0228_00", "origin_question": "Using your orientation at the end of turn 4 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "50.47", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 4 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "scene0228_00/01000.jpg", "scene0228_00/01040.jpg", "scene0228_00/01080.jpg", "scene0228_00/01120.jpg", "scene0228_00/01160.jpg" ], "turn_id": 6, "scan_id": "scene0228_00", "origin_question": "Assuming the direction you are facing at the end of the turn 5 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Right", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 5 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0228_00/01200.jpg", "scene0228_00/01240.jpg", "scene0228_00/01280.jpg", "scene0228_00/01320.jpg", "scene0228_00/01360.jpg" ], "turn_id": 7, "scan_id": "scene0228_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0228_00/01400.jpg", "scene0228_00/01440.jpg", "scene0228_00/01480.jpg", "scene0228_00/01520.jpg", "scene0228_00/01560.jpg" ], "turn_id": 8, "scan_id": "scene0228_00", "origin_question": "Using your orientation at the end of turn 7 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 7 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0228_00/01600.jpg", "scene0228_00/01640.jpg", "scene0228_00/01680.jpg", "scene0228_00/01720.jpg", "scene0228_00/01760.jpg" ], "turn_id": 9, "scan_id": "scene0228_00", "origin_question": "Remember, have you seen any plunger(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any plunger(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0228_00/01800.jpg", "scene0228_00/01840.jpg", "scene0228_00/01880.jpg", "scene0228_00/01920.jpg", "scene0228_00/01960.jpg" ], "turn_id": 10, "scan_id": "scene0228_00", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0057_region23/231d8560066e49dbb56c6e2920cd6c67_i1_3.jpg", "1mp3d_0057_region23/231d8560066e49dbb56c6e2920cd6c67_i2_3.jpg", "1mp3d_0057_region23/231d8560066e49dbb56c6e2920cd6c67_i1_2.jpg", "1mp3d_0057_region23/231d8560066e49dbb56c6e2920cd6c67_i2_2.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0057_region23", "origin_question": "Remember, have you seen any bed(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any bed(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "1mp3d_0057_region23/92d5c09481694d5f8792cda91271428f_i1_3.jpg", "1mp3d_0057_region23/92d5c09481694d5f8792cda91271428f_i2_3.jpg", "1mp3d_0057_region23/4f39c10703bb432baf1cef0c935037ea_i1_4.jpg", "1mp3d_0057_region23/4f39c10703bb432baf1cef0c935037ea_i2_4.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0057_region23", "origin_question": "When did you first discover \"the dark blue rectangular pillow made of soft fabric\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the dark blue rectangular pillow made of soft fabric\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0057_region23/df8abc262daa4ab391ecd1d10ba8fa32_i1_3.jpg", "1mp3d_0057_region23/df8abc262daa4ab391ecd1d10ba8fa32_i2_3.jpg", "1mp3d_0057_region23/90c0daf91e49430994f7b4196cf24064_i1_3.jpg", "1mp3d_0057_region23/90c0daf91e49430994f7b4196cf24064_i2_3.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0057_region23", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the light brown circular wooden rail\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the light brown circular wooden rail\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "1mp3d_0057_region23/90c0daf91e49430994f7b4196cf24064_i1_4.jpg", "1mp3d_0057_region23/90c0daf91e49430994f7b4196cf24064_i2_4.jpg", "1mp3d_0057_region23/df8abc262daa4ab391ecd1d10ba8fa32_i1_4.jpg", "1mp3d_0057_region23/df8abc262daa4ab391ecd1d10ba8fa32_i2_4.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0057_region23", "origin_question": "Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "1mp3d_0057_region23/df8abc262daa4ab391ecd1d10ba8fa32_i1_5.jpg", "1mp3d_0057_region23/df8abc262daa4ab391ecd1d10ba8fa32_i2_5.jpg", "1mp3d_0057_region23/4f39c10703bb432baf1cef0c935037ea_i1_5.jpg", "1mp3d_0057_region23/4f39c10703bb432baf1cef0c935037ea_i2_5.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0057_region23", "origin_question": "When did you first discover \"the light brown rectangular wooden table\"? (index of the turn)", "option": [], "answer": "5", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the light brown rectangular wooden table\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "1mp3d_0057_region23/92d5c09481694d5f8792cda91271428f_i1_0.jpg", "1mp3d_0057_region23/92d5c09481694d5f8792cda91271428f_i2_0.jpg", "1mp3d_0057_region23/1d51077088474881bd87b9853f0f558c_i1_1.jpg", "1mp3d_0057_region23/1d51077088474881bd87b9853f0f558c_i2_1.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0057_region23", "origin_question": "Using your orientation at the end of turn 4 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "108.03", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 4 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "1mp3d_0057_region23/1d51077088474881bd87b9853f0f558c_i1_2.jpg", "1mp3d_0057_region23/1d51077088474881bd87b9853f0f558c_i2_2.jpg", "1mp3d_0057_region23/80330cded0b349aaaa6aa036393d393b_i1_1.jpg", "1mp3d_0057_region23/80330cded0b349aaaa6aa036393d393b_i2_1.jpg" ], "turn_id": 7, "scan_id": "1mp3d_0057_region23", "origin_question": "Assuming the direction you are facing at the end of the turn 4 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Left", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 4 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "1mp3d_0057_region23/eff407e362f84686a488383e8134e1b3_i1_0.jpg", "1mp3d_0057_region23/eff407e362f84686a488383e8134e1b3_i2_0.jpg", "1mp3d_0057_region23/eff407e362f84686a488383e8134e1b3_i1_1.jpg", "1mp3d_0057_region23/eff407e362f84686a488383e8134e1b3_i2_1.jpg" ], "turn_id": 8, "scan_id": "1mp3d_0057_region23", "origin_question": "Which two objects are on the same side of you now? \"the new flat-screen television with a standard frame\", \"the black lamp with a beige cylindrical lampshade\", \"the light brown rectangular wooden table\". ", "option": [ "the new flat-screen television with a standard frame, the black lamp with a beige cylindrical lampshade", "the new flat-screen television with a standard frame, the light brown rectangular wooden table", "the black lamp with a beige cylindrical lampshade, the light brown rectangular wooden table" ], "answer": "the black lamp with a beige cylindrical lampshade, the light brown rectangular wooden table", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the new flat-screen television with a standard frame\", \"the black lamp with a beige cylindrical lampshade\", \"the light brown rectangular wooden table\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the new flat-screen television with a standard frame, the black lamp with a beige cylindrical lampshade', 'the new flat-screen television with a standard frame, the light brown rectangular wooden table', 'the black lamp with a beige cylindrical lampshade, the light brown rectangular wooden table'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0565_00/00000.jpg", "scene0565_00/00040.jpg", "scene0565_00/00080.jpg", "scene0565_00/00120.jpg" ], "turn_id": 1, "scan_id": "scene0565_00", "origin_question": "Remember, have you seen any bin(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any bin(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0565_00/00160.jpg", "scene0565_00/00200.jpg", "scene0565_00/00240.jpg", "scene0565_00/00280.jpg" ], "turn_id": 2, "scan_id": "scene0565_00", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the computer\"; \"the large rectangular white plastic board\"; \"the rectangular piece of smooth white paper\" ", "option": [ "the computer", "the large rectangular white plastic board", "the rectangular piece of smooth white paper" ], "answer": "the large rectangular white plastic board", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the computer\"; \"the large rectangular white plastic board\"; \"the rectangular piece of smooth white paper\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the computer', 'the large rectangular white plastic board', 'the rectangular piece of smooth white paper'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0565_00/00320.jpg", "scene0565_00/00360.jpg", "scene0565_00/00400.jpg", "scene0565_00/00440.jpg" ], "turn_id": 3, "scan_id": "scene0565_00", "origin_question": "In which turn were you farthest from \"the rectangular piece of smooth white paper\"? ", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the rectangular piece of smooth white paper\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0565_00/00480.jpg", "scene0565_00/00520.jpg", "scene0565_00/00560.jpg", "scene0565_00/00600.jpg" ], "turn_id": 4, "scan_id": "scene0565_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0565_00/00640.jpg", "scene0565_00/00680.jpg", "scene0565_00/00720.jpg", "scene0565_00/00760.jpg" ], "turn_id": 5, "scan_id": "scene0565_00", "origin_question": "When did you first discover \"the small, rectangular white wooden cabinet\"? (index of the turn)", "option": [], "answer": "5", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the small, rectangular white wooden cabinet\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0565_00/00800.jpg", "scene0565_00/00840.jpg", "scene0565_00/00880.jpg", "scene0565_00/00920.jpg" ], "turn_id": 6, "scan_id": "scene0565_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0072_02/00000.jpg", "scene0072_02/00040.jpg", "scene0072_02/00080.jpg", "scene0072_02/00120.jpg" ], "turn_id": 1, "scan_id": "scene0072_02", "origin_question": "Remember, how many door knob(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many door knob(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0072_02/00160.jpg", "scene0072_02/00200.jpg", "scene0072_02/00240.jpg", "scene0072_02/00280.jpg" ], "turn_id": 2, "scan_id": "scene0072_02", "origin_question": "Which direction is the cylindrical metal light in soft brown in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the cylindrical metal light in soft brown in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0072_02/00320.jpg", "scene0072_02/00360.jpg", "scene0072_02/00400.jpg", "scene0072_02/00440.jpg" ], "turn_id": 3, "scan_id": "scene0072_02", "origin_question": "Which direction is the single-sized bed with a wooden frame and yellow bedsheet in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the single-sized bed with a wooden frame and yellow bedsheet in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0072_02/00480.jpg", "scene0072_02/00520.jpg", "scene0072_02/00560.jpg", "scene0072_02/00600.jpg" ], "turn_id": 4, "scan_id": "scene0072_02", "origin_question": "Is the printed black and white cotton clothes to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the printed black and white cotton clothes to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0072_02/00640.jpg", "scene0072_02/00680.jpg", "scene0072_02/00720.jpg", "scene0072_02/00760.jpg" ], "turn_id": 5, "scan_id": "scene0072_02", "origin_question": "By the end of the 3 turn, were you getting closer to or farther away from these two objects: X: \"the printed black and white cotton clothes\", Y: \"the cylindrical metal light in soft brown\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to Y and farther away from X", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 3 turn, were you getting closer to or farther away from these two objects: X: \"the printed black and white cotton clothes\", Y: \"the cylindrical metal light in soft brown\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0072_02/00800.jpg", "scene0072_02/00840.jpg", "scene0072_02/00880.jpg", "scene0072_02/00920.jpg" ], "turn_id": 6, "scan_id": "scene0072_02", "origin_question": "Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "142.9", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0072_02/00960.jpg", "scene0072_02/01000.jpg", "scene0072_02/01040.jpg", "scene0072_02/01080.jpg" ], "turn_id": 7, "scan_id": "scene0072_02", "origin_question": "Which two objects are on the same side of you now? \"the rectangular, light-colored shelf\", \"the medium cylindrical woven basket of brown and beige\", \"the bin\". ", "option": [ "the rectangular, light-colored shelf, the medium cylindrical woven basket of brown and beige", "the rectangular, light-colored shelf, the bin", "the medium cylindrical woven basket of brown and beige, the bin" ], "answer": "the medium cylindrical woven basket of brown and beige, the bin", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the rectangular, light-colored shelf\", \"the medium cylindrical woven basket of brown and beige\", \"the bin\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular, light-colored shelf, the medium cylindrical woven basket of brown and beige', 'the rectangular, light-colored shelf, the bin', 'the medium cylindrical woven basket of brown and beige, the bin'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0057_region15/94abc807216d4b71ae3a1a2c6a39bd23_i1_4.jpg", "1mp3d_0057_region15/94abc807216d4b71ae3a1a2c6a39bd23_i2_4.jpg", "1mp3d_0057_region15/39ce98c3166d4e30a22cae30ee7ba0db_i1_1.jpg", "1mp3d_0057_region15/39ce98c3166d4e30a22cae30ee7ba0db_i2_1.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0057_region15", "origin_question": "Remember, have you seen any couch(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any couch(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0057_region15/e21f4a6f659642f6afff4610f15031e7_i1_2.jpg", "1mp3d_0057_region15/e21f4a6f659642f6afff4610f15031e7_i2_2.jpg", "1mp3d_0057_region15/f00970b67a46474cb616cd98cda9dee6_i1_2.jpg", "1mp3d_0057_region15/f00970b67a46474cb616cd98cda9dee6_i2_2.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0057_region15", "origin_question": "Remember, have you seen any washbasin(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any washbasin(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "1mp3d_0057_region15/e21f4a6f659642f6afff4610f15031e7_i1_1.jpg", "1mp3d_0057_region15/e21f4a6f659642f6afff4610f15031e7_i2_1.jpg", "1mp3d_0057_region15/39ce98c3166d4e30a22cae30ee7ba0db_i1_0.jpg", "1mp3d_0057_region15/39ce98c3166d4e30a22cae30ee7ba0db_i2_0.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0057_region15", "origin_question": "In which turn were you closest to \"the warm brown rectangular couch with patterned fabric\"? ", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the warm brown rectangular couch with patterned fabric\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "1mp3d_0057_region15/94abc807216d4b71ae3a1a2c6a39bd23_i1_3.jpg", "1mp3d_0057_region15/94abc807216d4b71ae3a1a2c6a39bd23_i2_3.jpg", "1mp3d_0057_region15/94abc807216d4b71ae3a1a2c6a39bd23_i1_2.jpg", "1mp3d_0057_region15/94abc807216d4b71ae3a1a2c6a39bd23_i2_2.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0057_region15", "origin_question": "In which turn did you see both \"the warm brown rectangular couch with patterned fabric\" and \"the black flat-screen television of moderate size\" simultaneously? ", "option": [], "answer": "3", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the warm brown rectangular couch with patterned fabric\" and \"the black flat-screen television of moderate size\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0057_region15/3fce8d41890d45c0aaf6b38aad66be51_i1_4.jpg", "1mp3d_0057_region15/3fce8d41890d45c0aaf6b38aad66be51_i2_4.jpg", "1mp3d_0057_region15/dd0a228d7f3c4630bdac948b986c2046_i1_2.jpg", "1mp3d_0057_region15/dd0a228d7f3c4630bdac948b986c2046_i2_2.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0057_region15", "origin_question": "Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the black flat-screen television of moderate size\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the black flat-screen television of moderate size\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0057_region15/e5e991d56b1040c697d713ea1e9fdf9e_i1_1.jpg", "1mp3d_0057_region15/e5e991d56b1040c697d713ea1e9fdf9e_i2_1.jpg", "1mp3d_0057_region15/deaca0e4393847c5921593c4aa7392cc_i1_2.jpg", "1mp3d_0057_region15/deaca0e4393847c5921593c4aa7392cc_i2_2.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0057_region15", "origin_question": "Compared to your position at the end of the 5 turn, are you now closer or farther away from \"the black flat-screen television of moderate size\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 5 turn, are you now closer or farther away from \"the black flat-screen television of moderate size\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "1mp3d_0057_region15/a94e4a04e9844299a87fa2776ba442e3_i1_2.jpg", "1mp3d_0057_region15/a94e4a04e9844299a87fa2776ba442e3_i2_2.jpg", "1mp3d_0057_region15/a94e4a04e9844299a87fa2776ba442e3_i1_3.jpg", "1mp3d_0057_region15/a94e4a04e9844299a87fa2776ba442e3_i2_3.jpg" ], "turn_id": 7, "scan_id": "1mp3d_0057_region15", "origin_question": "How far is your current position from where you were at the end of the turn 4? (in meters)", "option": [], "answer": "3.33", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 4? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0057_region15/b2b6f9bc6773421a90f965ba369a4786_i1_3.jpg", "1mp3d_0057_region15/b2b6f9bc6773421a90f965ba369a4786_i2_3.jpg", "1mp3d_0057_region15/293078a983fa45eebfc063ec680637d8_i1_2.jpg", "1mp3d_0057_region15/293078a983fa45eebfc063ec680637d8_i2_2.jpg" ], "turn_id": 8, "scan_id": "1mp3d_0057_region15", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the warm brown rectangular couch with patterned fabric\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the warm brown rectangular couch with patterned fabric\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "1mp3d_0057_region15/13d9c9e942dc41d681f97bf0989b4bc8_i1_3.jpg", "1mp3d_0057_region15/13d9c9e942dc41d681f97bf0989b4bc8_i2_3.jpg", "1mp3d_0057_region15/47094ff79bbe4111a8e2a6be34cac487_i1_3.jpg", "1mp3d_0057_region15/47094ff79bbe4111a8e2a6be34cac487_i2_3.jpg" ], "turn_id": 9, "scan_id": "1mp3d_0057_region15", "origin_question": "Remember, how many globe(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 9 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many globe(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0030_region0/949a6c58781d4e80bff19ca32493efcd_i2_0.jpg", "1mp3d_0030_region0/9ccf363f5cee403b91fd038d8e3a78dc_i1_1.jpg", "1mp3d_0030_region0/9ccf363f5cee403b91fd038d8e3a78dc_i2_1.jpg", "1mp3d_0030_region0/6576c9319a91404ea9653091b86b93bd_i1_0.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0030_region0", "origin_question": "Remember, have you seen any bed(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any bed(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "1mp3d_0030_region0/6576c9319a91404ea9653091b86b93bd_i2_0.jpg", "1mp3d_0030_region0/6576c9319a91404ea9653091b86b93bd_i2_5.jpg", "1mp3d_0030_region0/9ccf363f5cee403b91fd038d8e3a78dc_i1_0.jpg", "1mp3d_0030_region0/9ccf363f5cee403b91fd038d8e3a78dc_i2_0.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0030_region0", "origin_question": "In which turn did you see both \"the bed\" and \"then abstract picture in deep blue and brown tones\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the bed\" and \"then abstract picture in deep blue and brown tones\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "1mp3d_0030_region0/6576c9319a91404ea9653091b86b93bd_i1_3.jpg", "1mp3d_0030_region0/6576c9319a91404ea9653091b86b93bd_i2_3.jpg", "1mp3d_0030_region0/6576c9319a91404ea9653091b86b93bd_i1_2.jpg", "1mp3d_0030_region0/6576c9319a91404ea9653091b86b93bd_i2_2.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0030_region0", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the bed\"; \"the cylindrical brown bin without handles\"; \"the simple, white rectangular door with a round handle\" ", "option": [ "the bed", "the cylindrical brown bin without handles", "the simple, white rectangular door with a round handle" ], "answer": "the simple, white rectangular door with a round handle", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the bed\"; \"the cylindrical brown bin without handles\"; \"the simple, white rectangular door with a round handle\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the bed', 'the cylindrical brown bin without handles', 'the simple, white rectangular door with a round handle'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "1mp3d_0030_region0/9ccf363f5cee403b91fd038d8e3a78dc_i1_2.jpg", "1mp3d_0030_region0/9ccf363f5cee403b91fd038d8e3a78dc_i2_2.jpg", "1mp3d_0030_region0/9ccf363f5cee403b91fd038d8e3a78dc_i1_3.jpg", "1mp3d_0030_region0/9ccf363f5cee403b91fd038d8e3a78dc_i2_3.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0030_region0", "origin_question": "Among these three objects(\"the dresser\",\"the cylindrical brown bin without handles\",\"the simple, white rectangular door with a round handle\" ), which one is the closest to you?", "option": [ "the dresser", "the cylindrical brown bin without handles", "the simple, white rectangular door with a round handle" ], "answer": "the simple, white rectangular door with a round handle", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects(\"the dresser\",\"the cylindrical brown bin without handles\",\"the simple, white rectangular door with a round handle\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the dresser', 'the cylindrical brown bin without handles', 'the simple, white rectangular door with a round handle'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "1mp3d_0030_region0/9ccf363f5cee403b91fd038d8e3a78dc_i1_4.jpg", "1mp3d_0030_region0/9ccf363f5cee403b91fd038d8e3a78dc_i2_4.jpg", "1mp3d_0030_region0/949a6c58781d4e80bff19ca32493efcd_i1_2.jpg", "1mp3d_0030_region0/949a6c58781d4e80bff19ca32493efcd_i2_2.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0030_region0", "origin_question": "Among these three objects(\"the bed\",\"the dresser\",\"the cylindrical brown bin without handles\" ), which one is the farthest from you?", "option": [ "the bed", "the dresser", "the cylindrical brown bin without handles" ], "answer": "the bed", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects(\"the bed\",\"the dresser\",\"the cylindrical brown bin without handles\" ), which one is the farthest from you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the bed', 'the dresser', 'the cylindrical brown bin without handles'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0029_region29/93292ae9057244519e582c2c53abfb3e_i1_1.jpg", "1mp3d_0029_region29/93292ae9057244519e582c2c53abfb3e_i2_1.jpg", "1mp3d_0029_region29/93292ae9057244519e582c2c53abfb3e_i1_0.jpg", "1mp3d_0029_region29/93292ae9057244519e582c2c53abfb3e_i2_0.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0029_region29", "origin_question": "Remember, have you seen any carpet(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any carpet(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "1mp3d_0029_region29/ccb593339b944f0a84caa0a19133b3e5_i1_4.jpg", "1mp3d_0029_region29/ccb593339b944f0a84caa0a19133b3e5_i2_4.jpg", "1mp3d_0029_region29/93292ae9057244519e582c2c53abfb3e_i1_2.jpg", "1mp3d_0029_region29/93292ae9057244519e582c2c53abfb3e_i2_2.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0029_region29", "origin_question": "At the end of which turn were both of \"the small deep blue modern chair with a curved backrest\" and \"the rectangular, light-toned woven carpet\" on your left side?", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the small deep blue modern chair with a curved backrest\" and \"the rectangular, light-toned woven carpet\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "1mp3d_0029_region29/ccb593339b944f0a84caa0a19133b3e5_i1_3.jpg", "1mp3d_0029_region29/ccb593339b944f0a84caa0a19133b3e5_i2_3.jpg", "1mp3d_0029_region29/fe045dba6a244fefbc9950293c087287_i1_2.jpg", "1mp3d_0029_region29/fe045dba6a244fefbc9950293c087287_i2_2.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0029_region29", "origin_question": "Which direction is the small deep blue modern chair with a curved backrest in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the small deep blue modern chair with a curved backrest in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0029_region29/fe045dba6a244fefbc9950293c087287_i1_3.jpg", "1mp3d_0029_region29/fe045dba6a244fefbc9950293c087287_i2_3.jpg", "1mp3d_0029_region29/ccb593339b944f0a84caa0a19133b3e5_i1_5.jpg", "1mp3d_0029_region29/ccb593339b944f0a84caa0a19133b3e5_i2_5.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0029_region29", "origin_question": "", "option": [], "answer": "", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0029_region29/ccb593339b944f0a84caa0a19133b3e5_i1_0.jpg", "1mp3d_0029_region29/ccb593339b944f0a84caa0a19133b3e5_i2_0.jpg", "1mp3d_0029_region29/ccb593339b944f0a84caa0a19133b3e5_i1_1.jpg", "1mp3d_0029_region29/ccb593339b944f0a84caa0a19133b3e5_i2_1.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0029_region29", "origin_question": "", "option": [], "answer": "", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0029_region29/fe045dba6a244fefbc9950293c087287_i1_1.jpg", "1mp3d_0029_region29/fe045dba6a244fefbc9950293c087287_i2_1.jpg", "1mp3d_0029_region29/fe045dba6a244fefbc9950293c087287_i1_0.jpg", "1mp3d_0029_region29/fe045dba6a244fefbc9950293c087287_i2_0.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0029_region29", "origin_question": "Compared to your position at the end of the 5 turn, are you now closer or farther away from \"the rectangular white window frame made of glass\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 5 turn, are you now closer or farther away from \"the rectangular white window frame made of glass\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "1mp3d_0067_region29/5e0cbbfd7dd742c2ae7028e73e881161_i1_1.jpg", "1mp3d_0067_region29/5e0cbbfd7dd742c2ae7028e73e881161_i2_1.jpg", "1mp3d_0067_region29/e4f3d7d539964b0f9a4f1eec24fa137e_i1_1.jpg", "1mp3d_0067_region29/e4f3d7d539964b0f9a4f1eec24fa137e_i2_1.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0067_region29", "origin_question": "Remember, how many remote control(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many remote control(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "1mp3d_0067_region29/5e0cbbfd7dd742c2ae7028e73e881161_i1_5.jpg", "1mp3d_0067_region29/5e0cbbfd7dd742c2ae7028e73e881161_i2_5.jpg", "1mp3d_0067_region29/e4f3d7d539964b0f9a4f1eec24fa137e_i1_2.jpg", "1mp3d_0067_region29/e4f3d7d539964b0f9a4f1eec24fa137e_i2_2.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0067_region29", "origin_question": "At the end of which turn were both of \"the round brown ceramic lamp with a white striped shade\" and \"the rectangular light brown wooden tray with handles\" on your right side?", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the round brown ceramic lamp with a white striped shade\" and \"the rectangular light brown wooden tray with handles\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "1mp3d_0067_region29/cb13e2147d6c4c569608bb8b2428173a_i1_1.jpg", "1mp3d_0067_region29/cb13e2147d6c4c569608bb8b2428173a_i2_1.jpg", "1mp3d_0067_region29/fb32301e479c48bb8aefae8ee568ad09_i1_4.jpg", "1mp3d_0067_region29/fb32301e479c48bb8aefae8ee568ad09_i2_4.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0067_region29", "origin_question": "What will be the first-time appearance order of the following categories: couch, tray, tv?", "option": [ "tray, tv, couch", "tv, tray, couch", "couch, tv, tray", "couch, tray, tv" ], "answer": "couch, tv, tray", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: couch, tray, tv? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['tray, tv, couch', 'tv, tray, couch', 'couch, tv, tray', 'couch, tray, tv'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0067_region29/cb13e2147d6c4c569608bb8b2428173a_i1_2.jpg", "1mp3d_0067_region29/cb13e2147d6c4c569608bb8b2428173a_i2_2.jpg", "1mp3d_0067_region29/fb32301e479c48bb8aefae8ee568ad09_i1_3.jpg", "1mp3d_0067_region29/fb32301e479c48bb8aefae8ee568ad09_i2_3.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0067_region29", "origin_question": "Remember, have you seen any window(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any window(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "1mp3d_0067_region29/fb32301e479c48bb8aefae8ee568ad09_i1_2.jpg", "1mp3d_0067_region29/fb32301e479c48bb8aefae8ee568ad09_i2_2.jpg", "1mp3d_0067_region29/c8e6c588d25d42cd8400a07db98c40fe_i1_3.jpg", "1mp3d_0067_region29/c8e6c588d25d42cd8400a07db98c40fe_i2_3.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0067_region29", "origin_question": "At the end of which turn were both of \"the medium-sized black television with a thin frame\" and \"the rectangular light brown wooden tray with handles\" on your right side?", "option": [], "answer": "5", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the medium-sized black television with a thin frame\" and \"the rectangular light brown wooden tray with handles\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0067_region29/c8e6c588d25d42cd8400a07db98c40fe_i1_4.jpg", "1mp3d_0067_region29/c8e6c588d25d42cd8400a07db98c40fe_i2_4.jpg", "1mp3d_0067_region29/c8e6c588d25d42cd8400a07db98c40fe_i1_5.jpg", "1mp3d_0067_region29/c8e6c588d25d42cd8400a07db98c40fe_i2_5.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0067_region29", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the medium-sized black television with a thin frame\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the medium-sized black television with a thin frame\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0607_00/00000.jpg", "scene0607_00/00040.jpg", "scene0607_00/00080.jpg", "scene0607_00/00120.jpg" ], "turn_id": 1, "scan_id": "scene0607_00", "origin_question": "Remember, have you seen any table(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any table(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0607_00/00160.jpg", "scene0607_00/00200.jpg", "scene0607_00/00240.jpg", "scene0607_00/00280.jpg" ], "turn_id": 2, "scan_id": "scene0607_00", "origin_question": "When did you first discover \"the sleek rectangular grey dispenser\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the sleek rectangular grey dispenser\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0607_00/00320.jpg", "scene0607_00/00360.jpg", "scene0607_00/00400.jpg", "scene0607_00/00440.jpg" ], "turn_id": 3, "scan_id": "scene0607_00", "origin_question": "Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "32.77", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0607_00/00480.jpg", "scene0607_00/00520.jpg", "scene0607_00/00560.jpg", "scene0607_00/00600.jpg" ], "turn_id": 4, "scan_id": "scene0607_00", "origin_question": "Which direction is the white oval-shaped plastic soap dispenser in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the white oval-shaped plastic soap dispenser in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0607_00/00640.jpg", "scene0607_00/00680.jpg", "scene0607_00/00720.jpg", "scene0607_00/00760.jpg" ], "turn_id": 5, "scan_id": "scene0607_00", "origin_question": "At the end of which turn were both of \"the refrigerator\" and \"the white oval-shaped plastic soap dispenser\" on your left side?", "option": [], "answer": "4", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the refrigerator\" and \"the white oval-shaped plastic soap dispenser\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0607_00/00800.jpg", "scene0607_00/00840.jpg", "scene0607_00/00880.jpg", "scene0607_00/00920.jpg" ], "turn_id": 6, "scan_id": "scene0607_00", "origin_question": "Which two objects are on the same side of you now? \"the rectangular wooden table with a polished surface\", \"the sleek rectangular grey dispenser\", \"the white oval-shaped plastic soap dispenser\". ", "option": [ "the rectangular wooden table with a polished surface, the sleek rectangular grey dispenser", "the rectangular wooden table with a polished surface, the white oval-shaped plastic soap dispenser", "the sleek rectangular grey dispenser, the white oval-shaped plastic soap dispenser" ], "answer": "the sleek rectangular grey dispenser, the white oval-shaped plastic soap dispenser", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the rectangular wooden table with a polished surface\", \"the sleek rectangular grey dispenser\", \"the white oval-shaped plastic soap dispenser\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular wooden table with a polished surface, the sleek rectangular grey dispenser', 'the rectangular wooden table with a polished surface, the white oval-shaped plastic soap dispenser', 'the sleek rectangular grey dispenser, the white oval-shaped plastic soap dispenser'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0047_region30/a206c84102884dda82c5cbdd4e471bcd_i1_5.jpg", "1mp3d_0047_region30/a206c84102884dda82c5cbdd4e471bcd_i2_5.jpg", "1mp3d_0047_region30/5c1880dd58d34e68abf88d6551f0f449_i1_1.jpg", "1mp3d_0047_region30/5c1880dd58d34e68abf88d6551f0f449_i2_1.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0047_region30", "origin_question": "Remember, have you seen any chair(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any chair(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0047_region30/0e7a7b8f99a84a719128d3d0982e9b59_i1_0.jpg", "1mp3d_0047_region30/0e7a7b8f99a84a719128d3d0982e9b59_i2_0.jpg", "1mp3d_0047_region30/36a2419151544f299c99cc0f247b2bb2_i1_5.jpg", "1mp3d_0047_region30/36a2419151544f299c99cc0f247b2bb2_i2_5.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0047_region30", "origin_question": "Remember, have you seen any cabinet(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any cabinet(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "1mp3d_0047_region30/0e7a7b8f99a84a719128d3d0982e9b59_i1_5.jpg", "1mp3d_0047_region30/0e7a7b8f99a84a719128d3d0982e9b59_i2_5.jpg", "1mp3d_0047_region30/5c1880dd58d34e68abf88d6551f0f449_i1_0.jpg", "1mp3d_0047_region30/5c1880dd58d34e68abf88d6551f0f449_i2_0.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0047_region30", "origin_question": "Remember, how many shower(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many shower(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "1mp3d_0047_region30/5c1880dd58d34e68abf88d6551f0f449_i1_5.jpg", "1mp3d_0047_region30/5c1880dd58d34e68abf88d6551f0f449_i2_5.jpg", "1mp3d_0047_region30/a206c84102884dda82c5cbdd4e471bcd_i1_1.jpg", "1mp3d_0047_region30/a206c84102884dda82c5cbdd4e471bcd_i2_1.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0047_region30", "origin_question": "Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance forward or backward from that position?", "option": [ "Forward", "Backward" ], "answer": "Forward", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance forward or backward from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Forward', 'Backward'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "1mp3d_0047_region30/a206c84102884dda82c5cbdd4e471bcd_i1_2.jpg", "1mp3d_0047_region30/a206c84102884dda82c5cbdd4e471bcd_i2_2.jpg", "1mp3d_0047_region30/8e2151a34ab345c9af23baa214326de7_i1_3.jpg", "1mp3d_0047_region30/8e2151a34ab345c9af23baa214326de7_i2_3.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0047_region30", "origin_question": "Assuming the direction you are facing at the end of the turn 3 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Right", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 3 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0047_region30/8fe0b78b55964b8dabd6ea8693f87ec8_i1_4.jpg", "1mp3d_0047_region30/8fe0b78b55964b8dabd6ea8693f87ec8_i2_4.jpg", "1mp3d_0047_region30/791d70f94a124a9e8013f173ce51ff08_i1_2.jpg", "1mp3d_0047_region30/791d70f94a124a9e8013f173ce51ff08_i2_2.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0047_region30", "origin_question": "", "option": [], "answer": "", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "1mp3d_0047_region30/98f2fab16dc44b82aef16383127ee674_i1_3.jpg", "1mp3d_0047_region30/98f2fab16dc44b82aef16383127ee674_i2_3.jpg", "1mp3d_0047_region30/791d70f94a124a9e8013f173ce51ff08_i1_1.jpg", "1mp3d_0047_region30/791d70f94a124a9e8013f173ce51ff08_i2_1.jpg" ], "turn_id": 7, "scan_id": "1mp3d_0047_region30", "origin_question": "By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular black flat-screen television\", Y: \"the large modern metal desk with a black tabletop\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular black flat-screen television\", Y: \"the large modern metal desk with a black tabletop\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "1mp3d_0047_region30/8fe0b78b55964b8dabd6ea8693f87ec8_i1_2.jpg", "1mp3d_0047_region30/8fe0b78b55964b8dabd6ea8693f87ec8_i2_2.jpg", "1mp3d_0047_region30/8e2151a34ab345c9af23baa214326de7_i1_1.jpg", "1mp3d_0047_region30/8e2151a34ab345c9af23baa214326de7_i2_1.jpg" ], "turn_id": 8, "scan_id": "1mp3d_0047_region30", "origin_question": "Is the rectangular black flat-screen television to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the rectangular black flat-screen television to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "1mp3d_0047_region30/bbfd4eb3b8d646a1bcb815d63dbf0fe2_i1_3.jpg", "1mp3d_0047_region30/bbfd4eb3b8d646a1bcb815d63dbf0fe2_i2_3.jpg", "1mp3d_0047_region30/af7ed758b8c04875a5ae383252e35332_i1_3.jpg", "1mp3d_0047_region30/af7ed758b8c04875a5ae383252e35332_i2_3.jpg" ], "turn_id": 9, "scan_id": "1mp3d_0047_region30", "origin_question": "Please recall \"the large modern metal desk with a black tabletop\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "7.19", "user_message": "For the 9 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the large modern metal desk with a black tabletop\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0164_03/00000.jpg", "scene0164_03/00040.jpg", "scene0164_03/00080.jpg", "scene0164_03/00120.jpg" ], "turn_id": 1, "scan_id": "scene0164_03", "origin_question": "Remember, have you seen any soap(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any soap(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0164_03/00160.jpg", "scene0164_03/00200.jpg", "scene0164_03/00240.jpg", "scene0164_03/00280.jpg" ], "turn_id": 2, "scan_id": "scene0164_03", "origin_question": "At the end of which turn were both of \"the sleek black rectangular microwave with a digital display\" and \"the modern, compact black metal toaster with a digital display\" on your left side?", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the sleek black rectangular microwave with a digital display\" and \"the modern, compact black metal toaster with a digital display\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0164_03/00320.jpg", "scene0164_03/00360.jpg", "scene0164_03/00400.jpg", "scene0164_03/00440.jpg" ], "turn_id": 3, "scan_id": "scene0164_03", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the small rectangular white plastic dish rack\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the small rectangular white plastic dish rack\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0164_03/00480.jpg", "scene0164_03/00520.jpg", "scene0164_03/00560.jpg", "scene0164_03/00600.jpg" ], "turn_id": 4, "scan_id": "scene0164_03", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "91.18", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0164_03/00640.jpg", "scene0164_03/00680.jpg", "scene0164_03/00720.jpg", "scene0164_03/00760.jpg" ], "turn_id": 5, "scan_id": "scene0164_03", "origin_question": "What will be the first-time appearance order of the following categories: coffee maker, microwave, tissue?", "option": [ "microwave, tissue, coffee maker", "microwave, coffee maker, tissue", "tissue, coffee maker, microwave", "coffee maker, microwave, tissue" ], "answer": "coffee maker, microwave, tissue", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: coffee maker, microwave, tissue? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['microwave, tissue, coffee maker', 'microwave, coffee maker, tissue', 'tissue, coffee maker, microwave', 'coffee maker, microwave, tissue'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0164_03/00800.jpg", "scene0164_03/00840.jpg", "scene0164_03/00880.jpg", "scene0164_03/00920.jpg" ], "turn_id": 6, "scan_id": "scene0164_03", "origin_question": "Which two objects are on the same side of you now? \"the rectangular light-colored stainless steel range hood\", \"the rectangular cotton towel with blue and brown stripes\", \"the small rectangular white plastic dish rack\". ", "option": [ "the rectangular light-colored stainless steel range hood, the rectangular cotton towel with blue and brown stripes", "the rectangular light-colored stainless steel range hood, the small rectangular white plastic dish rack", "the rectangular cotton towel with blue and brown stripes, the small rectangular white plastic dish rack" ], "answer": "the rectangular light-colored stainless steel range hood, the rectangular cotton towel with blue and brown stripes", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the rectangular light-colored stainless steel range hood\", \"the rectangular cotton towel with blue and brown stripes\", \"the small rectangular white plastic dish rack\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular light-colored stainless steel range hood, the rectangular cotton towel with blue and brown stripes', 'the rectangular light-colored stainless steel range hood, the small rectangular white plastic dish rack', 'the rectangular cotton towel with blue and brown stripes, the small rectangular white plastic dish rack'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0164_03/00960.jpg", "scene0164_03/01000.jpg", "scene0164_03/01040.jpg", "scene0164_03/01080.jpg" ], "turn_id": 7, "scan_id": "scene0164_03", "origin_question": "Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "32.58", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0047_00/00000.jpg", "scene0047_00/00040.jpg", "scene0047_00/00080.jpg", "scene0047_00/00120.jpg" ], "turn_id": 1, "scan_id": "scene0047_00", "origin_question": "Remember, have you seen any table(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any table(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0047_00/00160.jpg", "scene0047_00/00200.jpg", "scene0047_00/00240.jpg", "scene0047_00/00280.jpg" ], "turn_id": 2, "scan_id": "scene0047_00", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the compact, rectangular white soap dispenser\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the compact, rectangular white soap dispenser\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0047_00/00320.jpg", "scene0047_00/00360.jpg", "scene0047_00/00400.jpg", "scene0047_00/00440.jpg" ], "turn_id": 3, "scan_id": "scene0047_00", "origin_question": "Is the compact, rectangular white soap dispenser to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the compact, rectangular white soap dispenser to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0047_00/00480.jpg", "scene0047_00/00520.jpg", "scene0047_00/00560.jpg", "scene0047_00/00600.jpg" ], "turn_id": 4, "scan_id": "scene0047_00", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the white double-door refrigerator with a black handle\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the white double-door refrigerator with a black handle\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0047_00/00640.jpg", "scene0047_00/00680.jpg", "scene0047_00/00720.jpg", "scene0047_00/00760.jpg" ], "turn_id": 5, "scan_id": "scene0047_00", "origin_question": "Which two objects are on the same side of you now? \"the grey vertical wall-mounted dispenser with a metal base\", \"the small, modern black and silver coffee maker\", \"the medium rectangular wooden table with a smooth surface\". ", "option": [ "the grey vertical wall-mounted dispenser with a metal base, the small, modern black and silver coffee maker", "the grey vertical wall-mounted dispenser with a metal base, the medium rectangular wooden table with a smooth surface", "the small, modern black and silver coffee maker, the medium rectangular wooden table with a smooth surface" ], "answer": "the grey vertical wall-mounted dispenser with a metal base, the small, modern black and silver coffee maker", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the grey vertical wall-mounted dispenser with a metal base\", \"the small, modern black and silver coffee maker\", \"the medium rectangular wooden table with a smooth surface\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the grey vertical wall-mounted dispenser with a metal base, the small, modern black and silver coffee maker', 'the grey vertical wall-mounted dispenser with a metal base, the medium rectangular wooden table with a smooth surface', 'the small, modern black and silver coffee maker, the medium rectangular wooden table with a smooth surface'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0047_00/00800.jpg", "scene0047_00/00840.jpg", "scene0047_00/00880.jpg", "scene0047_00/00920.jpg" ], "turn_id": 6, "scan_id": "scene0047_00", "origin_question": "Which two objects are on the same side of you now? \"the white double-door refrigerator with a black handle\", \"the cylindrical dark blue bin made of sturdy material\", \"the bottle\". ", "option": [ "the white double-door refrigerator with a black handle, the cylindrical dark blue bin made of sturdy material", "the white double-door refrigerator with a black handle, the bottle", "the cylindrical dark blue bin made of sturdy material, the bottle" ], "answer": "the cylindrical dark blue bin made of sturdy material, the bottle", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the white double-door refrigerator with a black handle\", \"the cylindrical dark blue bin made of sturdy material\", \"the bottle\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the white double-door refrigerator with a black handle, the cylindrical dark blue bin made of sturdy material', 'the white double-door refrigerator with a black handle, the bottle', 'the cylindrical dark blue bin made of sturdy material, the bottle'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0047_00/00960.jpg", "scene0047_00/01000.jpg", "scene0047_00/01040.jpg", "scene0047_00/01080.jpg" ], "turn_id": 7, "scan_id": "scene0047_00", "origin_question": "How far is your current position from where you were at the end of the turn 3? (in meters)", "option": [], "answer": "2.3", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 3? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0047_00/01120.jpg", "scene0047_00/01160.jpg", "scene0047_00/01200.jpg", "scene0047_00/01240.jpg" ], "turn_id": 8, "scan_id": "scene0047_00", "origin_question": "In which turn did you see both \"the grey vertical wall-mounted dispenser with a metal base\" and \"the bottle\" simultaneously? ", "option": [], "answer": "4", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the grey vertical wall-mounted dispenser with a metal base\" and \"the bottle\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0040_region59/25db7ff144954535ba3aa8d5b9ac54ed_i1_3.jpg", "1mp3d_0040_region59/25db7ff144954535ba3aa8d5b9ac54ed_i2_3.jpg", "1mp3d_0040_region59/c331f23cf14b4de697fb5a3542c6b28c_i1_3.jpg", "1mp3d_0040_region59/c331f23cf14b4de697fb5a3542c6b28c_i2_3.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0040_region59", "origin_question": "Remember, have you seen any candlestick(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any candlestick(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0040_region59/e211c7a914cc4880a869fd9a4de997e1_i1_5.jpg", "1mp3d_0040_region59/e211c7a914cc4880a869fd9a4de997e1_i2_5.jpg", "1mp3d_0040_region59/a8f4d86fe29043e3b6eee87bd8612432_i1_5.jpg", "1mp3d_0040_region59/a8f4d86fe29043e3b6eee87bd8612432_i2_5.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0040_region59", "origin_question": "Remember, have you seen any dvd(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any dvd(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "1mp3d_0040_region59/8761de5366b14140bb6c5b3ee12d1f3b_i1_5.jpg", "1mp3d_0040_region59/8761de5366b14140bb6c5b3ee12d1f3b_i1_0.jpg", "1mp3d_0040_region59/a8f4d86fe29043e3b6eee87bd8612432_i1_0.jpg", "1mp3d_0040_region59/a8f4d86fe29043e3b6eee87bd8612432_i2_0.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0040_region59", "origin_question": "At the end of which turn were both of \"the tall arrangement of red roses and white lilies\" and \"the small dark brown metal rail with intricate patterns\" on your right side?", "option": [], "answer": "2", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the tall arrangement of red roses and white lilies\" and \"the small dark brown metal rail with intricate patterns\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "1mp3d_0040_region59/e211c7a914cc4880a869fd9a4de997e1_i1_1.jpg", "1mp3d_0040_region59/e211c7a914cc4880a869fd9a4de997e1_i2_1.jpg", "1mp3d_0040_region59/c331f23cf14b4de697fb5a3542c6b28c_i1_1.jpg", "1mp3d_0040_region59/c331f23cf14b4de697fb5a3542c6b28c_i2_1.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0040_region59", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "1.76", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "1mp3d_0040_region59/25db7ff144954535ba3aa8d5b9ac54ed_i1_2.jpg", "1mp3d_0040_region59/25db7ff144954535ba3aa8d5b9ac54ed_i2_2.jpg", "1mp3d_0040_region59/25db7ff144954535ba3aa8d5b9ac54ed_i1_1.jpg", "1mp3d_0040_region59/25db7ff144954535ba3aa8d5b9ac54ed_i2_1.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0040_region59", "origin_question": "By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the framed rectangular picture of a bridge\", Y: \"the rectangular, traditional couch with dark patterned fabric\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the framed rectangular picture of a bridge\", Y: \"the rectangular, traditional couch with dark patterned fabric\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0568_01/00000.jpg", "scene0568_01/00040.jpg", "scene0568_01/00080.jpg", "scene0568_01/00120.jpg" ], "turn_id": 1, "scan_id": "scene0568_01", "origin_question": "Remember, have you seen any table(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any table(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0568_01/00160.jpg", "scene0568_01/00200.jpg", "scene0568_01/00240.jpg", "scene0568_01/00280.jpg" ], "turn_id": 2, "scan_id": "scene0568_01", "origin_question": "Please recall \"the small rectangular white plastic tissue box\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.86", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the small rectangular white plastic tissue box\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0568_01/00320.jpg", "scene0568_01/00360.jpg", "scene0568_01/00400.jpg", "scene0568_01/00440.jpg" ], "turn_id": 3, "scan_id": "scene0568_01", "origin_question": "How far is your current position from where you were at the end of the turn 1? (in meters)", "option": [], "answer": "2.44", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 1? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0568_01/00480.jpg", "scene0568_01/00520.jpg", "scene0568_01/00560.jpg", "scene0568_01/00600.jpg" ], "turn_id": 4, "scan_id": "scene0568_01", "origin_question": "Please recall \"the rectangular glass table with a wooden frame\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "3.26", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the rectangular glass table with a wooden frame\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0568_01/00640.jpg", "scene0568_01/00680.jpg", "scene0568_01/00720.jpg", "scene0568_01/00760.jpg" ], "turn_id": 5, "scan_id": "scene0568_01", "origin_question": "Is the rectangular glass table with a wooden frame to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the rectangular glass table with a wooden frame to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0568_01/00800.jpg", "scene0568_01/00840.jpg", "scene0568_01/00880.jpg", "scene0568_01/00920.jpg" ], "turn_id": 6, "scan_id": "scene0568_01", "origin_question": "Which direction is the small rectangular white plastic tissue box in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the small rectangular white plastic tissue box in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0568_01/00960.jpg", "scene0568_01/01000.jpg", "scene0568_01/01040.jpg", "scene0568_01/01080.jpg" ], "turn_id": 7, "scan_id": "scene0568_01", "origin_question": "Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0568_01/01120.jpg", "scene0568_01/01160.jpg", "scene0568_01/01200.jpg", "scene0568_01/01240.jpg" ], "turn_id": 8, "scan_id": "scene0568_01", "origin_question": "When did you first discover \"the small rectangular white plastic tissue box\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the small rectangular white plastic tissue box\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0498_01/00000.jpg", "scene0498_01/00030.jpg", "scene0498_01/00060.jpg", "scene0498_01/00090.jpg" ], "turn_id": 1, "scan_id": "scene0498_01", "origin_question": "Remember, have you seen any radiator(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any radiator(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0498_01/00120.jpg", "scene0498_01/00150.jpg", "scene0498_01/00180.jpg", "scene0498_01/00210.jpg" ], "turn_id": 2, "scan_id": "scene0498_01", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the windowsill\"; \"the stainless steel double-door refrigerator with a minimalistic design\"; \"the rectangular gray metal radiator in good condition\" ", "option": [ "the windowsill", "the stainless steel double-door refrigerator with a minimalistic design", "the rectangular gray metal radiator in good condition" ], "answer": "the stainless steel double-door refrigerator with a minimalistic design", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the windowsill\"; \"the stainless steel double-door refrigerator with a minimalistic design\"; \"the rectangular gray metal radiator in good condition\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the windowsill', 'the stainless steel double-door refrigerator with a minimalistic design', 'the rectangular gray metal radiator in good condition'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0498_01/00240.jpg", "scene0498_01/00270.jpg", "scene0498_01/00300.jpg", "scene0498_01/00330.jpg" ], "turn_id": 3, "scan_id": "scene0498_01", "origin_question": "Which direction is the dispenser in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the dispenser in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0498_01/00360.jpg", "scene0498_01/00390.jpg", "scene0498_01/00420.jpg", "scene0498_01/00450.jpg" ], "turn_id": 4, "scan_id": "scene0498_01", "origin_question": "What will be the first-time appearance order of the following categories: cup, door, radiator?", "option": [ "door, cup, radiator", "radiator, door, cup", "door, radiator, cup", "radiator, cup, door" ], "answer": "radiator, cup, door", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: cup, door, radiator? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['door, cup, radiator', 'radiator, door, cup', 'door, radiator, cup', 'radiator, cup, door'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0498_01/00480.jpg", "scene0498_01/00510.jpg", "scene0498_01/00540.jpg", "scene0498_01/00570.jpg" ], "turn_id": 5, "scan_id": "scene0498_01", "origin_question": "Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "82.15", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0131_02/00000.jpg", "scene0131_02/00030.jpg", "scene0131_02/00060.jpg", "scene0131_02/00090.jpg" ], "turn_id": 1, "scan_id": "scene0131_02", "origin_question": "Remember, how many chair(s) have you seen so far? ", "option": [], "answer": "3", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many chair(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0131_02/00120.jpg", "scene0131_02/00150.jpg", "scene0131_02/00180.jpg", "scene0131_02/00210.jpg" ], "turn_id": 2, "scan_id": "scene0131_02", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular wooden shelf with a clean design\", Y: \"the simple white cylindrical ceramic cup\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular wooden shelf with a clean design\", Y: \"the simple white cylindrical ceramic cup\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0131_02/00240.jpg", "scene0131_02/00270.jpg", "scene0131_02/00300.jpg", "scene0131_02/00330.jpg" ], "turn_id": 3, "scan_id": "scene0131_02", "origin_question": "", "option": [], "answer": "", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0131_02/00360.jpg", "scene0131_02/00390.jpg", "scene0131_02/00420.jpg", "scene0131_02/00450.jpg" ], "turn_id": 4, "scan_id": "scene0131_02", "origin_question": "Is the simple white cylindrical ceramic cup to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the simple white cylindrical ceramic cup to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0131_02/00480.jpg", "scene0131_02/00510.jpg", "scene0131_02/00540.jpg", "scene0131_02/00570.jpg" ], "turn_id": 5, "scan_id": "scene0131_02", "origin_question": "At the end of which turn were both of \"the rectangular wooden office desk with drawers\" and \"the bright orange modern couch with clean lines\" on your left side?", "option": [], "answer": "3", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the rectangular wooden office desk with drawers\" and \"the bright orange modern couch with clean lines\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0131_02/00600.jpg", "scene0131_02/00630.jpg", "scene0131_02/00660.jpg", "scene0131_02/00690.jpg" ], "turn_id": 6, "scan_id": "scene0131_02", "origin_question": "Which two objects are on the same side of you now? \"the rectangular wooden shelf with a clean design\", \"the simple white cylindrical ceramic cup\", \"the cylindrical black plastic bin\". ", "option": [ "the rectangular wooden shelf with a clean design, the simple white cylindrical ceramic cup", "the rectangular wooden shelf with a clean design, the cylindrical black plastic bin", "the simple white cylindrical ceramic cup, the cylindrical black plastic bin" ], "answer": "the rectangular wooden shelf with a clean design, the simple white cylindrical ceramic cup", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the rectangular wooden shelf with a clean design\", \"the simple white cylindrical ceramic cup\", \"the cylindrical black plastic bin\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular wooden shelf with a clean design, the simple white cylindrical ceramic cup', 'the rectangular wooden shelf with a clean design, the cylindrical black plastic bin', 'the simple white cylindrical ceramic cup, the cylindrical black plastic bin'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0663_01/00000.jpg", "scene0663_01/00040.jpg", "scene0663_01/00080.jpg", "scene0663_01/00120.jpg", "scene0663_01/00160.jpg" ], "turn_id": 1, "scan_id": "scene0663_01", "origin_question": "Remember, have you seen any bin(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any bin(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0663_01/00200.jpg", "scene0663_01/00240.jpg", "scene0663_01/00280.jpg", "scene0663_01/00320.jpg", "scene0663_01/00360.jpg" ], "turn_id": 2, "scan_id": "scene0663_01", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the rectangular black monitor with a metal stand\"; \"the gray rectangular metal and plastic shelf with layers\"; \"the modern black telephone with a rectangular shape\" ", "option": [ "the rectangular black monitor with a metal stand", "the gray rectangular metal and plastic shelf with layers", "the modern black telephone with a rectangular shape" ], "answer": "the gray rectangular metal and plastic shelf with layers", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the rectangular black monitor with a metal stand\"; \"the gray rectangular metal and plastic shelf with layers\"; \"the modern black telephone with a rectangular shape\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular black monitor with a metal stand', 'the gray rectangular metal and plastic shelf with layers', 'the modern black telephone with a rectangular shape'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0663_01/00400.jpg", "scene0663_01/00440.jpg", "scene0663_01/00480.jpg", "scene0663_01/00520.jpg", "scene0663_01/00560.jpg" ], "turn_id": 3, "scan_id": "scene0663_01", "origin_question": "In which turn did you see both \"the blue cylindrical adhesive tape made of paper or plastic\" and \"the black rectangular desktop computer with an LED light\" simultaneously? ", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the blue cylindrical adhesive tape made of paper or plastic\" and \"the black rectangular desktop computer with an LED light\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0663_01/00600.jpg", "scene0663_01/00640.jpg", "scene0663_01/00680.jpg", "scene0663_01/00720.jpg", "scene0663_01/00760.jpg" ], "turn_id": 4, "scan_id": "scene0663_01", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the blue cylindrical ceramic cup\" from your position?", "option": [], "answer": "98.24", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the blue cylindrical ceramic cup\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0663_01/00800.jpg", "scene0663_01/00840.jpg", "scene0663_01/00880.jpg", "scene0663_01/00920.jpg", "scene0663_01/00960.jpg" ], "turn_id": 5, "scan_id": "scene0663_01", "origin_question": "What will be the first-time appearance order of the following categories: backpack, chair, shelf?", "option": [ "chair, backpack, shelf", "backpack, chair, shelf", "shelf, backpack, chair", "chair, shelf, backpack" ], "answer": "chair, shelf, backpack", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: backpack, chair, shelf? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['chair, backpack, shelf', 'backpack, chair, shelf', 'shelf, backpack, chair', 'chair, shelf, backpack'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0663_01/01000.jpg", "scene0663_01/01040.jpg", "scene0663_01/01080.jpg", "scene0663_01/01120.jpg", "scene0663_01/01160.jpg" ], "turn_id": 6, "scan_id": "scene0663_01", "origin_question": "At the end of which turn were both of \"the rectangular black monitor with a metal stand\" and \"the dark nylon clothes with a front zipper\" on your right side?", "option": [], "answer": "1", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the rectangular black monitor with a metal stand\" and \"the dark nylon clothes with a front zipper\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0663_01/01200.jpg", "scene0663_01/01240.jpg", "scene0663_01/01280.jpg", "scene0663_01/01320.jpg", "scene0663_01/01360.jpg" ], "turn_id": 7, "scan_id": "scene0663_01", "origin_question": "Remember, have you seen any shelf(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any shelf(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0663_01/01400.jpg", "scene0663_01/01440.jpg", "scene0663_01/01480.jpg", "scene0663_01/01520.jpg", "scene0663_01/01560.jpg" ], "turn_id": 8, "scan_id": "scene0663_01", "origin_question": "Which two objects are on the same side of you now? \"the blue cylindrical adhesive tape made of paper or plastic\", \"the medium-sized rectangular black plastic bin\", \"the dark nylon clothes with a front zipper\". ", "option": [ "the blue cylindrical adhesive tape made of paper or plastic, the medium-sized rectangular black plastic bin", "the blue cylindrical adhesive tape made of paper or plastic, the dark nylon clothes with a front zipper", "the medium-sized rectangular black plastic bin, the dark nylon clothes with a front zipper" ], "answer": "the blue cylindrical adhesive tape made of paper or plastic, the medium-sized rectangular black plastic bin", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the blue cylindrical adhesive tape made of paper or plastic\", \"the medium-sized rectangular black plastic bin\", \"the dark nylon clothes with a front zipper\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the blue cylindrical adhesive tape made of paper or plastic, the medium-sized rectangular black plastic bin', 'the blue cylindrical adhesive tape made of paper or plastic, the dark nylon clothes with a front zipper', 'the medium-sized rectangular black plastic bin, the dark nylon clothes with a front zipper'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0663_01/01600.jpg", "scene0663_01/01640.jpg", "scene0663_01/01680.jpg", "scene0663_01/01720.jpg", "scene0663_01/01760.jpg" ], "turn_id": 9, "scan_id": "scene0663_01", "origin_question": "Is the blue cylindrical adhesive tape made of paper or plastic to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the blue cylindrical adhesive tape made of paper or plastic to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0663_01/01800.jpg", "scene0663_01/01840.jpg", "scene0663_01/01880.jpg", "scene0663_01/01920.jpg", "scene0663_01/01960.jpg" ], "turn_id": 10, "scan_id": "scene0663_01", "origin_question": "How far is your current position from where you were at the end of the turn 3? (in meters)", "option": [], "answer": "1.89", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 3? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0274_02/00000.jpg", "scene0274_02/00040.jpg", "scene0274_02/00080.jpg", "scene0274_02/00120.jpg", "scene0274_02/00160.jpg" ], "turn_id": 1, "scan_id": "scene0274_02", "origin_question": "Remember, have you seen any shoe(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any shoe(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0274_02/00200.jpg", "scene0274_02/00240.jpg", "scene0274_02/00280.jpg", "scene0274_02/00320.jpg", "scene0274_02/00360.jpg" ], "turn_id": 2, "scan_id": "scene0274_02", "origin_question": "When did you first discover \"the white corded landline telephone made of plastic\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the white corded landline telephone made of plastic\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0274_02/00400.jpg", "scene0274_02/00440.jpg", "scene0274_02/00480.jpg", "scene0274_02/00520.jpg", "scene0274_02/00560.jpg" ], "turn_id": 3, "scan_id": "scene0274_02", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0274_02/00600.jpg", "scene0274_02/00640.jpg", "scene0274_02/00680.jpg", "scene0274_02/00720.jpg", "scene0274_02/00760.jpg" ], "turn_id": 4, "scan_id": "scene0274_02", "origin_question": "Which direction is the green plastic standard-sized bin in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the green plastic standard-sized bin in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0274_02/00800.jpg", "scene0274_02/00840.jpg", "scene0274_02/00880.jpg", "scene0274_02/00920.jpg", "scene0274_02/00960.jpg" ], "turn_id": 5, "scan_id": "scene0274_02", "origin_question": "Please recall \"the large rectangular orange cardboard box\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.46", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the large rectangular orange cardboard box\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0274_02/01000.jpg", "scene0274_02/01040.jpg", "scene0274_02/01080.jpg", "scene0274_02/01120.jpg", "scene0274_02/01160.jpg" ], "turn_id": 6, "scan_id": "scene0274_02", "origin_question": "Which direction is the green plastic standard-sized bin in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the green plastic standard-sized bin in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0274_02/01200.jpg", "scene0274_02/01240.jpg", "scene0274_02/01280.jpg", "scene0274_02/01320.jpg", "scene0274_02/01360.jpg" ], "turn_id": 7, "scan_id": "scene0274_02", "origin_question": "In which turn were you closest to \"the green plastic standard-sized bin\"? ", "option": [], "answer": "3", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the green plastic standard-sized bin\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0274_02/01400.jpg", "scene0274_02/01440.jpg", "scene0274_02/01480.jpg", "scene0274_02/01520.jpg", "scene0274_02/01560.jpg" ], "turn_id": 8, "scan_id": "scene0274_02", "origin_question": "What will be the first-time appearance order of the following categories: eraser, light, table?", "option": [ "light, eraser, table", "eraser, light, table", "eraser, table, light", "table, eraser, light" ], "answer": "table, eraser, light", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: eraser, light, table? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['light, eraser, table', 'eraser, light, table', 'eraser, table, light', 'table, eraser, light'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0274_02/01610.jpg", "scene0274_02/01650.jpg", "scene0274_02/01690.jpg", "scene0274_02/01730.jpg", "scene0274_02/01770.jpg" ], "turn_id": 9, "scan_id": "scene0274_02", "origin_question": "Which direction is the small rectangular dark gray eraser made of rubber in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the small rectangular dark gray eraser made of rubber in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0274_02/01810.jpg", "scene0274_02/01850.jpg", "scene0274_02/01890.jpg", "scene0274_02/01930.jpg", "scene0274_02/01970.jpg" ], "turn_id": 10, "scan_id": "scene0274_02", "origin_question": "What will be the first-time appearance order of the following categories: door, eraser, light?", "option": [ "door, light, eraser", "light, eraser, door", "door, eraser, light", "light, door, eraser" ], "answer": "door, eraser, light", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: door, eraser, light? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['door, light, eraser', 'light, eraser, door', 'door, eraser, light', 'light, door, eraser'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0675_01/00000.jpg", "scene0675_01/00030.jpg", "scene0675_01/00060.jpg", "scene0675_01/00090.jpg" ], "turn_id": 1, "scan_id": "scene0675_01", "origin_question": "Remember, have you seen any bin(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any bin(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0675_01/00120.jpg", "scene0675_01/00150.jpg", "scene0675_01/00180.jpg", "scene0675_01/00210.jpg" ], "turn_id": 2, "scan_id": "scene0675_01", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the light brown rectangular wooden desk with a drawer\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the light brown rectangular wooden desk with a drawer\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0675_01/00240.jpg", "scene0675_01/00270.jpg", "scene0675_01/00300.jpg", "scene0675_01/00330.jpg" ], "turn_id": 3, "scan_id": "scene0675_01", "origin_question": "Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "109.57", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0675_01/00360.jpg", "scene0675_01/00390.jpg", "scene0675_01/00420.jpg", "scene0675_01/00450.jpg" ], "turn_id": 4, "scan_id": "scene0675_01", "origin_question": "In which turn did you see both \"the light brown rectangular wooden desk with a drawer\" and \"the light brown rectangular wooden dresser with three drawers\" simultaneously? ", "option": [], "answer": "3", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the light brown rectangular wooden desk with a drawer\" and \"the light brown rectangular wooden dresser with three drawers\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0675_01/00480.jpg", "scene0675_01/00510.jpg", "scene0675_01/00540.jpg", "scene0675_01/00570.jpg" ], "turn_id": 5, "scan_id": "scene0675_01", "origin_question": "Is the door to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the door to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0675_01/00600.jpg", "scene0675_01/00630.jpg", "scene0675_01/00660.jpg", "scene0675_01/00690.jpg" ], "turn_id": 6, "scan_id": "scene0675_01", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular window with a white frame and blinds\", Y: \"the light brown rectangular wooden desk with a drawer\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular window with a white frame and blinds\", Y: \"the light brown rectangular wooden desk with a drawer\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0040_region24/fd7bc314137242a498ccec4264834f8a_i1_0.jpg", "1mp3d_0040_region24/fd7bc314137242a498ccec4264834f8a_i2_0.jpg", "1mp3d_0040_region24/a1d210d6a26b45528017394d195a7019_i1_5.jpg", "1mp3d_0040_region24/a1d210d6a26b45528017394d195a7019_i2_5.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0040_region24", "origin_question": "Remember, have you seen any light(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any light(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0040_region24/bfe100fb6ef24661831ea5d65e4e62ce_i1_5.jpg", "1mp3d_0040_region24/bfe100fb6ef24661831ea5d65e4e62ce_i2_5.jpg", "1mp3d_0040_region24/bc9804993e6c4e4a8d2b3c9569276946_i1_5.jpg", "1mp3d_0040_region24/bc9804993e6c4e4a8d2b3c9569276946_i2_5.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0040_region24", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "1mp3d_0040_region24/bc9804993e6c4e4a8d2b3c9569276946_i1_0.jpg", "1mp3d_0040_region24/bc9804993e6c4e4a8d2b3c9569276946_i2_0.jpg", "1mp3d_0040_region24/e93d23ba6ebd41469ac71306bc57fb71_i1_0.jpg", "1mp3d_0040_region24/e93d23ba6ebd41469ac71306bc57fb71_i2_0.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0040_region24", "origin_question": "What will be the first-time appearance order of the following categories: car, garage door, light?", "option": [ "light, car, garage door", "garage door, light, car", "garage door, car, light", "car, garage door, light" ], "answer": "light, car, garage door", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: car, garage door, light? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['light, car, garage door', 'garage door, light, car', 'garage door, car, light', 'car, garage door, light'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "1mp3d_0040_region24/e0e69f99ad6347b2aaf144db43f53ee5_i1_2.jpg", "1mp3d_0040_region24/e0e69f99ad6347b2aaf144db43f53ee5_i2_2.jpg", "1mp3d_0040_region24/7cb43b9d9c6e4f04bfd9e52ef67d9201_i1_0.jpg", "1mp3d_0040_region24/7cb43b9d9c6e4f04bfd9e52ef67d9201_i2_0.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0040_region24", "origin_question": "Please recall \"the large white vertical garage door with metal bars\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "7.82", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the large white vertical garage door with metal bars\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "1mp3d_0040_region24/7cb43b9d9c6e4f04bfd9e52ef67d9201_i1_5.jpg", "1mp3d_0040_region24/7cb43b9d9c6e4f04bfd9e52ef67d9201_i2_5.jpg", "1mp3d_0040_region24/290c17506be44249ba792f9c7ccd42e0_i1_5.jpg", "1mp3d_0040_region24/290c17506be44249ba792f9c7ccd42e0_i2_5.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0040_region24", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the large white vertical garage door with metal bars\" from your position?", "option": [], "answer": "93.61", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the large white vertical garage door with metal bars\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "1mp3d_0040_region24/77e04f8332b34768a3208bdceb7eb866_i1_3.jpg", "1mp3d_0040_region24/77e04f8332b34768a3208bdceb7eb866_i2_3.jpg", "1mp3d_0040_region24/65e411cc696a40f49d89a3c278889c3c_i1_3.jpg", "1mp3d_0040_region24/65e411cc696a40f49d89a3c278889c3c_i2_3.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0040_region24", "origin_question": "Please recall \"the sleek black metal car with a smooth body\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.76", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the sleek black metal car with a smooth body\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "1mp3d_0040_region24/65e411cc696a40f49d89a3c278889c3c_i1_4.jpg", "1mp3d_0040_region24/65e411cc696a40f49d89a3c278889c3c_i2_4.jpg", "1mp3d_0040_region24/65e411cc696a40f49d89a3c278889c3c_i1_5.jpg", "1mp3d_0040_region24/65e411cc696a40f49d89a3c278889c3c_i2_5.jpg" ], "turn_id": 7, "scan_id": "1mp3d_0040_region24", "origin_question": "Please recall \"the cylindrical, pale yellow metal pillar with black spots\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "6.06", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the cylindrical, pale yellow metal pillar with black spots\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0040_region24/77e04f8332b34768a3208bdceb7eb866_i1_5.jpg", "1mp3d_0040_region24/77e04f8332b34768a3208bdceb7eb866_i2_5.jpg", "1mp3d_0040_region24/0c8b82d8a76b4adeb6b4be225a38aaf4_i1_3.jpg", "1mp3d_0040_region24/0c8b82d8a76b4adeb6b4be225a38aaf4_i2_3.jpg" ], "turn_id": 8, "scan_id": "1mp3d_0040_region24", "origin_question": "Compared to your position at the end of the 5 turn, are you now closer or farther away from \"the sleek black metal car with a smooth body\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 5 turn, are you now closer or farther away from \"the sleek black metal car with a smooth body\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0542_00/00000.jpg", "scene0542_00/00040.jpg", "scene0542_00/00080.jpg", "scene0542_00/00120.jpg", "scene0542_00/00160.jpg" ], "turn_id": 1, "scan_id": "scene0542_00", "origin_question": "Remember, how many counter(s) have you seen so far? ", "option": [], "answer": "2", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many counter(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0542_00/00200.jpg", "scene0542_00/00240.jpg", "scene0542_00/00280.jpg", "scene0542_00/00320.jpg", "scene0542_00/00360.jpg" ], "turn_id": 2, "scan_id": "scene0542_00", "origin_question": "Which direction is the modern stainless steel refrigerator with a sleek exterior in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the modern stainless steel refrigerator with a sleek exterior in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0542_00/00400.jpg", "scene0542_00/00440.jpg", "scene0542_00/00480.jpg", "scene0542_00/00520.jpg", "scene0542_00/00560.jpg" ], "turn_id": 3, "scan_id": "scene0542_00", "origin_question": "Is the medium-sized light brown wooden table to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the medium-sized light brown wooden table to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0542_00/00600.jpg", "scene0542_00/00640.jpg", "scene0542_00/00680.jpg", "scene0542_00/00720.jpg", "scene0542_00/00760.jpg" ], "turn_id": 4, "scan_id": "scene0542_00", "origin_question": "Which direction is the modern stainless steel refrigerator with a sleek exterior in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the modern stainless steel refrigerator with a sleek exterior in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0542_00/00800.jpg", "scene0542_00/00840.jpg", "scene0542_00/00880.jpg", "scene0542_00/00920.jpg", "scene0542_00/00960.jpg" ], "turn_id": 5, "scan_id": "scene0542_00", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the sleek stainless steel triangular range hood\" from your position?", "option": [], "answer": "43.66", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the sleek stainless steel triangular range hood\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0542_00/01000.jpg", "scene0542_00/01040.jpg", "scene0542_00/01080.jpg", "scene0542_00/01120.jpg", "scene0542_00/01160.jpg" ], "turn_id": 6, "scan_id": "scene0542_00", "origin_question": "Which direction is the rectangular wooden shelf with a natural brown color in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the rectangular wooden shelf with a natural brown color in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0542_00/01200.jpg", "scene0542_00/01240.jpg", "scene0542_00/01280.jpg", "scene0542_00/01320.jpg", "scene0542_00/01360.jpg" ], "turn_id": 7, "scan_id": "scene0542_00", "origin_question": "How far is your current position from where you were at the end of the turn 6? (in meters)", "option": [], "answer": "1.33", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 6? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0542_00/01400.jpg", "scene0542_00/01440.jpg", "scene0542_00/01480.jpg", "scene0542_00/01520.jpg", "scene0542_00/01560.jpg" ], "turn_id": 8, "scan_id": "scene0542_00", "origin_question": "Which two objects are on the same side of you now? \"the rectangular wooden shelf with a natural brown color\", \"the circular wall clock with a black wooden frame\", \"the medium-sized black freestanding oven with a glass door\". ", "option": [ "the rectangular wooden shelf with a natural brown color, the circular wall clock with a black wooden frame", "the rectangular wooden shelf with a natural brown color, the medium-sized black freestanding oven with a glass door", "the circular wall clock with a black wooden frame, the medium-sized black freestanding oven with a glass door" ], "answer": "the rectangular wooden shelf with a natural brown color, the circular wall clock with a black wooden frame", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the rectangular wooden shelf with a natural brown color\", \"the circular wall clock with a black wooden frame\", \"the medium-sized black freestanding oven with a glass door\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular wooden shelf with a natural brown color, the circular wall clock with a black wooden frame', 'the rectangular wooden shelf with a natural brown color, the medium-sized black freestanding oven with a glass door', 'the circular wall clock with a black wooden frame, the medium-sized black freestanding oven with a glass door'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0542_00/01600.jpg", "scene0542_00/01640.jpg", "scene0542_00/01680.jpg", "scene0542_00/01720.jpg", "scene0542_00/01760.jpg" ], "turn_id": 9, "scan_id": "scene0542_00", "origin_question": "Which two objects are on the same side of you now? \"the modern stainless steel refrigerator with a sleek exterior\", \"the small rectangular window with a white frame\", \"the sleek stainless steel triangular range hood\". ", "option": [ "the modern stainless steel refrigerator with a sleek exterior, the small rectangular window with a white frame", "the modern stainless steel refrigerator with a sleek exterior, the sleek stainless steel triangular range hood", "the small rectangular window with a white frame, the sleek stainless steel triangular range hood" ], "answer": "the modern stainless steel refrigerator with a sleek exterior, the small rectangular window with a white frame", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the modern stainless steel refrigerator with a sleek exterior\", \"the small rectangular window with a white frame\", \"the sleek stainless steel triangular range hood\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the modern stainless steel refrigerator with a sleek exterior, the small rectangular window with a white frame', 'the modern stainless steel refrigerator with a sleek exterior, the sleek stainless steel triangular range hood', 'the small rectangular window with a white frame, the sleek stainless steel triangular range hood'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0542_00/01800.jpg", "scene0542_00/01840.jpg", "scene0542_00/01880.jpg", "scene0542_00/01920.jpg", "scene0542_00/01960.jpg" ], "turn_id": 10, "scan_id": "scene0542_00", "origin_question": "Is the modern stainless steel refrigerator with a sleek exterior to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the modern stainless steel refrigerator with a sleek exterior to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0257_00/00000.jpg", "scene0257_00/00040.jpg", "scene0257_00/00080.jpg", "scene0257_00/00120.jpg" ], "turn_id": 1, "scan_id": "scene0257_00", "origin_question": "Remember, how many toothbrush(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many toothbrush(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0257_00/00160.jpg", "scene0257_00/00200.jpg", "scene0257_00/00240.jpg", "scene0257_00/00280.jpg" ], "turn_id": 2, "scan_id": "scene0257_00", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the bright red cylindrical metal fire extinguisher\"; \"the sleek silver faucet with a green handle\"; \"Medium concrete or stone stairs with a simple design\" ", "option": [ "the bright red cylindrical metal fire extinguisher", "the sleek silver faucet with a green handle", "Medium concrete or stone stairs with a simple design" ], "answer": "Medium concrete or stone stairs with a simple design", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the bright red cylindrical metal fire extinguisher\"; \"the sleek silver faucet with a green handle\"; \"Medium concrete or stone stairs with a simple design\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the bright red cylindrical metal fire extinguisher', 'the sleek silver faucet with a green handle', 'Medium concrete or stone stairs with a simple design'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0257_00/00320.jpg", "scene0257_00/00360.jpg", "scene0257_00/00400.jpg", "scene0257_00/00440.jpg" ], "turn_id": 3, "scan_id": "scene0257_00", "origin_question": "In which turn were you farthest from \"the rectangular white folding table with black legs\"? ", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the rectangular white folding table with black legs\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0257_00/00480.jpg", "scene0257_00/00520.jpg", "scene0257_00/00560.jpg", "scene0257_00/00600.jpg" ], "turn_id": 4, "scan_id": "scene0257_00", "origin_question": "Which direction is the control in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the control in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0257_00/00640.jpg", "scene0257_00/00680.jpg", "scene0257_00/00720.jpg", "scene0257_00/00760.jpg" ], "turn_id": 5, "scan_id": "scene0257_00", "origin_question": "In which turn did you see both \"the bright red cylindrical metal fire extinguisher\" and \"the sturdy, medium-sized brown metal rail\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the bright red cylindrical metal fire extinguisher\" and \"the sturdy, medium-sized brown metal rail\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0257_00/00800.jpg", "scene0257_00/00840.jpg", "scene0257_00/00880.jpg", "scene0257_00/00920.jpg" ], "turn_id": 6, "scan_id": "scene0257_00", "origin_question": "Which direction is Medium concrete or stone stairs with a simple design in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is Medium concrete or stone stairs with a simple design in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0257_00/00960.jpg", "scene0257_00/01000.jpg", "scene0257_00/01040.jpg", "scene0257_00/01080.jpg" ], "turn_id": 7, "scan_id": "scene0257_00", "origin_question": "When did you first discover \"Medium concrete or stone stairs with a simple design\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"Medium concrete or stone stairs with a simple design\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0257_00/01120.jpg", "scene0257_00/01160.jpg", "scene0257_00/01200.jpg", "scene0257_00/01240.jpg" ], "turn_id": 8, "scan_id": "scene0257_00", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the sleek silver faucet with a green handle\" from your position?", "option": [], "answer": "72.38", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the sleek silver faucet with a green handle\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0257_00/01280.jpg", "scene0257_00/01320.jpg", "scene0257_00/01360.jpg", "scene0257_00/01400.jpg" ], "turn_id": 9, "scan_id": "scene0257_00", "origin_question": "Is the control to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 9 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the control to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0078_region3/dea81734bad44c9389a908251f9b70d0_i1_0.jpg", "1mp3d_0078_region3/dea81734bad44c9389a908251f9b70d0_i2_0.jpg", "1mp3d_0078_region3/bf218ea3168f4a7e9ab0385613d2ed19_i1_3.jpg", "1mp3d_0078_region3/bf218ea3168f4a7e9ab0385613d2ed19_i2_3.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0078_region3", "origin_question": "Remember, have you seen any notebook(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any notebook(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 2 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0078_region3/bf218ea3168f4a7e9ab0385613d2ed19_i1_4.jpg", "1mp3d_0078_region3/bf218ea3168f4a7e9ab0385613d2ed19_i2_4.jpg", "1mp3d_0078_region3/dea81734bad44c9389a908251f9b70d0_i1_5.jpg", "1mp3d_0078_region3/dea81734bad44c9389a908251f9b70d0_i2_5.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0078_region3", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 2 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0430_00/00000.jpg", "scene0430_00/00040.jpg", "scene0430_00/00080.jpg", "scene0430_00/00120.jpg", "scene0430_00/00160.jpg" ], "turn_id": 1, "scan_id": "scene0430_00", "origin_question": "Remember, how many cosmetics(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many cosmetics(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0430_00/00200.jpg", "scene0430_00/00240.jpg", "scene0430_00/00280.jpg", "scene0430_00/00320.jpg", "scene0430_00/00360.jpg" ], "turn_id": 2, "scan_id": "scene0430_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0430_00/00400.jpg", "scene0430_00/00440.jpg", "scene0430_00/00480.jpg", "scene0430_00/00520.jpg", "scene0430_00/00560.jpg" ], "turn_id": 3, "scan_id": "scene0430_00", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the medium-sized black wooden door with a golden handle\", Y: \"the retractable white rectangular screen made of durable material\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to Y and farther away from X", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the medium-sized black wooden door with a golden handle\", Y: \"the retractable white rectangular screen made of durable material\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0430_00/00600.jpg", "scene0430_00/00640.jpg", "scene0430_00/00680.jpg", "scene0430_00/00720.jpg", "scene0430_00/00760.jpg" ], "turn_id": 4, "scan_id": "scene0430_00", "origin_question": "Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "72.02", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0430_00/00800.jpg", "scene0430_00/00840.jpg", "scene0430_00/00880.jpg", "scene0430_00/00920.jpg", "scene0430_00/00960.jpg" ], "turn_id": 5, "scan_id": "scene0430_00", "origin_question": "Please recall \"the medium-sized black wooden door with a golden handle\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "5.44", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the medium-sized black wooden door with a golden handle\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0430_00/01000.jpg", "scene0430_00/01040.jpg", "scene0430_00/01080.jpg", "scene0430_00/01120.jpg", "scene0430_00/01160.jpg" ], "turn_id": 6, "scan_id": "scene0430_00", "origin_question": "Is the retractable white rectangular screen made of durable material to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the retractable white rectangular screen made of durable material to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0430_00/01200.jpg", "scene0430_00/01240.jpg", "scene0430_00/01280.jpg", "scene0430_00/01320.jpg", "scene0430_00/01360.jpg" ], "turn_id": 7, "scan_id": "scene0430_00", "origin_question": "Which direction is the retractable white rectangular screen made of durable material in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the retractable white rectangular screen made of durable material in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0430_00/01400.jpg", "scene0430_00/01440.jpg", "scene0430_00/01480.jpg", "scene0430_00/01520.jpg", "scene0430_00/01560.jpg" ], "turn_id": 8, "scan_id": "scene0430_00", "origin_question": "In which turn did you see both \"the medium-sized black wooden door with a golden handle\" and \"the medium red and black nylon backpack\" simultaneously? ", "option": [], "answer": "7", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the medium-sized black wooden door with a golden handle\" and \"the medium red and black nylon backpack\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0430_00/01600.jpg", "scene0430_00/01640.jpg", "scene0430_00/01680.jpg", "scene0430_00/01720.jpg", "scene0430_00/01760.jpg" ], "turn_id": 9, "scan_id": "scene0430_00", "origin_question": "How far is your current position from where you were at the end of the turn 4? (in meters)", "option": [], "answer": "4.97", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 4? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0430_00/01800.jpg", "scene0430_00/01840.jpg", "scene0430_00/01880.jpg", "scene0430_00/01920.jpg", "scene0430_00/01960.jpg" ], "turn_id": 10, "scan_id": "scene0430_00", "origin_question": "Please recall \"the medium-sized black wooden door with a golden handle\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.69", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the medium-sized black wooden door with a golden handle\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0100_02/00000.jpg", "scene0100_02/00030.jpg", "scene0100_02/00060.jpg", "scene0100_02/00090.jpg" ], "turn_id": 1, "scan_id": "scene0100_02", "origin_question": "Remember, have you seen any hair dryer(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any hair dryer(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0100_02/00120.jpg", "scene0100_02/00150.jpg", "scene0100_02/00180.jpg", "scene0100_02/00210.jpg" ], "turn_id": 2, "scan_id": "scene0100_02", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the counter\"; \"the towel\"; \"the small round white plastic bin with a lid\" ", "option": [ "the counter", "the towel", "the small round white plastic bin with a lid" ], "answer": "the counter", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the counter\"; \"the towel\"; \"the small round white plastic bin with a lid\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the counter', 'the towel', 'the small round white plastic bin with a lid'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0100_02/00240.jpg", "scene0100_02/00270.jpg", "scene0100_02/00300.jpg", "scene0100_02/00330.jpg" ], "turn_id": 3, "scan_id": "scene0100_02", "origin_question": "Is the small round white plastic bin with a lid to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the small round white plastic bin with a lid to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0100_02/00360.jpg", "scene0100_02/00390.jpg", "scene0100_02/00420.jpg", "scene0100_02/00450.jpg" ], "turn_id": 4, "scan_id": "scene0100_02", "origin_question": "Which direction is the light brown oval-shaped ceramic sink in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the light brown oval-shaped ceramic sink in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0100_02/00480.jpg", "scene0100_02/00510.jpg", "scene0100_02/00540.jpg", "scene0100_02/00570.jpg" ], "turn_id": 5, "scan_id": "scene0100_02", "origin_question": "Please recall \"the light-colored wooden rectangular holder\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.52", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the light-colored wooden rectangular holder\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0100_02/00600.jpg", "scene0100_02/00630.jpg", "scene0100_02/00660.jpg", "scene0100_02/00690.jpg" ], "turn_id": 6, "scan_id": "scene0100_02", "origin_question": "Which two objects are on the same side of you now? \"the light brown oval-shaped ceramic sink\", \"the round blue and grey shoe made of soft material\", \"the bathtub\". ", "option": [ "the light brown oval-shaped ceramic sink, the round blue and grey shoe made of soft material", "the light brown oval-shaped ceramic sink, the bathtub", "the round blue and grey shoe made of soft material, the bathtub" ], "answer": "the round blue and grey shoe made of soft material, the bathtub", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the light brown oval-shaped ceramic sink\", \"the round blue and grey shoe made of soft material\", \"the bathtub\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the light brown oval-shaped ceramic sink, the round blue and grey shoe made of soft material', 'the light brown oval-shaped ceramic sink, the bathtub', 'the round blue and grey shoe made of soft material, the bathtub'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0100_02/00720.jpg", "scene0100_02/00750.jpg", "scene0100_02/00780.jpg", "scene0100_02/00810.jpg" ], "turn_id": 7, "scan_id": "scene0100_02", "origin_question": "Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0100_02/00840.jpg", "scene0100_02/00870.jpg", "scene0100_02/00900.jpg", "scene0100_02/00930.jpg" ], "turn_id": 8, "scan_id": "scene0100_02", "origin_question": "In which turn did you see both \"the classic, rounded beige ceramic toilet\" and \"the counter\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the classic, rounded beige ceramic toilet\" and \"the counter\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0040_region37/afbb23c155354c97a0b7064049f93ac7_i1_1.jpg", "1mp3d_0040_region37/afbb23c155354c97a0b7064049f93ac7_i2_1.jpg", "1mp3d_0040_region37/e8b5fb3a96e140398b0bc3600c04c7e9_i1_5.jpg", "1mp3d_0040_region37/e8b5fb3a96e140398b0bc3600c04c7e9_i2_5.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0040_region37", "origin_question": "Remember, have you seen any ladder(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any ladder(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 2 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0040_region37/afbb23c155354c97a0b7064049f93ac7_i1_0.jpg", "1mp3d_0040_region37/afbb23c155354c97a0b7064049f93ac7_i2_0.jpg", "1mp3d_0040_region37/e8b5fb3a96e140398b0bc3600c04c7e9_i1_0.jpg", "1mp3d_0040_region37/e8b5fb3a96e140398b0bc3600c04c7e9_i2_0.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0040_region37", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 2 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0148_00/00000.jpg", "scene0148_00/00040.jpg", "scene0148_00/00080.jpg", "scene0148_00/00120.jpg" ], "turn_id": 1, "scan_id": "scene0148_00", "origin_question": "Remember, have you seen any couch(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any couch(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0148_00/00160.jpg", "scene0148_00/00200.jpg", "scene0148_00/00240.jpg", "scene0148_00/00280.jpg" ], "turn_id": 2, "scan_id": "scene0148_00", "origin_question": "When did you first discover \"the rectangular, neutral-colored soft mat\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the rectangular, neutral-colored soft mat\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "scene0148_00/00320.jpg", "scene0148_00/00360.jpg", "scene0148_00/00400.jpg", "scene0148_00/00440.jpg" ], "turn_id": 3, "scan_id": "scene0148_00", "origin_question": "Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance forward or backward from that position?", "option": [ "Forward", "Backward" ], "answer": "Forward", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance forward or backward from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Forward', 'Backward'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0148_00/00480.jpg", "scene0148_00/00520.jpg", "scene0148_00/00560.jpg", "scene0148_00/00600.jpg" ], "turn_id": 4, "scan_id": "scene0148_00", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the rectangular, neutral-colored soft mat\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the rectangular, neutral-colored soft mat\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0148_00/00640.jpg", "scene0148_00/00680.jpg", "scene0148_00/00720.jpg", "scene0148_00/00760.jpg" ], "turn_id": 5, "scan_id": "scene0148_00", "origin_question": "Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "89.57", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0148_00/00800.jpg", "scene0148_00/00840.jpg", "scene0148_00/00880.jpg", "scene0148_00/00920.jpg" ], "turn_id": 6, "scan_id": "scene0148_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0355_01/00000.jpg", "scene0355_01/00030.jpg", "scene0355_01/00060.jpg", "scene0355_01/00090.jpg" ], "turn_id": 1, "scan_id": "scene0355_01", "origin_question": "Remember, have you seen any dress(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any dress(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0355_01/00120.jpg", "scene0355_01/00150.jpg", "scene0355_01/00180.jpg", "scene0355_01/00210.jpg" ], "turn_id": 2, "scan_id": "scene0355_01", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0355_01/00240.jpg", "scene0355_01/00270.jpg", "scene0355_01/00300.jpg", "scene0355_01/00330.jpg" ], "turn_id": 3, "scan_id": "scene0355_01", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the round wooden table with a smooth surface\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the round wooden table with a smooth surface\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0355_01/00360.jpg", "scene0355_01/00390.jpg", "scene0355_01/00420.jpg", "scene0355_01/00880.jpg" ], "turn_id": 4, "scan_id": "scene0355_01", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "2.74", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0044_region20/bc8ddbf1b6254074842fba693e7f81d1_i1_2.jpg", "1mp3d_0044_region20/bc8ddbf1b6254074842fba693e7f81d1_i2_2.jpg", "1mp3d_0044_region20/4a508feb2f784f1094cd33c96053657b_i1_2.jpg", "1mp3d_0044_region20/4a508feb2f784f1094cd33c96053657b_i2_2.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0044_region20", "origin_question": "Remember, have you seen any door(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any door(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "1mp3d_0044_region20/4a508feb2f784f1094cd33c96053657b_i1_1.jpg", "1mp3d_0044_region20/4a508feb2f784f1094cd33c96053657b_i2_1.jpg", "1mp3d_0044_region20/57fba128d2f042f7a59793c665a3f587_i1_5.jpg", "1mp3d_0044_region20/57fba128d2f042f7a59793c665a3f587_i2_5.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0044_region20", "origin_question": "In which turn did you see both \"the long silver metal bar mounted on the wall\" and \"the rectangular window with a dark wooden frame\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the long silver metal bar mounted on the wall\" and \"the rectangular window with a dark wooden frame\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "1mp3d_0044_region20/0e27bb2cbb55402c8186e9830d21147c_i1_0.jpg", "1mp3d_0044_region20/0e27bb2cbb55402c8186e9830d21147c_i2_0.jpg", "1mp3d_0044_region20/0e27bb2cbb55402c8186e9830d21147c_i1_1.jpg", "1mp3d_0044_region20/0e27bb2cbb55402c8186e9830d21147c_i2_1.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0044_region20", "origin_question": "Please recall \"the rectangular window with a dark wooden frame\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.54", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the rectangular window with a dark wooden frame\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "1mp3d_0044_region20/57fba128d2f042f7a59793c665a3f587_i1_1.jpg", "1mp3d_0044_region20/57fba128d2f042f7a59793c665a3f587_i2_1.jpg", "1mp3d_0044_region20/57fba128d2f042f7a59793c665a3f587_i1_0.jpg", "1mp3d_0044_region20/57fba128d2f042f7a59793c665a3f587_i2_0.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0044_region20", "origin_question": "In which turn did you see both \"the roll of standard white toilet paper\" and \"the rectangular deep brown wooden vanity with a black countertop\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the roll of standard white toilet paper\" and \"the rectangular deep brown wooden vanity with a black countertop\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "1mp3d_0044_region20/c9dc3f6946b4459baa32204af20ba935_i1_5.jpg", "1mp3d_0044_region20/c9dc3f6946b4459baa32204af20ba935_i2_5.jpg", "1mp3d_0044_region20/c9dc3f6946b4459baa32204af20ba935_i1_0.jpg", "1mp3d_0044_region20/c9dc3f6946b4459baa32204af20ba935_i2_0.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0044_region20", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the rectangular black ornate mirror with intricate carvings\"; \"the small shiny silver faucet with a simple design\"; \"the rectangular dark-colored shelf with a simple design\" ", "option": [ "the rectangular black ornate mirror with intricate carvings", "the small shiny silver faucet with a simple design", "the rectangular dark-colored shelf with a simple design" ], "answer": "the rectangular dark-colored shelf with a simple design", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the rectangular black ornate mirror with intricate carvings\"; \"the small shiny silver faucet with a simple design\"; \"the rectangular dark-colored shelf with a simple design\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular black ornate mirror with intricate carvings', 'the small shiny silver faucet with a simple design', 'the rectangular dark-colored shelf with a simple design'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "1mp3d_0044_region20/c9dc3f6946b4459baa32204af20ba935_i1_1.jpg", "1mp3d_0044_region20/c9dc3f6946b4459baa32204af20ba935_i2_1.jpg", "1mp3d_0044_region20/a94d862351b34730bd743442fc32cdab_i1_2.jpg", "1mp3d_0044_region20/a94d862351b34730bd743442fc32cdab_i2_2.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0044_region20", "origin_question": "How far is your current position from where you were at the end of the turn 4? (in meters)", "option": [], "answer": "1.18", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 4? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "1mp3d_0044_region20/4fd7ad2ab317402bbab5d08b012e98f3_i1_2.jpg", "1mp3d_0044_region20/4fd7ad2ab317402bbab5d08b012e98f3_i2_2.jpg", "1mp3d_0044_region20/a94d862351b34730bd743442fc32cdab_i2_1.jpg", "1mp3d_0044_region20/4fd7ad2ab317402bbab5d08b012e98f3_i1_1.jpg" ], "turn_id": 7, "scan_id": "1mp3d_0044_region20", "origin_question": "Which direction is the rectangular dark-colored shelf with a simple design in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the rectangular dark-colored shelf with a simple design in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0392_02/00000.jpg", "scene0392_02/00030.jpg", "scene0392_02/00060.jpg", "scene0392_02/00090.jpg" ], "turn_id": 1, "scan_id": "scene0392_02", "origin_question": "Remember, have you seen any chair(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any chair(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 3 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0392_02/00120.jpg", "scene0392_02/00150.jpg", "scene0392_02/00180.jpg", "scene0392_02/00210.jpg" ], "turn_id": 2, "scan_id": "scene0392_02", "origin_question": "In which turn did you see both \"the light brown rectangular wooden shelf with a simple design\" and \"the rectangular light wooden bed with a deep blue mattress\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the light brown rectangular wooden shelf with a simple design\" and \"the rectangular light wooden bed with a deep blue mattress\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 3 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0392_02/00240.jpg", "scene0392_02/00310.jpg", "scene0392_02/00340.jpg", "scene0392_02/00370.jpg" ], "turn_id": 3, "scan_id": "scene0392_02", "origin_question": "", "option": [], "answer": "", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 3 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0573_00/00000.jpg", "scene0573_00/00030.jpg", "scene0573_00/00060.jpg", "scene0573_00/00090.jpg" ], "turn_id": 1, "scan_id": "scene0573_00", "origin_question": "Remember, have you seen any refrigerator(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any refrigerator(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0573_00/00120.jpg", "scene0573_00/00150.jpg", "scene0573_00/00180.jpg", "scene0573_00/00210.jpg" ], "turn_id": 2, "scan_id": "scene0573_00", "origin_question": "Which two objects are on the same side of you now? \"the white counter with black plastic burners\", \"the white double-door stainless steel refrigerator\", \"the rectangular white plastic switch\". ", "option": [ "the white counter with black plastic burners, the white double-door stainless steel refrigerator", "the white counter with black plastic burners, the rectangular white plastic switch", "the white double-door stainless steel refrigerator, the rectangular white plastic switch" ], "answer": "the white counter with black plastic burners, the white double-door stainless steel refrigerator", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the white counter with black plastic burners\", \"the white double-door stainless steel refrigerator\", \"the rectangular white plastic switch\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the white counter with black plastic burners, the white double-door stainless steel refrigerator', 'the white counter with black plastic burners, the rectangular white plastic switch', 'the white double-door stainless steel refrigerator, the rectangular white plastic switch'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0573_00/00240.jpg", "scene0573_00/00270.jpg", "scene0573_00/00300.jpg", "scene0573_00/00330.jpg" ], "turn_id": 3, "scan_id": "scene0573_00", "origin_question": "In which turn did you see both \"the white rectangular electric stove with black burners\" and \"the white double-door stainless steel refrigerator\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the white rectangular electric stove with black burners\" and \"the white double-door stainless steel refrigerator\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0573_00/00360.jpg", "scene0573_00/00390.jpg", "scene0573_00/00440.jpg", "scene0573_00/00470.jpg" ], "turn_id": 4, "scan_id": "scene0573_00", "origin_question": "Which direction is the rectangular white plastic switch in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the rectangular white plastic switch in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0573_00/00500.jpg", "scene0573_00/00530.jpg", "scene0573_00/00560.jpg", "scene0573_00/00590.jpg" ], "turn_id": 5, "scan_id": "scene0573_00", "origin_question": "At the end of which turn were both of \"the rectangular white plastic switch\" and \"the standard kitchen sink faucet with a smooth metal surface\" on your right side?", "option": [], "answer": "2", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the rectangular white plastic switch\" and \"the standard kitchen sink faucet with a smooth metal surface\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0573_00/00620.jpg", "scene0573_00/00650.jpg", "scene0573_00/00680.jpg", "scene0573_00/00710.jpg" ], "turn_id": 6, "scan_id": "scene0573_00", "origin_question": "Using your orientation at the end of turn 5 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "75.43", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 5 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0426_00/00000.jpg", "scene0426_00/00040.jpg", "scene0426_00/00080.jpg", "scene0426_00/00120.jpg", "scene0426_00/00160.jpg" ], "turn_id": 1, "scan_id": "scene0426_00", "origin_question": "Remember, have you seen any piano(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any piano(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0426_00/00200.jpg", "scene0426_00/00240.jpg", "scene0426_00/00280.jpg", "scene0426_00/00320.jpg", "scene0426_00/00360.jpg" ], "turn_id": 2, "scan_id": "scene0426_00", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the medium-sized rectangular wooden desk in deep brown\" from your position?", "option": [], "answer": "118.48", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the medium-sized rectangular wooden desk in deep brown\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0426_00/00400.jpg", "scene0426_00/00440.jpg", "scene0426_00/00480.jpg", "scene0426_00/00520.jpg", "scene0426_00/00560.jpg" ], "turn_id": 3, "scan_id": "scene0426_00", "origin_question": "Is the medium-sized rectangular wooden desk in deep brown to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the medium-sized rectangular wooden desk in deep brown to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0426_00/00600.jpg", "scene0426_00/00640.jpg", "scene0426_00/00680.jpg", "scene0426_00/00720.jpg", "scene0426_00/00760.jpg" ], "turn_id": 4, "scan_id": "scene0426_00", "origin_question": "Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "144.77", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0426_00/00800.jpg", "scene0426_00/00840.jpg", "scene0426_00/00880.jpg", "scene0426_00/00920.jpg", "scene0426_00/00960.jpg" ], "turn_id": 5, "scan_id": "scene0426_00", "origin_question": "Please recall \"the medium-sized rectangular wooden desk in deep brown\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.39", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the medium-sized rectangular wooden desk in deep brown\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0426_00/01000.jpg", "scene0426_00/01040.jpg", "scene0426_00/01080.jpg", "scene0426_00/01120.jpg", "scene0426_00/01160.jpg" ], "turn_id": 6, "scan_id": "scene0426_00", "origin_question": "Please recall \"the compact dark piano with white keys\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.61", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the compact dark piano with white keys\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0426_00/01200.jpg", "scene0426_00/01240.jpg", "scene0426_00/01280.jpg", "scene0426_00/01320.jpg", "scene0426_00/01360.jpg" ], "turn_id": 7, "scan_id": "scene0426_00", "origin_question": "Please recall \"the tall, elegant lamp with a white conical shade\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.35", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the tall, elegant lamp with a white conical shade\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0426_00/01400.jpg", "scene0426_00/01440.jpg", "scene0426_00/01480.jpg", "scene0426_00/01520.jpg", "scene0426_00/01560.jpg" ], "turn_id": 8, "scan_id": "scene0426_00", "origin_question": "Which two objects are on the same side of you now? \"the medium-sized rectangular wooden desk in deep brown\", \"the standard-sized white paper sheet\", \"the flat-panel TV with a dark matte surface\". ", "option": [ "the medium-sized rectangular wooden desk in deep brown, the standard-sized white paper sheet", "the medium-sized rectangular wooden desk in deep brown, the flat-panel TV with a dark matte surface", "the standard-sized white paper sheet, the flat-panel TV with a dark matte surface" ], "answer": "the medium-sized rectangular wooden desk in deep brown, the standard-sized white paper sheet", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the medium-sized rectangular wooden desk in deep brown\", \"the standard-sized white paper sheet\", \"the flat-panel TV with a dark matte surface\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the medium-sized rectangular wooden desk in deep brown, the standard-sized white paper sheet', 'the medium-sized rectangular wooden desk in deep brown, the flat-panel TV with a dark matte surface', 'the standard-sized white paper sheet, the flat-panel TV with a dark matte surface'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0426_00/01600.jpg", "scene0426_00/01640.jpg", "scene0426_00/01680.jpg", "scene0426_00/01720.jpg", "scene0426_00/01760.jpg" ], "turn_id": 9, "scan_id": "scene0426_00", "origin_question": "In which turn did you see both \"the medium rectangular tissue box with a blue pattern\" and \"the window\" simultaneously? ", "option": [], "answer": "6", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the medium rectangular tissue box with a blue pattern\" and \"the window\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0426_00/01800.jpg", "scene0426_00/01840.jpg", "scene0426_00/01880.jpg", "scene0426_00/01920.jpg", "scene0426_00/01960.jpg" ], "turn_id": 10, "scan_id": "scene0426_00", "origin_question": "Compared to your position at the end of the 6 turn, are you now closer or farther away from \"the tall, elegant lamp with a white conical shade\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 6 turn, are you now closer or farther away from \"the tall, elegant lamp with a white conical shade\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0029_region33/93318cd6a48a4eb59eb59d2481095044_i1_3.jpg", "1mp3d_0029_region33/93318cd6a48a4eb59eb59d2481095044_i2_3.jpg", "1mp3d_0029_region33/bbcc19afdad742eba89290ad2b5dbf6b_i1_3.jpg", "1mp3d_0029_region33/bbcc19afdad742eba89290ad2b5dbf6b_i2_3.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0029_region33", "origin_question": "Remember, have you seen any towel(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any towel(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 2 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0029_region33/bbcc19afdad742eba89290ad2b5dbf6b_i1_4.jpg", "1mp3d_0029_region33/bbcc19afdad742eba89290ad2b5dbf6b_i2_4.jpg", "1mp3d_0029_region33/bbcc19afdad742eba89290ad2b5dbf6b_i1_5.jpg", "1mp3d_0029_region33/bbcc19afdad742eba89290ad2b5dbf6b_i2_5.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0029_region33", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 2 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0280_01/00000.jpg", "scene0280_01/00040.jpg", "scene0280_01/00080.jpg", "scene0280_01/00120.jpg", "scene0280_01/00160.jpg" ], "turn_id": 1, "scan_id": "scene0280_01", "origin_question": "Remember, have you seen any pillow(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any pillow(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0280_01/00200.jpg", "scene0280_01/00240.jpg", "scene0280_01/00280.jpg", "scene0280_01/00320.jpg", "scene0280_01/00360.jpg" ], "turn_id": 2, "scan_id": "scene0280_01", "origin_question": "Please recall \"the black rectangular plastic computer with a white keyboard\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.54", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the black rectangular plastic computer with a white keyboard\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0280_01/00400.jpg", "scene0280_01/00440.jpg", "scene0280_01/00480.jpg", "scene0280_01/00520.jpg", "scene0280_01/00560.jpg" ], "turn_id": 3, "scan_id": "scene0280_01", "origin_question": "In which turn did you see both \"the rectangular dark brown wooden dresser with multiple drawers\" and \"the rectangular green tissue box made of cardboard\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the rectangular dark brown wooden dresser with multiple drawers\" and \"the rectangular green tissue box made of cardboard\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0280_01/00600.jpg", "scene0280_01/00640.jpg", "scene0280_01/00680.jpg", "scene0280_01/00720.jpg", "scene0280_01/00760.jpg" ], "turn_id": 4, "scan_id": "scene0280_01", "origin_question": "Which two objects are on the same side of you now? \"the rectangular, striped pillow in various colors\", \"the massive green smooth ball\", \"the book\". ", "option": [ "the rectangular, striped pillow in various colors, the massive green smooth ball", "the rectangular, striped pillow in various colors, the book", "the massive green smooth ball, the book" ], "answer": "the massive green smooth ball, the book", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the rectangular, striped pillow in various colors\", \"the massive green smooth ball\", \"the book\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular, striped pillow in various colors, the massive green smooth ball', 'the rectangular, striped pillow in various colors, the book', 'the massive green smooth ball, the book'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0280_01/00800.jpg", "scene0280_01/00840.jpg", "scene0280_01/00880.jpg", "scene0280_01/00920.jpg", "scene0280_01/00960.jpg" ], "turn_id": 5, "scan_id": "scene0280_01", "origin_question": "Is the rectangular green tissue box made of cardboard to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the rectangular green tissue box made of cardboard to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0280_01/01000.jpg", "scene0280_01/01040.jpg", "scene0280_01/01080.jpg", "scene0280_01/01120.jpg", "scene0280_01/01160.jpg" ], "turn_id": 6, "scan_id": "scene0280_01", "origin_question": "Remember, how many commode(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many commode(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0280_01/01200.jpg", "scene0280_01/01240.jpg", "scene0280_01/01280.jpg", "scene0280_01/01320.jpg", "scene0280_01/01360.jpg" ], "turn_id": 7, "scan_id": "scene0280_01", "origin_question": "In which turn did you see both \"the rectangular, striped pillow in various colors\" and \"the massive green smooth ball\" simultaneously? ", "option": [], "answer": "7", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the rectangular, striped pillow in various colors\" and \"the massive green smooth ball\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0280_01/01400.jpg", "scene0280_01/01440.jpg", "scene0280_01/01480.jpg", "scene0280_01/01520.jpg", "scene0280_01/01560.jpg" ], "turn_id": 8, "scan_id": "scene0280_01", "origin_question": "Please recall \"the rectangular green tissue box made of cardboard\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.84", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the rectangular green tissue box made of cardboard\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0280_01/01600.jpg", "scene0280_01/01640.jpg", "scene0280_01/01680.jpg", "scene0280_01/01720.jpg", "scene0280_01/01760.jpg" ], "turn_id": 9, "scan_id": "scene0280_01", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the rectangular green tissue box made of cardboard\" from your position?", "option": [], "answer": "95.1", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the rectangular green tissue box made of cardboard\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0678_00/00000.jpg", "scene0678_00/00040.jpg", "scene0678_00/00080.jpg", "scene0678_00/00120.jpg", "scene0678_00/00160.jpg" ], "turn_id": 1, "scan_id": "scene0678_00", "origin_question": "Remember, how many bin(s) have you seen so far? ", "option": [], "answer": "2", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many bin(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0678_00/00200.jpg", "scene0678_00/00240.jpg", "scene0678_00/00280.jpg", "scene0678_00/00320.jpg", "scene0678_00/00360.jpg" ], "turn_id": 2, "scan_id": "scene0678_00", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the cylindrical white plastic fan\"; \"the small cylindrical red mesh hamper\"; \"the clothes\" ", "option": [ "the cylindrical white plastic fan", "the small cylindrical red mesh hamper", "the clothes" ], "answer": "the clothes", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the cylindrical white plastic fan\"; \"the small cylindrical red mesh hamper\"; \"the clothes\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the cylindrical white plastic fan', 'the small cylindrical red mesh hamper', 'the clothes'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0678_00/00400.jpg", "scene0678_00/00440.jpg", "scene0678_00/00480.jpg", "scene0678_00/00520.jpg", "scene0678_00/00560.jpg" ], "turn_id": 3, "scan_id": "scene0678_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0678_00/00600.jpg", "scene0678_00/00640.jpg", "scene0678_00/00680.jpg", "scene0678_00/00720.jpg", "scene0678_00/00760.jpg" ], "turn_id": 4, "scan_id": "scene0678_00", "origin_question": "Which direction is the door in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the door in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0678_00/00800.jpg", "scene0678_00/00840.jpg", "scene0678_00/00880.jpg", "scene0678_00/00920.jpg", "scene0678_00/00960.jpg" ], "turn_id": 5, "scan_id": "scene0678_00", "origin_question": "Please recall \"the small cylindrical red mesh hamper\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "3.18", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the small cylindrical red mesh hamper\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0678_00/01000.jpg", "scene0678_00/01040.jpg", "scene0678_00/01080.jpg", "scene0678_00/01120.jpg", "scene0678_00/01160.jpg" ], "turn_id": 6, "scan_id": "scene0678_00", "origin_question": "Which direction is then orange plastic bottle resembling a four-sided pyramid in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is then orange plastic bottle resembling a four-sided pyramid in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0678_00/01200.jpg", "scene0678_00/01240.jpg", "scene0678_00/01280.jpg", "scene0678_00/01320.jpg", "scene0678_00/01360.jpg" ], "turn_id": 7, "scan_id": "scene0678_00", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the small cylindrical red mesh hamper\" from your position?", "option": [], "answer": "43.6", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the small cylindrical red mesh hamper\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0678_00/01400.jpg", "scene0678_00/01440.jpg", "scene0678_00/01480.jpg", "scene0678_00/01520.jpg", "scene0678_00/01560.jpg" ], "turn_id": 8, "scan_id": "scene0678_00", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the rectangular white towel made of soft fabric\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the rectangular white towel made of soft fabric\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0678_00/01600.jpg", "scene0678_00/01640.jpg", "scene0678_00/01680.jpg", "scene0678_00/01720.jpg", "scene0678_00/01760.jpg" ], "turn_id": 9, "scan_id": "scene0678_00", "origin_question": "In which turn did you see both \"the small cylindrical red mesh hamper\" and \"the large, flat stone or concrete ledge\" simultaneously? ", "option": [], "answer": "9", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the small cylindrical red mesh hamper\" and \"the large, flat stone or concrete ledge\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0678_00/01800.jpg", "scene0678_00/01840.jpg", "scene0678_00/01880.jpg", "scene0678_00/01920.jpg", "scene0678_00/01960.jpg" ], "turn_id": 10, "scan_id": "scene0678_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0605_00/00000.jpg", "scene0605_00/00030.jpg", "scene0605_00/00060.jpg", "scene0605_00/00090.jpg" ], "turn_id": 1, "scan_id": "scene0605_00", "origin_question": "Remember, have you seen any bin(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any bin(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0605_00/00120.jpg", "scene0605_00/00150.jpg", "scene0605_00/00180.jpg", "scene0605_00/00210.jpg" ], "turn_id": 2, "scan_id": "scene0605_00", "origin_question": "At the end of which turn were both of \"the wooden rectangular shelf in good condition\" and \"the rectangular soft brown wooden drawer\" on your right side?", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the wooden rectangular shelf in good condition\" and \"the rectangular soft brown wooden drawer\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0605_00/00240.jpg", "scene0605_00/00270.jpg", "scene0605_00/00300.jpg", "scene0605_00/00330.jpg" ], "turn_id": 3, "scan_id": "scene0605_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0605_00/00360.jpg", "scene0605_00/00390.jpg", "scene0605_00/00420.jpg", "scene0605_00/00450.jpg" ], "turn_id": 4, "scan_id": "scene0605_00", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the silver or gray car with an elegant front grille\"; \"the medium-sized black handheld umbrella made of nylon\"; \"the wooden rectangular shelf in good condition\" ", "option": [ "the silver or gray car with an elegant front grille", "the medium-sized black handheld umbrella made of nylon", "the wooden rectangular shelf in good condition" ], "answer": "the medium-sized black handheld umbrella made of nylon", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the silver or gray car with an elegant front grille\"; \"the medium-sized black handheld umbrella made of nylon\"; \"the wooden rectangular shelf in good condition\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the silver or gray car with an elegant front grille', 'the medium-sized black handheld umbrella made of nylon', 'the wooden rectangular shelf in good condition'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0605_00/00480.jpg", "scene0605_00/00510.jpg", "scene0605_00/00540.jpg", "scene0605_00/00570.jpg" ], "turn_id": 5, "scan_id": "scene0605_00", "origin_question": "Is the wooden rectangular shelf in good condition to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the wooden rectangular shelf in good condition to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0314_00/00000.jpg", "scene0314_00/00030.jpg", "scene0314_00/00060.jpg", "scene0314_00/00090.jpg" ], "turn_id": 1, "scan_id": "scene0314_00", "origin_question": "Remember, have you seen any wardrobe(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any wardrobe(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0314_00/00120.jpg", "scene0314_00/00150.jpg", "scene0314_00/00180.jpg", "scene0314_00/00210.jpg" ], "turn_id": 2, "scan_id": "scene0314_00", "origin_question": "At the end of which turn were both of \"the tall, elegant lamp with a white lampshade\" and \"the black, sleek, modern mouse made of plastic\" on your left side?", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the tall, elegant lamp with a white lampshade\" and \"the black, sleek, modern mouse made of plastic\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0314_00/00240.jpg", "scene0314_00/00270.jpg", "scene0314_00/00300.jpg", "scene0314_00/00330.jpg" ], "turn_id": 3, "scan_id": "scene0314_00", "origin_question": "Remember, how many toaster(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many toaster(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0314_00/00360.jpg", "scene0314_00/00390.jpg", "scene0314_00/00420.jpg", "scene0314_00/00450.jpg" ], "turn_id": 4, "scan_id": "scene0314_00", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "2.74", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "scene0314_00/00480.jpg", "scene0314_00/00510.jpg", "scene0314_00/00540.jpg", "scene0314_00/00570.jpg" ], "turn_id": 5, "scan_id": "scene0314_00", "origin_question": "Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Left", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0314_00/00600.jpg", "scene0314_00/00630.jpg", "scene0314_00/00660.jpg", "scene0314_00/00690.jpg" ], "turn_id": 6, "scan_id": "scene0314_00", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "2.2", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0314_00/00720.jpg", "scene0314_00/00750.jpg", "scene0314_00/00780.jpg", "scene0314_00/00810.jpg" ], "turn_id": 7, "scan_id": "scene0314_00", "origin_question": "Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the round wooden table with a smooth brown surface\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the round wooden table with a smooth brown surface\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0432_00/00000.jpg", "scene0432_00/00030.jpg", "scene0432_00/00060.jpg", "scene0432_00/00090.jpg" ], "turn_id": 1, "scan_id": "scene0432_00", "origin_question": "Remember, how many fence(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many fence(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 2 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0432_00/00120.jpg", "scene0432_00/00150.jpg", "scene0432_00/00180.jpg", "scene0432_00/00210.jpg" ], "turn_id": 2, "scan_id": "scene0432_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 2 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0471_01/00000.jpg", "scene0471_01/00030.jpg", "scene0471_01/00060.jpg", "scene0471_01/00090.jpg" ], "turn_id": 1, "scan_id": "scene0471_01", "origin_question": "Remember, have you seen any curtain(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any curtain(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0471_01/00120.jpg", "scene0471_01/00150.jpg", "scene0471_01/00180.jpg", "scene0471_01/00210.jpg" ], "turn_id": 2, "scan_id": "scene0471_01", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0471_01/00240.jpg", "scene0471_01/00270.jpg", "scene0471_01/00300.jpg", "scene0471_01/00330.jpg" ], "turn_id": 3, "scan_id": "scene0471_01", "origin_question": "In which turn were you closest to \"the small rectangular beige restroom stall\"? ", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the small rectangular beige restroom stall\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0471_01/00360.jpg", "scene0471_01/00390.jpg", "scene0471_01/00420.jpg", "scene0471_01/00450.jpg" ], "turn_id": 4, "scan_id": "scene0471_01", "origin_question": "Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0471_01/00480.jpg", "scene0471_01/00510.jpg", "scene0471_01/00540.jpg", "scene0471_01/00570.jpg" ], "turn_id": 5, "scan_id": "scene0471_01", "origin_question": "Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the dispenser\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the dispenser\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "1mp3d_0030_region15/33ff25e81c9b4834a37b1bb799c4b26f_i1_0.jpg", "1mp3d_0030_region15/33ff25e81c9b4834a37b1bb799c4b26f_i2_0.jpg", "1mp3d_0030_region15/33ff25e81c9b4834a37b1bb799c4b26f_i1_1.jpg", "1mp3d_0030_region15/33ff25e81c9b4834a37b1bb799c4b26f_i2_1.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0030_region15", "origin_question": "Remember, how many globe(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many globe(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "1mp3d_0030_region15/e850bf9ddf434263bdce31b4029c5a97_i1_1.jpg", "1mp3d_0030_region15/e850bf9ddf434263bdce31b4029c5a97_i2_1.jpg", "1mp3d_0030_region15/e3eeb7b8deed4a2fa0000daec3538b75_i1_4.jpg", "1mp3d_0030_region15/e3eeb7b8deed4a2fa0000daec3538b75_i2_4.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0030_region15", "origin_question": "Which direction is the stainless steel rectangular double-door refrigerator in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the stainless steel rectangular double-door refrigerator in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "1mp3d_0030_region15/e850bf9ddf434263bdce31b4029c5a97_i1_5.jpg", "1mp3d_0030_region15/e850bf9ddf434263bdce31b4029c5a97_i2_5.jpg", "1mp3d_0030_region15/e850bf9ddf434263bdce31b4029c5a97_i1_4.jpg", "1mp3d_0030_region15/e850bf9ddf434263bdce31b4029c5a97_i2_4.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0030_region15", "origin_question": "In which turn were you farthest from \"the rectangular black microwave of standard size\"? ", "option": [], "answer": "2", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the rectangular black microwave of standard size\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "1mp3d_0030_region15/e850bf9ddf434263bdce31b4029c5a97_i1_3.jpg", "1mp3d_0030_region15/e850bf9ddf434263bdce31b4029c5a97_i2_3.jpg", "1mp3d_0030_region15/e3eeb7b8deed4a2fa0000daec3538b75_i1_1.jpg", "1mp3d_0030_region15/e3eeb7b8deed4a2fa0000daec3538b75_i2_1.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0030_region15", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the rectangular black microwave of standard size\"; \"the dark rectangular kitchen island with a wood countertop\"; \"the modern silver rectangular gas stove with black accents\" ", "option": [ "the rectangular black microwave of standard size", "the dark rectangular kitchen island with a wood countertop", "the modern silver rectangular gas stove with black accents" ], "answer": "the modern silver rectangular gas stove with black accents", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the rectangular black microwave of standard size\"; \"the dark rectangular kitchen island with a wood countertop\"; \"the modern silver rectangular gas stove with black accents\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular black microwave of standard size', 'the dark rectangular kitchen island with a wood countertop', 'the modern silver rectangular gas stove with black accents'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "1mp3d_0030_region15/e3eeb7b8deed4a2fa0000daec3538b75_i1_0.jpg", "1mp3d_0030_region15/e3eeb7b8deed4a2fa0000daec3538b75_i2_0.jpg", "1mp3d_0030_region15/e3eeb7b8deed4a2fa0000daec3538b75_i1_5.jpg", "1mp3d_0030_region15/e3eeb7b8deed4a2fa0000daec3538b75_i2_5.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0030_region15", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the modern silver rectangular gas stove with black accents\", Y: \"the cylindrical, transparent soap with a white pump head\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "farther away from both X and Y", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the modern silver rectangular gas stove with black accents\", Y: \"the cylindrical, transparent soap with a white pump head\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "1mp3d_0030_region15/f4032225a0c6462ba02f728d031c2638_i1_5.jpg", "1mp3d_0030_region15/f4032225a0c6462ba02f728d031c2638_i2_5.jpg", "1mp3d_0030_region15/f4032225a0c6462ba02f728d031c2638_i1_0.jpg", "1mp3d_0030_region15/f4032225a0c6462ba02f728d031c2638_i2_0.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0030_region15", "origin_question": "Which direction is the rectangular silver stainless steel dishwasher in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the rectangular silver stainless steel dishwasher in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0106_00/00000.jpg", "scene0106_00/00040.jpg", "scene0106_00/00080.jpg", "scene0106_00/00120.jpg", "scene0106_00/00160.jpg" ], "turn_id": 1, "scan_id": "scene0106_00", "origin_question": "Remember, how many water heater(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many water heater(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0106_00/00200.jpg", "scene0106_00/00240.jpg", "scene0106_00/00280.jpg", "scene0106_00/00320.jpg", "scene0106_00/00360.jpg" ], "turn_id": 2, "scan_id": "scene0106_00", "origin_question": "When did you first discover \"the small rectangular white ceramic sink\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the small rectangular white ceramic sink\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0106_00/00400.jpg", "scene0106_00/00440.jpg", "scene0106_00/00480.jpg", "scene0106_00/00520.jpg", "scene0106_00/00560.jpg" ], "turn_id": 3, "scan_id": "scene0106_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0106_00/00600.jpg", "scene0106_00/00640.jpg", "scene0106_00/00680.jpg", "scene0106_00/00720.jpg", "scene0106_00/00760.jpg" ], "turn_id": 4, "scan_id": "scene0106_00", "origin_question": "Is the large black flat-screen television to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the large black flat-screen television to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0106_00/00800.jpg", "scene0106_00/00840.jpg", "scene0106_00/00880.jpg", "scene0106_00/00920.jpg", "scene0106_00/00960.jpg" ], "turn_id": 5, "scan_id": "scene0106_00", "origin_question": "In which turn were you closest to \"the small rectangular white ceramic sink\"? ", "option": [], "answer": "1", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the small rectangular white ceramic sink\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0106_00/01000.jpg", "scene0106_00/01040.jpg", "scene0106_00/01080.jpg", "scene0106_00/01120.jpg", "scene0106_00/01160.jpg" ], "turn_id": 6, "scan_id": "scene0106_00", "origin_question": "By the end of the 4 turn, were you getting closer to or farther away from these two objects: X: \"the small rectangular white ceramic sink\", Y: \"the window\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to X and farther away from Y", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 4 turn, were you getting closer to or farther away from these two objects: X: \"the small rectangular white ceramic sink\", Y: \"the window\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "scene0106_00/01200.jpg", "scene0106_00/01240.jpg", "scene0106_00/01280.jpg", "scene0106_00/01320.jpg", "scene0106_00/01360.jpg" ], "turn_id": 7, "scan_id": "scene0106_00", "origin_question": "Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Left", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0106_00/01400.jpg", "scene0106_00/01440.jpg", "scene0106_00/01480.jpg", "scene0106_00/01520.jpg", "scene0106_00/01560.jpg" ], "turn_id": 8, "scan_id": "scene0106_00", "origin_question": "Is the small rectangular white ceramic sink to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the small rectangular white ceramic sink to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0106_00/01600.jpg", "scene0106_00/01640.jpg", "scene0106_00/01680.jpg", "scene0106_00/01720.jpg", "scene0106_00/01760.jpg" ], "turn_id": 9, "scan_id": "scene0106_00", "origin_question": "In which turn were you closest to \"the window\"? ", "option": [], "answer": "3", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the window\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "scene0106_00/01800.jpg", "scene0106_00/01840.jpg", "scene0106_00/01880.jpg", "scene0106_00/01920.jpg", "scene0106_00/01960.jpg" ], "turn_id": 10, "scan_id": "scene0106_00", "origin_question": "Among these three objects(\"the small rectangular white ceramic sink\",\"the window\",\"the large black flat-screen television\" ), which one is the closest to you?", "option": [ "the small rectangular white ceramic sink", "the window", "the large black flat-screen television" ], "answer": "the large black flat-screen television", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the small rectangular white ceramic sink\",\"the window\",\"the large black flat-screen television\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the small rectangular white ceramic sink', 'the window', 'the large black flat-screen television'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "1mp3d_0030_region9/c4b53f73a3734bc7a5b51747e5d1d5f7_i1_0.jpg", "1mp3d_0030_region9/c4b53f73a3734bc7a5b51747e5d1d5f7_i2_0.jpg", "1mp3d_0030_region9/3de5de73b782437291628651d994bfcf_i1_3.jpg", "1mp3d_0030_region9/3de5de73b782437291628651d994bfcf_i2_3.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0030_region9", "origin_question": "Remember, how many fire extinguisher(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many fire extinguisher(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "1mp3d_0030_region9/c4b53f73a3734bc7a5b51747e5d1d5f7_i1_4.jpg", "1mp3d_0030_region9/c4b53f73a3734bc7a5b51747e5d1d5f7_i2_4.jpg", "1mp3d_0030_region9/3de5de73b782437291628651d994bfcf_i1_4.jpg", "1mp3d_0030_region9/3de5de73b782437291628651d994bfcf_i2_4.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0030_region9", "origin_question": "In which turn did you see both \"the medium-sized rectangular black plastic remote control\" and \"the medium conical lampshade with a dark cylindrical base\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the medium-sized rectangular black plastic remote control\" and \"the medium conical lampshade with a dark cylindrical base\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "1mp3d_0030_region9/b03eb3aa3b6e4c5abb5e78830372ff89_i2_5.jpg", "1mp3d_0030_region9/b03eb3aa3b6e4c5abb5e78830372ff89_i1_4.jpg", "1mp3d_0030_region9/b03eb3aa3b6e4c5abb5e78830372ff89_i2_4.jpg", "1mp3d_0030_region9/b03eb3aa3b6e4c5abb5e78830372ff89_i1_2.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0030_region9", "origin_question": "Which two objects are on the same side of you now? \"the rectangular red, beige, and white carpet with designs\", \"the medium conical lampshade with a dark cylindrical base\", \"the rectangular wooden cabinet in rich brown color\". ", "option": [ "the rectangular red, beige, and white carpet with designs, the medium conical lampshade with a dark cylindrical base", "the rectangular red, beige, and white carpet with designs, the rectangular wooden cabinet in rich brown color", "the medium conical lampshade with a dark cylindrical base, the rectangular wooden cabinet in rich brown color" ], "answer": "the rectangular red, beige, and white carpet with designs, the medium conical lampshade with a dark cylindrical base", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the rectangular red, beige, and white carpet with designs\", \"the medium conical lampshade with a dark cylindrical base\", \"the rectangular wooden cabinet in rich brown color\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular red, beige, and white carpet with designs, the medium conical lampshade with a dark cylindrical base', 'the rectangular red, beige, and white carpet with designs, the rectangular wooden cabinet in rich brown color', 'the medium conical lampshade with a dark cylindrical base, the rectangular wooden cabinet in rich brown color'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "1mp3d_0030_region9/b03eb3aa3b6e4c5abb5e78830372ff89_i2_2.jpg", "1mp3d_0030_region9/b03eb3aa3b6e4c5abb5e78830372ff89_i1_3.jpg", "1mp3d_0030_region9/b03eb3aa3b6e4c5abb5e78830372ff89_i2_3.jpg", "1mp3d_0030_region9/3de5de73b782437291628651d994bfcf_i1_0.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0030_region9", "origin_question": "In which turn did you see both \"the small rectangular yellow box with flower patterns\" and \"the two-seater couch with light blue cushions and curved armrests\" simultaneously? ", "option": [], "answer": "4", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the small rectangular yellow box with flower patterns\" and \"the two-seater couch with light blue cushions and curved armrests\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "1mp3d_0030_region9/3de5de73b782437291628651d994bfcf_i2_0.jpg", "1mp3d_0030_region9/3de5de73b782437291628651d994bfcf_i1_1.jpg", "1mp3d_0030_region9/3de5de73b782437291628651d994bfcf_i2_1.jpg", "1mp3d_0030_region9/3de5de73b782437291628651d994bfcf_i1_5.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0030_region9", "origin_question": "Which direction is the medium-sized rectangular black plastic remote control in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the medium-sized rectangular black plastic remote control in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0030_01/00000.jpg", "scene0030_01/00040.jpg", "scene0030_01/00080.jpg", "scene0030_01/00120.jpg", "scene0030_01/00160.jpg" ], "turn_id": 1, "scan_id": "scene0030_01", "origin_question": "Remember, how many coat hanger(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many coat hanger(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0030_01/00200.jpg", "scene0030_01/00240.jpg", "scene0030_01/00280.jpg", "scene0030_01/00320.jpg", "scene0030_01/00360.jpg" ], "turn_id": 2, "scan_id": "scene0030_01", "origin_question": "Remember, how many projector(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many projector(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0030_01/00400.jpg", "scene0030_01/00440.jpg", "scene0030_01/00480.jpg", "scene0030_01/00520.jpg", "scene0030_01/00560.jpg" ], "turn_id": 3, "scan_id": "scene0030_01", "origin_question": "What will be the first-time appearance order of the following categories: book, chair, stapler?", "option": [ "chair, stapler, book", "book, chair, stapler", "chair, book, stapler", "stapler, book, chair" ], "answer": "chair, book, stapler", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: book, chair, stapler? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['chair, stapler, book', 'book, chair, stapler', 'chair, book, stapler', 'stapler, book, chair'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0030_01/00600.jpg", "scene0030_01/00640.jpg", "scene0030_01/00680.jpg", "scene0030_01/00720.jpg", "scene0030_01/00760.jpg" ], "turn_id": 4, "scan_id": "scene0030_01", "origin_question": "What will be the first-time appearance order of the following categories: book, stapler, table?", "option": [ "book, stapler, table", "table, book, stapler", "stapler, table, book", "table, stapler, book" ], "answer": "table, book, stapler", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: book, stapler, table? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['book, stapler, table', 'table, book, stapler', 'stapler, table, book', 'table, stapler, book'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0030_01/00800.jpg", "scene0030_01/00840.jpg", "scene0030_01/00880.jpg", "scene0030_01/00920.jpg", "scene0030_01/00960.jpg" ], "turn_id": 5, "scan_id": "scene0030_01", "origin_question": "In which turn were you closest to \"the windowsill\"? ", "option": [], "answer": "5", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the windowsill\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0030_01/01000.jpg", "scene0030_01/01040.jpg", "scene0030_01/01080.jpg", "scene0030_01/01120.jpg", "scene0030_01/01160.jpg" ], "turn_id": 6, "scan_id": "scene0030_01", "origin_question": "Please recall \"the brown rectangular paper bag with a red logo\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.21", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the brown rectangular paper bag with a red logo\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0030_01/01200.jpg", "scene0030_01/01240.jpg", "scene0030_01/01280.jpg", "scene0030_01/01320.jpg", "scene0030_01/01360.jpg" ], "turn_id": 7, "scan_id": "scene0030_01", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the brown rectangular paper bag with a red logo\" from your position?", "option": [], "answer": "154.82", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the brown rectangular paper bag with a red logo\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0030_01/01400.jpg", "scene0030_01/01440.jpg", "scene0030_01/01480.jpg", "scene0030_01/01520.jpg", "scene0030_01/01560.jpg" ], "turn_id": 8, "scan_id": "scene0030_01", "origin_question": "In which turn did you see both \"the windowsill\" and \"the white rectangular door made of wood or composite\" simultaneously? ", "option": [], "answer": "5", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the windowsill\" and \"the white rectangular door made of wood or composite\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0044_region41/368d034f98ce4de7a03c5c31b5792680_i1_0.jpg", "1mp3d_0044_region41/368d034f98ce4de7a03c5c31b5792680_i2_0.jpg", "1mp3d_0044_region41/7491e2abb84849aab17c063a2f4902b3_i1_1.jpg", "1mp3d_0044_region41/368d034f98ce4de7a03c5c31b5792680_i1_1.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0044_region41", "origin_question": "", "option": [], "answer": "", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 2 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "1mp3d_0044_region41/368d034f98ce4de7a03c5c31b5792680_i2_1.jpg", "1mp3d_0044_region41/7491e2abb84849aab17c063a2f4902b3_i1_2.jpg", "1mp3d_0044_region41/368d034f98ce4de7a03c5c31b5792680_i1_5.jpg", "1mp3d_0044_region41/368d034f98ce4de7a03c5c31b5792680_i2_5.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0044_region41", "origin_question": "When did you first discover \"the solitary plant in a large round brown pot\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the solitary plant in a large round brown pot\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 2 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0111_02/00000.jpg", "scene0111_02/00040.jpg", "scene0111_02/00080.jpg", "scene0111_02/00120.jpg", "scene0111_02/00160.jpg" ], "turn_id": 1, "scan_id": "scene0111_02", "origin_question": "Remember, have you seen any pen(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any pen(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0111_02/00200.jpg", "scene0111_02/00240.jpg", "scene0111_02/00280.jpg", "scene0111_02/00320.jpg", "scene0111_02/00360.jpg" ], "turn_id": 2, "scan_id": "scene0111_02", "origin_question": "Which two objects are on the same side of you now? \"the sleek, black rectangular microwave with silver trim\", \"the sink\", \"the transparent glass bottle with a brown cap\". ", "option": [ "the sleek, black rectangular microwave with silver trim, the sink", "the sleek, black rectangular microwave with silver trim, the transparent glass bottle with a brown cap", "the sink, the transparent glass bottle with a brown cap" ], "answer": "the sleek, black rectangular microwave with silver trim, the sink", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the sleek, black rectangular microwave with silver trim\", \"the sink\", \"the transparent glass bottle with a brown cap\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the sleek, black rectangular microwave with silver trim, the sink', 'the sleek, black rectangular microwave with silver trim, the transparent glass bottle with a brown cap', 'the sink, the transparent glass bottle with a brown cap'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0111_02/00400.jpg", "scene0111_02/00440.jpg", "scene0111_02/00480.jpg", "scene0111_02/00520.jpg", "scene0111_02/00560.jpg" ], "turn_id": 3, "scan_id": "scene0111_02", "origin_question": "In which turn did you see both \"the small white rectangular range hood made of plastic or metal\" and \"the transparent glass bottle with a brown cap\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the small white rectangular range hood made of plastic or metal\" and \"the transparent glass bottle with a brown cap\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "scene0111_02/00600.jpg", "scene0111_02/00640.jpg", "scene0111_02/00710.jpg", "scene0111_02/00750.jpg", "scene0111_02/00790.jpg" ], "turn_id": 4, "scan_id": "scene0111_02", "origin_question": "Among these three objects(\"the white rectangular stove with four burners\",\"the small white rectangular range hood made of plastic or metal\",\"the transparent glass bottle with a brown cap\" ), which one is the closest to you?", "option": [ "the white rectangular stove with four burners", "the small white rectangular range hood made of plastic or metal", "the transparent glass bottle with a brown cap" ], "answer": "the transparent glass bottle with a brown cap", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the white rectangular stove with four burners\",\"the small white rectangular range hood made of plastic or metal\",\"the transparent glass bottle with a brown cap\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the white rectangular stove with four burners', 'the small white rectangular range hood made of plastic or metal', 'the transparent glass bottle with a brown cap'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0111_02/00830.jpg", "scene0111_02/00870.jpg", "scene0111_02/00910.jpg", "scene0111_02/00950.jpg", "scene0111_02/00990.jpg" ], "turn_id": 5, "scan_id": "scene0111_02", "origin_question": "In which turn were you farthest from \"the stand\"? ", "option": [], "answer": "5", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the stand\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0111_02/01030.jpg", "scene0111_02/01070.jpg", "scene0111_02/01110.jpg", "scene0111_02/01150.jpg", "scene0111_02/01190.jpg" ], "turn_id": 6, "scan_id": "scene0111_02", "origin_question": "Which direction is the stand in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the stand in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0111_02/01230.jpg", "scene0111_02/01270.jpg", "scene0111_02/01310.jpg", "scene0111_02/01350.jpg", "scene0111_02/01390.jpg" ], "turn_id": 7, "scan_id": "scene0111_02", "origin_question": "When did you first discover \"the rectangular light brown wooden table\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the rectangular light brown wooden table\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0111_02/01430.jpg", "scene0111_02/01470.jpg", "scene0111_02/01510.jpg", "scene0111_02/01550.jpg", "scene0111_02/01590.jpg" ], "turn_id": 8, "scan_id": "scene0111_02", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the scale\" from your position?", "option": [], "answer": "35.26", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the scale\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0111_02/01630.jpg", "scene0111_02/01670.jpg", "scene0111_02/01710.jpg", "scene0111_02/01750.jpg", "scene0111_02/01790.jpg" ], "turn_id": 9, "scan_id": "scene0111_02", "origin_question": "What will be the first-time appearance order of the following categories: picture, stool, tv?", "option": [ "tv, stool, picture", "stool, tv, picture", "tv, picture, stool", "picture, stool, tv" ], "answer": "tv, picture, stool", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: picture, stool, tv? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['tv, stool, picture', 'stool, tv, picture', 'tv, picture, stool', 'picture, stool, tv'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "scene0111_02/01830.jpg", "scene0111_02/01870.jpg", "scene0111_02/01910.jpg", "scene0111_02/01950.jpg", "scene0111_02/01990.jpg" ], "turn_id": 10, "scan_id": "scene0111_02", "origin_question": "Among these three objects(\"the couch\",\"the rectangular light brown wooden table\",\"the small white rectangular range hood made of plastic or metal\" ), which one is the farthest from you?", "option": [ "the couch", "the rectangular light brown wooden table", "the small white rectangular range hood made of plastic or metal" ], "answer": "the couch", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the couch\",\"the rectangular light brown wooden table\",\"the small white rectangular range hood made of plastic or metal\" ), which one is the farthest from you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the couch', 'the rectangular light brown wooden table', 'the small white rectangular range hood made of plastic or metal'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0040_region60/880fada214cb4df7bd77378c1d9d6fcd_i1_4.jpg", "1mp3d_0040_region60/880fada214cb4df7bd77378c1d9d6fcd_i2_4.jpg", "1mp3d_0040_region60/391f8e2041304ceb9d090f0f49a8441d_i1_2.jpg", "1mp3d_0040_region60/391f8e2041304ceb9d090f0f49a8441d_i2_2.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0040_region60", "origin_question": "Remember, have you seen any kitchenware(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any kitchenware(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "1mp3d_0040_region60/01eccced6298441a94543957fff684cc_i1_1.jpg", "1mp3d_0040_region60/01eccced6298441a94543957fff684cc_i2_1.jpg", "1mp3d_0040_region60/bf12f8713f7d44a9b2d4dee1c8d43ba7_i1_0.jpg", "1mp3d_0040_region60/bf12f8713f7d44a9b2d4dee1c8d43ba7_i2_0.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0040_region60", "origin_question": "Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance forward or backward from that position?", "option": [ "Forward", "Backward" ], "answer": "Backward", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance forward or backward from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Forward', 'Backward'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0040_region60/bf12f8713f7d44a9b2d4dee1c8d43ba7_i1_1.jpg", "1mp3d_0040_region60/bf12f8713f7d44a9b2d4dee1c8d43ba7_i2_1.jpg", "1mp3d_0040_region60/01eccced6298441a94543957fff684cc_i1_2.jpg", "1mp3d_0040_region60/01eccced6298441a94543957fff684cc_i2_2.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0040_region60", "origin_question": "", "option": [], "answer": "", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0040_region60/391f8e2041304ceb9d090f0f49a8441d_i1_4.jpg", "1mp3d_0040_region60/391f8e2041304ceb9d090f0f49a8441d_i2_4.jpg", "1mp3d_0040_region60/880fada214cb4df7bd77378c1d9d6fcd_i1_5.jpg", "1mp3d_0040_region60/880fada214cb4df7bd77378c1d9d6fcd_i2_5.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0040_region60", "origin_question": "", "option": [], "answer": "", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "1mp3d_0040_region60/391f8e2041304ceb9d090f0f49a8441d_i1_3.jpg", "1mp3d_0040_region60/391f8e2041304ceb9d090f0f49a8441d_i2_3.jpg", "1mp3d_0040_region60/391f8e2041304ceb9d090f0f49a8441d_i1_5.jpg", "1mp3d_0040_region60/391f8e2041304ceb9d090f0f49a8441d_i2_5.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0040_region60", "origin_question": "In which turn did you see both \"the picture\" and \"the door\" simultaneously? ", "option": [], "answer": "5", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the picture\" and \"the door\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "1mp3d_0040_region60/01eccced6298441a94543957fff684cc_i1_3.jpg", "1mp3d_0040_region60/01eccced6298441a94543957fff684cc_i2_3.jpg", "1mp3d_0040_region60/bf12f8713f7d44a9b2d4dee1c8d43ba7_i1_5.jpg", "1mp3d_0040_region60/bf12f8713f7d44a9b2d4dee1c8d43ba7_i2_5.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0040_region60", "origin_question": "At the end of which turn were both of \"the glossy black flat-screen TV with rounded corners\" and \"the picture\" on your right side?", "option": [], "answer": "6", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the glossy black flat-screen TV with rounded corners\" and \"the picture\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0017_01/00000.jpg", "scene0017_01/00040.jpg", "scene0017_01/00080.jpg", "scene0017_01/00120.jpg" ], "turn_id": 1, "scan_id": "scene0017_01", "origin_question": "Remember, how many keyboard(s) have you seen so far? ", "option": [], "answer": "2", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many keyboard(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0017_01/00160.jpg", "scene0017_01/00200.jpg", "scene0017_01/00240.jpg", "scene0017_01/00280.jpg" ], "turn_id": 2, "scan_id": "scene0017_01", "origin_question": "At the end of which turn were both of \"the rectangular light-colored wooden desk\" and \"the windowsill\" on your left side?", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the rectangular light-colored wooden desk\" and \"the windowsill\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0017_01/00320.jpg", "scene0017_01/00360.jpg", "scene0017_01/00400.jpg", "scene0017_01/00440.jpg" ], "turn_id": 3, "scan_id": "scene0017_01", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the small rectangular white plastic switch\" from your position?", "option": [], "answer": "147.35", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the small rectangular white plastic switch\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0017_01/00480.jpg", "scene0017_01/00520.jpg", "scene0017_01/00560.jpg", "scene0017_01/00600.jpg" ], "turn_id": 4, "scan_id": "scene0017_01", "origin_question": "At the end of which turn were both of \"the small rectangular white plastic switch\" and \"the small white corded telephone made of plastic\" on your left side?", "option": [], "answer": "3", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the small rectangular white plastic switch\" and \"the small white corded telephone made of plastic\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0017_01/00640.jpg", "scene0017_01/00680.jpg", "scene0017_01/00720.jpg", "scene0017_01/00760.jpg" ], "turn_id": 5, "scan_id": "scene0017_01", "origin_question": "At the end of which turn were both of \"the small white corded telephone made of plastic\" and \"the round black metal fan\" on your left side?", "option": [], "answer": "1", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the small white corded telephone made of plastic\" and \"the round black metal fan\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0017_01/00800.jpg", "scene0017_01/00840.jpg", "scene0017_01/00880.jpg", "scene0017_01/00920.jpg" ], "turn_id": 6, "scan_id": "scene0017_01", "origin_question": "", "option": [], "answer": "", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0017_01/00960.jpg", "scene0017_01/01000.jpg", "scene0017_01/01040.jpg", "scene0017_01/01080.jpg" ], "turn_id": 7, "scan_id": "scene0017_01", "origin_question": "When did you first discover \"the blue wooden door with a glass panel\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the blue wooden door with a glass panel\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0017_01/01120.jpg", "scene0017_01/01160.jpg", "scene0017_01/01200.jpg", "scene0017_01/01240.jpg" ], "turn_id": 8, "scan_id": "scene0017_01", "origin_question": "Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0046_02/00000.jpg", "scene0046_02/00040.jpg", "scene0046_02/00080.jpg", "scene0046_02/00120.jpg", "scene0046_02/00160.jpg" ], "turn_id": 1, "scan_id": "scene0046_02", "origin_question": "Remember, have you seen any faucet(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any faucet(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0046_02/00200.jpg", "scene0046_02/00240.jpg", "scene0046_02/00280.jpg", "scene0046_02/00320.jpg", "scene0046_02/00360.jpg" ], "turn_id": 2, "scan_id": "scene0046_02", "origin_question": "Which direction is the black standard laptop made of plastic in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the black standard laptop made of plastic in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0046_02/00400.jpg", "scene0046_02/00440.jpg", "scene0046_02/00480.jpg", "scene0046_02/00520.jpg", "scene0046_02/00560.jpg" ], "turn_id": 3, "scan_id": "scene0046_02", "origin_question": "Please recall \"the small light brown wooden stand\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.83", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the small light brown wooden stand\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0046_02/00600.jpg", "scene0046_02/00640.jpg", "scene0046_02/00680.jpg", "scene0046_02/00720.jpg", "scene0046_02/00760.jpg" ], "turn_id": 4, "scan_id": "scene0046_02", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the medium black plastic telephone with a standard design\" from your position?", "option": [], "answer": "151.53", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the medium black plastic telephone with a standard design\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0046_02/00800.jpg", "scene0046_02/00840.jpg", "scene0046_02/00880.jpg", "scene0046_02/00920.jpg", "scene0046_02/00960.jpg" ], "turn_id": 5, "scan_id": "scene0046_02", "origin_question": "What will be the first-time appearance order of the following categories: desk, remote control, shoe?", "option": [ "shoe, remote control, desk", "remote control, shoe, desk", "desk, shoe, remote control", "shoe, desk, remote control" ], "answer": "desk, shoe, remote control", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: desk, remote control, shoe? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['shoe, remote control, desk', 'remote control, shoe, desk', 'desk, shoe, remote control', 'shoe, desk, remote control'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0046_02/01000.jpg", "scene0046_02/01040.jpg", "scene0046_02/01080.jpg", "scene0046_02/01120.jpg", "scene0046_02/01160.jpg" ], "turn_id": 6, "scan_id": "scene0046_02", "origin_question": "Is the blue and white medium bag made of sturdy material to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the blue and white medium bag made of sturdy material to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0046_02/01200.jpg", "scene0046_02/01240.jpg", "scene0046_02/01280.jpg", "scene0046_02/01320.jpg", "scene0046_02/01360.jpg" ], "turn_id": 7, "scan_id": "scene0046_02", "origin_question": "", "option": [], "answer": "", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "scene0046_02/01400.jpg", "scene0046_02/01440.jpg", "scene0046_02/01480.jpg", "scene0046_02/01520.jpg", "scene0046_02/01560.jpg" ], "turn_id": 8, "scan_id": "scene0046_02", "origin_question": "Assuming the direction you are facing at the end of the turn 3 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Left", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 3 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0046_02/01600.jpg", "scene0046_02/01640.jpg", "scene0046_02/01680.jpg", "scene0046_02/01720.jpg", "scene0046_02/01760.jpg" ], "turn_id": 9, "scan_id": "scene0046_02", "origin_question": "What will be the first-time appearance order of the following categories: bathtub, stand, toothbrush?", "option": [ "toothbrush, stand, bathtub", "bathtub, stand, toothbrush", "stand, bathtub, toothbrush", "toothbrush, bathtub, stand" ], "answer": "stand, bathtub, toothbrush", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bathtub, stand, toothbrush? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['toothbrush, stand, bathtub', 'bathtub, stand, toothbrush', 'stand, bathtub, toothbrush', 'toothbrush, bathtub, stand'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "scene0046_02/01800.jpg", "scene0046_02/01840.jpg", "scene0046_02/01880.jpg", "scene0046_02/01920.jpg", "scene0046_02/01960.jpg" ], "turn_id": 10, "scan_id": "scene0046_02", "origin_question": "Assuming the direction you are facing at the end of the turn 7 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Right", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 7 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0625_00/00000.jpg", "scene0625_00/00040.jpg", "scene0625_00/00080.jpg", "scene0625_00/00120.jpg" ], "turn_id": 1, "scan_id": "scene0625_00", "origin_question": "Remember, have you seen any toilet paper(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any toilet paper(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0625_00/00160.jpg", "scene0625_00/00200.jpg", "scene0625_00/00240.jpg", "scene0625_00/00280.jpg" ], "turn_id": 2, "scan_id": "scene0625_00", "origin_question": "In which turn were you closest to \"the white, smooth, rounded ceramic toilet\"? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the white, smooth, rounded ceramic toilet\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0625_00/00320.jpg", "scene0625_00/00360.jpg", "scene0625_00/00400.jpg", "scene0625_00/00440.jpg" ], "turn_id": 3, "scan_id": "scene0625_00", "origin_question": "Is the bin to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the bin to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0625_00/00480.jpg", "scene0625_00/00520.jpg", "scene0625_00/00560.jpg", "scene0625_00/00600.jpg" ], "turn_id": 4, "scan_id": "scene0625_00", "origin_question": "In which turn did you see both \"the white, smooth, rounded ceramic toilet\" and \"the vertical silver metal bar\" simultaneously? ", "option": [], "answer": "3", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the white, smooth, rounded ceramic toilet\" and \"the vertical silver metal bar\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0625_00/00640.jpg", "scene0625_00/00680.jpg", "scene0625_00/00720.jpg", "scene0625_00/00760.jpg" ], "turn_id": 5, "scan_id": "scene0625_00", "origin_question": "In which turn were you closest to \"the beige rectangular wall-mounted telephone\"? ", "option": [], "answer": "2", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the beige rectangular wall-mounted telephone\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0625_00/00800.jpg", "scene0625_00/00840.jpg", "scene0625_00/00880.jpg", "scene0625_00/00920.jpg" ], "turn_id": 6, "scan_id": "scene0625_00", "origin_question": "In which turn were you closest to \"the white rectangular semi-transparent curtain\"? ", "option": [], "answer": "6", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the white rectangular semi-transparent curtain\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0625_00/00960.jpg", "scene0625_00/01000.jpg", "scene0625_00/01040.jpg", "scene0625_00/01100.jpg" ], "turn_id": 7, "scan_id": "scene0625_00", "origin_question": "In which turn did you see both \"the beige rectangular wall-mounted telephone\" and \"the standard white rectangular ceramic bathtub\" simultaneously? ", "option": [], "answer": "6", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the beige rectangular wall-mounted telephone\" and \"the standard white rectangular ceramic bathtub\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0692_02/00000.jpg", "scene0692_02/00040.jpg", "scene0692_02/00080.jpg", "scene0692_02/00120.jpg" ], "turn_id": 1, "scan_id": "scene0692_02", "origin_question": "Remember, have you seen any book(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any book(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0692_02/00160.jpg", "scene0692_02/00200.jpg", "scene0692_02/00250.jpg", "scene0692_02/00290.jpg" ], "turn_id": 2, "scan_id": "scene0692_02", "origin_question": "When did you first discover \"the door\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the door\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0692_02/00330.jpg", "scene0692_02/00370.jpg", "scene0692_02/00410.jpg", "scene0692_02/00450.jpg" ], "turn_id": 3, "scan_id": "scene0692_02", "origin_question": "Remember, have you seen any shelf(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any shelf(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0692_02/00490.jpg", "scene0692_02/00530.jpg", "scene0692_02/00570.jpg", "scene0692_02/00610.jpg" ], "turn_id": 4, "scan_id": "scene0692_02", "origin_question": "", "option": [], "answer": "", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0692_02/00650.jpg", "scene0692_02/00690.jpg", "scene0692_02/00730.jpg", "scene0692_02/00770.jpg" ], "turn_id": 5, "scan_id": "scene0692_02", "origin_question": "When did you first discover \"the medium-sized lamp with a conical white shade\"? (index of the turn)", "option": [], "answer": "5", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the medium-sized lamp with a conical white shade\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0692_02/00810.jpg", "scene0692_02/00850.jpg", "scene0692_02/00890.jpg", "scene0692_02/00930.jpg" ], "turn_id": 6, "scan_id": "scene0692_02", "origin_question": "What will be the first-time appearance order of the following categories: door, light, shelf?", "option": [ "light, shelf, door", "door, light, shelf", "shelf, door, light", "door, shelf, light" ], "answer": "shelf, door, light", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: door, light, shelf? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['light, shelf, door', 'door, light, shelf', 'shelf, door, light', 'door, shelf, light'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0680_00/00000.jpg", "scene0680_00/00030.jpg", "scene0680_00/00060.jpg", "scene0680_00/00090.jpg" ], "turn_id": 1, "scan_id": "scene0680_00", "origin_question": "Remember, have you seen any chair(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any chair(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0680_00/00120.jpg", "scene0680_00/00150.jpg", "scene0680_00/00180.jpg", "scene0680_00/00210.jpg" ], "turn_id": 2, "scan_id": "scene0680_00", "origin_question": "In which turn were you closest to \"the table\"? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the table\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0680_00/00240.jpg", "scene0680_00/00270.jpg", "scene0680_00/00300.jpg", "scene0680_00/00330.jpg" ], "turn_id": 3, "scan_id": "scene0680_00", "origin_question": "At the end of which turn were both of \"the table\" and \"the large flat-panel TV with a black screen\" on your right side?", "option": [], "answer": "2", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the table\" and \"the large flat-panel TV with a black screen\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0680_00/00360.jpg", "scene0680_00/00390.jpg", "scene0680_00/00420.jpg", "scene0680_00/00450.jpg" ], "turn_id": 4, "scan_id": "scene0680_00", "origin_question": "Which direction is the table in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the table in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0680_00/00480.jpg", "scene0680_00/00510.jpg", "scene0680_00/00540.jpg", "scene0680_00/00570.jpg" ], "turn_id": 5, "scan_id": "scene0680_00", "origin_question": "At the end of which turn were both of \"the red and black sturdy backpack with multiple pockets\" and \"the large flat-panel TV with a black screen\" on your right side?", "option": [], "answer": "2", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the red and black sturdy backpack with multiple pockets\" and \"the large flat-panel TV with a black screen\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0680_00/00600.jpg", "scene0680_00/00630.jpg", "scene0680_00/00660.jpg", "scene0680_00/00690.jpg" ], "turn_id": 6, "scan_id": "scene0680_00", "origin_question": "Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "149.34", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0040_00/00000.jpg", "scene0040_00/00040.jpg", "scene0040_00/00080.jpg", "scene0040_00/00120.jpg", "scene0040_00/00160.jpg" ], "turn_id": 1, "scan_id": "scene0040_00", "origin_question": "Remember, how many monitor(s) have you seen so far? ", "option": [], "answer": "2", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many monitor(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0040_00/00200.jpg", "scene0040_00/00240.jpg", "scene0040_00/00280.jpg", "scene0040_00/00320.jpg", "scene0040_00/00360.jpg" ], "turn_id": 2, "scan_id": "scene0040_00", "origin_question": "When did you first discover \"the table\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the table\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0040_00/00400.jpg", "scene0040_00/00440.jpg", "scene0040_00/00480.jpg", "scene0040_00/00520.jpg", "scene0040_00/00560.jpg" ], "turn_id": 3, "scan_id": "scene0040_00", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the rectangular, bright orange, fluffy pillow\" from your position?", "option": [], "answer": "165.77", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the rectangular, bright orange, fluffy pillow\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0040_00/00600.jpg", "scene0040_00/00640.jpg", "scene0040_00/00680.jpg", "scene0040_00/00720.jpg", "scene0040_00/00760.jpg" ], "turn_id": 4, "scan_id": "scene0040_00", "origin_question": "Which direction is the table in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the table in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "scene0040_00/00800.jpg", "scene0040_00/00840.jpg", "scene0040_00/00880.jpg", "scene0040_00/00920.jpg", "scene0040_00/00960.jpg" ], "turn_id": 5, "scan_id": "scene0040_00", "origin_question": "Among these three objects(\"the large white plush toy with a round head\",\"the round, colorful plastic clock\",\"the table\" ), which one is the closest to you?", "option": [ "the large white plush toy with a round head", "the round, colorful plastic clock", "the table" ], "answer": "the table", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the large white plush toy with a round head\",\"the round, colorful plastic clock\",\"the table\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the large white plush toy with a round head', 'the round, colorful plastic clock', 'the table'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0040_00/01000.jpg", "scene0040_00/01040.jpg", "scene0040_00/01080.jpg", "scene0040_00/01120.jpg", "scene0040_00/01160.jpg" ], "turn_id": 6, "scan_id": "scene0040_00", "origin_question": "Which direction is the clear cylindrical glass container with a smooth surface in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the clear cylindrical glass container with a smooth surface in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0040_00/01200.jpg", "scene0040_00/01240.jpg", "scene0040_00/01280.jpg", "scene0040_00/01320.jpg", "scene0040_00/01360.jpg" ], "turn_id": 7, "scan_id": "scene0040_00", "origin_question": "Using your orientation at the end of turn 6 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 6 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0040_00/01400.jpg", "scene0040_00/01440.jpg", "scene0040_00/01480.jpg", "scene0040_00/01520.jpg", "scene0040_00/01560.jpg" ], "turn_id": 8, "scan_id": "scene0040_00", "origin_question": "Please recall \"the table\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.1", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the table\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0040_00/01600.jpg", "scene0040_00/01640.jpg", "scene0040_00/01680.jpg", "scene0040_00/01720.jpg", "scene0040_00/01760.jpg" ], "turn_id": 9, "scan_id": "scene0040_00", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the round, colorful plastic clock\"; \"the blue rectangular soft pad made of foam\"; \"the large brown corrugated cardboard tube\" ", "option": [ "the round, colorful plastic clock", "the blue rectangular soft pad made of foam", "the large brown corrugated cardboard tube" ], "answer": "the large brown corrugated cardboard tube", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the round, colorful plastic clock\"; \"the blue rectangular soft pad made of foam\"; \"the large brown corrugated cardboard tube\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the round, colorful plastic clock', 'the blue rectangular soft pad made of foam', 'the large brown corrugated cardboard tube'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0040_00/01800.jpg", "scene0040_00/01840.jpg", "scene0040_00/01880.jpg", "scene0040_00/01920.jpg", "scene0040_00/01960.jpg" ], "turn_id": 10, "scan_id": "scene0040_00", "origin_question": "How far is your current position from where you were at the end of the turn 5? (in meters)", "option": [], "answer": "2.19", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 5? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0435_00/00000.jpg", "scene0435_00/00040.jpg", "scene0435_00/00080.jpg", "scene0435_00/00120.jpg", "scene0435_00/00160.jpg" ], "turn_id": 1, "scan_id": "scene0435_00", "origin_question": "Remember, have you seen any excercise equipment(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any excercise equipment(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0435_00/00200.jpg", "scene0435_00/00240.jpg", "scene0435_00/00280.jpg", "scene0435_00/00320.jpg", "scene0435_00/00360.jpg" ], "turn_id": 2, "scan_id": "scene0435_00", "origin_question": "Remember, have you seen any bed(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any bed(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0435_00/00400.jpg", "scene0435_00/00440.jpg", "scene0435_00/00480.jpg", "scene0435_00/00520.jpg", "scene0435_00/00560.jpg" ], "turn_id": 3, "scan_id": "scene0435_00", "origin_question": "Which two objects are on the same side of you now? \"the small rectangular dark wooden desk for work\", \"the bin\", \"the light brown rectangular fabric couch with a curved backrest\". ", "option": [ "the small rectangular dark wooden desk for work, the bin", "the small rectangular dark wooden desk for work, the light brown rectangular fabric couch with a curved backrest", "the bin, the light brown rectangular fabric couch with a curved backrest" ], "answer": "the small rectangular dark wooden desk for work, the bin", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the small rectangular dark wooden desk for work\", \"the bin\", \"the light brown rectangular fabric couch with a curved backrest\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the small rectangular dark wooden desk for work, the bin', 'the small rectangular dark wooden desk for work, the light brown rectangular fabric couch with a curved backrest', 'the bin, the light brown rectangular fabric couch with a curved backrest'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0435_00/00600.jpg", "scene0435_00/00640.jpg", "scene0435_00/00680.jpg", "scene0435_00/00720.jpg", "scene0435_00/00760.jpg" ], "turn_id": 4, "scan_id": "scene0435_00", "origin_question": "Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the light brown rectangular fabric couch with a curved backrest\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the light brown rectangular fabric couch with a curved backrest\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0435_00/00800.jpg", "scene0435_00/00840.jpg", "scene0435_00/00880.jpg", "scene0435_00/00920.jpg", "scene0435_00/00960.jpg" ], "turn_id": 5, "scan_id": "scene0435_00", "origin_question": "Which two objects are on the same side of you now? \"the modern office chair with a black mesh backrest\", \"the light brown rectangular fabric couch with a curved backrest\", \"the blue cotton shirt casually draped over a chair\". ", "option": [ "the modern office chair with a black mesh backrest, the light brown rectangular fabric couch with a curved backrest", "the modern office chair with a black mesh backrest, the blue cotton shirt casually draped over a chair", "the light brown rectangular fabric couch with a curved backrest, the blue cotton shirt casually draped over a chair" ], "answer": "the light brown rectangular fabric couch with a curved backrest, the blue cotton shirt casually draped over a chair", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the modern office chair with a black mesh backrest\", \"the light brown rectangular fabric couch with a curved backrest\", \"the blue cotton shirt casually draped over a chair\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the modern office chair with a black mesh backrest, the light brown rectangular fabric couch with a curved backrest', 'the modern office chair with a black mesh backrest, the blue cotton shirt casually draped over a chair', 'the light brown rectangular fabric couch with a curved backrest, the blue cotton shirt casually draped over a chair'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0435_00/01000.jpg", "scene0435_00/01040.jpg", "scene0435_00/01080.jpg", "scene0435_00/01120.jpg", "scene0435_00/01160.jpg" ], "turn_id": 6, "scan_id": "scene0435_00", "origin_question": "How far is your current position from where you were at the end of the turn 4? (in meters)", "option": [], "answer": "2.26", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 4? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0435_00/01200.jpg", "scene0435_00/01240.jpg", "scene0435_00/01280.jpg", "scene0435_00/01320.jpg", "scene0435_00/01360.jpg" ], "turn_id": 7, "scan_id": "scene0435_00", "origin_question": "Which two objects are on the same side of you now? \"the small rectangular dark wooden desk for work\", \"the medium black rectangular nylon backpack\", \"the rectangular, dark-colored wooden stand\". ", "option": [ "the small rectangular dark wooden desk for work, the medium black rectangular nylon backpack", "the small rectangular dark wooden desk for work, the rectangular, dark-colored wooden stand", "the medium black rectangular nylon backpack, the rectangular, dark-colored wooden stand" ], "answer": "the medium black rectangular nylon backpack, the rectangular, dark-colored wooden stand", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the small rectangular dark wooden desk for work\", \"the medium black rectangular nylon backpack\", \"the rectangular, dark-colored wooden stand\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the small rectangular dark wooden desk for work, the medium black rectangular nylon backpack', 'the small rectangular dark wooden desk for work, the rectangular, dark-colored wooden stand', 'the medium black rectangular nylon backpack, the rectangular, dark-colored wooden stand'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0435_00/01400.jpg", "scene0435_00/01440.jpg", "scene0435_00/01480.jpg", "scene0435_00/01520.jpg", "scene0435_00/01560.jpg" ], "turn_id": 8, "scan_id": "scene0435_00", "origin_question": "In which turn did you see both \"the rectangular, dark-colored wooden stand\" and \"the light brown rectangular wooden door\" simultaneously? ", "option": [], "answer": "8", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the rectangular, dark-colored wooden stand\" and \"the light brown rectangular wooden door\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0435_00/01600.jpg", "scene0435_00/01640.jpg", "scene0435_00/01680.jpg", "scene0435_00/01720.jpg", "scene0435_00/01760.jpg" ], "turn_id": 9, "scan_id": "scene0435_00", "origin_question": "Compared to your position at the end of the 6 turn, are you now closer or farther away from \"the medium-sized rectangular wooden cabinet with a deep brown coating\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 6 turn, are you now closer or farther away from \"the medium-sized rectangular wooden cabinet with a deep brown coating\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0435_00/01800.jpg", "scene0435_00/01840.jpg", "scene0435_00/01880.jpg", "scene0435_00/01920.jpg", "scene0435_00/01960.jpg" ], "turn_id": 10, "scan_id": "scene0435_00", "origin_question": "Remember, have you seen any kitchenware(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any kitchenware(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "1mp3d_0040_region38/e3fd312ce2a2403bb5c3e3fed63bc42f_i1_5.jpg", "1mp3d_0040_region38/e3fd312ce2a2403bb5c3e3fed63bc42f_i2_5.jpg", "1mp3d_0040_region38/9866369b1ec044b7817222f3568a362a_i1_4.jpg", "1mp3d_0040_region38/9866369b1ec044b7817222f3568a362a_i2_4.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0040_region38", "origin_question": "Remember, how many cosmetics(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many cosmetics(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "1mp3d_0040_region38/e3fd312ce2a2403bb5c3e3fed63bc42f_i1_4.jpg", "1mp3d_0040_region38/e3fd312ce2a2403bb5c3e3fed63bc42f_i2_4.jpg", "1mp3d_0040_region38/e3fd312ce2a2403bb5c3e3fed63bc42f_i1_0.jpg", "1mp3d_0040_region38/e3fd312ce2a2403bb5c3e3fed63bc42f_i2_0.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0040_region38", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular beige stone fireplace with a speckled appearance\", Y: \"the brown wooden door with glass panels\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "farther away from both X and Y", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular beige stone fireplace with a speckled appearance\", Y: \"the brown wooden door with glass panels\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "1mp3d_0040_region38/9866369b1ec044b7817222f3568a362a_i1_5.jpg", "1mp3d_0040_region38/9866369b1ec044b7817222f3568a362a_i2_5.jpg", "1mp3d_0040_region38/dc30c1404c44455ba1f72a0cd51d536c_i1_5.jpg", "1mp3d_0040_region38/dc30c1404c44455ba1f72a0cd51d536c_i2_5.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0040_region38", "origin_question": "When did you first discover \"the tall, lush plant in a black flower pot\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the tall, lush plant in a black flower pot\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "1mp3d_0040_region38/dc30c1404c44455ba1f72a0cd51d536c_i1_4.jpg", "1mp3d_0040_region38/dc30c1404c44455ba1f72a0cd51d536c_i2_4.jpg", "1mp3d_0040_region38/9866369b1ec044b7817222f3568a362a_i1_0.jpg", "1mp3d_0040_region38/9866369b1ec044b7817222f3568a362a_i2_0.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0040_region38", "origin_question": "Which two objects are on the same side of you now? \"the rectangular beige stone fireplace with a speckled appearance\", \"the large wooden table with a glass top\", \"the brown wooden door with glass panels\". ", "option": [ "the rectangular beige stone fireplace with a speckled appearance, the large wooden table with a glass top", "the rectangular beige stone fireplace with a speckled appearance, the brown wooden door with glass panels", "the large wooden table with a glass top, the brown wooden door with glass panels" ], "answer": "the rectangular beige stone fireplace with a speckled appearance, the large wooden table with a glass top", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the rectangular beige stone fireplace with a speckled appearance\", \"the large wooden table with a glass top\", \"the brown wooden door with glass panels\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular beige stone fireplace with a speckled appearance, the large wooden table with a glass top', 'the rectangular beige stone fireplace with a speckled appearance, the brown wooden door with glass panels', 'the large wooden table with a glass top, the brown wooden door with glass panels'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "1mp3d_0040_region38/dc30c1404c44455ba1f72a0cd51d536c_i1_3.jpg", "1mp3d_0040_region38/dc30c1404c44455ba1f72a0cd51d536c_i2_3.jpg", "1mp3d_0040_region38/dc30c1404c44455ba1f72a0cd51d536c_i1_2.jpg", "1mp3d_0040_region38/dc30c1404c44455ba1f72a0cd51d536c_i2_2.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0040_region38", "origin_question": "Is the tall, lush plant in a black flower pot to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the tall, lush plant in a black flower pot to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "1mp3d_0040_region38/dc30c1404c44455ba1f72a0cd51d536c_i1_1.jpg", "1mp3d_0040_region38/dc30c1404c44455ba1f72a0cd51d536c_i2_1.jpg", "1mp3d_0040_region38/dc30c1404c44455ba1f72a0cd51d536c_i1_0.jpg", "1mp3d_0040_region38/dc30c1404c44455ba1f72a0cd51d536c_i2_0.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0040_region38", "origin_question": "In which turn did you see both \"the large, ornate metal and glass chandelier\" and \"the tall, lush plant in a black flower pot\" simultaneously? ", "option": [], "answer": "3", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the large, ornate metal and glass chandelier\" and \"the tall, lush plant in a black flower pot\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0072_region9/c6ade0195d9e4ca0abbb20fb0eb2fb66_i1_2.jpg", "1mp3d_0072_region9/c6ade0195d9e4ca0abbb20fb0eb2fb66_i2_2.jpg", "1mp3d_0072_region9/c6ade0195d9e4ca0abbb20fb0eb2fb66_i1_3.jpg", "1mp3d_0072_region9/c6ade0195d9e4ca0abbb20fb0eb2fb66_i2_3.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0072_region9", "origin_question": "Remember, have you seen any bench(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any bench(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 1 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0047_region9/be6ba50cd124475583f264066eff5e90_i1_1.jpg", "1mp3d_0047_region9/be6ba50cd124475583f264066eff5e90_i2_1.jpg", "1mp3d_0047_region9/16bc4c4f0681469bb55147cef79a2346_i1_1.jpg", "1mp3d_0047_region9/16bc4c4f0681469bb55147cef79a2346_i2_1.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0047_region9", "origin_question": "Remember, have you seen any sink(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any sink(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 2 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "1mp3d_0047_region9/16bc4c4f0681469bb55147cef79a2346_i1_0.jpg", "1mp3d_0047_region9/16bc4c4f0681469bb55147cef79a2346_i2_0.jpg", "1mp3d_0047_region9/be6ba50cd124475583f264066eff5e90_i1_5.jpg", "1mp3d_0047_region9/be6ba50cd124475583f264066eff5e90_i2_5.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0047_region9", "origin_question": "Which two objects are on the same side of you now? \"the small rectangular shiny metal cabinet\", \"the round, shiny stainless steel sink\", \"the light\". ", "option": [ "the small rectangular shiny metal cabinet, the round, shiny stainless steel sink", "the small rectangular shiny metal cabinet, the light", "the round, shiny stainless steel sink, the light" ], "answer": "the small rectangular shiny metal cabinet, the round, shiny stainless steel sink", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the small rectangular shiny metal cabinet\", \"the round, shiny stainless steel sink\", \"the light\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the small rectangular shiny metal cabinet, the round, shiny stainless steel sink', 'the small rectangular shiny metal cabinet, the light', 'the round, shiny stainless steel sink, the light'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 2 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0510_01/00000.jpg", "scene0510_01/00040.jpg", "scene0510_01/00080.jpg", "scene0510_01/00120.jpg" ], "turn_id": 1, "scan_id": "scene0510_01", "origin_question": "Remember, have you seen any refrigerator(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any refrigerator(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0510_01/00160.jpg", "scene0510_01/00200.jpg", "scene0510_01/00240.jpg", "scene0510_01/00280.jpg" ], "turn_id": 2, "scan_id": "scene0510_01", "origin_question": "Is the modern black wall-mounted dispenser to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the modern black wall-mounted dispenser to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0510_01/00320.jpg", "scene0510_01/00360.jpg", "scene0510_01/00400.jpg", "scene0510_01/00440.jpg" ], "turn_id": 3, "scan_id": "scene0510_01", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the white cylindrical metal or plastic pipe\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the white cylindrical metal or plastic pipe\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0510_01/00480.jpg", "scene0510_01/00520.jpg", "scene0510_01/00560.jpg", "scene0510_01/00600.jpg" ], "turn_id": 4, "scan_id": "scene0510_01", "origin_question": "Which direction is the black padded jacket made of thick fabric in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the black padded jacket made of thick fabric in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0510_01/00640.jpg", "scene0510_01/00680.jpg", "scene0510_01/00720.jpg", "scene0510_01/00760.jpg" ], "turn_id": 5, "scan_id": "scene0510_01", "origin_question": "Which two objects are on the same side of you now? \"the black padded jacket made of thick fabric\", \"the light-colored, smooth cotton curtain\", \"the shelf\". ", "option": [ "the black padded jacket made of thick fabric, the light-colored, smooth cotton curtain", "the black padded jacket made of thick fabric, the shelf", "the light-colored, smooth cotton curtain, the shelf" ], "answer": "the black padded jacket made of thick fabric, the light-colored, smooth cotton curtain", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the black padded jacket made of thick fabric\", \"the light-colored, smooth cotton curtain\", \"the shelf\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the black padded jacket made of thick fabric, the light-colored, smooth cotton curtain', 'the black padded jacket made of thick fabric, the shelf', 'the light-colored, smooth cotton curtain, the shelf'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0510_01/00800.jpg", "scene0510_01/00840.jpg", "scene0510_01/00880.jpg", "scene0510_01/00920.jpg" ], "turn_id": 6, "scan_id": "scene0510_01", "origin_question": "In which turn did you see both \"the sign\" and \"the cylindrical, transparent plastic bottle with a narrow neck\" simultaneously? ", "option": [], "answer": "3", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the sign\" and \"the cylindrical, transparent plastic bottle with a narrow neck\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0510_01/00960.jpg", "scene0510_01/01000.jpg", "scene0510_01/01040.jpg", "scene0510_01/01080.jpg" ], "turn_id": 7, "scan_id": "scene0510_01", "origin_question": "How far is your current position from where you were at the end of the turn 4? (in meters)", "option": [], "answer": "1.8", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 4? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0510_01/01120.jpg", "scene0510_01/01160.jpg", "scene0510_01/01200.jpg", "scene0510_01/01240.jpg" ], "turn_id": 8, "scan_id": "scene0510_01", "origin_question": "When did you first discover \"the cylindrical, transparent plastic bottle with a narrow neck\"? (index of the turn)", "option": [], "answer": "3", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the cylindrical, transparent plastic bottle with a narrow neck\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0510_01/01280.jpg", "scene0510_01/01320.jpg", "scene0510_01/01360.jpg", "scene0510_01/01400.jpg" ], "turn_id": 9, "scan_id": "scene0510_01", "origin_question": "Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "46.07", "user_message": "For the 9 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0065_00/00000.jpg", "scene0065_00/00040.jpg", "scene0065_00/00080.jpg", "scene0065_00/00120.jpg" ], "turn_id": 1, "scan_id": "scene0065_00", "origin_question": "Remember, have you seen any towel(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any towel(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0065_00/00160.jpg", "scene0065_00/00200.jpg", "scene0065_00/00240.jpg", "scene0065_00/00280.jpg" ], "turn_id": 2, "scan_id": "scene0065_00", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the medium-sized plunger with a wooden handle and red cup\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the medium-sized plunger with a wooden handle and red cup\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0065_00/00320.jpg", "scene0065_00/00360.jpg", "scene0065_00/00400.jpg", "scene0065_00/00440.jpg" ], "turn_id": 3, "scan_id": "scene0065_00", "origin_question": "Please recall \"the white rectangular bathtub made of glossy material\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.04", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the white rectangular bathtub made of glossy material\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0065_00/00480.jpg", "scene0065_00/00520.jpg", "scene0065_00/00560.jpg", "scene0065_00/00600.jpg" ], "turn_id": 4, "scan_id": "scene0065_00", "origin_question": "Please recall \"the medium-sized plunger with a wooden handle and red cup\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.2", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the medium-sized plunger with a wooden handle and red cup\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0065_00/00640.jpg", "scene0065_00/00680.jpg", "scene0065_00/00740.jpg", "scene0065_00/00780.jpg" ], "turn_id": 5, "scan_id": "scene0065_00", "origin_question": "Which direction is the rectangular, light brown wooden cabinet with a door in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the rectangular, light brown wooden cabinet with a door in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0065_00/00820.jpg", "scene0065_00/00860.jpg", "scene0065_00/00900.jpg", "scene0065_00/00940.jpg" ], "turn_id": 6, "scan_id": "scene0065_00", "origin_question": "Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the rectangular, light brown wooden cabinet with a door\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the rectangular, light brown wooden cabinet with a door\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0361_01/00000.jpg", "scene0361_01/00030.jpg", "scene0361_01/00060.jpg", "scene0361_01/00090.jpg" ], "turn_id": 1, "scan_id": "scene0361_01", "origin_question": "Remember, have you seen any book(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any book(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0361_01/00120.jpg", "scene0361_01/00150.jpg", "scene0361_01/00180.jpg", "scene0361_01/00210.jpg" ], "turn_id": 2, "scan_id": "scene0361_01", "origin_question": "Remember, how many boots(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many boots(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0361_01/00240.jpg", "scene0361_01/00270.jpg", "scene0361_01/00300.jpg", "scene0361_01/00330.jpg" ], "turn_id": 3, "scan_id": "scene0361_01", "origin_question": "", "option": [], "answer": "", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0361_01/00360.jpg", "scene0361_01/00390.jpg", "scene0361_01/00420.jpg", "scene0361_01/00450.jpg" ], "turn_id": 4, "scan_id": "scene0361_01", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0361_01/00480.jpg", "scene0361_01/00510.jpg", "scene0361_01/00540.jpg", "scene0361_01/00570.jpg" ], "turn_id": 5, "scan_id": "scene0361_01", "origin_question": "In which turn did you see both \"the cylindrical shiny metal bin without a lid\" and \"the medium-sized reddish-brown carpet with golden patterns\" simultaneously? ", "option": [], "answer": "5", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the cylindrical shiny metal bin without a lid\" and \"the medium-sized reddish-brown carpet with golden patterns\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0361_01/00600.jpg", "scene0361_01/00630.jpg", "scene0361_01/00660.jpg", "scene0361_01/00690.jpg" ], "turn_id": 6, "scan_id": "scene0361_01", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the medium-sized reddish-brown carpet with golden patterns\" from your position?", "option": [], "answer": "166.44", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the medium-sized reddish-brown carpet with golden patterns\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0361_01/00720.jpg", "scene0361_01/00750.jpg", "scene0361_01/00780.jpg", "scene0361_01/00810.jpg" ], "turn_id": 7, "scan_id": "scene0361_01", "origin_question": "In which turn did you see both \"the chair\" and \"the medium-sized reddish-brown carpet with golden patterns\" simultaneously? ", "option": [], "answer": "4", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the chair\" and \"the medium-sized reddish-brown carpet with golden patterns\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0361_01/00840.jpg", "scene0361_01/00870.jpg", "scene0361_01/00900.jpg", "scene0361_01/00930.jpg" ], "turn_id": 8, "scan_id": "scene0361_01", "origin_question": "Please recall \"the medium-sized reddish-brown carpet with golden patterns\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.96", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the medium-sized reddish-brown carpet with golden patterns\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0448_00/00000.jpg", "scene0448_00/00040.jpg", "scene0448_00/00080.jpg", "scene0448_00/00120.jpg", "scene0448_00/00160.jpg" ], "turn_id": 1, "scan_id": "scene0448_00", "origin_question": "Remember, how many thermostat(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many thermostat(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0448_00/00200.jpg", "scene0448_00/00240.jpg", "scene0448_00/00280.jpg", "scene0448_00/00320.jpg", "scene0448_00/00360.jpg" ], "turn_id": 2, "scan_id": "scene0448_00", "origin_question": "Which two objects are on the same side of you now? \"the small round wooden table with slim curved legs\", \"the silver cylindrical metal rod\", \"the round, off-white countertop sink made of granite\". ", "option": [ "the small round wooden table with slim curved legs, the silver cylindrical metal rod", "the small round wooden table with slim curved legs, the round, off-white countertop sink made of granite", "the silver cylindrical metal rod, the round, off-white countertop sink made of granite" ], "answer": "the small round wooden table with slim curved legs, the silver cylindrical metal rod", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the small round wooden table with slim curved legs\", \"the silver cylindrical metal rod\", \"the round, off-white countertop sink made of granite\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the small round wooden table with slim curved legs, the silver cylindrical metal rod', 'the small round wooden table with slim curved legs, the round, off-white countertop sink made of granite', 'the silver cylindrical metal rod, the round, off-white countertop sink made of granite'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0448_00/00400.jpg", "scene0448_00/00440.jpg", "scene0448_00/00480.jpg", "scene0448_00/00520.jpg", "scene0448_00/00560.jpg" ], "turn_id": 3, "scan_id": "scene0448_00", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the round, off-white countertop sink made of granite\"; \"the rectangular white porcelain bathtub\"; \"the silver metal stopcock with two handles\" ", "option": [ "the round, off-white countertop sink made of granite", "the rectangular white porcelain bathtub", "the silver metal stopcock with two handles" ], "answer": "the silver metal stopcock with two handles", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the round, off-white countertop sink made of granite\"; \"the rectangular white porcelain bathtub\"; \"the silver metal stopcock with two handles\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the round, off-white countertop sink made of granite', 'the rectangular white porcelain bathtub', 'the silver metal stopcock with two handles'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0448_00/00600.jpg", "scene0448_00/00640.jpg", "scene0448_00/00680.jpg", "scene0448_00/00720.jpg", "scene0448_00/00760.jpg" ], "turn_id": 4, "scan_id": "scene0448_00", "origin_question": "Which direction is the small round wooden table with slim curved legs in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the small round wooden table with slim curved legs in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0448_00/00800.jpg", "scene0448_00/00840.jpg", "scene0448_00/00880.jpg", "scene0448_00/00920.jpg", "scene0448_00/00960.jpg" ], "turn_id": 5, "scan_id": "scene0448_00", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the blue and white rectangular toothpaste tube\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the blue and white rectangular toothpaste tube\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0448_00/01000.jpg", "scene0448_00/01040.jpg", "scene0448_00/01080.jpg", "scene0448_00/01120.jpg", "scene0448_00/01160.jpg" ], "turn_id": 6, "scan_id": "scene0448_00", "origin_question": "Using your orientation at the end of turn 5 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 5 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0448_00/01200.jpg", "scene0448_00/01240.jpg", "scene0448_00/01280.jpg", "scene0448_00/01320.jpg", "scene0448_00/01360.jpg" ], "turn_id": 7, "scan_id": "scene0448_00", "origin_question": "Which two objects are on the same side of you now? \"the small round wooden table with slim curved legs\", \"the silver cylindrical metal rod\", \"the curtain\". ", "option": [ "the small round wooden table with slim curved legs, the silver cylindrical metal rod", "the small round wooden table with slim curved legs, the curtain", "the silver cylindrical metal rod, the curtain" ], "answer": "the small round wooden table with slim curved legs, the silver cylindrical metal rod", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the small round wooden table with slim curved legs\", \"the silver cylindrical metal rod\", \"the curtain\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the small round wooden table with slim curved legs, the silver cylindrical metal rod', 'the small round wooden table with slim curved legs, the curtain', 'the silver cylindrical metal rod, the curtain'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0448_00/01400.jpg", "scene0448_00/01440.jpg", "scene0448_00/01480.jpg", "scene0448_00/01520.jpg", "scene0448_00/01560.jpg" ], "turn_id": 8, "scan_id": "scene0448_00", "origin_question": "Which two objects are on the same side of you now? \"the rectangular white porcelain bathtub\", \"the silver cylindrical metal rod\", \"the cylindrical, transparent plastic bin\". ", "option": [ "the rectangular white porcelain bathtub, the silver cylindrical metal rod", "the rectangular white porcelain bathtub, the cylindrical, transparent plastic bin", "the silver cylindrical metal rod, the cylindrical, transparent plastic bin" ], "answer": "the rectangular white porcelain bathtub, the silver cylindrical metal rod", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the rectangular white porcelain bathtub\", \"the silver cylindrical metal rod\", \"the cylindrical, transparent plastic bin\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular white porcelain bathtub, the silver cylindrical metal rod', 'the rectangular white porcelain bathtub, the cylindrical, transparent plastic bin', 'the silver cylindrical metal rod, the cylindrical, transparent plastic bin'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0448_00/01600.jpg", "scene0448_00/01640.jpg", "scene0448_00/01680.jpg", "scene0448_00/01720.jpg", "scene0448_00/01760.jpg" ], "turn_id": 9, "scan_id": "scene0448_00", "origin_question": "Which two objects are on the same side of you now? \"the silver cylindrical metal rod\", \"the window\", \"the blue and white rectangular toothpaste tube\". ", "option": [ "the silver cylindrical metal rod, the window", "the silver cylindrical metal rod, the blue and white rectangular toothpaste tube", "the window, the blue and white rectangular toothpaste tube" ], "answer": "the silver cylindrical metal rod, the window", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the silver cylindrical metal rod\", \"the window\", \"the blue and white rectangular toothpaste tube\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the silver cylindrical metal rod, the window', 'the silver cylindrical metal rod, the blue and white rectangular toothpaste tube', 'the window, the blue and white rectangular toothpaste tube'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0448_00/01800.jpg", "scene0448_00/01840.jpg", "scene0448_00/01880.jpg", "scene0448_00/01920.jpg", "scene0448_00/01960.jpg" ], "turn_id": 10, "scan_id": "scene0448_00", "origin_question": "Is the curtain to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the curtain to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0078_region14/af23bd7af9a14d7494fdd8733cb7fa5e_i1_2.jpg", "1mp3d_0078_region14/af23bd7af9a14d7494fdd8733cb7fa5e_i2_2.jpg", "1mp3d_0078_region14/ab93f352d8e341b08971ee8a7faed6e1_i1_2.jpg", "1mp3d_0078_region14/ab93f352d8e341b08971ee8a7faed6e1_i2_2.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0078_region14", "origin_question": "Remember, have you seen any heater(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any heater(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0078_region14/ab93f352d8e341b08971ee8a7faed6e1_i1_1.jpg", "1mp3d_0078_region14/ab93f352d8e341b08971ee8a7faed6e1_i2_1.jpg", "1mp3d_0078_region14/25a9232a33ad4d739ce1d6b0dc8add81_i1_1.jpg", "1mp3d_0078_region14/25a9232a33ad4d739ce1d6b0dc8add81_i2_1.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0078_region14", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0078_region14/b95c51cfbd884525b995ea4234be83c2_i1_1.jpg", "1mp3d_0078_region14/b95c51cfbd884525b995ea4234be83c2_i2_1.jpg", "1mp3d_0078_region14/07e011fca68e4d6bbda2c6083c8a134c_i1_3.jpg", "1mp3d_0078_region14/07e011fca68e4d6bbda2c6083c8a134c_i2_3.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0078_region14", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the light brown wooden decoration with a red and yellow handle\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the light brown wooden decoration with a red and yellow handle\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "1mp3d_0078_region14/9b9a964d89cf4b29b5f3bbfb4329ff65_i1_5.jpg", "1mp3d_0078_region14/9b9a964d89cf4b29b5f3bbfb4329ff65_i2_5.jpg", "1mp3d_0078_region14/c2b53ec055544828bb80579ec1f4d105_i1_1.jpg", "1mp3d_0078_region14/c2b53ec055544828bb80579ec1f4d105_i2_1.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0078_region14", "origin_question": "In which turn were you farthest from \"the light brown wooden decoration with a red and yellow handle\"? ", "option": [], "answer": "4", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the light brown wooden decoration with a red and yellow handle\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0078_region14/9b9a964d89cf4b29b5f3bbfb4329ff65_i1_4.jpg", "1mp3d_0078_region14/9b9a964d89cf4b29b5f3bbfb4329ff65_i2_4.jpg", "1mp3d_0078_region14/07e011fca68e4d6bbda2c6083c8a134c_i1_2.jpg", "1mp3d_0078_region14/07e011fca68e4d6bbda2c6083c8a134c_i2_2.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0078_region14", "origin_question": "", "option": [], "answer": "", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "1mp3d_0078_region14/b95c51cfbd884525b995ea4234be83c2_i1_0.jpg", "1mp3d_0078_region14/b95c51cfbd884525b995ea4234be83c2_i2_0.jpg", "1mp3d_0078_region14/25a9232a33ad4d739ce1d6b0dc8add81_i1_5.jpg", "1mp3d_0078_region14/25a9232a33ad4d739ce1d6b0dc8add81_i2_5.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0078_region14", "origin_question": "In which turn did you see both \"the cylindrical chair with natural wood finishing\" and \"the large rectangular light beige wooden window\" simultaneously? ", "option": [], "answer": "6", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the cylindrical chair with natural wood finishing\" and \"the large rectangular light beige wooden window\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "1mp3d_0078_region14/25a9232a33ad4d739ce1d6b0dc8add81_i1_0.jpg", "1mp3d_0078_region14/25a9232a33ad4d739ce1d6b0dc8add81_i2_0.jpg", "1mp3d_0078_region14/e5b8baa3c6474af1a2bc739d92c480e5_i1_0.jpg", "1mp3d_0078_region14/e5b8baa3c6474af1a2bc739d92c480e5_i2_0.jpg" ], "turn_id": 7, "scan_id": "1mp3d_0078_region14", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "1mp3d_0078_region14/dce1006e4277469ab535543055b169fd_i1_0.jpg", "1mp3d_0078_region14/dce1006e4277469ab535543055b169fd_i2_0.jpg", "1mp3d_0078_region14/842c473dcc3a41c5be26b8818d96db7c_i1_1.jpg", "1mp3d_0078_region14/842c473dcc3a41c5be26b8818d96db7c_i2_1.jpg" ], "turn_id": 8, "scan_id": "1mp3d_0078_region14", "origin_question": "Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "112.77", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "1mp3d_0078_region14/82e54e797e8844e289647e8da04f956a_i1_1.jpg", "1mp3d_0078_region14/82e54e797e8844e289647e8da04f956a_i2_1.jpg", "1mp3d_0078_region14/82e54e797e8844e289647e8da04f956a_i1_0.jpg", "1mp3d_0078_region14/82e54e797e8844e289647e8da04f956a_i2_0.jpg" ], "turn_id": 9, "scan_id": "1mp3d_0078_region14", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "1.89", "user_message": "For the 9 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0081_01/00000.jpg", "scene0081_01/00040.jpg", "scene0081_01/00080.jpg", "scene0081_01/00120.jpg" ], "turn_id": 1, "scan_id": "scene0081_01", "origin_question": "Remember, how many chair(s) have you seen so far? ", "option": [], "answer": "16", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many chair(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0081_01/00160.jpg", "scene0081_01/00200.jpg", "scene0081_01/00240.jpg", "scene0081_01/00280.jpg" ], "turn_id": 2, "scan_id": "scene0081_01", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0081_01/00320.jpg", "scene0081_01/00360.jpg", "scene0081_01/00400.jpg", "scene0081_01/00440.jpg" ], "turn_id": 3, "scan_id": "scene0081_01", "origin_question": "", "option": [], "answer": "", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0081_01/00480.jpg", "scene0081_01/00520.jpg", "scene0081_01/00560.jpg", "scene0081_01/00600.jpg" ], "turn_id": 4, "scan_id": "scene0081_01", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "2.17", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0081_01/00640.jpg", "scene0081_01/00680.jpg", "scene0081_01/00720.jpg", "scene0081_01/00760.jpg" ], "turn_id": 5, "scan_id": "scene0081_01", "origin_question": "Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0081_01/00800.jpg", "scene0081_01/00840.jpg", "scene0081_01/00880.jpg", "scene0081_01/00920.jpg" ], "turn_id": 6, "scan_id": "scene0081_01", "origin_question": "When did you first discover \"the white rectangular piece of thin paper\"? (index of the turn)", "option": [], "answer": "5", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the white rectangular piece of thin paper\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0081_01/00960.jpg", "scene0081_01/01000.jpg", "scene0081_01/01040.jpg", "scene0081_01/01080.jpg" ], "turn_id": 7, "scan_id": "scene0081_01", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "4.06", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0081_01/01120.jpg", "scene0081_01/01160.jpg", "scene0081_01/01200.jpg", "scene0081_01/01240.jpg" ], "turn_id": 8, "scan_id": "scene0081_01", "origin_question": "Which direction is the white rectangular piece of thin paper in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the white rectangular piece of thin paper in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0406_00/00000.jpg", "scene0406_00/00040.jpg", "scene0406_00/00080.jpg", "scene0406_00/00120.jpg" ], "turn_id": 1, "scan_id": "scene0406_00", "origin_question": "Remember, have you seen any frame(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any frame(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0406_00/00160.jpg", "scene0406_00/00200.jpg", "scene0406_00/00240.jpg", "scene0406_00/00280.jpg" ], "turn_id": 2, "scan_id": "scene0406_00", "origin_question": "Which direction is the standard roll of white toilet paper in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the standard roll of white toilet paper in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0406_00/00320.jpg", "scene0406_00/00360.jpg", "scene0406_00/00400.jpg", "scene0406_00/00440.jpg" ], "turn_id": 3, "scan_id": "scene0406_00", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "28.9", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0406_00/00480.jpg", "scene0406_00/00520.jpg", "scene0406_00/00560.jpg", "scene0406_00/00600.jpg" ], "turn_id": 4, "scan_id": "scene0406_00", "origin_question": "Please recall \"the dark silver cylindrical stainless steel bin\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.65", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the dark silver cylindrical stainless steel bin\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0406_00/00640.jpg", "scene0406_00/00680.jpg", "scene0406_00/00720.jpg", "scene0406_00/00760.jpg" ], "turn_id": 5, "scan_id": "scene0406_00", "origin_question": "Please recall \"the framed picture of tree leaves in vivid colors\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.97", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the framed picture of tree leaves in vivid colors\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0406_00/00800.jpg", "scene0406_00/00840.jpg", "scene0406_00/00880.jpg", "scene0406_00/00920.jpg" ], "turn_id": 6, "scan_id": "scene0406_00", "origin_question": "Please recall \"the framed picture of tree leaves in vivid colors\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.75", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the framed picture of tree leaves in vivid colors\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0406_00/00960.jpg", "scene0406_00/01000.jpg", "scene0406_00/01040.jpg", "scene0406_00/01080.jpg" ], "turn_id": 7, "scan_id": "scene0406_00", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the white rectangular porcelain bathtub\" from your position?", "option": [], "answer": "89.72", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the white rectangular porcelain bathtub\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0406_00/01120.jpg", "scene0406_00/01160.jpg", "scene0406_00/01200.jpg", "scene0406_00/01240.jpg" ], "turn_id": 8, "scan_id": "scene0406_00", "origin_question": "What will be the first-time appearance order of the following categories: door, picture, toilet?", "option": [ "door, toilet, picture", "picture, toilet, door", "toilet, picture, door", "picture, door, toilet" ], "answer": "toilet, picture, door", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: door, picture, toilet? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['door, toilet, picture', 'picture, toilet, door', 'toilet, picture, door', 'picture, door, toilet'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0078_region5/3c05cbcf2b30493592c04628faa3a0d7_i1_1.jpg", "1mp3d_0078_region5/3c05cbcf2b30493592c04628faa3a0d7_i2_1.jpg", "1mp3d_0078_region5/3c05cbcf2b30493592c04628faa3a0d7_i1_0.jpg", "1mp3d_0078_region5/3c05cbcf2b30493592c04628faa3a0d7_i2_0.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0078_region5", "origin_question": "Remember, have you seen any bin(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any bin(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "1mp3d_0078_region5/d61b99f8f134414085b3d2b6f46fee37_i1_3.jpg", "1mp3d_0078_region5/d61b99f8f134414085b3d2b6f46fee37_i2_3.jpg", "1mp3d_0078_region5/ca23f08d5698489da6a12607ebccecfc_i1_3.jpg", "1mp3d_0078_region5/ca23f08d5698489da6a12607ebccecfc_i2_3.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0078_region5", "origin_question": "Which direction is the cylindrical brown plastic bin in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the cylindrical brown plastic bin in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "1mp3d_0078_region5/d61b99f8f134414085b3d2b6f46fee37_i1_4.jpg", "1mp3d_0078_region5/d61b99f8f134414085b3d2b6f46fee37_i2_4.jpg", "1mp3d_0078_region5/51b17bc333674eb48d81ea7624118783_i1_3.jpg", "1mp3d_0078_region5/51b17bc333674eb48d81ea7624118783_i2_3.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0078_region5", "origin_question": "In which turn were you closest to \"the white oval claw-foot bathtub with a glossy surface\"? ", "option": [], "answer": "2", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the white oval claw-foot bathtub with a glossy surface\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "1mp3d_0078_region5/51b17bc333674eb48d81ea7624118783_i1_2.jpg", "1mp3d_0078_region5/51b17bc333674eb48d81ea7624118783_i2_2.jpg", "1mp3d_0078_region5/06d28e95541746838c132c796ece97a3_i1_2.jpg", "1mp3d_0078_region5/06d28e95541746838c132c796ece97a3_i2_2.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0078_region5", "origin_question": "How far is your current position from where you were at the end of the turn 1? (in meters)", "option": [], "answer": "2.04", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 1? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0078_region5/b95ef00660ea4ea89162e31370ba6558_i1_2.jpg", "1mp3d_0078_region5/b95ef00660ea4ea89162e31370ba6558_i2_2.jpg", "1mp3d_0078_region5/b95ef00660ea4ea89162e31370ba6558_i1_3.jpg", "1mp3d_0078_region5/b95ef00660ea4ea89162e31370ba6558_i2_3.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0078_region5", "origin_question": "", "option": [], "answer": "", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "1mp3d_0078_region5/b95ef00660ea4ea89162e31370ba6558_i2_4.jpg", "1mp3d_0078_region5/06d28e95541746838c132c796ece97a3_i1_4.jpg", "1mp3d_0078_region5/06d28e95541746838c132c796ece97a3_i2_4.jpg", "1mp3d_0078_region5/15b25adcf9bb4a0f8c751138a651dd8b_i1_4.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0078_region5", "origin_question": "Which two objects are on the same side of you now? \"the white oval claw-foot bathtub with a glossy surface\", \"the white ceramic toilet with a rectangular tank\", \"the wooden door with a rich brown, horizontal design\". ", "option": [ "the white oval claw-foot bathtub with a glossy surface, the white ceramic toilet with a rectangular tank", "the white oval claw-foot bathtub with a glossy surface, the wooden door with a rich brown, horizontal design", "the white ceramic toilet with a rectangular tank, the wooden door with a rich brown, horizontal design" ], "answer": "the white oval claw-foot bathtub with a glossy surface, the white ceramic toilet with a rectangular tank", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the white oval claw-foot bathtub with a glossy surface\", \"the white ceramic toilet with a rectangular tank\", \"the wooden door with a rich brown, horizontal design\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the white oval claw-foot bathtub with a glossy surface, the white ceramic toilet with a rectangular tank', 'the white oval claw-foot bathtub with a glossy surface, the wooden door with a rich brown, horizontal design', 'the white ceramic toilet with a rectangular tank, the wooden door with a rich brown, horizontal design'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "1mp3d_0078_region5/15b25adcf9bb4a0f8c751138a651dd8b_i2_4.jpg", "1mp3d_0078_region5/5d9879e590c74d83b038b280448c69e5_i1_5.jpg", "1mp3d_0078_region5/5d9879e590c74d83b038b280448c69e5_i2_5.jpg", "1mp3d_0078_region5/5d9879e590c74d83b038b280448c69e5_i1_0.jpg" ], "turn_id": 7, "scan_id": "1mp3d_0078_region5", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the large rectangular white wooden frame window\"; \"the white oval claw-foot bathtub with a glossy surface\"; \"the large white oval sink made of glossy material\" ", "option": [ "the large rectangular white wooden frame window", "the white oval claw-foot bathtub with a glossy surface", "the large white oval sink made of glossy material" ], "answer": "the large rectangular white wooden frame window", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the large rectangular white wooden frame window\"; \"the white oval claw-foot bathtub with a glossy surface\"; \"the large white oval sink made of glossy material\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the large rectangular white wooden frame window', 'the white oval claw-foot bathtub with a glossy surface', 'the large white oval sink made of glossy material'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "1mp3d_0078_region5/5d9879e590c74d83b038b280448c69e5_i2_0.jpg", "1mp3d_0078_region5/15b25adcf9bb4a0f8c751138a651dd8b_i1_0.jpg", "1mp3d_0078_region5/15b25adcf9bb4a0f8c751138a651dd8b_i2_0.jpg", "1mp3d_0078_region5/15b25adcf9bb4a0f8c751138a651dd8b_i1_5.jpg" ], "turn_id": 8, "scan_id": "1mp3d_0078_region5", "origin_question": "Which two objects are on the same side of you now? \"the cylindrical brown plastic bin\", \"the rectangular mirror with a dark wooden frame\", \"the wooden door with a rich brown, horizontal design\". ", "option": [ "the cylindrical brown plastic bin, the rectangular mirror with a dark wooden frame", "the cylindrical brown plastic bin, the wooden door with a rich brown, horizontal design", "the rectangular mirror with a dark wooden frame, the wooden door with a rich brown, horizontal design" ], "answer": "the cylindrical brown plastic bin, the rectangular mirror with a dark wooden frame", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the cylindrical brown plastic bin\", \"the rectangular mirror with a dark wooden frame\", \"the wooden door with a rich brown, horizontal design\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the cylindrical brown plastic bin, the rectangular mirror with a dark wooden frame', 'the cylindrical brown plastic bin, the wooden door with a rich brown, horizontal design', 'the rectangular mirror with a dark wooden frame, the wooden door with a rich brown, horizontal design'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0655_01/00000.jpg", "scene0655_01/00030.jpg", "scene0655_01/00060.jpg", "scene0655_01/00090.jpg" ], "turn_id": 1, "scan_id": "scene0655_01", "origin_question": "Remember, have you seen any chair(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any chair(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0655_01/00120.jpg", "scene0655_01/00150.jpg", "scene0655_01/00180.jpg", "scene0655_01/00210.jpg" ], "turn_id": 2, "scan_id": "scene0655_01", "origin_question": "In which turn did you see both \"the rectangular white wooden table\" and \"the blackboard\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the rectangular white wooden table\" and \"the blackboard\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0655_01/00240.jpg", "scene0655_01/00270.jpg", "scene0655_01/00300.jpg", "scene0655_01/00330.jpg" ], "turn_id": 3, "scan_id": "scene0655_01", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "63.9", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0655_01/00360.jpg", "scene0655_01/00390.jpg", "scene0655_01/00420.jpg", "scene0655_01/00450.jpg" ], "turn_id": 4, "scan_id": "scene0655_01", "origin_question": "When did you first discover \"the blackboard\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the blackboard\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0655_01/00480.jpg", "scene0655_01/00510.jpg", "scene0655_01/00540.jpg", "scene0655_01/00570.jpg" ], "turn_id": 5, "scan_id": "scene0655_01", "origin_question": "At the end of which turn were both of \"the rectangular white wooden table\" and \"the blackboard\" on your left side?", "option": [], "answer": "4", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the rectangular white wooden table\" and \"the blackboard\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0655_01/00600.jpg", "scene0655_01/00630.jpg", "scene0655_01/00660.jpg", "scene0655_01/00690.jpg" ], "turn_id": 6, "scan_id": "scene0655_01", "origin_question": "In which turn were you closest to \"the rectangular white wooden table\"? ", "option": [], "answer": "3", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the rectangular white wooden table\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0655_01/00720.jpg", "scene0655_01/00750.jpg", "scene0655_01/00780.jpg", "scene0655_01/00810.jpg" ], "turn_id": 7, "scan_id": "scene0655_01", "origin_question": "Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the blackboard\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the blackboard\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0054_00/00000.jpg", "scene0054_00/00040.jpg", "scene0054_00/00080.jpg", "scene0054_00/00120.jpg", "scene0054_00/00160.jpg" ], "turn_id": 1, "scan_id": "scene0054_00", "origin_question": "Remember, how many window(s) have you seen so far? ", "option": [], "answer": "2", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many window(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0054_00/00200.jpg", "scene0054_00/00240.jpg", "scene0054_00/00280.jpg", "scene0054_00/00320.jpg", "scene0054_00/00360.jpg" ], "turn_id": 2, "scan_id": "scene0054_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0054_00/00400.jpg", "scene0054_00/00440.jpg", "scene0054_00/00480.jpg", "scene0054_00/00520.jpg", "scene0054_00/00560.jpg" ], "turn_id": 3, "scan_id": "scene0054_00", "origin_question": "In which turn did you see both \"the sturdy rectangular notebook with a black exterior\" and \"the rectangular table with a smooth, white surface\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the sturdy rectangular notebook with a black exterior\" and \"the rectangular table with a smooth, white surface\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0054_00/00600.jpg", "scene0054_00/00640.jpg", "scene0054_00/00680.jpg", "scene0054_00/00720.jpg", "scene0054_00/00760.jpg" ], "turn_id": 4, "scan_id": "scene0054_00", "origin_question": "Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "20.83", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0054_00/00800.jpg", "scene0054_00/00840.jpg", "scene0054_00/00880.jpg", "scene0054_00/00920.jpg", "scene0054_00/00960.jpg" ], "turn_id": 5, "scan_id": "scene0054_00", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the classic lamp with a white pleated lampshade\" from your position?", "option": [], "answer": "125.35", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the classic lamp with a white pleated lampshade\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0054_00/01000.jpg", "scene0054_00/01040.jpg", "scene0054_00/01080.jpg", "scene0054_00/01120.jpg", "scene0054_00/01160.jpg" ], "turn_id": 6, "scan_id": "scene0054_00", "origin_question": "At the end of which turn were both of \"the modern black multi-functional printer with a touchscreen\" and \"the board\" on your right side?", "option": [], "answer": "1", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the modern black multi-functional printer with a touchscreen\" and \"the board\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "scene0054_00/01200.jpg", "scene0054_00/01240.jpg", "scene0054_00/01280.jpg", "scene0054_00/01320.jpg", "scene0054_00/01360.jpg" ], "turn_id": 7, "scan_id": "scene0054_00", "origin_question": "Assuming the direction you are facing at the end of the turn 6 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Right", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 6 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0054_00/01400.jpg", "scene0054_00/01440.jpg", "scene0054_00/01480.jpg", "scene0054_00/01520.jpg", "scene0054_00/01560.jpg" ], "turn_id": 8, "scan_id": "scene0054_00", "origin_question": "Is the board to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the board to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0054_00/01600.jpg", "scene0054_00/01640.jpg", "scene0054_00/01680.jpg", "scene0054_00/01720.jpg", "scene0054_00/01760.jpg" ], "turn_id": 9, "scan_id": "scene0054_00", "origin_question": "Using your orientation at the end of turn 8 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 8 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0054_00/01800.jpg", "scene0054_00/01840.jpg", "scene0054_00/01880.jpg", "scene0054_00/01920.jpg", "scene0054_00/01960.jpg" ], "turn_id": 10, "scan_id": "scene0054_00", "origin_question": "At the end of which turn were both of \"the modern black multi-functional printer with a touchscreen\" and \"the rectangular dark velvet pillow\" on your right side?", "option": [], "answer": "1", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the modern black multi-functional printer with a touchscreen\" and \"the rectangular dark velvet pillow\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0044_region10/54bad10553c44721a9b6d5bb822e1d3f_i1_2.jpg", "1mp3d_0044_region10/54bad10553c44721a9b6d5bb822e1d3f_i2_2.jpg", "1mp3d_0044_region10/63dcba1c6ee545ffb1909bf229d0dfa9_i1_1.jpg", "1mp3d_0044_region10/63dcba1c6ee545ffb1909bf229d0dfa9_i2_1.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0044_region10", "origin_question": "Remember, have you seen any window(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any window(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "1mp3d_0044_region10/e20f207faf0c43fd9f6682b651aaaae2_i1_5.jpg", "1mp3d_0044_region10/e20f207faf0c43fd9f6682b651aaaae2_i2_5.jpg", "1mp3d_0044_region10/145632964ccd4f27af3d581c0fd52e7d_i1_0.jpg", "1mp3d_0044_region10/145632964ccd4f27af3d581c0fd52e7d_i2_0.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0044_region10", "origin_question": "When did you first discover \"the rectangular glass window with a deep brown frame\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the rectangular glass window with a deep brown frame\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "1mp3d_0044_region10/145632964ccd4f27af3d581c0fd52e7d_i1_1.jpg", "1mp3d_0044_region10/145632964ccd4f27af3d581c0fd52e7d_i2_1.jpg", "1mp3d_0044_region10/e20f207faf0c43fd9f6682b651aaaae2_i1_0.jpg", "1mp3d_0044_region10/e20f207faf0c43fd9f6682b651aaaae2_i2_0.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0044_region10", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "3.48", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "1mp3d_0044_region10/63dcba1c6ee545ffb1909bf229d0dfa9_i1_2.jpg", "1mp3d_0044_region10/63dcba1c6ee545ffb1909bf229d0dfa9_i2_2.jpg", "1mp3d_0044_region10/63dcba1c6ee545ffb1909bf229d0dfa9_i1_3.jpg", "1mp3d_0044_region10/63dcba1c6ee545ffb1909bf229d0dfa9_i2_3.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0044_region10", "origin_question": "When did you first discover \"the pink metal bicycle with black seat and handlebars\"? (index of the turn)", "option": [], "answer": "3", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the pink metal bicycle with black seat and handlebars\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "1mp3d_0044_region10/63dcba1c6ee545ffb1909bf229d0dfa9_i1_4.jpg", "1mp3d_0044_region10/63dcba1c6ee545ffb1909bf229d0dfa9_i2_4.jpg", "1mp3d_0044_region10/54bad10553c44721a9b6d5bb822e1d3f_i1_4.jpg", "1mp3d_0044_region10/54bad10553c44721a9b6d5bb822e1d3f_i2_4.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0044_region10", "origin_question": "Which direction is the rectangular glass window with a deep brown frame in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the rectangular glass window with a deep brown frame in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "1mp3d_0044_region10/54bad10553c44721a9b6d5bb822e1d3f_i1_3.jpg", "1mp3d_0044_region10/54bad10553c44721a9b6d5bb822e1d3f_i2_3.jpg", "1mp3d_0044_region10/79e57827894143058552acf4e69af464_i1_2.jpg", "1mp3d_0044_region10/79e57827894143058552acf4e69af464_i2_2.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0044_region10", "origin_question": "Which direction is the standard yellow plastic bucket in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the standard yellow plastic bucket in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "1mp3d_0044_region10/79e57827894143058552acf4e69af464_i1_1.jpg", "1mp3d_0044_region10/79e57827894143058552acf4e69af464_i2_1.jpg", "1mp3d_0044_region10/d1ab0cfff0ca43c680c41e95bf65bd3e_i1_0.jpg", "1mp3d_0044_region10/d1ab0cfff0ca43c680c41e95bf65bd3e_i2_0.jpg" ], "turn_id": 7, "scan_id": "1mp3d_0044_region10", "origin_question": "What will be the first-time appearance order of the following categories: bicycle, light, window?", "option": [ "light, window, bicycle", "window, bicycle, light", "light, bicycle, window", "window, light, bicycle" ], "answer": "window, light, bicycle", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bicycle, light, window? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['light, window, bicycle', 'window, bicycle, light', 'light, bicycle, window', 'window, light, bicycle'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0044_region10/6d865c2bb466490f98ce4ca45ce3dd67_i1_4.jpg", "1mp3d_0044_region10/6d865c2bb466490f98ce4ca45ce3dd67_i2_4.jpg", "1mp3d_0044_region10/6d865c2bb466490f98ce4ca45ce3dd67_i1_3.jpg", "1mp3d_0044_region10/6d865c2bb466490f98ce4ca45ce3dd67_i2_3.jpg" ], "turn_id": 8, "scan_id": "1mp3d_0044_region10", "origin_question": "", "option": [], "answer": "", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "1mp3d_0044_region10/d1ab0cfff0ca43c680c41e95bf65bd3e_i1_4.jpg", "1mp3d_0044_region10/d1ab0cfff0ca43c680c41e95bf65bd3e_i2_4.jpg", "1mp3d_0044_region10/d1ab0cfff0ca43c680c41e95bf65bd3e_i1_3.jpg", "1mp3d_0044_region10/d1ab0cfff0ca43c680c41e95bf65bd3e_i2_3.jpg" ], "turn_id": 9, "scan_id": "1mp3d_0044_region10", "origin_question": "By the end of the 3 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular, soft, dense carpet in darker colors\", Y: \"the pink metal bicycle with black seat and handlebars\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to Y and farther away from X", "user_message": "For the 9 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 3 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular, soft, dense carpet in darker colors\", Y: \"the pink metal bicycle with black seat and handlebars\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0474_01/00000.jpg", "scene0474_01/00040.jpg", "scene0474_01/00080.jpg", "scene0474_01/00120.jpg", "scene0474_01/00160.jpg" ], "turn_id": 1, "scan_id": "scene0474_01", "origin_question": "Remember, how many kitchen island(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many kitchen island(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0474_01/00200.jpg", "scene0474_01/00240.jpg", "scene0474_01/00280.jpg", "scene0474_01/00320.jpg", "scene0474_01/00360.jpg" ], "turn_id": 2, "scan_id": "scene0474_01", "origin_question": "At the end of which turn were both of \"the small rectangular black plastic speaker\" and \"the rectangular dark gray telephone\" on your left side?", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the small rectangular black plastic speaker\" and \"the rectangular dark gray telephone\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0474_01/00400.jpg", "scene0474_01/00440.jpg", "scene0474_01/00480.jpg", "scene0474_01/00520.jpg", "scene0474_01/00560.jpg" ], "turn_id": 3, "scan_id": "scene0474_01", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the small rectangular black plastic speaker\" from your position?", "option": [], "answer": "48.35", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the small rectangular black plastic speaker\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0474_01/00600.jpg", "scene0474_01/00640.jpg", "scene0474_01/00680.jpg", "scene0474_01/00720.jpg", "scene0474_01/00760.jpg" ], "turn_id": 4, "scan_id": "scene0474_01", "origin_question": "Remember, how many statue(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many statue(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0474_01/00800.jpg", "scene0474_01/00840.jpg", "scene0474_01/00880.jpg", "scene0474_01/00920.jpg", "scene0474_01/00960.jpg" ], "turn_id": 5, "scan_id": "scene0474_01", "origin_question": "In which turn were you closest to \"the dark casual denim jacket with a tailored fit\"? ", "option": [], "answer": "3", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the dark casual denim jacket with a tailored fit\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0474_01/01000.jpg", "scene0474_01/01040.jpg", "scene0474_01/01080.jpg", "scene0474_01/01120.jpg", "scene0474_01/01160.jpg" ], "turn_id": 6, "scan_id": "scene0474_01", "origin_question": "", "option": [], "answer": "", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0474_01/01200.jpg", "scene0474_01/01240.jpg", "scene0474_01/01280.jpg", "scene0474_01/01320.jpg", "scene0474_01/01360.jpg" ], "turn_id": 7, "scan_id": "scene0474_01", "origin_question": "Using your orientation at the end of turn 5 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "104.68", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 5 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0474_01/01400.jpg", "scene0474_01/01440.jpg", "scene0474_01/01480.jpg", "scene0474_01/01520.jpg", "scene0474_01/01560.jpg" ], "turn_id": 8, "scan_id": "scene0474_01", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the brown wooden door with a silver handle\"; \"the rectangular blackboard with a smooth white surface\"; \"the brown wooden shelf with visible grains\" ", "option": [ "the brown wooden door with a silver handle", "the rectangular blackboard with a smooth white surface", "the brown wooden shelf with visible grains" ], "answer": "the brown wooden shelf with visible grains", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the brown wooden door with a silver handle\"; \"the rectangular blackboard with a smooth white surface\"; \"the brown wooden shelf with visible grains\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the brown wooden door with a silver handle', 'the rectangular blackboard with a smooth white surface', 'the brown wooden shelf with visible grains'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0474_01/01600.jpg", "scene0474_01/01640.jpg", "scene0474_01/01680.jpg", "scene0474_01/01720.jpg", "scene0474_01/01760.jpg" ], "turn_id": 9, "scan_id": "scene0474_01", "origin_question": "Which two objects are on the same side of you now? \"the brown wooden door with a silver handle\", \"the standard white ceramic cylindrical cup\", \"the dark casual denim jacket with a tailored fit\". ", "option": [ "the brown wooden door with a silver handle, the standard white ceramic cylindrical cup", "the brown wooden door with a silver handle, the dark casual denim jacket with a tailored fit", "the standard white ceramic cylindrical cup, the dark casual denim jacket with a tailored fit" ], "answer": "the brown wooden door with a silver handle, the standard white ceramic cylindrical cup", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the brown wooden door with a silver handle\", \"the standard white ceramic cylindrical cup\", \"the dark casual denim jacket with a tailored fit\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the brown wooden door with a silver handle, the standard white ceramic cylindrical cup', 'the brown wooden door with a silver handle, the dark casual denim jacket with a tailored fit', 'the standard white ceramic cylindrical cup, the dark casual denim jacket with a tailored fit'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0474_01/01800.jpg", "scene0474_01/01840.jpg", "scene0474_01/01880.jpg", "scene0474_01/01920.jpg", "scene0474_01/01960.jpg" ], "turn_id": 10, "scan_id": "scene0474_01", "origin_question": "By the end of the 6 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular dark gray telephone\", Y: \"the dark brown and black nylon backpack with pockets\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to Y and farther away from X", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 6 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular dark gray telephone\", Y: \"the dark brown and black nylon backpack with pockets\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0670_01/00000.jpg", "scene0670_01/00040.jpg", "scene0670_01/00080.jpg", "scene0670_01/00120.jpg", "scene0670_01/00160.jpg" ], "turn_id": 1, "scan_id": "scene0670_01", "origin_question": "Remember, have you seen any kitchenware(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any kitchenware(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0670_01/00200.jpg", "scene0670_01/00240.jpg", "scene0670_01/00280.jpg", "scene0670_01/00320.jpg", "scene0670_01/00360.jpg" ], "turn_id": 2, "scan_id": "scene0670_01", "origin_question": "Is the rectangular light brown wooden kitchenware to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the rectangular light brown wooden kitchenware to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0670_01/00400.jpg", "scene0670_01/00440.jpg", "scene0670_01/00480.jpg", "scene0670_01/00520.jpg", "scene0670_01/00560.jpg" ], "turn_id": 3, "scan_id": "scene0670_01", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the large rectangular stainless steel sink\"; \"the round wicker basket made of light brown material\"; \"the stainless steel built-in dishwasher with a modern design\" ", "option": [ "the large rectangular stainless steel sink", "the round wicker basket made of light brown material", "the stainless steel built-in dishwasher with a modern design" ], "answer": "the round wicker basket made of light brown material", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the large rectangular stainless steel sink\"; \"the round wicker basket made of light brown material\"; \"the stainless steel built-in dishwasher with a modern design\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the large rectangular stainless steel sink', 'the round wicker basket made of light brown material', 'the stainless steel built-in dishwasher with a modern design'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0670_01/00600.jpg", "scene0670_01/00640.jpg", "scene0670_01/00680.jpg", "scene0670_01/00720.jpg", "scene0670_01/00760.jpg" ], "turn_id": 4, "scan_id": "scene0670_01", "origin_question": "When did you first discover \"the counter\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the counter\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "scene0670_01/00800.jpg", "scene0670_01/00840.jpg", "scene0670_01/00880.jpg", "scene0670_01/00920.jpg", "scene0670_01/00960.jpg" ], "turn_id": 5, "scan_id": "scene0670_01", "origin_question": "Among these three objects(\"the large rectangular white glass window\",\"the rectangular brown wooden board\",\"the rectangular light brown wooden kitchenware\" ), which one is the closest to you?", "option": [ "the large rectangular white glass window", "the rectangular brown wooden board", "the rectangular light brown wooden kitchenware" ], "answer": "the large rectangular white glass window", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the large rectangular white glass window\",\"the rectangular brown wooden board\",\"the rectangular light brown wooden kitchenware\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the large rectangular white glass window', 'the rectangular brown wooden board', 'the rectangular light brown wooden kitchenware'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0670_01/01000.jpg", "scene0670_01/01040.jpg", "scene0670_01/01080.jpg", "scene0670_01/01120.jpg", "scene0670_01/01160.jpg" ], "turn_id": 6, "scan_id": "scene0670_01", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0670_01/01200.jpg", "scene0670_01/01240.jpg", "scene0670_01/01280.jpg", "scene0670_01/01320.jpg", "scene0670_01/01360.jpg" ], "turn_id": 7, "scan_id": "scene0670_01", "origin_question": "Please recall \"the rectangular brown wooden board\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "3.55", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the rectangular brown wooden board\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0670_01/01400.jpg", "scene0670_01/01440.jpg", "scene0670_01/01480.jpg", "scene0670_01/01520.jpg", "scene0670_01/01560.jpg" ], "turn_id": 8, "scan_id": "scene0670_01", "origin_question": "Is the round wicker basket made of light brown material to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the round wicker basket made of light brown material to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0670_01/01600.jpg", "scene0670_01/01640.jpg", "scene0670_01/01680.jpg", "scene0670_01/01720.jpg", "scene0670_01/01760.jpg" ], "turn_id": 9, "scan_id": "scene0670_01", "origin_question": "Which direction is the rectangular light brown wooden kitchenware in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the rectangular light brown wooden kitchenware in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "scene0670_01/01800.jpg", "scene0670_01/01840.jpg", "scene0670_01/01880.jpg", "scene0670_01/01920.jpg", "scene0670_01/01960.jpg" ], "turn_id": 10, "scan_id": "scene0670_01", "origin_question": "Among these three objects(\"the stainless steel built-in dishwasher with a modern design\",\"the round wicker basket made of light brown material\",\"the round, brown leather ball\" ), which one is the closest to you?", "option": [ "the stainless steel built-in dishwasher with a modern design", "the round wicker basket made of light brown material", "the round, brown leather ball" ], "answer": "the round, brown leather ball", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the stainless steel built-in dishwasher with a modern design\",\"the round wicker basket made of light brown material\",\"the round, brown leather ball\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the stainless steel built-in dishwasher with a modern design', 'the round wicker basket made of light brown material', 'the round, brown leather ball'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0700_01/00000.jpg", "scene0700_01/00040.jpg", "scene0700_01/00080.jpg", "scene0700_01/00120.jpg", "scene0700_01/00160.jpg" ], "turn_id": 1, "scan_id": "scene0700_01", "origin_question": "Remember, have you seen any bag(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any bag(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "scene0700_01/00200.jpg", "scene0700_01/00240.jpg", "scene0700_01/00280.jpg", "scene0700_01/00320.jpg", "scene0700_01/00360.jpg" ], "turn_id": 2, "scan_id": "scene0700_01", "origin_question": "Among these three objects(\"the medium-sized rectangular white desk made of smooth material\",\"the black ergonomic mouse made of plastic\",\"Black over-ear headphones with leather ear cups\" ), which one is the closest to you?", "option": [ "the medium-sized rectangular white desk made of smooth material", "the black ergonomic mouse made of plastic", "Black over-ear headphones with leather ear cups" ], "answer": "the medium-sized rectangular white desk made of smooth material", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the medium-sized rectangular white desk made of smooth material\",\"the black ergonomic mouse made of plastic\",\"Black over-ear headphones with leather ear cups\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the medium-sized rectangular white desk made of smooth material', 'the black ergonomic mouse made of plastic', 'Black over-ear headphones with leather ear cups'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0700_01/00400.jpg", "scene0700_01/00440.jpg", "scene0700_01/00480.jpg", "scene0700_01/00520.jpg", "scene0700_01/00560.jpg" ], "turn_id": 3, "scan_id": "scene0700_01", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the black ergonomic mouse made of plastic\"; \"the rectangular black plastic keyboard with raised keys\"; \"the window\" ", "option": [ "the black ergonomic mouse made of plastic", "the rectangular black plastic keyboard with raised keys", "the window" ], "answer": "the window", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the black ergonomic mouse made of plastic\"; \"the rectangular black plastic keyboard with raised keys\"; \"the window\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the black ergonomic mouse made of plastic', 'the rectangular black plastic keyboard with raised keys', 'the window'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0700_01/00600.jpg", "scene0700_01/00640.jpg", "scene0700_01/00680.jpg", "scene0700_01/00720.jpg", "scene0700_01/00760.jpg" ], "turn_id": 4, "scan_id": "scene0700_01", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "115.12", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0700_01/00800.jpg", "scene0700_01/00840.jpg", "scene0700_01/00880.jpg", "scene0700_01/00920.jpg", "scene0700_01/00960.jpg" ], "turn_id": 5, "scan_id": "scene0700_01", "origin_question": "Which two objects are on the same side of you now? \"the medium-sized rectangular white desk made of smooth material\", \"the medium-sized cylindrical pen with a yellow body\", \"the small, glossy orange cylindrical toy\". ", "option": [ "the medium-sized rectangular white desk made of smooth material, the medium-sized cylindrical pen with a yellow body", "the medium-sized rectangular white desk made of smooth material, the small, glossy orange cylindrical toy", "the medium-sized cylindrical pen with a yellow body, the small, glossy orange cylindrical toy" ], "answer": "the medium-sized rectangular white desk made of smooth material, the medium-sized cylindrical pen with a yellow body", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the medium-sized rectangular white desk made of smooth material\", \"the medium-sized cylindrical pen with a yellow body\", \"the small, glossy orange cylindrical toy\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the medium-sized rectangular white desk made of smooth material, the medium-sized cylindrical pen with a yellow body', 'the medium-sized rectangular white desk made of smooth material, the small, glossy orange cylindrical toy', 'the medium-sized cylindrical pen with a yellow body, the small, glossy orange cylindrical toy'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0700_01/01000.jpg", "scene0700_01/01040.jpg", "scene0700_01/01080.jpg", "scene0700_01/01120.jpg", "scene0700_01/01160.jpg" ], "turn_id": 6, "scan_id": "scene0700_01", "origin_question": "Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the windowsill\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the windowsill\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0700_01/01200.jpg", "scene0700_01/01240.jpg", "scene0700_01/01280.jpg", "scene0700_01/01320.jpg", "scene0700_01/01360.jpg" ], "turn_id": 7, "scan_id": "scene0700_01", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the medium-sized black leather glove\" from your position?", "option": [], "answer": "74.87", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the medium-sized black leather glove\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0700_01/01400.jpg", "scene0700_01/01440.jpg", "scene0700_01/01480.jpg", "scene0700_01/01520.jpg", "scene0700_01/01560.jpg" ], "turn_id": 8, "scan_id": "scene0700_01", "origin_question": "What will be the first-time appearance order of the following categories: blackboard, sponge, toy?", "option": [ "toy, blackboard, sponge", "sponge, toy, blackboard", "blackboard, sponge, toy", "toy, sponge, blackboard" ], "answer": "sponge, toy, blackboard", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: blackboard, sponge, toy? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['toy, blackboard, sponge', 'sponge, toy, blackboard', 'blackboard, sponge, toy', 'toy, sponge, blackboard'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0700_01/01600.jpg", "scene0700_01/01640.jpg", "scene0700_01/01680.jpg", "scene0700_01/01720.jpg", "scene0700_01/01760.jpg" ], "turn_id": 9, "scan_id": "scene0700_01", "origin_question": "Which two objects are on the same side of you now? \"the medium-sized cylindrical pen with a yellow body\", \"the rectangular paperback book with a white cover\", \"the clothes\". ", "option": [ "the medium-sized cylindrical pen with a yellow body, the rectangular paperback book with a white cover", "the medium-sized cylindrical pen with a yellow body, the clothes", "the rectangular paperback book with a white cover, the clothes" ], "answer": "the medium-sized cylindrical pen with a yellow body, the rectangular paperback book with a white cover", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the medium-sized cylindrical pen with a yellow body\", \"the rectangular paperback book with a white cover\", \"the clothes\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the medium-sized cylindrical pen with a yellow body, the rectangular paperback book with a white cover', 'the medium-sized cylindrical pen with a yellow body, the clothes', 'the rectangular paperback book with a white cover, the clothes'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0700_01/01800.jpg", "scene0700_01/01840.jpg", "scene0700_01/01880.jpg", "scene0700_01/01920.jpg", "scene0700_01/01960.jpg" ], "turn_id": 10, "scan_id": "scene0700_01", "origin_question": "Is the rectangular yellow-green sponge with a rough surface to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the rectangular yellow-green sponge with a rough surface to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0377_01/00000.jpg", "scene0377_01/00040.jpg", "scene0377_01/00080.jpg", "scene0377_01/00120.jpg" ], "turn_id": 1, "scan_id": "scene0377_01", "origin_question": "Remember, have you seen any bucket(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any bucket(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0377_01/00160.jpg", "scene0377_01/00200.jpg", "scene0377_01/00240.jpg", "scene0377_01/00280.jpg" ], "turn_id": 2, "scan_id": "scene0377_01", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the small rectangular white box\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the small rectangular white box\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0377_01/00320.jpg", "scene0377_01/00360.jpg", "scene0377_01/00400.jpg", "scene0377_01/00440.jpg" ], "turn_id": 3, "scan_id": "scene0377_01", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the small rectangular white box\", Y: \"the large, shiny red cylindrical jar\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the small rectangular white box\", Y: \"the large, shiny red cylindrical jar\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0377_01/00480.jpg", "scene0377_01/00520.jpg", "scene0377_01/00560.jpg", "scene0377_01/00600.jpg" ], "turn_id": 4, "scan_id": "scene0377_01", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the window\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the window\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0377_01/00640.jpg", "scene0377_01/00680.jpg", "scene0377_01/00720.jpg", "scene0377_01/00760.jpg" ], "turn_id": 5, "scan_id": "scene0377_01", "origin_question": "When did you first discover \"the small rectangular brown wooden crate with a grid top\"? (index of the turn)", "option": [], "answer": "4", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the small rectangular brown wooden crate with a grid top\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0377_01/00800.jpg", "scene0377_01/00840.jpg", "scene0377_01/00880.jpg", "scene0377_01/00920.jpg" ], "turn_id": 6, "scan_id": "scene0377_01", "origin_question": "In which turn were you closest to \"the paper\"? ", "option": [], "answer": "3", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the paper\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0136_02/00000.jpg", "scene0136_02/00040.jpg", "scene0136_02/00080.jpg", "scene0136_02/00120.jpg" ], "turn_id": 1, "scan_id": "scene0136_02", "origin_question": "Remember, have you seen any copier(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any copier(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0136_02/00160.jpg", "scene0136_02/00200.jpg", "scene0136_02/00240.jpg", "scene0136_02/00280.jpg" ], "turn_id": 2, "scan_id": "scene0136_02", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the large white multi-functional metal copier\", Y: \"the cylindrical white plastic bucket with a round lid\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the large white multi-functional metal copier\", Y: \"the cylindrical white plastic bucket with a round lid\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0136_02/00320.jpg", "scene0136_02/00360.jpg", "scene0136_02/00400.jpg", "scene0136_02/00440.jpg" ], "turn_id": 3, "scan_id": "scene0136_02", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the cylindrical white plastic bucket with a round lid\" from your position?", "option": [], "answer": "113.42", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the cylindrical white plastic bucket with a round lid\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0136_02/00480.jpg", "scene0136_02/00520.jpg", "scene0136_02/00560.jpg", "scene0136_02/00600.jpg" ], "turn_id": 4, "scan_id": "scene0136_02", "origin_question": "In which turn did you see both \"the sturdy rectangular table with a black metal frame\" and \"the rectangular white board with pinned papers\" simultaneously? ", "option": [], "answer": "4", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the sturdy rectangular table with a black metal frame\" and \"the rectangular white board with pinned papers\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0136_02/00640.jpg", "scene0136_02/00680.jpg", "scene0136_02/00720.jpg", "scene0136_02/00760.jpg" ], "turn_id": 5, "scan_id": "scene0136_02", "origin_question": "Which two objects are on the same side of you now? \"the cylindrical white plastic bucket with a round lid\", \"the paper cutter\", \"the rectangular white board with pinned papers\". ", "option": [ "the cylindrical white plastic bucket with a round lid, the paper cutter", "the cylindrical white plastic bucket with a round lid, the rectangular white board with pinned papers", "the paper cutter, the rectangular white board with pinned papers" ], "answer": "the cylindrical white plastic bucket with a round lid, the paper cutter", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the cylindrical white plastic bucket with a round lid\", \"the paper cutter\", \"the rectangular white board with pinned papers\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the cylindrical white plastic bucket with a round lid, the paper cutter', 'the cylindrical white plastic bucket with a round lid, the rectangular white board with pinned papers', 'the paper cutter, the rectangular white board with pinned papers'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0136_02/00800.jpg", "scene0136_02/00840.jpg", "scene0136_02/00880.jpg", "scene0136_02/00920.jpg" ], "turn_id": 6, "scan_id": "scene0136_02", "origin_question": "Which direction is the rectangular white board with pinned papers in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the rectangular white board with pinned papers in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0136_02/00960.jpg", "scene0136_02/01000.jpg", "scene0136_02/01040.jpg", "scene0136_02/01080.jpg" ], "turn_id": 7, "scan_id": "scene0136_02", "origin_question": "Is the cylindrical white plastic bucket with a round lid to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the cylindrical white plastic bucket with a round lid to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0136_02/01120.jpg", "scene0136_02/01160.jpg", "scene0136_02/01200.jpg", "scene0136_02/01240.jpg" ], "turn_id": 8, "scan_id": "scene0136_02", "origin_question": "", "option": [], "answer": "", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0136_02/01280.jpg", "scene0136_02/01320.jpg", "scene0136_02/01360.jpg", "scene0136_02/01400.jpg" ], "turn_id": 9, "scan_id": "scene0136_02", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the cylindrical white plastic bucket with a round lid\" from your position?", "option": [], "answer": "108.27", "user_message": "For the 9 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the cylindrical white plastic bucket with a round lid\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0633_00/00000.jpg", "scene0633_00/00040.jpg", "scene0633_00/00080.jpg", "scene0633_00/00120.jpg", "scene0633_00/00160.jpg" ], "turn_id": 1, "scan_id": "scene0633_00", "origin_question": "Remember, how many stand(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many stand(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0633_00/00200.jpg", "scene0633_00/00240.jpg", "scene0633_00/00280.jpg", "scene0633_00/00320.jpg", "scene0633_00/00360.jpg" ], "turn_id": 2, "scan_id": "scene0633_00", "origin_question": "In which turn were you farthest from \"the carpet\"? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the carpet\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0633_00/00400.jpg", "scene0633_00/00440.jpg", "scene0633_00/00480.jpg", "scene0633_00/00520.jpg", "scene0633_00/00560.jpg" ], "turn_id": 3, "scan_id": "scene0633_00", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0633_00/00600.jpg", "scene0633_00/00640.jpg", "scene0633_00/00680.jpg", "scene0633_00/00720.jpg", "scene0633_00/00760.jpg" ], "turn_id": 4, "scan_id": "scene0633_00", "origin_question": "In which turn did you see both \"the carpet\" and \"then orange plush ball with black lines\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the carpet\" and \"then orange plush ball with black lines\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0633_00/00800.jpg", "scene0633_00/00840.jpg", "scene0633_00/00880.jpg", "scene0633_00/00920.jpg", "scene0633_00/00960.jpg" ], "turn_id": 5, "scan_id": "scene0633_00", "origin_question": "Which direction is then orange plush ball with black lines in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is then orange plush ball with black lines in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0633_00/01000.jpg", "scene0633_00/01040.jpg", "scene0633_00/01080.jpg", "scene0633_00/01120.jpg", "scene0633_00/01160.jpg" ], "turn_id": 6, "scan_id": "scene0633_00", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the cup\" from your position?", "option": [], "answer": "72.31", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the cup\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0633_00/01200.jpg", "scene0633_00/01240.jpg", "scene0633_00/01280.jpg", "scene0633_00/01320.jpg", "scene0633_00/01360.jpg" ], "turn_id": 7, "scan_id": "scene0633_00", "origin_question": "Which direction is the large rectangular white wooden wardrobe with silver handles in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the large rectangular white wooden wardrobe with silver handles in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0633_00/01400.jpg", "scene0633_00/01440.jpg", "scene0633_00/01480.jpg", "scene0633_00/01520.jpg", "scene0633_00/01560.jpg" ], "turn_id": 8, "scan_id": "scene0633_00", "origin_question": "Which two objects are on the same side of you now? \"the medium-sized lamp with a white lampshade\", \"the cup\", \"the carpet\". ", "option": [ "the medium-sized lamp with a white lampshade, the cup", "the medium-sized lamp with a white lampshade, the carpet", "the cup, the carpet" ], "answer": "the medium-sized lamp with a white lampshade, the cup", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the medium-sized lamp with a white lampshade\", \"the cup\", \"the carpet\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the medium-sized lamp with a white lampshade, the cup', 'the medium-sized lamp with a white lampshade, the carpet', 'the cup, the carpet'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0628_01/00000.jpg", "scene0628_01/00040.jpg", "scene0628_01/00080.jpg", "scene0628_01/00120.jpg" ], "turn_id": 1, "scan_id": "scene0628_01", "origin_question": "Remember, have you seen any chair(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any chair(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0628_01/00160.jpg", "scene0628_01/00200.jpg", "scene0628_01/00240.jpg", "scene0628_01/00280.jpg" ], "turn_id": 2, "scan_id": "scene0628_01", "origin_question": "In which turn were you farthest from \"the rectangular, dark-toned canvas or nylon backpack\"? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the rectangular, dark-toned canvas or nylon backpack\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0628_01/00320.jpg", "scene0628_01/00360.jpg", "scene0628_01/00400.jpg", "scene0628_01/00440.jpg" ], "turn_id": 3, "scan_id": "scene0628_01", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the casual deep blue long-sleeved jacket\"; \"the window\"; \"the rectangular, dark-toned canvas or nylon backpack\" ", "option": [ "the casual deep blue long-sleeved jacket", "the window", "the rectangular, dark-toned canvas or nylon backpack" ], "answer": "the casual deep blue long-sleeved jacket", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the casual deep blue long-sleeved jacket\"; \"the window\"; \"the rectangular, dark-toned canvas or nylon backpack\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the casual deep blue long-sleeved jacket', 'the window', 'the rectangular, dark-toned canvas or nylon backpack'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0628_01/00480.jpg", "scene0628_01/00520.jpg", "scene0628_01/00560.jpg", "scene0628_01/00600.jpg" ], "turn_id": 4, "scan_id": "scene0628_01", "origin_question": "In which turn were you closest to \"the casual deep blue long-sleeved jacket\"? ", "option": [], "answer": "2", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the casual deep blue long-sleeved jacket\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0628_01/00640.jpg", "scene0628_01/00680.jpg", "scene0628_01/00720.jpg", "scene0628_01/00760.jpg" ], "turn_id": 5, "scan_id": "scene0628_01", "origin_question": "Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "148.11", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0628_01/00800.jpg", "scene0628_01/00840.jpg", "scene0628_01/00880.jpg", "scene0628_01/00920.jpg" ], "turn_id": 6, "scan_id": "scene0628_01", "origin_question": "Please recall \"the rectangular, dark-toned canvas or nylon backpack\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.35", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the rectangular, dark-toned canvas or nylon backpack\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0557_02/00000.jpg", "scene0557_02/00040.jpg", "scene0557_02/00080.jpg", "scene0557_02/00120.jpg" ], "turn_id": 1, "scan_id": "scene0557_02", "origin_question": "Remember, have you seen any lamp(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any lamp(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0557_02/00160.jpg", "scene0557_02/00200.jpg", "scene0557_02/00240.jpg", "scene0557_02/00280.jpg" ], "turn_id": 2, "scan_id": "scene0557_02", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the bed\"; \"the rectangular dark brown wooden dresser with six drawers\"; \"the large, white plush blanket made of cotton\" ", "option": [ "the bed", "the rectangular dark brown wooden dresser with six drawers", "the large, white plush blanket made of cotton" ], "answer": "the rectangular dark brown wooden dresser with six drawers", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the bed\"; \"the rectangular dark brown wooden dresser with six drawers\"; \"the large, white plush blanket made of cotton\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the bed', 'the rectangular dark brown wooden dresser with six drawers', 'the large, white plush blanket made of cotton'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0557_02/00320.jpg", "scene0557_02/00360.jpg", "scene0557_02/00400.jpg", "scene0557_02/00440.jpg" ], "turn_id": 3, "scan_id": "scene0557_02", "origin_question": "When did you first discover \"the rectangular dark brown wooden dresser with six drawers\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the rectangular dark brown wooden dresser with six drawers\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0557_02/00480.jpg", "scene0557_02/00520.jpg", "scene0557_02/00560.jpg", "scene0557_02/00600.jpg" ], "turn_id": 4, "scan_id": "scene0557_02", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the rectangular dark brown wooden dresser with six drawers\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the rectangular dark brown wooden dresser with six drawers\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "scene0557_02/00640.jpg", "scene0557_02/00680.jpg", "scene0557_02/00720.jpg", "scene0557_02/00760.jpg" ], "turn_id": 5, "scan_id": "scene0557_02", "origin_question": "Assuming the direction you are facing at the end of the turn 3 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Right", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 3 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0557_02/00800.jpg", "scene0557_02/00840.jpg", "scene0557_02/00880.jpg", "scene0557_02/00920.jpg" ], "turn_id": 6, "scan_id": "scene0557_02", "origin_question": "Please recall \"the circular dark clock in digital or analog style\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.96", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the circular dark clock in digital or analog style\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0610_02/00000.jpg", "scene0610_02/00040.jpg", "scene0610_02/00080.jpg", "scene0610_02/00120.jpg", "scene0610_02/00160.jpg" ], "turn_id": 1, "scan_id": "scene0610_02", "origin_question": "Remember, have you seen any stool(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any stool(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0610_02/00200.jpg", "scene0610_02/00240.jpg", "scene0610_02/00280.jpg", "scene0610_02/00320.jpg", "scene0610_02/00550.jpg" ], "turn_id": 2, "scan_id": "scene0610_02", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the rectangular white ceramic bathtub\"; \"the door\"; \"the deep brown rectangular wooden cabinet with a silver handle\" ", "option": [ "the rectangular white ceramic bathtub", "the door", "the deep brown rectangular wooden cabinet with a silver handle" ], "answer": "the door", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the rectangular white ceramic bathtub\"; \"the door\"; \"the deep brown rectangular wooden cabinet with a silver handle\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular white ceramic bathtub', 'the door', 'the deep brown rectangular wooden cabinet with a silver handle'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "scene0610_02/00590.jpg", "scene0610_02/00630.jpg", "scene0610_02/00670.jpg", "scene0610_02/00710.jpg", "scene0610_02/00750.jpg" ], "turn_id": 3, "scan_id": "scene0610_02", "origin_question": "Among these three objects(\"the standard white ceramic toilet with a matching tank\",\"the fluffy white bath towel made of cotton\",\"the rectangular white sheet of paper with a red stripe\" ), which one is the farthest from you?", "option": [ "the standard white ceramic toilet with a matching tank", "the fluffy white bath towel made of cotton", "the rectangular white sheet of paper with a red stripe" ], "answer": "the standard white ceramic toilet with a matching tank", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the standard white ceramic toilet with a matching tank\",\"the fluffy white bath towel made of cotton\",\"the rectangular white sheet of paper with a red stripe\" ), which one is the farthest from you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the standard white ceramic toilet with a matching tank', 'the fluffy white bath towel made of cotton', 'the rectangular white sheet of paper with a red stripe'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0610_02/00790.jpg", "scene0610_02/00840.jpg", "scene0610_02/00880.jpg", "scene0610_02/00920.jpg", "scene0610_02/00960.jpg" ], "turn_id": 4, "scan_id": "scene0610_02", "origin_question": "Please recall \"the standard white ceramic toilet with a matching tank\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.56", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the standard white ceramic toilet with a matching tank\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0610_02/01000.jpg", "scene0610_02/01040.jpg", "scene0610_02/01080.jpg", "scene0610_02/01120.jpg", "scene0610_02/01160.jpg" ], "turn_id": 5, "scan_id": "scene0610_02", "origin_question": "Please recall \"the white oval-shaped sink with a shiny silver faucet\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.56", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the white oval-shaped sink with a shiny silver faucet\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0610_02/01200.jpg", "scene0610_02/01240.jpg", "scene0610_02/01280.jpg", "scene0610_02/01320.jpg", "scene0610_02/01360.jpg" ], "turn_id": 6, "scan_id": "scene0610_02", "origin_question": "", "option": [], "answer": "", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0610_02/01400.jpg", "scene0610_02/01440.jpg", "scene0610_02/01480.jpg", "scene0610_02/01520.jpg", "scene0610_02/01560.jpg" ], "turn_id": 7, "scan_id": "scene0610_02", "origin_question": "In which turn were you closest to \"the standard white ceramic toilet with a matching tank\"? ", "option": [], "answer": "2", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the standard white ceramic toilet with a matching tank\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0610_02/01600.jpg", "scene0610_02/01640.jpg", "scene0610_02/01680.jpg", "scene0610_02/01720.jpg", "scene0610_02/01760.jpg" ], "turn_id": 8, "scan_id": "scene0610_02", "origin_question": "Is the white oval-shaped sink with a shiny silver faucet to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the white oval-shaped sink with a shiny silver faucet to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0610_02/01800.jpg", "scene0610_02/01840.jpg", "scene0610_02/01880.jpg", "scene0610_02/01920.jpg", "scene0610_02/01960.jpg" ], "turn_id": 9, "scan_id": "scene0610_02", "origin_question": "By the end of the 5 turn, were you getting closer to or farther away from these two objects: X: \"the white semi-transparent curtain with colorful stars\", Y: \"the deep brown rectangular wooden cabinet with a silver handle\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to X and farther away from Y", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 5 turn, were you getting closer to or farther away from these two objects: X: \"the white semi-transparent curtain with colorful stars\", Y: \"the deep brown rectangular wooden cabinet with a silver handle\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "scene0610_02/02000.jpg", "scene0610_02/02040.jpg", "scene0610_02/02080.jpg", "scene0610_02/02120.jpg", "scene0610_02/02160.jpg" ], "turn_id": 10, "scan_id": "scene0610_02", "origin_question": "Assuming the direction you are facing at the end of the turn 5 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Right", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 5 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0029_region26/155fac2d50764bf09feb6c8f33e8fb76_i1_4.jpg", "1mp3d_0029_region26/155fac2d50764bf09feb6c8f33e8fb76_i2_4.jpg", "1mp3d_0029_region26/155fac2d50764bf09feb6c8f33e8fb76_i1_5.jpg", "1mp3d_0029_region26/155fac2d50764bf09feb6c8f33e8fb76_i2_5.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0029_region26", "origin_question": "Remember, have you seen any sink(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any sink(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "1mp3d_0029_region26/155fac2d50764bf09feb6c8f33e8fb76_i1_0.jpg", "1mp3d_0029_region26/155fac2d50764bf09feb6c8f33e8fb76_i2_0.jpg", "1mp3d_0029_region26/155fac2d50764bf09feb6c8f33e8fb76_i1_1.jpg", "1mp3d_0029_region26/155fac2d50764bf09feb6c8f33e8fb76_i2_1.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0029_region26", "origin_question": "Among these three objects(\"the rectangular white door with a classic design\",\"the rectangular, framed picture on canvas\",\"the rectangular white marble sink with modern design\" ), which one is the closest to you?", "option": [ "the rectangular white door with a classic design", "the rectangular, framed picture on canvas", "the rectangular white marble sink with modern design" ], "answer": "the rectangular white door with a classic design", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects(\"the rectangular white door with a classic design\",\"the rectangular, framed picture on canvas\",\"the rectangular white marble sink with modern design\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular white door with a classic design', 'the rectangular, framed picture on canvas', 'the rectangular white marble sink with modern design'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "1mp3d_0029_region26/bfc3e00afca648f4b0b6945f4f7cffe0_i1_3.jpg", "1mp3d_0029_region26/bfc3e00afca648f4b0b6945f4f7cffe0_i2_3.jpg", "1mp3d_0029_region26/bfc3e00afca648f4b0b6945f4f7cffe0_i1_4.jpg", "1mp3d_0029_region26/bfc3e00afca648f4b0b6945f4f7cffe0_i2_4.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0029_region26", "origin_question": "Which direction is the rectangular white marble sink with modern design in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the rectangular white marble sink with modern design in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "1mp3d_0029_region26/6f6a10929810436f894094cbcd00e9e2_i1_4.jpg", "1mp3d_0029_region26/6f6a10929810436f894094cbcd00e9e2_i2_4.jpg", "1mp3d_0029_region26/6f6a10929810436f894094cbcd00e9e2_i1_3.jpg", "1mp3d_0029_region26/6f6a10929810436f894094cbcd00e9e2_i2_3.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0029_region26", "origin_question": "In which turn were you farthest from \"the white oval standalone bathtub with a glossy surface\"? ", "option": [], "answer": "4", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the white oval standalone bathtub with a glossy surface\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "1mp3d_0029_region26/bfc3e00afca648f4b0b6945f4f7cffe0_i1_2.jpg", "1mp3d_0029_region26/bfc3e00afca648f4b0b6945f4f7cffe0_i2_2.jpg", "1mp3d_0029_region26/bfc3e00afca648f4b0b6945f4f7cffe0_i1_1.jpg", "1mp3d_0029_region26/bfc3e00afca648f4b0b6945f4f7cffe0_i2_1.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0029_region26", "origin_question": "In which turn did you see both \"the tall transparent glass flowerpot with a wide base\" and \"the sleek, silver stainless steel faucet with two handles\" simultaneously? ", "option": [], "answer": "5", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the tall transparent glass flowerpot with a wide base\" and \"the sleek, silver stainless steel faucet with two handles\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "1mp3d_0029_region26/bfc3e00afca648f4b0b6945f4f7cffe0_i1_0.jpg", "1mp3d_0029_region26/bfc3e00afca648f4b0b6945f4f7cffe0_i2_0.jpg", "1mp3d_0029_region26/6f6a10929810436f894094cbcd00e9e2_i1_5.jpg", "1mp3d_0029_region26/6f6a10929810436f894094cbcd00e9e2_i2_5.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0029_region26", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular, framed picture on canvas\", Y: \"the fluffy, white cotton towel\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "farther away from both X and Y", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular, framed picture on canvas\", Y: \"the fluffy, white cotton towel\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0207_01/00000.jpg", "scene0207_01/00040.jpg", "scene0207_01/00080.jpg", "scene0207_01/00120.jpg", "scene0207_01/00160.jpg" ], "turn_id": 1, "scan_id": "scene0207_01", "origin_question": "Remember, how many cosmetics(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many cosmetics(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0207_01/00200.jpg", "scene0207_01/00240.jpg", "scene0207_01/00280.jpg", "scene0207_01/00320.jpg", "scene0207_01/00360.jpg" ], "turn_id": 2, "scan_id": "scene0207_01", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the medium-sized rectangular transparent box made of plastic\"; \"the slim, vertical laser printer with a modern design\"; \"the rectangular cream plastic switch\" ", "option": [ "the medium-sized rectangular transparent box made of plastic", "the slim, vertical laser printer with a modern design", "the rectangular cream plastic switch" ], "answer": "the rectangular cream plastic switch", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the medium-sized rectangular transparent box made of plastic\"; \"the slim, vertical laser printer with a modern design\"; \"the rectangular cream plastic switch\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the medium-sized rectangular transparent box made of plastic', 'the slim, vertical laser printer with a modern design', 'the rectangular cream plastic switch'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0207_01/00400.jpg", "scene0207_01/00440.jpg", "scene0207_01/00480.jpg", "scene0207_01/00520.jpg", "scene0207_01/00560.jpg" ], "turn_id": 3, "scan_id": "scene0207_01", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the white cylindrical paper tissue\" from your position?", "option": [], "answer": "146.71", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the white cylindrical paper tissue\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "scene0207_01/00600.jpg", "scene0207_01/00640.jpg", "scene0207_01/00680.jpg", "scene0207_01/00720.jpg", "scene0207_01/00760.jpg" ], "turn_id": 4, "scan_id": "scene0207_01", "origin_question": "Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Right", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0207_01/00800.jpg", "scene0207_01/00840.jpg", "scene0207_01/00880.jpg", "scene0207_01/00920.jpg", "scene0207_01/00960.jpg" ], "turn_id": 5, "scan_id": "scene0207_01", "origin_question": "How far is your current position from where you were at the end of the turn 3? (in meters)", "option": [], "answer": "2.56", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 3? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0207_01/01000.jpg", "scene0207_01/01040.jpg", "scene0207_01/01080.jpg", "scene0207_01/01120.jpg", "scene0207_01/01160.jpg" ], "turn_id": 6, "scan_id": "scene0207_01", "origin_question": "Which direction is the rectangular white double bed in good condition in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the rectangular white double bed in good condition in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0207_01/01200.jpg", "scene0207_01/01240.jpg", "scene0207_01/01280.jpg", "scene0207_01/01320.jpg", "scene0207_01/01360.jpg" ], "turn_id": 7, "scan_id": "scene0207_01", "origin_question": "Is the slim, vertical laser printer with a modern design to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the slim, vertical laser printer with a modern design to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0207_01/01400.jpg", "scene0207_01/01440.jpg", "scene0207_01/01480.jpg", "scene0207_01/01520.jpg", "scene0207_01/01560.jpg" ], "turn_id": 8, "scan_id": "scene0207_01", "origin_question": "What will be the first-time appearance order of the following categories: couch, printer, vacuum cleaner?", "option": [ "couch, printer, vacuum cleaner", "vacuum cleaner, printer, couch", "couch, vacuum cleaner, printer", "printer, vacuum cleaner, couch" ], "answer": "couch, printer, vacuum cleaner", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: couch, printer, vacuum cleaner? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['couch, printer, vacuum cleaner', 'vacuum cleaner, printer, couch', 'couch, vacuum cleaner, printer', 'printer, vacuum cleaner, couch'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0207_01/01600.jpg", "scene0207_01/01640.jpg", "scene0207_01/01680.jpg", "scene0207_01/01720.jpg", "scene0207_01/01760.jpg" ], "turn_id": 9, "scan_id": "scene0207_01", "origin_question": "In which turn did you see both \"the large white plastic case\" and \"the rectangular glass mirror with a light beige color\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the large white plastic case\" and \"the rectangular glass mirror with a light beige color\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0291_01/00000.jpg", "scene0291_01/00040.jpg", "scene0291_01/00080.jpg", "scene0291_01/00120.jpg", "scene0291_01/00160.jpg" ], "turn_id": 1, "scan_id": "scene0291_01", "origin_question": "Remember, have you seen any water heater(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any water heater(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0291_01/00200.jpg", "scene0291_01/00240.jpg", "scene0291_01/00280.jpg", "scene0291_01/00320.jpg", "scene0291_01/00360.jpg" ], "turn_id": 2, "scan_id": "scene0291_01", "origin_question": "When did you first discover \"the light brown wooden rectangular door with a panel\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the light brown wooden rectangular door with a panel\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0291_01/00400.jpg", "scene0291_01/00440.jpg", "scene0291_01/00480.jpg", "scene0291_01/00520.jpg", "scene0291_01/00560.jpg" ], "turn_id": 3, "scan_id": "scene0291_01", "origin_question": "In which turn did you see both \"the large rectangular wooden table with a smooth surface\" and \"the light brown wooden rectangular door with a panel\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the large rectangular wooden table with a smooth surface\" and \"the light brown wooden rectangular door with a panel\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0291_01/00600.jpg", "scene0291_01/00640.jpg", "scene0291_01/00680.jpg", "scene0291_01/00720.jpg", "scene0291_01/00760.jpg" ], "turn_id": 4, "scan_id": "scene0291_01", "origin_question": "Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "108.13", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0291_01/00800.jpg", "scene0291_01/00840.jpg", "scene0291_01/00880.jpg", "scene0291_01/00920.jpg", "scene0291_01/00960.jpg" ], "turn_id": 5, "scan_id": "scene0291_01", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the medium-sized rectangular black box\" from your position?", "option": [], "answer": "172.28", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the medium-sized rectangular black box\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0291_01/01000.jpg", "scene0291_01/01040.jpg", "scene0291_01/01080.jpg", "scene0291_01/01120.jpg", "scene0291_01/01160.jpg" ], "turn_id": 6, "scan_id": "scene0291_01", "origin_question": "Please recall \"the medium-sized rectangular black box\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "5.43", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the medium-sized rectangular black box\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0291_01/01200.jpg", "scene0291_01/01240.jpg", "scene0291_01/01280.jpg", "scene0291_01/01320.jpg", "scene0291_01/01360.jpg" ], "turn_id": 7, "scan_id": "scene0291_01", "origin_question": "In which turn were you closest to \"the large rectangular black blackboard with a white frame\"? ", "option": [], "answer": "4", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the large rectangular black blackboard with a white frame\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0291_01/01400.jpg", "scene0291_01/01440.jpg", "scene0291_01/01480.jpg", "scene0291_01/01520.jpg", "scene0291_01/01560.jpg" ], "turn_id": 8, "scan_id": "scene0291_01", "origin_question": "By the end of the 4 turn, were you getting closer to or farther away from these two objects: X: \"the large rectangular black blackboard with a white frame\", Y: \"the medium-sized rectangular black box\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "farther away from both X and Y", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 4 turn, were you getting closer to or farther away from these two objects: X: \"the large rectangular black blackboard with a white frame\", Y: \"the medium-sized rectangular black box\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0044_region35/795af89129d44ba9a2513be917aa2e75_i1_0.jpg", "1mp3d_0044_region35/795af89129d44ba9a2513be917aa2e75_i2_0.jpg", "1mp3d_0044_region35/f4054160a6f445239913e0d2e20a8448_i1_0.jpg", "1mp3d_0044_region35/f4054160a6f445239913e0d2e20a8448_i2_0.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0044_region35", "origin_question": "Remember, have you seen any laptop(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any laptop(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "1mp3d_0044_region35/2de328d1e8954da6a4f8027a2d6092f0_i1_4.jpg", "1mp3d_0044_region35/2de328d1e8954da6a4f8027a2d6092f0_i2_4.jpg", "1mp3d_0044_region35/2de328d1e8954da6a4f8027a2d6092f0_i1_3.jpg", "1mp3d_0044_region35/2de328d1e8954da6a4f8027a2d6092f0_i2_3.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0044_region35", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular window with a white frame\", Y: \"the long, soft-colored floral curtain\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular window with a white frame\", Y: \"the long, soft-colored floral curtain\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "1mp3d_0044_region35/f4054160a6f445239913e0d2e20a8448_i1_1.jpg", "1mp3d_0044_region35/f4054160a6f445239913e0d2e20a8448_i2_1.jpg", "1mp3d_0044_region35/f4054160a6f445239913e0d2e20a8448_i1_2.jpg", "1mp3d_0044_region35/795af89129d44ba9a2513be917aa2e75_i1_5.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0044_region35", "origin_question": "By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the white and pink decoration of two animated characters\", Y: \"the medium-sized picture in a golden frame\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the white and pink decoration of two animated characters\", Y: \"the medium-sized picture in a golden frame\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "1mp3d_0044_region35/795af89129d44ba9a2513be917aa2e75_i2_5.jpg", "1mp3d_0044_region35/44258be9bf664fccb0e728d319ce2638_i1_3.jpg", "1mp3d_0044_region35/44258be9bf664fccb0e728d319ce2638_i2_3.jpg", "1mp3d_0044_region35/44258be9bf664fccb0e728d319ce2638_i1_2.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0044_region35", "origin_question": "Please recall \"the white and pink decoration of two animated characters\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.71", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the white and pink decoration of two animated characters\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "1mp3d_0044_region35/44258be9bf664fccb0e728d319ce2638_i2_2.jpg", "1mp3d_0044_region35/44258be9bf664fccb0e728d319ce2638_i1_1.jpg", "1mp3d_0044_region35/44258be9bf664fccb0e728d319ce2638_i2_1.jpg", "1mp3d_0044_region35/795af89129d44ba9a2513be917aa2e75_i1_2.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0044_region35", "origin_question": "At the end of which turn were both of \"the long, soft-colored floral curtain\" and \"the white and pink decoration of two animated characters\" on your left side?", "option": [], "answer": "5", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the long, soft-colored floral curtain\" and \"the white and pink decoration of two animated characters\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0378_00/00000.jpg", "scene0378_00/00040.jpg", "scene0378_00/00080.jpg", "scene0378_00/00120.jpg", "scene0378_00/00160.jpg" ], "turn_id": 1, "scan_id": "scene0378_00", "origin_question": "Remember, have you seen any bin(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any bin(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0378_00/00200.jpg", "scene0378_00/00240.jpg", "scene0378_00/00280.jpg", "scene0378_00/00320.jpg", "scene0378_00/00360.jpg" ], "turn_id": 2, "scan_id": "scene0378_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0378_00/00400.jpg", "scene0378_00/00440.jpg", "scene0378_00/00480.jpg", "scene0378_00/00520.jpg", "scene0378_00/00560.jpg" ], "turn_id": 3, "scan_id": "scene0378_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0378_00/00600.jpg", "scene0378_00/00640.jpg", "scene0378_00/00680.jpg", "scene0378_00/00720.jpg", "scene0378_00/00760.jpg" ], "turn_id": 4, "scan_id": "scene0378_00", "origin_question": "Which direction is the rectangular, colorful picture in white in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the rectangular, colorful picture in white in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0378_00/00800.jpg", "scene0378_00/00840.jpg", "scene0378_00/00880.jpg", "scene0378_00/00920.jpg", "scene0378_00/00960.jpg" ], "turn_id": 5, "scan_id": "scene0378_00", "origin_question": "Is the sleek silver monitor with black borders to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the sleek silver monitor with black borders to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0378_00/01000.jpg", "scene0378_00/01040.jpg", "scene0378_00/01080.jpg", "scene0378_00/01120.jpg", "scene0378_00/01160.jpg" ], "turn_id": 6, "scan_id": "scene0378_00", "origin_question": "Is the rectangular blue plastic garbage bin to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the rectangular blue plastic garbage bin to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0378_00/01200.jpg", "scene0378_00/01240.jpg", "scene0378_00/01280.jpg", "scene0378_00/01320.jpg", "scene0378_00/01360.jpg" ], "turn_id": 7, "scan_id": "scene0378_00", "origin_question": "Please recall \"the rectangular, colorful picture in white\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.55", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the rectangular, colorful picture in white\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0378_00/01400.jpg", "scene0378_00/01440.jpg", "scene0378_00/01480.jpg", "scene0378_00/01520.jpg", "scene0378_00/01560.jpg" ], "turn_id": 8, "scan_id": "scene0378_00", "origin_question": "In which turn were you closest to \"Blue athletic shoes with white stripes and logo\"? ", "option": [], "answer": "8", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"Blue athletic shoes with white stripes and logo\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0378_00/01600.jpg", "scene0378_00/01640.jpg", "scene0378_00/01680.jpg", "scene0378_00/01720.jpg", "scene0378_00/01760.jpg" ], "turn_id": 9, "scan_id": "scene0378_00", "origin_question": "Which direction is the rectangular yellow door made of sturdy material in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the rectangular yellow door made of sturdy material in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0072_region0/227ed4cadd89473b93bfdde29decd566_i1_1.jpg", "1mp3d_0072_region0/227ed4cadd89473b93bfdde29decd566_i2_1.jpg", "1mp3d_0072_region0/227ed4cadd89473b93bfdde29decd566_i1_0.jpg", "1mp3d_0072_region0/227ed4cadd89473b93bfdde29decd566_i2_0.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0072_region0", "origin_question": "Remember, have you seen any rope(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any rope(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "1mp3d_0072_region0/ae03b51ef39a43088e0e6300c4ea97a0_i1_0.jpg", "1mp3d_0072_region0/ae03b51ef39a43088e0e6300c4ea97a0_i2_0.jpg", "1mp3d_0072_region0/bf6f080d1a064a2eb641e3ec699ed354_i1_1.jpg", "1mp3d_0072_region0/bf6f080d1a064a2eb641e3ec699ed354_i2_1.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0072_region0", "origin_question": "In which turn were you closest to \"the standard single bed with a deep blue bedspread\"? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the standard single bed with a deep blue bedspread\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0072_region0/1a5c7ec82b6e4a83afb6e76b4d2118c7_i1_0.jpg", "1mp3d_0072_region0/1a5c7ec82b6e4a83afb6e76b4d2118c7_i2_0.jpg", "1mp3d_0072_region0/bf6f080d1a064a2eb641e3ec699ed354_i1_0.jpg", "1mp3d_0072_region0/bf6f080d1a064a2eb641e3ec699ed354_i2_0.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0072_region0", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the white wooden chair with a slatted seat and backrest\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the white wooden chair with a slatted seat and backrest\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "1mp3d_0072_region0/bf6f080d1a064a2eb641e3ec699ed354_i1_5.jpg", "1mp3d_0072_region0/bf6f080d1a064a2eb641e3ec699ed354_i2_5.jpg", "1mp3d_0072_region0/bf6f080d1a064a2eb641e3ec699ed354_i1_4.jpg", "1mp3d_0072_region0/bf6f080d1a064a2eb641e3ec699ed354_i2_4.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0072_region0", "origin_question": "What will be the first-time appearance order of the following categories: couch, desk, pillow?", "option": [ "desk, pillow, couch", "couch, desk, pillow", "pillow, desk, couch", "couch, pillow, desk" ], "answer": "desk, pillow, couch", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: couch, desk, pillow? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['desk, pillow, couch', 'couch, desk, pillow', 'pillow, desk, couch', 'couch, pillow, desk'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0072_region0/bf6f080d1a064a2eb641e3ec699ed354_i1_3.jpg", "1mp3d_0072_region0/bf6f080d1a064a2eb641e3ec699ed354_i2_3.jpg", "1mp3d_0072_region0/1a5c7ec82b6e4a83afb6e76b4d2118c7_i1_1.jpg", "1mp3d_0072_region0/1a5c7ec82b6e4a83afb6e76b4d2118c7_i2_1.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0072_region0", "origin_question": "", "option": [], "answer": "", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "1mp3d_0072_region0/1a5c7ec82b6e4a83afb6e76b4d2118c7_i1_2.jpg", "1mp3d_0072_region0/1a5c7ec82b6e4a83afb6e76b4d2118c7_i2_2.jpg", "1mp3d_0072_region0/1a5c7ec82b6e4a83afb6e76b4d2118c7_i1_3.jpg", "1mp3d_0072_region0/1a5c7ec82b6e4a83afb6e76b4d2118c7_i2_3.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0072_region0", "origin_question": "Is the rectangular black marble table with a smooth surface to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the rectangular black marble table with a smooth surface to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "1mp3d_0072_region0/1a5c7ec82b6e4a83afb6e76b4d2118c7_i1_4.jpg", "1mp3d_0072_region0/1a5c7ec82b6e4a83afb6e76b4d2118c7_i2_4.jpg", "1mp3d_0072_region0/1a5c7ec82b6e4a83afb6e76b4d2118c7_i1_5.jpg", "1mp3d_0072_region0/1a5c7ec82b6e4a83afb6e76b4d2118c7_i2_5.jpg" ], "turn_id": 7, "scan_id": "1mp3d_0072_region0", "origin_question": "In which turn did you see both \"the rectangular weathered white wooden desk\" and \"the rectangular black marble table with a smooth surface\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the rectangular weathered white wooden desk\" and \"the rectangular black marble table with a smooth surface\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "1mp3d_0072_region0/b7eae2b57c19455991ae3242798a8b05_i1_0.jpg", "1mp3d_0072_region0/b7eae2b57c19455991ae3242798a8b05_i2_0.jpg", "1mp3d_0072_region0/b7eae2b57c19455991ae3242798a8b05_i1_1.jpg", "1mp3d_0072_region0/b7eae2b57c19455991ae3242798a8b05_i2_1.jpg" ], "turn_id": 8, "scan_id": "1mp3d_0072_region0", "origin_question": "Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "119.67", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "1mp3d_0072_region0/b7eae2b57c19455991ae3242798a8b05_i1_2.jpg", "1mp3d_0072_region0/b7eae2b57c19455991ae3242798a8b05_i2_2.jpg", "1mp3d_0072_region0/b7eae2b57c19455991ae3242798a8b05_i1_3.jpg", "1mp3d_0072_region0/b7eae2b57c19455991ae3242798a8b05_i2_3.jpg" ], "turn_id": 9, "scan_id": "1mp3d_0072_region0", "origin_question": "Which direction is the white wooden chair with a slatted seat and backrest in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 9 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the white wooden chair with a slatted seat and backrest in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0307_00/00000.jpg", "scene0307_00/00040.jpg", "scene0307_00/00080.jpg", "scene0307_00/00120.jpg", "scene0307_00/00160.jpg" ], "turn_id": 1, "scan_id": "scene0307_00", "origin_question": "Remember, how many lamp(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many lamp(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0307_00/00200.jpg", "scene0307_00/00240.jpg", "scene0307_00/00280.jpg", "scene0307_00/00320.jpg", "scene0307_00/00360.jpg" ], "turn_id": 2, "scan_id": "scene0307_00", "origin_question": "Is the white front-loading washing machine made of metal to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the white front-loading washing machine made of metal to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "scene0307_00/00400.jpg", "scene0307_00/00440.jpg", "scene0307_00/00480.jpg", "scene0307_00/00520.jpg", "scene0307_00/00560.jpg" ], "turn_id": 3, "scan_id": "scene0307_00", "origin_question": "Among these three objects(\"the wooden chair with a green fabric seat\",\"the shower\",\"the tall, white stainless steel refrigerator\" ), which one is the farthest from you?", "option": [ "the wooden chair with a green fabric seat", "the shower", "the tall, white stainless steel refrigerator" ], "answer": "the tall, white stainless steel refrigerator", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the wooden chair with a green fabric seat\",\"the shower\",\"the tall, white stainless steel refrigerator\" ), which one is the farthest from you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the wooden chair with a green fabric seat', 'the shower', 'the tall, white stainless steel refrigerator'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0307_00/00600.jpg", "scene0307_00/00640.jpg", "scene0307_00/00680.jpg", "scene0307_00/00720.jpg", "scene0307_00/00760.jpg" ], "turn_id": 4, "scan_id": "scene0307_00", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0307_00/00800.jpg", "scene0307_00/00840.jpg", "scene0307_00/00880.jpg", "scene0307_00/00920.jpg", "scene0307_00/00960.jpg" ], "turn_id": 5, "scan_id": "scene0307_00", "origin_question": "Which two objects are on the same side of you now? \"the book\", \"the shower\", \"the gray medium-sized clothes made of cotton and polyester\". ", "option": [ "the book, the shower", "the book, the gray medium-sized clothes made of cotton and polyester", "the shower, the gray medium-sized clothes made of cotton and polyester" ], "answer": "the shower, the gray medium-sized clothes made of cotton and polyester", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the book\", \"the shower\", \"the gray medium-sized clothes made of cotton and polyester\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the book, the shower', 'the book, the gray medium-sized clothes made of cotton and polyester', 'the shower, the gray medium-sized clothes made of cotton and polyester'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0307_00/01000.jpg", "scene0307_00/01040.jpg", "scene0307_00/01080.jpg", "scene0307_00/01120.jpg", "scene0307_00/01160.jpg" ], "turn_id": 6, "scan_id": "scene0307_00", "origin_question": "Is the rectangular green and white detergent box to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the rectangular green and white detergent box to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0307_00/01200.jpg", "scene0307_00/01240.jpg", "scene0307_00/01280.jpg", "scene0307_00/01320.jpg", "scene0307_00/01360.jpg" ], "turn_id": 7, "scan_id": "scene0307_00", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the tall, white stainless steel refrigerator\" from your position?", "option": [], "answer": "83.15", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the tall, white stainless steel refrigerator\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0307_00/01400.jpg", "scene0307_00/01440.jpg", "scene0307_00/01480.jpg", "scene0307_00/01520.jpg", "scene0307_00/01560.jpg" ], "turn_id": 8, "scan_id": "scene0307_00", "origin_question": "Which two objects are on the same side of you now? \"the rectangular green and white detergent box\", \"the rectangular white plastic switch\", \"the medium-sized cylindrical brown jar with a textured surface\". ", "option": [ "the rectangular green and white detergent box, the rectangular white plastic switch", "the rectangular green and white detergent box, the medium-sized cylindrical brown jar with a textured surface", "the rectangular white plastic switch, the medium-sized cylindrical brown jar with a textured surface" ], "answer": "the rectangular green and white detergent box, the rectangular white plastic switch", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the rectangular green and white detergent box\", \"the rectangular white plastic switch\", \"the medium-sized cylindrical brown jar with a textured surface\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular green and white detergent box, the rectangular white plastic switch', 'the rectangular green and white detergent box, the medium-sized cylindrical brown jar with a textured surface', 'the rectangular white plastic switch, the medium-sized cylindrical brown jar with a textured surface'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0307_00/01600.jpg", "scene0307_00/01640.jpg", "scene0307_00/01680.jpg", "scene0307_00/01720.jpg", "scene0307_00/01760.jpg" ], "turn_id": 9, "scan_id": "scene0307_00", "origin_question": "By the end of the 8 turn, were you getting closer to or farther away from these two objects: X: \"the white front-loading washing machine made of metal\", Y: \"the medium-sized rectangular white metal radiator\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 8 turn, were you getting closer to or farther away from these two objects: X: \"the white front-loading washing machine made of metal\", Y: \"the medium-sized rectangular white metal radiator\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "scene0307_00/01800.jpg", "scene0307_00/01840.jpg", "scene0307_00/01880.jpg", "scene0307_00/01920.jpg", "scene0307_00/01960.jpg" ], "turn_id": 10, "scan_id": "scene0307_00", "origin_question": "Assuming the direction you are facing at the end of the turn 8 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Right", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 8 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0040_region22/011ea5a962624e9abd2269d8fed44f11_i1_3.jpg", "1mp3d_0040_region22/011ea5a962624e9abd2269d8fed44f11_i2_3.jpg", "1mp3d_0040_region22/6c2c25242c9d47e28fe641a235a45acd_i1_0.jpg", "1mp3d_0040_region22/6c2c25242c9d47e28fe641a235a45acd_i2_0.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0040_region22", "origin_question": "Remember, have you seen any sheet(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any sheet(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "1mp3d_0040_region22/011ea5a962624e9abd2269d8fed44f11_i1_4.jpg", "1mp3d_0040_region22/011ea5a962624e9abd2269d8fed44f11_i2_4.jpg", "1mp3d_0040_region22/6c2c25242c9d47e28fe641a235a45acd_i1_1.jpg", "1mp3d_0040_region22/6c2c25242c9d47e28fe641a235a45acd_i2_1.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0040_region22", "origin_question": "In which turn did you see both \"the glossy black flat-panel TV with a slim profile\" and \"the medium-sized gray and black textured carpet\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the glossy black flat-panel TV with a slim profile\" and \"the medium-sized gray and black textured carpet\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "1mp3d_0040_region22/5cdd513110c54c72b6d9be6f351d32df_i1_0.jpg", "1mp3d_0040_region22/5cdd513110c54c72b6d9be6f351d32df_i2_0.jpg", "1mp3d_0040_region22/5cdd513110c54c72b6d9be6f351d32df_i1_5.jpg", "1mp3d_0040_region22/5cdd513110c54c72b6d9be6f351d32df_i2_5.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0040_region22", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the glossy black flat-panel TV with a slim profile\", Y: \"the black metal rack with a single layer\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "farther away from both X and Y", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the glossy black flat-panel TV with a slim profile\", Y: \"the black metal rack with a single layer\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "1mp3d_0040_region22/5cdd513110c54c72b6d9be6f351d32df_i1_4.jpg", "1mp3d_0040_region22/5cdd513110c54c72b6d9be6f351d32df_i2_4.jpg", "1mp3d_0040_region22/5cdd513110c54c72b6d9be6f351d32df_i1_3.jpg", "1mp3d_0040_region22/5cdd513110c54c72b6d9be6f351d32df_i2_3.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0040_region22", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "2.14", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "1mp3d_0040_region22/6c2c25242c9d47e28fe641a235a45acd_i1_3.jpg", "1mp3d_0040_region22/6c2c25242c9d47e28fe641a235a45acd_i2_3.jpg", "1mp3d_0040_region22/6c2c25242c9d47e28fe641a235a45acd_i1_4.jpg", "1mp3d_0040_region22/6c2c25242c9d47e28fe641a235a45acd_i2_4.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0040_region22", "origin_question": "How far is your current position from where you were at the end of the turn 3? (in meters)", "option": [], "answer": "2.13", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 3? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "1mp3d_0040_region22/011ea5a962624e9abd2269d8fed44f11_i1_2.jpg", "1mp3d_0040_region22/011ea5a962624e9abd2269d8fed44f11_i2_2.jpg", "1mp3d_0040_region22/011ea5a962624e9abd2269d8fed44f11_i1_1.jpg", "1mp3d_0040_region22/011ea5a962624e9abd2269d8fed44f11_i2_1.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0040_region22", "origin_question": "Remember, how many door(s) have you seen so far? ", "option": [], "answer": "3", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many door(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0071_00/00000.jpg", "scene0071_00/00040.jpg", "scene0071_00/00080.jpg", "scene0071_00/00120.jpg" ], "turn_id": 1, "scan_id": "scene0071_00", "origin_question": "Remember, have you seen any picture(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any picture(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0071_00/00160.jpg", "scene0071_00/00200.jpg", "scene0071_00/00240.jpg", "scene0071_00/00280.jpg" ], "turn_id": 2, "scan_id": "scene0071_00", "origin_question": "When did you first discover \"the red rectangular metal fire extinguisher\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the red rectangular metal fire extinguisher\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0071_00/00320.jpg", "scene0071_00/00360.jpg", "scene0071_00/00400.jpg", "scene0071_00/00440.jpg" ], "turn_id": 3, "scan_id": "scene0071_00", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the red rectangular metal fire extinguisher\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the red rectangular metal fire extinguisher\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0071_00/00480.jpg", "scene0071_00/00520.jpg", "scene0071_00/00560.jpg", "scene0071_00/00600.jpg" ], "turn_id": 4, "scan_id": "scene0071_00", "origin_question": "Which direction is the red rectangular metal fire extinguisher in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the red rectangular metal fire extinguisher in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0071_00/00640.jpg", "scene0071_00/00680.jpg", "scene0071_00/00720.jpg", "scene0071_00/00760.jpg" ], "turn_id": 5, "scan_id": "scene0071_00", "origin_question": "Please recall \"the red rectangular metal fire extinguisher\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.61", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the red rectangular metal fire extinguisher\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0071_00/00800.jpg", "scene0071_00/00840.jpg", "scene0071_00/00880.jpg", "scene0071_00/00920.jpg" ], "turn_id": 6, "scan_id": "scene0071_00", "origin_question": "Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0071_00/00960.jpg", "scene0071_00/01000.jpg", "scene0071_00/01040.jpg", "scene0071_00/01080.jpg" ], "turn_id": 7, "scan_id": "scene0071_00", "origin_question": "Which direction is the medium-sized orange chair with a round seat in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the medium-sized orange chair with a round seat in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0004_region22/60640d7394b94da68c549715eed7e83d_i1_1.jpg", "1mp3d_0004_region22/60640d7394b94da68c549715eed7e83d_i2_1.jpg", "1mp3d_0004_region22/21bc32e53a104942a278f5ecff75e896_i1_1.jpg", "1mp3d_0004_region22/21bc32e53a104942a278f5ecff75e896_i2_1.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0004_region22", "origin_question": "Remember, have you seen any towel(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any towel(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 3 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "1mp3d_0004_region22/60640d7394b94da68c549715eed7e83d_i1_0.jpg", "1mp3d_0004_region22/60640d7394b94da68c549715eed7e83d_i2_0.jpg", "1mp3d_0004_region22/60640d7394b94da68c549715eed7e83d_i1_5.jpg", "1mp3d_0004_region22/60640d7394b94da68c549715eed7e83d_i2_5.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0004_region22", "origin_question": "In which turn did you see both \"the plant\" and \"the window\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the plant\" and \"the window\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 3 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0004_region22/21bc32e53a104942a278f5ecff75e896_i1_5.jpg", "1mp3d_0004_region22/21bc32e53a104942a278f5ecff75e896_i2_5.jpg", "1mp3d_0004_region22/f68d84f143034de196b5b7f5b91fb28f_i1_3.jpg", "1mp3d_0004_region22/f68d84f143034de196b5b7f5b91fb28f_i2_3.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0004_region22", "origin_question": "", "option": [], "answer": "", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 3 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0246_00/00000.jpg", "scene0246_00/00040.jpg", "scene0246_00/00080.jpg", "scene0246_00/00120.jpg", "scene0246_00/00160.jpg" ], "turn_id": 1, "scan_id": "scene0246_00", "origin_question": "Remember, have you seen any dresser(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any dresser(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0246_00/00200.jpg", "scene0246_00/00240.jpg", "scene0246_00/00280.jpg", "scene0246_00/00320.jpg", "scene0246_00/00360.jpg" ], "turn_id": 2, "scan_id": "scene0246_00", "origin_question": "Which two objects are on the same side of you now? \"the rectangular white bed with a headboard and footboard\", \"the wardrobe\", \"the vibrant red curtain made from smooth fabric\". ", "option": [ "the rectangular white bed with a headboard and footboard, the wardrobe", "the rectangular white bed with a headboard and footboard, the vibrant red curtain made from smooth fabric", "the wardrobe, the vibrant red curtain made from smooth fabric" ], "answer": "the rectangular white bed with a headboard and footboard, the wardrobe", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the rectangular white bed with a headboard and footboard\", \"the wardrobe\", \"the vibrant red curtain made from smooth fabric\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular white bed with a headboard and footboard, the wardrobe', 'the rectangular white bed with a headboard and footboard, the vibrant red curtain made from smooth fabric', 'the wardrobe, the vibrant red curtain made from smooth fabric'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0246_00/00400.jpg", "scene0246_00/00440.jpg", "scene0246_00/00480.jpg", "scene0246_00/00520.jpg", "scene0246_00/00560.jpg" ], "turn_id": 3, "scan_id": "scene0246_00", "origin_question": "Which two objects are on the same side of you now? \"the large round white light\", \"the white rectangular minimalistic cabinet with a glossy finish\", \"the vibrant red curtain made from smooth fabric\". ", "option": [ "the large round white light, the white rectangular minimalistic cabinet with a glossy finish", "the large round white light, the vibrant red curtain made from smooth fabric", "the white rectangular minimalistic cabinet with a glossy finish, the vibrant red curtain made from smooth fabric" ], "answer": "the large round white light, the white rectangular minimalistic cabinet with a glossy finish", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the large round white light\", \"the white rectangular minimalistic cabinet with a glossy finish\", \"the vibrant red curtain made from smooth fabric\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the large round white light, the white rectangular minimalistic cabinet with a glossy finish', 'the large round white light, the vibrant red curtain made from smooth fabric', 'the white rectangular minimalistic cabinet with a glossy finish, the vibrant red curtain made from smooth fabric'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0246_00/00600.jpg", "scene0246_00/00640.jpg", "scene0246_00/00680.jpg", "scene0246_00/00720.jpg", "scene0246_00/00760.jpg" ], "turn_id": 4, "scan_id": "scene0246_00", "origin_question": "Which two objects are on the same side of you now? \"the vibrant red curtain made from smooth fabric\", \"the large rectangular white sink with a smooth surface\", \"the white rectangular nine-drawer dresser\". ", "option": [ "the vibrant red curtain made from smooth fabric, the large rectangular white sink with a smooth surface", "the vibrant red curtain made from smooth fabric, the white rectangular nine-drawer dresser", "the large rectangular white sink with a smooth surface, the white rectangular nine-drawer dresser" ], "answer": "the vibrant red curtain made from smooth fabric, the large rectangular white sink with a smooth surface", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the vibrant red curtain made from smooth fabric\", \"the large rectangular white sink with a smooth surface\", \"the white rectangular nine-drawer dresser\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the vibrant red curtain made from smooth fabric, the large rectangular white sink with a smooth surface', 'the vibrant red curtain made from smooth fabric, the white rectangular nine-drawer dresser', 'the large rectangular white sink with a smooth surface, the white rectangular nine-drawer dresser'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0246_00/00800.jpg", "scene0246_00/00840.jpg", "scene0246_00/00880.jpg", "scene0246_00/00920.jpg", "scene0246_00/00960.jpg" ], "turn_id": 5, "scan_id": "scene0246_00", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the white rectangular laminated table with a modern look\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the white rectangular laminated table with a modern look\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0246_00/01000.jpg", "scene0246_00/01040.jpg", "scene0246_00/01080.jpg", "scene0246_00/01120.jpg", "scene0246_00/01160.jpg" ], "turn_id": 6, "scan_id": "scene0246_00", "origin_question": "In which turn were you closest to \"the toilet\"? ", "option": [], "answer": "6", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the toilet\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0246_00/01200.jpg", "scene0246_00/01240.jpg", "scene0246_00/01280.jpg", "scene0246_00/01320.jpg", "scene0246_00/01360.jpg" ], "turn_id": 7, "scan_id": "scene0246_00", "origin_question": "Which direction is the white rectangular laminated table with a modern look in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the white rectangular laminated table with a modern look in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0246_00/01400.jpg", "scene0246_00/01440.jpg", "scene0246_00/01480.jpg", "scene0246_00/01520.jpg", "scene0246_00/01560.jpg" ], "turn_id": 8, "scan_id": "scene0246_00", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the rectangular white wooden stand\" from your position?", "option": [], "answer": "95.21", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the rectangular white wooden stand\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0246_00/01600.jpg", "scene0246_00/01640.jpg", "scene0246_00/01680.jpg", "scene0246_00/01720.jpg", "scene0246_00/01760.jpg" ], "turn_id": 9, "scan_id": "scene0246_00", "origin_question": "Is the wardrobe to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the wardrobe to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0246_00/01800.jpg", "scene0246_00/01840.jpg", "scene0246_00/01880.jpg", "scene0246_00/01920.jpg", "scene0246_00/01960.jpg" ], "turn_id": 10, "scan_id": "scene0246_00", "origin_question": "Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the large rectangular white sink with a smooth surface\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the large rectangular white sink with a smooth surface\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0664_01/00000.jpg", "scene0664_01/00040.jpg", "scene0664_01/00080.jpg", "scene0664_01/00120.jpg" ], "turn_id": 1, "scan_id": "scene0664_01", "origin_question": "Remember, have you seen any toilet paper(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any toilet paper(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0664_01/00160.jpg", "scene0664_01/00200.jpg", "scene0664_01/00240.jpg", "scene0664_01/00280.jpg" ], "turn_id": 2, "scan_id": "scene0664_01", "origin_question": "In which turn did you see both \"the standard white oval-shaped toilet made of porcelain\" and \"the modern round shower made of metal\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the standard white oval-shaped toilet made of porcelain\" and \"the modern round shower made of metal\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0664_01/00320.jpg", "scene0664_01/00360.jpg", "scene0664_01/00400.jpg", "scene0664_01/00440.jpg" ], "turn_id": 3, "scan_id": "scene0664_01", "origin_question": "Which direction is the cylindrical stainless steel bin with a curved top in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the cylindrical stainless steel bin with a curved top in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0664_01/00480.jpg", "scene0664_01/00520.jpg", "scene0664_01/00560.jpg", "scene0664_01/00600.jpg" ], "turn_id": 4, "scan_id": "scene0664_01", "origin_question": "What will be the first-time appearance order of the following categories: bin, curtain, picture?", "option": [ "bin, curtain, picture", "picture, curtain, bin", "curtain, bin, picture", "bin, picture, curtain" ], "answer": "bin, picture, curtain", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bin, curtain, picture? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['bin, curtain, picture', 'picture, curtain, bin', 'curtain, bin, picture', 'bin, picture, curtain'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "scene0664_01/00640.jpg", "scene0664_01/00680.jpg", "scene0664_01/00720.jpg", "scene0664_01/00760.jpg" ], "turn_id": 5, "scan_id": "scene0664_01", "origin_question": "Among these three objects(\"the standard white oval-shaped toilet made of porcelain\",\"the cylindrical stainless steel bin with a curved top\",\"the square, vividly colored picture with a black frame\" ), which one is the closest to you?", "option": [ "the standard white oval-shaped toilet made of porcelain", "the cylindrical stainless steel bin with a curved top", "the square, vividly colored picture with a black frame" ], "answer": "the square, vividly colored picture with a black frame", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects(\"the standard white oval-shaped toilet made of porcelain\",\"the cylindrical stainless steel bin with a curved top\",\"the square, vividly colored picture with a black frame\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the standard white oval-shaped toilet made of porcelain', 'the cylindrical stainless steel bin with a curved top', 'the square, vividly colored picture with a black frame'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0664_01/00800.jpg", "scene0664_01/00840.jpg", "scene0664_01/00880.jpg", "scene0664_01/00920.jpg" ], "turn_id": 6, "scan_id": "scene0664_01", "origin_question": "Is the square, vividly colored picture with a black frame to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the square, vividly colored picture with a black frame to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0664_01/00960.jpg", "scene0664_01/01000.jpg", "scene0664_01/01040.jpg", "scene0664_01/01080.jpg" ], "turn_id": 7, "scan_id": "scene0664_01", "origin_question": "Which direction is the square, vividly colored picture with a black frame in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the square, vividly colored picture with a black frame in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0664_01/01120.jpg", "scene0664_01/01160.jpg", "scene0664_01/01200.jpg", "scene0664_01/01240.jpg" ], "turn_id": 8, "scan_id": "scene0664_01", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the cylindrical stainless steel bin with a curved top\"; \"the counter\"; \"the tissue\" ", "option": [ "the cylindrical stainless steel bin with a curved top", "the counter", "the tissue" ], "answer": "the tissue", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the cylindrical stainless steel bin with a curved top\"; \"the counter\"; \"the tissue\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the cylindrical stainless steel bin with a curved top', 'the counter', 'the tissue'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0067_region3/0d258e8698254e6db674bd4bb11ed8d2_i1_5.jpg", "1mp3d_0067_region3/0d258e8698254e6db674bd4bb11ed8d2_i2_5.jpg", "1mp3d_0067_region3/fb6bdce12ffa4d87a740d226ae62c38c_i1_5.jpg", "1mp3d_0067_region3/fb6bdce12ffa4d87a740d226ae62c38c_i2_5.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0067_region3", "origin_question": "Remember, have you seen any stairs(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any stairs(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "1mp3d_0067_region3/0d258e8698254e6db674bd4bb11ed8d2_i1_1.jpg", "1mp3d_0067_region3/0d258e8698254e6db674bd4bb11ed8d2_i2_1.jpg", "1mp3d_0067_region3/fb6bdce12ffa4d87a740d226ae62c38c_i1_0.jpg", "1mp3d_0067_region3/fb6bdce12ffa4d87a740d226ae62c38c_i2_0.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0067_region3", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the computer\" from your position?", "option": [], "answer": "62.34", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the computer\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "1mp3d_0067_region3/fb6bdce12ffa4d87a740d226ae62c38c_i1_1.jpg", "1mp3d_0067_region3/fb6bdce12ffa4d87a740d226ae62c38c_i2_1.jpg", "1mp3d_0067_region3/fb6bdce12ffa4d87a740d226ae62c38c_i1_2.jpg", "1mp3d_0067_region3/fb6bdce12ffa4d87a740d226ae62c38c_i2_2.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0067_region3", "origin_question": "Please recall \"the computer\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.34", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the computer\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "1mp3d_0067_region3/0d258e8698254e6db674bd4bb11ed8d2_i1_3.jpg", "1mp3d_0067_region3/0d258e8698254e6db674bd4bb11ed8d2_i2_3.jpg", "1mp3d_0067_region3/5a958f240c5e420488d1129208451602_i1_1.jpg", "1mp3d_0067_region3/5a958f240c5e420488d1129208451602_i2_1.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0067_region3", "origin_question": "At the end of which turn were both of \"the standard brown wooden door with a dark handle\" and \"the silver rectangular ladder with orange handrails\" on your left side?", "option": [], "answer": "4", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the standard brown wooden door with a dark handle\" and \"the silver rectangular ladder with orange handrails\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "1mp3d_0067_region3/164adb0bf2464fa38e98e03b6cbfd3ff_i1_5.jpg", "1mp3d_0067_region3/164adb0bf2464fa38e98e03b6cbfd3ff_i2_5.jpg", "1mp3d_0067_region3/5a958f240c5e420488d1129208451602_i2_2.jpg", "1mp3d_0067_region3/164adb0bf2464fa38e98e03b6cbfd3ff_i1_2.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0067_region3", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the silver rectangular ladder with orange handrails\" from your position?", "option": [], "answer": "136.11", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the silver rectangular ladder with orange handrails\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0545_00/00000.jpg", "scene0545_00/00030.jpg", "scene0545_00/00060.jpg", "scene0545_00/00090.jpg" ], "turn_id": 1, "scan_id": "scene0545_00", "origin_question": "Remember, have you seen any shelf(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any shelf(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0545_00/00120.jpg", "scene0545_00/00150.jpg", "scene0545_00/00180.jpg", "scene0545_00/00210.jpg" ], "turn_id": 2, "scan_id": "scene0545_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0545_00/00240.jpg", "scene0545_00/00270.jpg", "scene0545_00/00300.jpg", "scene0545_00/00330.jpg" ], "turn_id": 3, "scan_id": "scene0545_00", "origin_question": "When did you first discover \"the rectangular white metal radiator with ventilation design\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the rectangular white metal radiator with ventilation design\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0545_00/00360.jpg", "scene0545_00/00390.jpg", "scene0545_00/00420.jpg", "scene0545_00/00450.jpg" ], "turn_id": 4, "scan_id": "scene0545_00", "origin_question": "Is the light brown rectangular wooden desk to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the light brown rectangular wooden desk to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0545_00/00480.jpg", "scene0545_00/00510.jpg", "scene0545_00/00540.jpg", "scene0545_00/00570.jpg" ], "turn_id": 5, "scan_id": "scene0545_00", "origin_question": "In which turn were you closest to \"the light brown rectangular wooden desk\"? ", "option": [], "answer": "2", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the light brown rectangular wooden desk\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0545_00/00600.jpg", "scene0545_00/00630.jpg", "scene0545_00/00660.jpg", "scene0545_00/00690.jpg" ], "turn_id": 6, "scan_id": "scene0545_00", "origin_question": "Please recall \"the window\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.58", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the window\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0545_00/00720.jpg", "scene0545_00/00750.jpg", "scene0545_00/00780.jpg", "scene0545_00/00810.jpg" ], "turn_id": 7, "scan_id": "scene0545_00", "origin_question": "Which direction is the curtain in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the curtain in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0095_01/00000.jpg", "scene0095_01/00040.jpg", "scene0095_01/00080.jpg", "scene0095_01/00120.jpg" ], "turn_id": 1, "scan_id": "scene0095_01", "origin_question": "Remember, have you seen any table(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any table(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0095_01/00160.jpg", "scene0095_01/00200.jpg", "scene0095_01/00240.jpg", "scene0095_01/00280.jpg" ], "turn_id": 2, "scan_id": "scene0095_01", "origin_question": "In which turn were you closest to \"the white, rectangular metal radiator with a vertical design\"? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the white, rectangular metal radiator with a vertical design\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0095_01/00320.jpg", "scene0095_01/00360.jpg", "scene0095_01/00400.jpg", "scene0095_01/00440.jpg" ], "turn_id": 3, "scan_id": "scene0095_01", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "1.4", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0095_01/00480.jpg", "scene0095_01/00520.jpg", "scene0095_01/00560.jpg", "scene0095_01/00600.jpg" ], "turn_id": 4, "scan_id": "scene0095_01", "origin_question": "Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the rectangular light brown wooden table with a smooth surface\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the rectangular light brown wooden table with a smooth surface\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0095_01/00640.jpg", "scene0095_01/00680.jpg", "scene0095_01/00720.jpg", "scene0095_01/00760.jpg" ], "turn_id": 5, "scan_id": "scene0095_01", "origin_question": "When did you first discover \"the rectangular light brown wooden table with a smooth surface\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the rectangular light brown wooden table with a smooth surface\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0095_01/00800.jpg", "scene0095_01/00840.jpg", "scene0095_01/00880.jpg", "scene0095_01/00920.jpg" ], "turn_id": 6, "scan_id": "scene0095_01", "origin_question": "In which turn did you see both \"the white, rectangular metal radiator with a vertical design\" and \"the window\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the white, rectangular metal radiator with a vertical design\" and \"the window\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0095_01/00960.jpg", "scene0095_01/01000.jpg", "scene0095_01/01040.jpg", "scene0095_01/01080.jpg" ], "turn_id": 7, "scan_id": "scene0095_01", "origin_question": "Which direction is the white, rectangular metal radiator with a vertical design in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the white, rectangular metal radiator with a vertical design in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0095_01/01120.jpg", "scene0095_01/01160.jpg", "scene0095_01/01200.jpg", "scene0095_01/01240.jpg" ], "turn_id": 8, "scan_id": "scene0095_01", "origin_question": "Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the white, rectangular metal radiator with a vertical design\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the white, rectangular metal radiator with a vertical design\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0095_01/01280.jpg", "scene0095_01/01320.jpg", "scene0095_01/01360.jpg", "scene0095_01/01400.jpg" ], "turn_id": 9, "scan_id": "scene0095_01", "origin_question": "By the end of the 7 turn, were you getting closer to or farther away from these two objects: X: \"the white, rectangular metal radiator with a vertical design\", Y: \"the window\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 9 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 7 turn, were you getting closer to or farther away from these two objects: X: \"the white, rectangular metal radiator with a vertical design\", Y: \"the window\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0469_02/00000.jpg", "scene0469_02/00030.jpg", "scene0469_02/00060.jpg", "scene0469_02/00090.jpg" ], "turn_id": 1, "scan_id": "scene0469_02", "origin_question": "Remember, have you seen any water cooler(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any water cooler(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0469_02/00120.jpg", "scene0469_02/00150.jpg", "scene0469_02/00180.jpg", "scene0469_02/00210.jpg" ], "turn_id": 2, "scan_id": "scene0469_02", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0469_02/00240.jpg", "scene0469_02/00270.jpg", "scene0469_02/00300.jpg", "scene0469_02/00330.jpg" ], "turn_id": 3, "scan_id": "scene0469_02", "origin_question": "What will be the first-time appearance order of the following categories: bag, bin, book?", "option": [ "bin, book, bag", "book, bag, bin", "bag, book, bin", "bin, bag, book" ], "answer": "bin, book, bag", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bag, bin, book? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['bin, book, bag', 'book, bag, bin', 'bag, book, bin', 'bin, bag, book'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0469_02/00360.jpg", "scene0469_02/00390.jpg", "scene0469_02/00420.jpg", "scene0469_02/00450.jpg" ], "turn_id": 4, "scan_id": "scene0469_02", "origin_question": "Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "119.82", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0469_02/00480.jpg", "scene0469_02/00510.jpg", "scene0469_02/00540.jpg", "scene0469_02/00570.jpg" ], "turn_id": 5, "scan_id": "scene0469_02", "origin_question": "Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the window\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the window\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0451_01/00000.jpg", "scene0451_01/00040.jpg", "scene0451_01/00080.jpg", "scene0451_01/00120.jpg", "scene0451_01/00160.jpg" ], "turn_id": 1, "scan_id": "scene0451_01", "origin_question": "Remember, how many jacket(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many jacket(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0451_01/00200.jpg", "scene0451_01/00240.jpg", "scene0451_01/00280.jpg", "scene0451_01/00320.jpg", "scene0451_01/00360.jpg" ], "turn_id": 2, "scan_id": "scene0451_01", "origin_question": "Which two objects are on the same side of you now? \"the rectangular floral-patterned couch\", \"the small rectangular white glass window\", \"the rectangular picture in an elegant brown wooden frame\". ", "option": [ "the rectangular floral-patterned couch, the small rectangular white glass window", "the rectangular floral-patterned couch, the rectangular picture in an elegant brown wooden frame", "the small rectangular white glass window, the rectangular picture in an elegant brown wooden frame" ], "answer": "the rectangular floral-patterned couch, the small rectangular white glass window", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the rectangular floral-patterned couch\", \"the small rectangular white glass window\", \"the rectangular picture in an elegant brown wooden frame\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular floral-patterned couch, the small rectangular white glass window', 'the rectangular floral-patterned couch, the rectangular picture in an elegant brown wooden frame', 'the small rectangular white glass window, the rectangular picture in an elegant brown wooden frame'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0451_01/00400.jpg", "scene0451_01/00440.jpg", "scene0451_01/00480.jpg", "scene0451_01/00520.jpg", "scene0451_01/00560.jpg" ], "turn_id": 3, "scan_id": "scene0451_01", "origin_question": "Please recall \"the white rectangular metal radiator\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.42", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the white rectangular metal radiator\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0451_01/00600.jpg", "scene0451_01/00640.jpg", "scene0451_01/00680.jpg", "scene0451_01/00720.jpg", "scene0451_01/00760.jpg" ], "turn_id": 4, "scan_id": "scene0451_01", "origin_question": "Which direction is the deep brown rectangular wooden bench in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the deep brown rectangular wooden bench in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0451_01/00800.jpg", "scene0451_01/00840.jpg", "scene0451_01/00880.jpg", "scene0451_01/00920.jpg", "scene0451_01/00960.jpg" ], "turn_id": 5, "scan_id": "scene0451_01", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "1.47", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0451_01/01000.jpg", "scene0451_01/01040.jpg", "scene0451_01/01080.jpg", "scene0451_01/01120.jpg", "scene0451_01/01160.jpg" ], "turn_id": 6, "scan_id": "scene0451_01", "origin_question": "Is the cup to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the cup to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0451_01/01200.jpg", "scene0451_01/01240.jpg", "scene0451_01/01280.jpg", "scene0451_01/01320.jpg", "scene0451_01/01360.jpg" ], "turn_id": 7, "scan_id": "scene0451_01", "origin_question": "By the end of the 3 turn, were you getting closer to or farther away from these two objects: X: \"the white plastic bag with two handles\", Y: \"the white rectangular metal radiator\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to X and farther away from Y", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 3 turn, were you getting closer to or farther away from these two objects: X: \"the white plastic bag with two handles\", Y: \"the white rectangular metal radiator\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0451_01/01400.jpg", "scene0451_01/01440.jpg", "scene0451_01/01480.jpg", "scene0451_01/01520.jpg", "scene0451_01/01560.jpg" ], "turn_id": 8, "scan_id": "scene0451_01", "origin_question": "Please recall \"the cup\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "4.73", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the cup\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0451_01/01600.jpg", "scene0451_01/01640.jpg", "scene0451_01/01680.jpg", "scene0451_01/01720.jpg", "scene0451_01/01760.jpg" ], "turn_id": 9, "scan_id": "scene0451_01", "origin_question": "What will be the first-time appearance order of the following categories: bag, couch, radiator?", "option": [ "bag, couch, radiator", "bag, radiator, couch", "couch, bag, radiator", "radiator, bag, couch" ], "answer": "couch, bag, radiator", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bag, couch, radiator? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['bag, couch, radiator', 'bag, radiator, couch', 'couch, bag, radiator', 'radiator, bag, couch'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "1mp3d_0084_region9/f0fa3bc756854f7280ca118d658bdaee_i1_1.jpg", "1mp3d_0084_region9/f0fa3bc756854f7280ca118d658bdaee_i2_1.jpg", "1mp3d_0084_region9/f0fa3bc756854f7280ca118d658bdaee_i1_0.jpg", "1mp3d_0084_region9/f0fa3bc756854f7280ca118d658bdaee_i2_0.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0084_region9", "origin_question": "Remember, how many stand(s) have you seen so far? ", "option": [], "answer": "3", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many stand(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "1mp3d_0084_region9/abc8f6c9abf44c2cb2509ddc331f948f_i1_5.jpg", "1mp3d_0084_region9/abc8f6c9abf44c2cb2509ddc331f948f_i2_5.jpg", "1mp3d_0084_region9/abc8f6c9abf44c2cb2509ddc331f948f_i1_4.jpg", "1mp3d_0084_region9/abc8f6c9abf44c2cb2509ddc331f948f_i2_4.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0084_region9", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular white wooden door with a white doorknob\", Y: \"the small rectangular wooden box in deep brown\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to Y and farther away from X", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular white wooden door with a white doorknob\", Y: \"the small rectangular wooden box in deep brown\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "1mp3d_0084_region9/8a27e5d9583e4459a0db21d42d8f0249_i1_5.jpg", "1mp3d_0084_region9/8a27e5d9583e4459a0db21d42d8f0249_i2_5.jpg", "1mp3d_0084_region9/8a27e5d9583e4459a0db21d42d8f0249_i1_4.jpg", "1mp3d_0084_region9/8a27e5d9583e4459a0db21d42d8f0249_i2_4.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0084_region9", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "1mp3d_0084_region9/8a27e5d9583e4459a0db21d42d8f0249_i1_3.jpg", "1mp3d_0084_region9/8a27e5d9583e4459a0db21d42d8f0249_i2_3.jpg", "1mp3d_0084_region9/8a27e5d9583e4459a0db21d42d8f0249_i1_2.jpg", "1mp3d_0084_region9/8a27e5d9583e4459a0db21d42d8f0249_i2_2.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0084_region9", "origin_question": "Which two objects are on the same side of you now? \"the cylindrical metal light with a warm amber glow\", \"the rectangular light brown wooden cabinet with drawers\", \"the small rectangular wooden box in deep brown\". ", "option": [ "the cylindrical metal light with a warm amber glow, the rectangular light brown wooden cabinet with drawers", "the cylindrical metal light with a warm amber glow, the small rectangular wooden box in deep brown", "the rectangular light brown wooden cabinet with drawers, the small rectangular wooden box in deep brown" ], "answer": "the rectangular light brown wooden cabinet with drawers, the small rectangular wooden box in deep brown", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the cylindrical metal light with a warm amber glow\", \"the rectangular light brown wooden cabinet with drawers\", \"the small rectangular wooden box in deep brown\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the cylindrical metal light with a warm amber glow, the rectangular light brown wooden cabinet with drawers', 'the cylindrical metal light with a warm amber glow, the small rectangular wooden box in deep brown', 'the rectangular light brown wooden cabinet with drawers, the small rectangular wooden box in deep brown'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "1mp3d_0084_region9/8a27e5d9583e4459a0db21d42d8f0249_i1_1.jpg", "1mp3d_0084_region9/8a27e5d9583e4459a0db21d42d8f0249_i2_1.jpg", "1mp3d_0084_region9/8a27e5d9583e4459a0db21d42d8f0249_i2_0.jpg", "1mp3d_0084_region9/c8a362a30cc24a5f952f1818a5f28048_i1_1.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0084_region9", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the rectangular light brown wooden cabinet with drawers\" from your position?", "option": [], "answer": "82.1", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the rectangular light brown wooden cabinet with drawers\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "1mp3d_0084_region9/c8a362a30cc24a5f952f1818a5f28048_i2_1.jpg", "1mp3d_0084_region9/5228fb701a24444792589b362ee1a1ed_i2_1.jpg", "1mp3d_0084_region9/5228fb701a24444792589b362ee1a1ed_i1_2.jpg", "1mp3d_0084_region9/5228fb701a24444792589b362ee1a1ed_i2_2.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0084_region9", "origin_question": "Among these three objects(\"the tall, slender plant with rich brown color\",\"the rectangular light brown wooden cabinet with drawers\",\"the small rectangular wooden box in deep brown\" ), which one is the closest to you?", "option": [ "the tall, slender plant with rich brown color", "the rectangular light brown wooden cabinet with drawers", "the small rectangular wooden box in deep brown" ], "answer": "the tall, slender plant with rich brown color", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects(\"the tall, slender plant with rich brown color\",\"the rectangular light brown wooden cabinet with drawers\",\"the small rectangular wooden box in deep brown\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the tall, slender plant with rich brown color', 'the rectangular light brown wooden cabinet with drawers', 'the small rectangular wooden box in deep brown'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "1mp3d_0084_region9/5228fb701a24444792589b362ee1a1ed_i1_3.jpg", "1mp3d_0084_region9/5228fb701a24444792589b362ee1a1ed_i2_3.jpg", "1mp3d_0084_region9/c8a362a30cc24a5f952f1818a5f28048_i1_2.jpg", "1mp3d_0084_region9/c8a362a30cc24a5f952f1818a5f28048_i2_2.jpg" ], "turn_id": 7, "scan_id": "1mp3d_0084_region9", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the cylindrical metal light with a warm amber glow\" from your position?", "option": [], "answer": "71.31", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the cylindrical metal light with a warm amber glow\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0118_01/00000.jpg", "scene0118_01/00030.jpg", "scene0118_01/00060.jpg", "scene0118_01/00090.jpg" ], "turn_id": 1, "scan_id": "scene0118_01", "origin_question": "Remember, have you seen any table(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any table(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0118_01/00120.jpg", "scene0118_01/00150.jpg", "scene0118_01/00180.jpg", "scene0118_01/00210.jpg" ], "turn_id": 2, "scan_id": "scene0118_01", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0118_01/00240.jpg", "scene0118_01/00270.jpg", "scene0118_01/00300.jpg", "scene0118_01/00330.jpg" ], "turn_id": 3, "scan_id": "scene0118_01", "origin_question": "Is the rectangular, deep orange hardcover book to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the rectangular, deep orange hardcover book to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0118_01/00360.jpg", "scene0118_01/00390.jpg", "scene0118_01/00420.jpg", "scene0118_01/00450.jpg" ], "turn_id": 4, "scan_id": "scene0118_01", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the rectangular, deep orange hardcover book\" from your position?", "option": [], "answer": "171.25", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the rectangular, deep orange hardcover book\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0208_00/00000.jpg", "scene0208_00/00040.jpg", "scene0208_00/00080.jpg", "scene0208_00/00120.jpg", "scene0208_00/00160.jpg" ], "turn_id": 1, "scan_id": "scene0208_00", "origin_question": "Remember, have you seen any screen(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any screen(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0208_00/00200.jpg", "scene0208_00/00240.jpg", "scene0208_00/00280.jpg", "scene0208_00/00320.jpg", "scene0208_00/00360.jpg" ], "turn_id": 2, "scan_id": "scene0208_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "scene0208_00/00400.jpg", "scene0208_00/00440.jpg", "scene0208_00/00480.jpg", "scene0208_00/00520.jpg", "scene0208_00/00560.jpg" ], "turn_id": 3, "scan_id": "scene0208_00", "origin_question": "Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance forward or backward from that position?", "option": [ "Forward", "Backward" ], "answer": "Forward", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance forward or backward from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Forward', 'Backward'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0208_00/00600.jpg", "scene0208_00/00640.jpg", "scene0208_00/00680.jpg", "scene0208_00/00720.jpg", "scene0208_00/00760.jpg" ], "turn_id": 4, "scan_id": "scene0208_00", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0208_00/00800.jpg", "scene0208_00/00840.jpg", "scene0208_00/00880.jpg", "scene0208_00/00920.jpg", "scene0208_00/00960.jpg" ], "turn_id": 5, "scan_id": "scene0208_00", "origin_question": "How far is your current position from where you were at the end of the turn 3? (in meters)", "option": [], "answer": "3.86", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 3? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0208_00/01000.jpg", "scene0208_00/01040.jpg", "scene0208_00/01080.jpg", "scene0208_00/01120.jpg", "scene0208_00/01160.jpg" ], "turn_id": 6, "scan_id": "scene0208_00", "origin_question": "Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the bench\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the bench\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0208_00/01200.jpg", "scene0208_00/01240.jpg", "scene0208_00/01280.jpg", "scene0208_00/01320.jpg", "scene0208_00/01360.jpg" ], "turn_id": 7, "scan_id": "scene0208_00", "origin_question": "How far is your current position from where you were at the end of the turn 4? (in meters)", "option": [], "answer": "1.94", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 4? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0208_00/01400.jpg", "scene0208_00/01440.jpg", "scene0208_00/01480.jpg", "scene0208_00/01520.jpg", "scene0208_00/01560.jpg" ], "turn_id": 8, "scan_id": "scene0208_00", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the bench\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the bench\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0208_00/01600.jpg", "scene0208_00/01640.jpg", "scene0208_00/01680.jpg", "scene0208_00/01720.jpg", "scene0208_00/01760.jpg" ], "turn_id": 9, "scan_id": "scene0208_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0208_00/01800.jpg", "scene0208_00/01840.jpg", "scene0208_00/01880.jpg", "scene0208_00/01920.jpg", "scene0208_00/01960.jpg" ], "turn_id": 10, "scan_id": "scene0208_00", "origin_question": "How far is your current position from where you were at the end of the turn 5? (in meters)", "option": [], "answer": "3.24", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 5? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "1mp3d_0040_region1/05a7eee165964cffb1a6ae6ace6a8c8a_i1_1.jpg", "1mp3d_0040_region1/05a7eee165964cffb1a6ae6ace6a8c8a_i2_1.jpg", "1mp3d_0040_region1/efd2f06dc9de45598b70828ba6168bd8_i1_0.jpg", "1mp3d_0040_region1/efd2f06dc9de45598b70828ba6168bd8_i2_0.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0040_region1", "origin_question": "Remember, how many door(s) have you seen so far? ", "option": [], "answer": "2", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many door(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "1mp3d_0040_region1/efd2f06dc9de45598b70828ba6168bd8_i1_1.jpg", "1mp3d_0040_region1/efd2f06dc9de45598b70828ba6168bd8_i2_1.jpg", "1mp3d_0040_region1/efd2f06dc9de45598b70828ba6168bd8_i1_3.jpg", "1mp3d_0040_region1/efd2f06dc9de45598b70828ba6168bd8_i2_3.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0040_region1", "origin_question": "Remember, how many pool table(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many pool table(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "1mp3d_0040_region1/efd2f06dc9de45598b70828ba6168bd8_i1_4.jpg", "1mp3d_0040_region1/efd2f06dc9de45598b70828ba6168bd8_i2_4.jpg", "1mp3d_0040_region1/05a7eee165964cffb1a6ae6ace6a8c8a_i1_3.jpg", "1mp3d_0040_region1/05a7eee165964cffb1a6ae6ace6a8c8a_i2_3.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0040_region1", "origin_question": "Remember, how many globe(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many globe(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0040_region1/b0e387fe3ac24aad80279f6485d31fd6_i1_4.jpg", "1mp3d_0040_region1/b0e387fe3ac24aad80279f6485d31fd6_i2_4.jpg", "1mp3d_0040_region1/b0e387fe3ac24aad80279f6485d31fd6_i1_3.jpg", "1mp3d_0040_region1/b0e387fe3ac24aad80279f6485d31fd6_i2_3.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0040_region1", "origin_question": "", "option": [], "answer": "", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0040_region1/b0e387fe3ac24aad80279f6485d31fd6_i1_2.jpg", "1mp3d_0040_region1/b0e387fe3ac24aad80279f6485d31fd6_i2_2.jpg", "1mp3d_0040_region1/4c9fab5cc3254cba83b3768b25f1b105_i1_3.jpg", "1mp3d_0040_region1/4c9fab5cc3254cba83b3768b25f1b105_i1_4.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0040_region1", "origin_question": "Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the long vertical metal fence with a dark finish\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the long vertical metal fence with a dark finish\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0040_region1/4c9fab5cc3254cba83b3768b25f1b105_i2_4.jpg", "1mp3d_0040_region1/8880585c8f53461fbe827765c2cb6fc7_i1_4.jpg", "1mp3d_0040_region1/8880585c8f53461fbe827765c2cb6fc7_i2_4.jpg", "1mp3d_0040_region1/9c242e45add84ffca6140c5eb7ae7d70_i1_4.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0040_region1", "origin_question": "", "option": [], "answer": "", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "1mp3d_0040_region1/9c242e45add84ffca6140c5eb7ae7d70_i2_4.jpg", "1mp3d_0040_region1/8880585c8f53461fbe827765c2cb6fc7_i1_3.jpg", "1mp3d_0040_region1/9c242e45add84ffca6140c5eb7ae7d70_i1_3.jpg", "1mp3d_0040_region1/86eba2e0b139463d9dc419ea6b71af87_i1_3.jpg" ], "turn_id": 7, "scan_id": "1mp3d_0040_region1", "origin_question": "Using your orientation at the end of turn 5 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "64.14", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 5 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0658_00/00000.jpg", "scene0658_00/00030.jpg", "scene0658_00/00060.jpg", "scene0658_00/00090.jpg" ], "turn_id": 1, "scan_id": "scene0658_00", "origin_question": "Remember, have you seen any bin(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any bin(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0658_00/00120.jpg", "scene0658_00/00150.jpg", "scene0658_00/00180.jpg", "scene0658_00/00210.jpg" ], "turn_id": 2, "scan_id": "scene0658_00", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the window\"; \"the sleek black desk with wood and metal materials\"; \"the single bed with a dark brown headboard and blue mattress\" ", "option": [ "the window", "the sleek black desk with wood and metal materials", "the single bed with a dark brown headboard and blue mattress" ], "answer": "the single bed with a dark brown headboard and blue mattress", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the window\"; \"the sleek black desk with wood and metal materials\"; \"the single bed with a dark brown headboard and blue mattress\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the window', 'the sleek black desk with wood and metal materials', 'the single bed with a dark brown headboard and blue mattress'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0658_00/00240.jpg", "scene0658_00/00270.jpg", "scene0658_00/00300.jpg", "scene0658_00/00330.jpg" ], "turn_id": 3, "scan_id": "scene0658_00", "origin_question": "In which turn were you closest to \"the rectangular white door with a smooth surface\"? ", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the rectangular white door with a smooth surface\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0658_00/00360.jpg", "scene0658_00/00390.jpg", "scene0658_00/00420.jpg", "scene0658_00/00450.jpg" ], "turn_id": 4, "scan_id": "scene0658_00", "origin_question": "How far is your current position from where you were at the end of the turn 1? (in meters)", "option": [], "answer": "1.47", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 1? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0579_01/00000.jpg", "scene0579_01/00030.jpg", "scene0579_01/00060.jpg", "scene0579_01/00090.jpg" ], "turn_id": 1, "scan_id": "scene0579_01", "origin_question": "Remember, how many sink(s) have you seen so far? ", "option": [], "answer": "3", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many sink(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0579_01/00120.jpg", "scene0579_01/00150.jpg", "scene0579_01/00180.jpg", "scene0579_01/00210.jpg" ], "turn_id": 2, "scan_id": "scene0579_01", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0579_01/00240.jpg", "scene0579_01/00270.jpg", "scene0579_01/00300.jpg", "scene0579_01/00330.jpg" ], "turn_id": 3, "scan_id": "scene0579_01", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0579_01/00370.jpg", "scene0579_01/00400.jpg", "scene0579_01/00430.jpg", "scene0579_01/00460.jpg" ], "turn_id": 4, "scan_id": "scene0579_01", "origin_question": "", "option": [], "answer": "", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0579_01/00490.jpg", "scene0579_01/00520.jpg", "scene0579_01/00550.jpg", "scene0579_01/00580.jpg" ], "turn_id": 5, "scan_id": "scene0579_01", "origin_question": "Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "152.31", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0579_01/00610.jpg", "scene0579_01/00640.jpg", "scene0579_01/00670.jpg", "scene0579_01/00700.jpg" ], "turn_id": 6, "scan_id": "scene0579_01", "origin_question": "How far is your current position from where you were at the end of the turn 5? (in meters)", "option": [], "answer": "1.56", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 5? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0030_region6/76dc52af56e842c5b53f25d93e938299_i1_4.jpg", "1mp3d_0030_region6/76dc52af56e842c5b53f25d93e938299_i2_4.jpg", "1mp3d_0030_region6/e4753c26daf64c8492bb3fad5d959fb3_i1_1.jpg", "1mp3d_0030_region6/e4753c26daf64c8492bb3fad5d959fb3_i2_1.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0030_region6", "origin_question": "Remember, have you seen any book(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any book(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "1mp3d_0030_region6/164618f724c64b43934cb95b11c89593_i1_2.jpg", "1mp3d_0030_region6/164618f724c64b43934cb95b11c89593_i2_2.jpg", "1mp3d_0030_region6/d89bee33427d45a5bd0cd5e1b4f8dd34_i1_4.jpg", "1mp3d_0030_region6/d89bee33427d45a5bd0cd5e1b4f8dd34_i2_4.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0030_region6", "origin_question": "At the end of which turn were both of \"the rectangular light brown wooden table with a smooth surface\" and \"the large white door made of wood\" on your right side?", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the rectangular light brown wooden table with a smooth surface\" and \"the large white door made of wood\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "1mp3d_0030_region6/164618f724c64b43934cb95b11c89593_i2_3.jpg", "1mp3d_0030_region6/e4753c26daf64c8492bb3fad5d959fb3_i1_2.jpg", "1mp3d_0030_region6/e4753c26daf64c8492bb3fad5d959fb3_i2_2.jpg", "1mp3d_0030_region6/e4753c26daf64c8492bb3fad5d959fb3_i1_3.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0030_region6", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "1mp3d_0030_region6/e4753c26daf64c8492bb3fad5d959fb3_i2_3.jpg", "1mp3d_0030_region6/e4753c26daf64c8492bb3fad5d959fb3_i1_4.jpg", "1mp3d_0030_region6/e4753c26daf64c8492bb3fad5d959fb3_i2_4.jpg", "1mp3d_0030_region6/76dc52af56e842c5b53f25d93e938299_i1_1.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0030_region6", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular light brown wooden table with a smooth surface\", Y: \"the large white door made of wood\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular light brown wooden table with a smooth surface\", Y: \"the large white door made of wood\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "1mp3d_0030_region6/76dc52af56e842c5b53f25d93e938299_i2_1.jpg", "1mp3d_0030_region6/76dc52af56e842c5b53f25d93e938299_i2_0.jpg", "1mp3d_0030_region6/be1a38f2bec74de79395d0a938234d38_i1_1.jpg", "1mp3d_0030_region6/be1a38f2bec74de79395d0a938234d38_i2_1.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0030_region6", "origin_question": "Please recall \"the vintage iron exercise equipment with a green frame\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.96", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the vintage iron exercise equipment with a green frame\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "1mp3d_0030_region6/2b4c37304fd54440a3d1e7fd223e26d5_i1_1.jpg", "1mp3d_0030_region6/2b4c37304fd54440a3d1e7fd223e26d5_i2_1.jpg", "1mp3d_0030_region6/be1a38f2bec74de79395d0a938234d38_i1_2.jpg", "1mp3d_0030_region6/be1a38f2bec74de79395d0a938234d38_i2_2.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0030_region6", "origin_question": "Please recall \"the large white door made of wood\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.81", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the large white door made of wood\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "1mp3d_0030_region6/2b4c37304fd54440a3d1e7fd223e26d5_i1_2.jpg", "1mp3d_0030_region6/2b4c37304fd54440a3d1e7fd223e26d5_i2_2.jpg", "1mp3d_0030_region6/448d573c3160494c8715d3e7314f4afb_i1_3.jpg", "1mp3d_0030_region6/448d573c3160494c8715d3e7314f4afb_i2_3.jpg" ], "turn_id": 7, "scan_id": "1mp3d_0030_region6", "origin_question": "Which direction is the deep brown wooden shelf with multiple tiers in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the deep brown wooden shelf with multiple tiers in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "1mp3d_0030_region6/448d573c3160494c8715d3e7314f4afb_i1_2.jpg", "1mp3d_0030_region6/448d573c3160494c8715d3e7314f4afb_i2_2.jpg", "1mp3d_0030_region6/32c98b5cfe804dd185a8cb6ecb91a63f_i1_2.jpg", "1mp3d_0030_region6/32c98b5cfe804dd185a8cb6ecb91a63f_i2_2.jpg" ], "turn_id": 8, "scan_id": "1mp3d_0030_region6", "origin_question": "Please recall \"the deep brown wooden shelf with multiple tiers\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.8", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the deep brown wooden shelf with multiple tiers\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0609_03/00000.jpg", "scene0609_03/00030.jpg", "scene0609_03/00060.jpg", "scene0609_03/00090.jpg" ], "turn_id": 1, "scan_id": "scene0609_03", "origin_question": "Remember, have you seen any chair(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any chair(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0609_03/00120.jpg", "scene0609_03/00150.jpg", "scene0609_03/00180.jpg", "scene0609_03/00210.jpg" ], "turn_id": 2, "scan_id": "scene0609_03", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0609_03/00240.jpg", "scene0609_03/00270.jpg", "scene0609_03/00300.jpg", "scene0609_03/00330.jpg" ], "turn_id": 3, "scan_id": "scene0609_03", "origin_question": "How far is your current position from where you were at the end of the turn 1? (in meters)", "option": [], "answer": "3.06", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 1? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0609_03/00360.jpg", "scene0609_03/00390.jpg", "scene0609_03/00420.jpg", "scene0609_03/00450.jpg" ], "turn_id": 4, "scan_id": "scene0609_03", "origin_question": "Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "42.79", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0609_03/00480.jpg", "scene0609_03/00510.jpg", "scene0609_03/00540.jpg", "scene0609_03/00570.jpg" ], "turn_id": 5, "scan_id": "scene0609_03", "origin_question": "Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the cylindrical, beige bin made of plastic or metal\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the cylindrical, beige bin made of plastic or metal\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0423_00/00000.jpg", "scene0423_00/00030.jpg", "scene0423_00/00060.jpg", "scene0423_00/00090.jpg" ], "turn_id": 1, "scan_id": "scene0423_00", "origin_question": "Remember, have you seen any body loofah(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any body loofah(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0423_00/00120.jpg", "scene0423_00/00150.jpg", "scene0423_00/00180.jpg", "scene0423_00/00210.jpg" ], "turn_id": 2, "scan_id": "scene0423_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "scene0423_00/00240.jpg", "scene0423_00/00270.jpg", "scene0423_00/00300.jpg", "scene0423_00/00330.jpg" ], "turn_id": 3, "scan_id": "scene0423_00", "origin_question": "Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Left", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0423_00/00360.jpg", "scene0423_00/00390.jpg", "scene0423_00/00420.jpg", "scene0423_00/00450.jpg" ], "turn_id": 4, "scan_id": "scene0423_00", "origin_question": "How far is your current position from where you were at the end of the turn 3? (in meters)", "option": [], "answer": "1.6", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 3? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0423_00/00480.jpg", "scene0423_00/00510.jpg", "scene0423_00/00540.jpg", "scene0423_00/00570.jpg" ], "turn_id": 5, "scan_id": "scene0423_00", "origin_question": "How far is your current position from where you were at the end of the turn 4? (in meters)", "option": [], "answer": "2.27", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 4? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0423_00/00600.jpg", "scene0423_00/00630.jpg", "scene0423_00/00660.jpg", "scene0423_00/00690.jpg" ], "turn_id": 6, "scan_id": "scene0423_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0538_00/00000.jpg", "scene0538_00/00040.jpg", "scene0538_00/00080.jpg", "scene0538_00/00120.jpg", "scene0538_00/00160.jpg" ], "turn_id": 1, "scan_id": "scene0538_00", "origin_question": "Remember, have you seen any bin(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any bin(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0538_00/00200.jpg", "scene0538_00/00240.jpg", "scene0538_00/00280.jpg", "scene0538_00/00320.jpg", "scene0538_00/00360.jpg" ], "turn_id": 2, "scan_id": "scene0538_00", "origin_question": "Which direction is the shower in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the shower in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0538_00/00400.jpg", "scene0538_00/00440.jpg", "scene0538_00/00480.jpg", "scene0538_00/00520.jpg", "scene0538_00/00560.jpg" ], "turn_id": 3, "scan_id": "scene0538_00", "origin_question": "In which turn were you closest to \"the soap dispenser\"? ", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the soap dispenser\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0538_00/00600.jpg", "scene0538_00/00640.jpg", "scene0538_00/00680.jpg", "scene0538_00/00720.jpg", "scene0538_00/00760.jpg" ], "turn_id": 4, "scan_id": "scene0538_00", "origin_question": "Remember, have you seen any sheet(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any sheet(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0538_00/00800.jpg", "scene0538_00/00840.jpg", "scene0538_00/00880.jpg", "scene0538_00/00920.jpg", "scene0538_00/00960.jpg" ], "turn_id": 5, "scan_id": "scene0538_00", "origin_question": "What will be the first-time appearance order of the following categories: curtain, sign, sink?", "option": [ "curtain, sign, sink", "sink, sign, curtain", "sign, curtain, sink", "curtain, sink, sign" ], "answer": "curtain, sink, sign", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: curtain, sign, sink? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['curtain, sign, sink', 'sink, sign, curtain', 'sign, curtain, sink', 'curtain, sink, sign'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0538_00/01000.jpg", "scene0538_00/01040.jpg", "scene0538_00/01080.jpg", "scene0538_00/01120.jpg", "scene0538_00/01160.jpg" ], "turn_id": 6, "scan_id": "scene0538_00", "origin_question": "Which direction is the smooth white rectangular ledge in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the smooth white rectangular ledge in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0538_00/01200.jpg", "scene0538_00/01240.jpg", "scene0538_00/01280.jpg", "scene0538_00/01320.jpg", "scene0538_00/01360.jpg" ], "turn_id": 7, "scan_id": "scene0538_00", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the rectangular yellow curtain made of soft fabric\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the rectangular yellow curtain made of soft fabric\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0538_00/01400.jpg", "scene0538_00/01440.jpg", "scene0538_00/01480.jpg", "scene0538_00/01520.jpg", "scene0538_00/01580.jpg" ], "turn_id": 8, "scan_id": "scene0538_00", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the gray oval plastic dispenser with a white handle\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the gray oval plastic dispenser with a white handle\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0538_00/01620.jpg", "scene0538_00/01660.jpg", "scene0538_00/01700.jpg", "scene0538_00/01740.jpg", "scene0538_00/01780.jpg" ], "turn_id": 9, "scan_id": "scene0538_00", "origin_question": "Which direction is the medium-sized rubber plunger with a yellow handle in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the medium-sized rubber plunger with a yellow handle in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0538_00/01820.jpg", "scene0538_00/01860.jpg", "scene0538_00/02050.jpg", "scene0538_00/02090.jpg", "scene0538_00/02130.jpg" ], "turn_id": 10, "scan_id": "scene0538_00", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0067_00/00000.jpg", "scene0067_00/00030.jpg", "scene0067_00/00060.jpg", "scene0067_00/00090.jpg" ], "turn_id": 1, "scan_id": "scene0067_00", "origin_question": "Remember, have you seen any toilet(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any toilet(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0067_00/00120.jpg", "scene0067_00/00150.jpg", "scene0067_00/00180.jpg", "scene0067_00/00210.jpg" ], "turn_id": 2, "scan_id": "scene0067_00", "origin_question": "When did you first discover \"the white ceramic toilet with a round bowl and rectangular tank\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the white ceramic toilet with a round bowl and rectangular tank\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0067_00/00240.jpg", "scene0067_00/00270.jpg", "scene0067_00/00300.jpg", "scene0067_00/00330.jpg" ], "turn_id": 3, "scan_id": "scene0067_00", "origin_question": "When did you first discover \"the medium-sized rectangular white door\"? (index of the turn)", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the medium-sized rectangular white door\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0067_00/00380.jpg", "scene0067_00/00410.jpg", "scene0067_00/00600.jpg", "scene0067_00/00630.jpg" ], "turn_id": 4, "scan_id": "scene0067_00", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the small rectangular white porcelain sink\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the small rectangular white porcelain sink\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0067_00/00660.jpg", "scene0067_00/00690.jpg", "scene0067_00/00720.jpg", "scene0067_00/00750.jpg" ], "turn_id": 5, "scan_id": "scene0067_00", "origin_question": "Is the grab bar to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the grab bar to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0067_00/00780.jpg", "scene0067_00/00810.jpg", "scene0067_00/00840.jpg", "scene0067_00/00870.jpg" ], "turn_id": 6, "scan_id": "scene0067_00", "origin_question": "In which turn did you see both \"the medium-sized rectangular white door\" and \"the small rectangular white porcelain sink\" simultaneously? ", "option": [], "answer": "4", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the medium-sized rectangular white door\" and \"the small rectangular white porcelain sink\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0067_00/00900.jpg", "scene0067_00/00930.jpg", "scene0067_00/00960.jpg", "scene0067_00/01000.jpg" ], "turn_id": 7, "scan_id": "scene0067_00", "origin_question": "Please recall \"the bin\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.63", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the bin\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0030_region13/bd01fe5cdc7f4d7580134b2a3ffd08a8_i1_3.jpg", "1mp3d_0030_region13/bd01fe5cdc7f4d7580134b2a3ffd08a8_i2_3.jpg", "1mp3d_0030_region13/38a021bef7634782b42c0583f84c6b8e_i1_1.jpg", "1mp3d_0030_region13/38a021bef7634782b42c0583f84c6b8e_i2_1.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0030_region13", "origin_question": "Remember, have you seen any chair(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any chair(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "1mp3d_0030_region13/f2c4c4feb71e4a2498a35dcef01aca5d_i1_4.jpg", "1mp3d_0030_region13/f2c4c4feb71e4a2498a35dcef01aca5d_i2_4.jpg", "1mp3d_0030_region13/ac32217f29704340bf467f51b7f15786_i1_4.jpg", "1mp3d_0030_region13/ac32217f29704340bf467f51b7f15786_i2_4.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0030_region13", "origin_question": "Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Right", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0030_region13/e58daa54fe36484984da0fc3e209ee39_i1_4.jpg", "1mp3d_0030_region13/e58daa54fe36484984da0fc3e209ee39_i2_4.jpg", "1mp3d_0030_region13/ac32217f29704340bf467f51b7f15786_i1_3.jpg", "1mp3d_0030_region13/ac32217f29704340bf467f51b7f15786_i2_3.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0030_region13", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the blue imitation horse toy with a rectangular shape\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the blue imitation horse toy with a rectangular shape\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "1mp3d_0030_region13/f2c4c4feb71e4a2498a35dcef01aca5d_i1_2.jpg", "1mp3d_0030_region13/f2c4c4feb71e4a2498a35dcef01aca5d_i2_2.jpg", "1mp3d_0030_region13/f2c4c4feb71e4a2498a35dcef01aca5d_i1_3.jpg", "1mp3d_0030_region13/f2c4c4feb71e4a2498a35dcef01aca5d_i2_3.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0030_region13", "origin_question": "In which turn were you closest to \"the round light brown wooden table\"? ", "option": [], "answer": "1", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the round light brown wooden table\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0030_region13/38a021bef7634782b42c0583f84c6b8e_i1_3.jpg", "1mp3d_0030_region13/38a021bef7634782b42c0583f84c6b8e_i2_3.jpg", "1mp3d_0030_region13/38a021bef7634782b42c0583f84c6b8e_i2_4.jpg", "1mp3d_0030_region13/bd01fe5cdc7f4d7580134b2a3ffd08a8_i1_4.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0030_region13", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the white ceiling fan with three plastic blades\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the white ceiling fan with three plastic blades\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "1mp3d_0030_region13/bd01fe5cdc7f4d7580134b2a3ffd08a8_i2_4.jpg", "1mp3d_0030_region13/4a3c031b6787433fb02da9d40a359c68_i1_4.jpg", "1mp3d_0030_region13/4a3c031b6787433fb02da9d40a359c68_i2_4.jpg", "1mp3d_0030_region13/4a3c031b6787433fb02da9d40a359c68_i1_3.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0030_region13", "origin_question": "When did you first discover \"the blue imitation horse toy with a rectangular shape\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the blue imitation horse toy with a rectangular shape\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0236_01/00000.jpg", "scene0236_01/00040.jpg", "scene0236_01/00080.jpg", "scene0236_01/00120.jpg", "scene0236_01/00160.jpg" ], "turn_id": 1, "scan_id": "scene0236_01", "origin_question": "Remember, how many treadmill(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many treadmill(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0236_01/00200.jpg", "scene0236_01/00240.jpg", "scene0236_01/00280.jpg", "scene0236_01/00320.jpg", "scene0236_01/00360.jpg" ], "turn_id": 2, "scan_id": "scene0236_01", "origin_question": "In which turn did you see both \"the rectangular light blue plastic tissue box\" and \"the rectangular, light-colored wooden dresser with multiple drawers\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the rectangular light blue plastic tissue box\" and \"the rectangular, light-colored wooden dresser with multiple drawers\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0236_01/00400.jpg", "scene0236_01/00440.jpg", "scene0236_01/00480.jpg", "scene0236_01/00520.jpg", "scene0236_01/00560.jpg" ], "turn_id": 3, "scan_id": "scene0236_01", "origin_question": "Is the rectangular picture with a pink border to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the rectangular picture with a pink border to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "scene0236_01/00600.jpg", "scene0236_01/00640.jpg", "scene0236_01/00680.jpg", "scene0236_01/00720.jpg", "scene0236_01/00760.jpg" ], "turn_id": 4, "scan_id": "scene0236_01", "origin_question": "Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Left", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0236_01/00800.jpg", "scene0236_01/00840.jpg", "scene0236_01/00880.jpg", "scene0236_01/00920.jpg", "scene0236_01/00960.jpg" ], "turn_id": 5, "scan_id": "scene0236_01", "origin_question": "When did you first discover \"the rectangular, light-colored wooden dresser with multiple drawers\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the rectangular, light-colored wooden dresser with multiple drawers\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0236_01/01000.jpg", "scene0236_01/01040.jpg", "scene0236_01/01080.jpg", "scene0236_01/01120.jpg", "scene0236_01/01160.jpg" ], "turn_id": 6, "scan_id": "scene0236_01", "origin_question": "When did you first discover \"the rectangular bed with gray and white bedding\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the rectangular bed with gray and white bedding\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0236_01/01200.jpg", "scene0236_01/01240.jpg", "scene0236_01/01280.jpg", "scene0236_01/01320.jpg", "scene0236_01/01360.jpg" ], "turn_id": 7, "scan_id": "scene0236_01", "origin_question": "When did you first discover \"the rectangular picture with a pink border\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the rectangular picture with a pink border\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0236_01/01400.jpg", "scene0236_01/01440.jpg", "scene0236_01/01480.jpg", "scene0236_01/01520.jpg", "scene0236_01/01560.jpg" ], "turn_id": 8, "scan_id": "scene0236_01", "origin_question": "In which turn did you see both \"the standard rectangular glass window with a light frame\" and \"the red, sleek plastic hair dryer\" simultaneously? ", "option": [], "answer": "8", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the standard rectangular glass window with a light frame\" and \"the red, sleek plastic hair dryer\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0040_region14/eea2e371b6ac43c4a72edf8546be9388_i2_5.jpg", "1mp3d_0040_region14/3e1701bd29544e47b96d79d370933af5_i2_0.jpg", "1mp3d_0040_region14/eea2e371b6ac43c4a72edf8546be9388_i1_4.jpg", "1mp3d_0040_region14/eea2e371b6ac43c4a72edf8546be9388_i2_4.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0040_region14", "origin_question": "Remember, have you seen any chair(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any chair(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "1mp3d_0040_region14/3e1701bd29544e47b96d79d370933af5_i1_2.jpg", "1mp3d_0040_region14/3e1701bd29544e47b96d79d370933af5_i2_2.jpg", "1mp3d_0040_region14/a4c2f102b23b4d4b87b21d3cb5d0cfa2_i1_3.jpg", "1mp3d_0040_region14/a4c2f102b23b4d4b87b21d3cb5d0cfa2_i2_3.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0040_region14", "origin_question": "Which direction is the rectangular wooden desk with a dark finish in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the rectangular wooden desk with a dark finish in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "1mp3d_0040_region14/77fa4b2db2cf45a0abff5833de1bc079_i1_2.jpg", "1mp3d_0040_region14/77fa4b2db2cf45a0abff5833de1bc079_i2_2.jpg", "1mp3d_0040_region14/a4c2f102b23b4d4b87b21d3cb5d0cfa2_i1_2.jpg", "1mp3d_0040_region14/a4c2f102b23b4d4b87b21d3cb5d0cfa2_i2_2.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0040_region14", "origin_question": "Which two objects are on the same side of you now? \"the bed\", \"the black chair with a mesh backrest\", \"the dark brown wooden door with an arched top\". ", "option": [ "the bed, the black chair with a mesh backrest", "the bed, the dark brown wooden door with an arched top", "the black chair with a mesh backrest, the dark brown wooden door with an arched top" ], "answer": "the bed, the black chair with a mesh backrest", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the bed\", \"the black chair with a mesh backrest\", \"the dark brown wooden door with an arched top\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the bed, the black chair with a mesh backrest', 'the bed, the dark brown wooden door with an arched top', 'the black chair with a mesh backrest, the dark brown wooden door with an arched top'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0040_region14/77fa4b2db2cf45a0abff5833de1bc079_i1_0.jpg", "1mp3d_0040_region14/77fa4b2db2cf45a0abff5833de1bc079_i2_0.jpg", "1mp3d_0040_region14/77fa4b2db2cf45a0abff5833de1bc079_i1_5.jpg", "1mp3d_0040_region14/77fa4b2db2cf45a0abff5833de1bc079_i2_5.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0040_region14", "origin_question": "", "option": [], "answer": "", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "1mp3d_0040_region14/77fa4b2db2cf45a0abff5833de1bc079_i1_1.jpg", "1mp3d_0040_region14/77fa4b2db2cf45a0abff5833de1bc079_i2_1.jpg", "1mp3d_0040_region14/b6169f1b5e0a42489ed145131e9ae486_i1_5.jpg", "1mp3d_0040_region14/b6169f1b5e0a42489ed145131e9ae486_i2_5.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0040_region14", "origin_question": "Which direction is the rectangular wooden desk with a dark finish in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the rectangular wooden desk with a dark finish in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0088_02/00000.jpg", "scene0088_02/00040.jpg", "scene0088_02/00080.jpg", "scene0088_02/00120.jpg", "scene0088_02/00160.jpg" ], "turn_id": 1, "scan_id": "scene0088_02", "origin_question": "Remember, have you seen any fireplace(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any fireplace(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0088_02/00200.jpg", "scene0088_02/00240.jpg", "scene0088_02/00280.jpg", "scene0088_02/00320.jpg", "scene0088_02/00360.jpg" ], "turn_id": 2, "scan_id": "scene0088_02", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the door\"; \"the rectangular black desktop computer made of plastic and metal\"; \"the rectangular wooden cabinet in good condition\" ", "option": [ "the door", "the rectangular black desktop computer made of plastic and metal", "the rectangular wooden cabinet in good condition" ], "answer": "the door", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the door\"; \"the rectangular black desktop computer made of plastic and metal\"; \"the rectangular wooden cabinet in good condition\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the door', 'the rectangular black desktop computer made of plastic and metal', 'the rectangular wooden cabinet in good condition'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0088_02/00400.jpg", "scene0088_02/00440.jpg", "scene0088_02/00480.jpg", "scene0088_02/00520.jpg", "scene0088_02/00560.jpg" ], "turn_id": 3, "scan_id": "scene0088_02", "origin_question": "Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "157.49", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "scene0088_02/00600.jpg", "scene0088_02/00640.jpg", "scene0088_02/00680.jpg", "scene0088_02/00720.jpg", "scene0088_02/00760.jpg" ], "turn_id": 4, "scan_id": "scene0088_02", "origin_question": "Assuming the direction you are facing at the end of the turn 3 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Left", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 3 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0088_02/00800.jpg", "scene0088_02/00840.jpg", "scene0088_02/00880.jpg", "scene0088_02/00920.jpg", "scene0088_02/00960.jpg" ], "turn_id": 5, "scan_id": "scene0088_02", "origin_question": "Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "135.05", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0088_02/01000.jpg", "scene0088_02/01040.jpg", "scene0088_02/01080.jpg", "scene0088_02/01120.jpg", "scene0088_02/01160.jpg" ], "turn_id": 6, "scan_id": "scene0088_02", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "scene0088_02/01200.jpg", "scene0088_02/01240.jpg", "scene0088_02/01280.jpg", "scene0088_02/01320.jpg", "scene0088_02/01360.jpg" ], "turn_id": 7, "scan_id": "scene0088_02", "origin_question": "Assuming the direction you are facing at the end of the turn 3 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Left", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 3 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0088_02/01400.jpg", "scene0088_02/01440.jpg", "scene0088_02/01480.jpg", "scene0088_02/01520.jpg", "scene0088_02/01560.jpg" ], "turn_id": 8, "scan_id": "scene0088_02", "origin_question": "In which turn were you closest to \"the eraser\"? ", "option": [], "answer": "8", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the eraser\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0088_02/01600.jpg", "scene0088_02/01640.jpg", "scene0088_02/01680.jpg", "scene0088_02/01720.jpg", "scene0088_02/01760.jpg" ], "turn_id": 9, "scan_id": "scene0088_02", "origin_question": "Please recall \"the eraser\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.69", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the eraser\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0088_02/01800.jpg", "scene0088_02/01840.jpg", "scene0088_02/01880.jpg", "scene0088_02/01920.jpg", "scene0088_02/01960.jpg" ], "turn_id": 10, "scan_id": "scene0088_02", "origin_question": "Compared to your position at the end of the 9 turn, are you now closer or farther away from \"the rectangular, standard-sized blackboard with a white surface\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 9 turn, are you now closer or farther away from \"the rectangular, standard-sized blackboard with a white surface\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0606_02/00000.jpg", "scene0606_02/00040.jpg", "scene0606_02/00080.jpg", "scene0606_02/00120.jpg", "scene0606_02/00160.jpg" ], "turn_id": 1, "scan_id": "scene0606_02", "origin_question": "Remember, have you seen any switch(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any switch(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0606_02/00200.jpg", "scene0606_02/00240.jpg", "scene0606_02/00280.jpg", "scene0606_02/00320.jpg", "scene0606_02/00360.jpg" ], "turn_id": 2, "scan_id": "scene0606_02", "origin_question": "In which turn were you closest to \"the brown rectangular wooden counter with a smooth surface\"? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the brown rectangular wooden counter with a smooth surface\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0606_02/00400.jpg", "scene0606_02/00440.jpg", "scene0606_02/00480.jpg", "scene0606_02/00520.jpg", "scene0606_02/00560.jpg" ], "turn_id": 3, "scan_id": "scene0606_02", "origin_question": "Remember, have you seen any bag(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any bag(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0606_02/00600.jpg", "scene0606_02/00640.jpg", "scene0606_02/00680.jpg", "scene0606_02/00720.jpg", "scene0606_02/00760.jpg" ], "turn_id": 4, "scan_id": "scene0606_02", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "44.9", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0606_02/00800.jpg", "scene0606_02/00840.jpg", "scene0606_02/00880.jpg", "scene0606_02/00920.jpg", "scene0606_02/00960.jpg" ], "turn_id": 5, "scan_id": "scene0606_02", "origin_question": "By the end of the 4 turn, were you getting closer to or farther away from these two objects: X: \"the large rectangular woven basket in brownish color\", Y: \"the moderate-sized rectangular metal tray with a matte surface\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "farther away from both X and Y", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 4 turn, were you getting closer to or farther away from these two objects: X: \"the large rectangular woven basket in brownish color\", Y: \"the moderate-sized rectangular metal tray with a matte surface\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0606_02/01000.jpg", "scene0606_02/01040.jpg", "scene0606_02/01080.jpg", "scene0606_02/01120.jpg", "scene0606_02/01160.jpg" ], "turn_id": 6, "scan_id": "scene0606_02", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the large rectangular woven basket in brownish color\" from your position?", "option": [], "answer": "153.58", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the large rectangular woven basket in brownish color\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0606_02/01200.jpg", "scene0606_02/01240.jpg", "scene0606_02/01280.jpg", "scene0606_02/01320.jpg", "scene0606_02/01360.jpg" ], "turn_id": 7, "scan_id": "scene0606_02", "origin_question": "In which turn were you closest to \"the rectangular, creamy soft pillow of standard size\"? ", "option": [], "answer": "7", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the rectangular, creamy soft pillow of standard size\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0606_02/01400.jpg", "scene0606_02/01440.jpg", "scene0606_02/01480.jpg", "scene0606_02/01520.jpg", "scene0606_02/01560.jpg" ], "turn_id": 8, "scan_id": "scene0606_02", "origin_question": "Please recall \"the soft, dark-colored couch with wooden legs\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.71", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the soft, dark-colored couch with wooden legs\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0606_02/01600.jpg", "scene0606_02/01640.jpg", "scene0606_02/01680.jpg", "scene0606_02/01720.jpg", "scene0606_02/01760.jpg" ], "turn_id": 9, "scan_id": "scene0606_02", "origin_question": "Is the blue rectangular plastic case to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the blue rectangular plastic case to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0606_02/01800.jpg", "scene0606_02/01840.jpg", "scene0606_02/01880.jpg", "scene0606_02/01920.jpg", "scene0606_02/01960.jpg" ], "turn_id": 10, "scan_id": "scene0606_02", "origin_question": "By the end of the 9 turn, were you getting closer to or farther away from these two objects: X: \"the large rectangular woven basket in brownish color\", Y: \"the cylindrical white tissue made of standard paper\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 9 turn, were you getting closer to or farther away from these two objects: X: \"the large rectangular woven basket in brownish color\", Y: \"the cylindrical white tissue made of standard paper\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0044_region7/4252403213214a11a8aaf4710d70fad6_i1_5.jpg", "1mp3d_0044_region7/4252403213214a11a8aaf4710d70fad6_i2_5.jpg", "1mp3d_0044_region7/93560cb960a34d2bbf43d45a18bbd62b_i1_2.jpg", "1mp3d_0044_region7/93560cb960a34d2bbf43d45a18bbd62b_i2_2.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0044_region7", "origin_question": "Remember, have you seen any arch(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any arch(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "1mp3d_0044_region7/4252403213214a11a8aaf4710d70fad6_i1_0.jpg", "1mp3d_0044_region7/4252403213214a11a8aaf4710d70fad6_i2_0.jpg", "1mp3d_0044_region7/4252403213214a11a8aaf4710d70fad6_i1_1.jpg", "1mp3d_0044_region7/4252403213214a11a8aaf4710d70fad6_i2_1.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0044_region7", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular, creamy white stone fireplace\", Y: \"the small circular metal decoration with black and green colors\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular, creamy white stone fireplace\", Y: \"the small circular metal decoration with black and green colors\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "1mp3d_0044_region7/4252403213214a11a8aaf4710d70fad6_i1_2.jpg", "1mp3d_0044_region7/4252403213214a11a8aaf4710d70fad6_i2_2.jpg", "1mp3d_0044_region7/e3ffb659fb744d33852508489d412649_i1_2.jpg", "1mp3d_0044_region7/e3ffb659fb744d33852508489d412649_i2_2.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0044_region7", "origin_question": "Which direction is the small circular metal decoration with black and green colors in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the small circular metal decoration with black and green colors in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "1mp3d_0044_region7/436cddf947a24e72834c4b8d18a138ca_i1_0.jpg", "1mp3d_0044_region7/436cddf947a24e72834c4b8d18a138ca_i2_0.jpg", "1mp3d_0044_region7/436cddf947a24e72834c4b8d18a138ca_i1_5.jpg", "1mp3d_0044_region7/436cddf947a24e72834c4b8d18a138ca_i2_5.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0044_region7", "origin_question": "Please recall \"the rectangular, creamy white stone fireplace\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.59", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the rectangular, creamy white stone fireplace\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "1mp3d_0044_region7/436cddf947a24e72834c4b8d18a138ca_i1_4.jpg", "1mp3d_0044_region7/436cddf947a24e72834c4b8d18a138ca_i2_4.jpg", "1mp3d_0044_region7/436cddf947a24e72834c4b8d18a138ca_i1_3.jpg", "1mp3d_0044_region7/436cddf947a24e72834c4b8d18a138ca_i2_3.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0044_region7", "origin_question": "In which turn did you see both \"the rectangular, soft light beige mat with floral patterns\" and \"the large brown wooden door with a glass panel\" simultaneously? ", "option": [], "answer": "3", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the rectangular, soft light beige mat with floral patterns\" and \"the large brown wooden door with a glass panel\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "1mp3d_0044_region7/e3ffb659fb744d33852508489d412649_i1_1.jpg", "1mp3d_0044_region7/e3ffb659fb744d33852508489d412649_i2_1.jpg", "1mp3d_0044_region7/e3ffb659fb744d33852508489d412649_i1_3.jpg", "1mp3d_0044_region7/e3ffb659fb744d33852508489d412649_i2_3.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0044_region7", "origin_question": "What will be the first-time appearance order of the following categories: mat, plant, tv?", "option": [ "mat, tv, plant", "mat, plant, tv", "tv, plant, mat", "tv, mat, plant" ], "answer": "tv, mat, plant", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: mat, plant, tv? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['mat, tv, plant', 'mat, plant, tv', 'tv, plant, mat', 'tv, mat, plant'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "1mp3d_0044_region7/e3ffb659fb744d33852508489d412649_i1_4.jpg", "1mp3d_0044_region7/e3ffb659fb744d33852508489d412649_i2_4.jpg", "1mp3d_0044_region7/e3ffb659fb744d33852508489d412649_i1_5.jpg", "1mp3d_0044_region7/e3ffb659fb744d33852508489d412649_i2_5.jpg" ], "turn_id": 7, "scan_id": "1mp3d_0044_region7", "origin_question": "Is the small circular metal decoration with black and green colors to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the small circular metal decoration with black and green colors to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "1mp3d_0044_region7/7b454c245c1f4cb0887ff561e26fced4_i1_2.jpg", "1mp3d_0044_region7/7b454c245c1f4cb0887ff561e26fced4_i2_2.jpg", "1mp3d_0044_region7/7b454c245c1f4cb0887ff561e26fced4_i1_3.jpg", "1mp3d_0044_region7/7b454c245c1f4cb0887ff561e26fced4_i2_3.jpg" ], "turn_id": 8, "scan_id": "1mp3d_0044_region7", "origin_question": "When did you first discover \"the sleek black flat-screen LCD TV with a rectangular shape\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the sleek black flat-screen LCD TV with a rectangular shape\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "1mp3d_0044_region7/7b454c245c1f4cb0887ff561e26fced4_i1_4.jpg", "1mp3d_0044_region7/7b454c245c1f4cb0887ff561e26fced4_i2_4.jpg", "1mp3d_0044_region7/7b454c245c1f4cb0887ff561e26fced4_i1_5.jpg", "1mp3d_0044_region7/7b454c245c1f4cb0887ff561e26fced4_i2_5.jpg" ], "turn_id": 9, "scan_id": "1mp3d_0044_region7", "origin_question": "How far is your current position from where you were at the end of the turn 7? (in meters)", "option": [], "answer": "1.29", "user_message": "For the 9 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 7? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0686_02/00000.jpg", "scene0686_02/00040.jpg", "scene0686_02/00080.jpg", "scene0686_02/00120.jpg" ], "turn_id": 1, "scan_id": "scene0686_02", "origin_question": "Remember, have you seen any sink(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any sink(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0686_02/00160.jpg", "scene0686_02/00200.jpg", "scene0686_02/00240.jpg", "scene0686_02/00280.jpg" ], "turn_id": 2, "scan_id": "scene0686_02", "origin_question": "When did you first discover \"the white, wall-mounted toilet with an elongated bowl\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the white, wall-mounted toilet with an elongated bowl\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0686_02/00320.jpg", "scene0686_02/00360.jpg", "scene0686_02/00400.jpg", "scene0686_02/00440.jpg" ], "turn_id": 3, "scan_id": "scene0686_02", "origin_question": "Please recall \"the dispenser\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.1", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the dispenser\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0686_02/00480.jpg", "scene0686_02/00520.jpg", "scene0686_02/00560.jpg", "scene0686_02/00600.jpg" ], "turn_id": 4, "scan_id": "scene0686_02", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the rectangular white wall-mounted porcelain sink\" from your position?", "option": [], "answer": "95.99", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the rectangular white wall-mounted porcelain sink\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0686_02/00640.jpg", "scene0686_02/00680.jpg", "scene0686_02/00720.jpg", "scene0686_02/00760.jpg" ], "turn_id": 5, "scan_id": "scene0686_02", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0686_02/00800.jpg", "scene0686_02/00840.jpg", "scene0686_02/00880.jpg", "scene0686_02/00920.jpg" ], "turn_id": 6, "scan_id": "scene0686_02", "origin_question": "Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the rectangular white wall-mounted porcelain sink\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the rectangular white wall-mounted porcelain sink\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0686_02/00960.jpg", "scene0686_02/01000.jpg", "scene0686_02/01040.jpg", "scene0686_02/01080.jpg" ], "turn_id": 7, "scan_id": "scene0686_02", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0686_02/01120.jpg", "scene0686_02/01160.jpg", "scene0686_02/01200.jpg", "scene0686_02/01240.jpg" ], "turn_id": 8, "scan_id": "scene0686_02", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "57.4", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0067_region19/684d9231bf93471a91638ee94f790c5b_i1_4.jpg", "1mp3d_0067_region19/684d9231bf93471a91638ee94f790c5b_i2_4.jpg", "1mp3d_0067_region19/a091468b0d6d43f099fe840ca10b1a14_i1_2.jpg", "1mp3d_0067_region19/a091468b0d6d43f099fe840ca10b1a14_i2_2.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0067_region19", "origin_question": "Remember, have you seen any clothes(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any clothes(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0067_region19/a091468b0d6d43f099fe840ca10b1a14_i1_4.jpg", "1mp3d_0067_region19/a091468b0d6d43f099fe840ca10b1a14_i2_4.jpg", "1mp3d_0067_region19/684d9231bf93471a91638ee94f790c5b_i1_5.jpg", "1mp3d_0067_region19/684d9231bf93471a91638ee94f790c5b_i2_5.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0067_region19", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0067_region19/6574e941f0be49afa9fd447b99b2e783_i1_4.jpg", "1mp3d_0067_region19/6574e941f0be49afa9fd447b99b2e783_i2_4.jpg", "1mp3d_0067_region19/6574e941f0be49afa9fd447b99b2e783_i1_5.jpg", "1mp3d_0067_region19/6574e941f0be49afa9fd447b99b2e783_i2_5.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0067_region19", "origin_question": "", "option": [], "answer": "", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0067_region19/7804a496622640ce9d5d79ecdc992254_i1_0.jpg", "1mp3d_0067_region19/7804a496622640ce9d5d79ecdc992254_i2_0.jpg", "1mp3d_0067_region19/6dfdde020dce4f68983f0382993b1c51_i1_1.jpg", "1mp3d_0067_region19/6dfdde020dce4f68983f0382993b1c51_i2_1.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0067_region19", "origin_question": "", "option": [], "answer": "", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "1mp3d_0067_region19/7804a496622640ce9d5d79ecdc992254_i1_5.jpg", "1mp3d_0067_region19/7804a496622640ce9d5d79ecdc992254_i2_5.jpg", "1mp3d_0067_region19/7804a496622640ce9d5d79ecdc992254_i1_3.jpg", "1mp3d_0067_region19/7804a496622640ce9d5d79ecdc992254_i2_3.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0067_region19", "origin_question": "What will be the first-time appearance order of the following categories: basket, clothes, shoe?", "option": [ "basket, clothes, shoe", "clothes, basket, shoe", "shoe, clothes, basket", "basket, shoe, clothes" ], "answer": "clothes, basket, shoe", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: basket, clothes, shoe? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['basket, clothes, shoe', 'clothes, basket, shoe', 'shoe, clothes, basket', 'basket, shoe, clothes'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "1mp3d_0067_region19/6dfdde020dce4f68983f0382993b1c51_i1_0.jpg", "1mp3d_0067_region19/6dfdde020dce4f68983f0382993b1c51_i2_0.jpg", "1mp3d_0067_region19/6dfdde020dce4f68983f0382993b1c51_i1_5.jpg", "1mp3d_0067_region19/6dfdde020dce4f68983f0382993b1c51_i2_5.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0067_region19", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "1mp3d_0067_region10/cc03fd369f254cd985894b77fd8babde_i1_2.jpg", "1mp3d_0067_region10/cc03fd369f254cd985894b77fd8babde_i2_2.jpg", "1mp3d_0067_region10/cc03fd369f254cd985894b77fd8babde_i1_3.jpg", "1mp3d_0067_region10/cc03fd369f254cd985894b77fd8babde_i2_3.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0067_region10", "origin_question": "Remember, how many pen(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many pen(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "1mp3d_0067_region10/067ec17ab4314effbaf5e67e5acfacad_i1_0.jpg", "1mp3d_0067_region10/067ec17ab4314effbaf5e67e5acfacad_i2_0.jpg", "1mp3d_0067_region10/067ec17ab4314effbaf5e67e5acfacad_i1_1.jpg", "1mp3d_0067_region10/067ec17ab4314effbaf5e67e5acfacad_i2_1.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0067_region10", "origin_question": "Which direction is the petite white shelf with a curved design in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the petite white shelf with a curved design in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0067_region10/0ffa6fd0a65a4462b2023f28b14741e5_i1_1.jpg", "1mp3d_0067_region10/0ffa6fd0a65a4462b2023f28b14741e5_i2_1.jpg", "1mp3d_0067_region10/0ffa6fd0a65a4462b2023f28b14741e5_i1_2.jpg", "1mp3d_0067_region10/0ffa6fd0a65a4462b2023f28b14741e5_i2_2.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0067_region10", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the large rectangular white fireplace with a black screen\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the large rectangular white fireplace with a black screen\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0067_region10/0ffa6fd0a65a4462b2023f28b14741e5_i1_3.jpg", "1mp3d_0067_region10/0ffa6fd0a65a4462b2023f28b14741e5_i2_3.jpg", "1mp3d_0067_region10/0ffa6fd0a65a4462b2023f28b14741e5_i1_4.jpg", "1mp3d_0067_region10/0ffa6fd0a65a4462b2023f28b14741e5_i2_4.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0067_region10", "origin_question": "", "option": [], "answer": "", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0067_region10/5816276565a94b17ba8c667db5829250_i1_4.jpg", "1mp3d_0067_region10/5816276565a94b17ba8c667db5829250_i2_4.jpg", "1mp3d_0067_region10/a9c7288e5225482aac6f99127fb581e0_i1_1.jpg", "1mp3d_0067_region10/a9c7288e5225482aac6f99127fb581e0_i2_1.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0067_region10", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the large rectangular picture with a black frame\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the large rectangular picture with a black frame\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "1mp3d_0067_region10/5816276565a94b17ba8c667db5829250_i1_0.jpg", "1mp3d_0067_region10/5816276565a94b17ba8c667db5829250_i2_0.jpg", "1mp3d_0067_region10/a9c7288e5225482aac6f99127fb581e0_i1_4.jpg", "1mp3d_0067_region10/a9c7288e5225482aac6f99127fb581e0_i2_4.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0067_region10", "origin_question": "Among these three objects(\"the rectangular stone plant pot with various plants\",\"the rectangular mat with intricate geometric patterns\",\"the rectangular blue sliding door with a wooden frame\" ), which one is the closest to you?", "option": [ "the rectangular stone plant pot with various plants", "the rectangular mat with intricate geometric patterns", "the rectangular blue sliding door with a wooden frame" ], "answer": "the rectangular mat with intricate geometric patterns", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects(\"the rectangular stone plant pot with various plants\",\"the rectangular mat with intricate geometric patterns\",\"the rectangular blue sliding door with a wooden frame\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular stone plant pot with various plants', 'the rectangular mat with intricate geometric patterns', 'the rectangular blue sliding door with a wooden frame'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "1mp3d_0067_region10/a9c7288e5225482aac6f99127fb581e0_i1_2.jpg", "1mp3d_0067_region10/a9c7288e5225482aac6f99127fb581e0_i2_2.jpg", "1mp3d_0067_region10/ffc3e703e412473da5216cca31f17292_i1_5.jpg", "1mp3d_0067_region10/ffc3e703e412473da5216cca31f17292_i2_5.jpg" ], "turn_id": 7, "scan_id": "1mp3d_0067_region10", "origin_question": "Among these three objects(\"the rectangular blue sliding door with a wooden frame\",\"the rectangular wooden window with a rich brown frame\",\"the shiny round metal bucket with a handle\" ), which one is the farthest from you?", "option": [ "the rectangular blue sliding door with a wooden frame", "the rectangular wooden window with a rich brown frame", "the shiny round metal bucket with a handle" ], "answer": "the rectangular blue sliding door with a wooden frame", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects(\"the rectangular blue sliding door with a wooden frame\",\"the rectangular wooden window with a rich brown frame\",\"the shiny round metal bucket with a handle\" ), which one is the farthest from you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular blue sliding door with a wooden frame', 'the rectangular wooden window with a rich brown frame', 'the shiny round metal bucket with a handle'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "1mp3d_0067_region10/ffc3e703e412473da5216cca31f17292_i1_0.jpg", "1mp3d_0067_region10/ffc3e703e412473da5216cca31f17292_i2_0.jpg", "1mp3d_0067_region10/ffc3e703e412473da5216cca31f17292_i1_1.jpg", "1mp3d_0067_region10/ffc3e703e412473da5216cca31f17292_i2_1.jpg" ], "turn_id": 8, "scan_id": "1mp3d_0067_region10", "origin_question": "Among these three objects(\"the rectangular stone plant pot with various plants\",\"the rectangular mat with intricate geometric patterns\",\"the large rectangular picture with a black frame\" ), which one is the farthest from you?", "option": [ "the rectangular stone plant pot with various plants", "the rectangular mat with intricate geometric patterns", "the large rectangular picture with a black frame" ], "answer": "the rectangular stone plant pot with various plants", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects(\"the rectangular stone plant pot with various plants\",\"the rectangular mat with intricate geometric patterns\",\"the large rectangular picture with a black frame\" ), which one is the farthest from you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular stone plant pot with various plants', 'the rectangular mat with intricate geometric patterns', 'the large rectangular picture with a black frame'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "1mp3d_0067_region10/ffc3e703e412473da5216cca31f17292_i1_2.jpg", "1mp3d_0067_region10/ffc3e703e412473da5216cca31f17292_i2_2.jpg", "1mp3d_0067_region10/ffc3e703e412473da5216cca31f17292_i1_3.jpg", "1mp3d_0067_region10/ffc3e703e412473da5216cca31f17292_i2_3.jpg" ], "turn_id": 9, "scan_id": "1mp3d_0067_region10", "origin_question": "Please recall \"the large rectangular picture with a black frame\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.08", "user_message": "For the 9 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the large rectangular picture with a black frame\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0256_02/00000.jpg", "scene0256_02/00040.jpg", "scene0256_02/00080.jpg", "scene0256_02/00120.jpg" ], "turn_id": 1, "scan_id": "scene0256_02", "origin_question": "Remember, have you seen any ottoman(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any ottoman(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0256_02/00160.jpg", "scene0256_02/00200.jpg", "scene0256_02/00240.jpg", "scene0256_02/00280.jpg" ], "turn_id": 2, "scan_id": "scene0256_02", "origin_question": "Is the rectangular white pillow made of soft material to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the rectangular white pillow made of soft material to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0256_02/00320.jpg", "scene0256_02/00360.jpg", "scene0256_02/00400.jpg", "scene0256_02/00440.jpg" ], "turn_id": 3, "scan_id": "scene0256_02", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the moderate-sized plant with white flowers and green leaves\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the moderate-sized plant with white flowers and green leaves\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0256_02/00480.jpg", "scene0256_02/00520.jpg", "scene0256_02/00560.jpg", "scene0256_02/00600.jpg" ], "turn_id": 4, "scan_id": "scene0256_02", "origin_question": "Which two objects are on the same side of you now? \"the white wooden cabinet with a minimalistic design\", \"the rectangular white pillow made of soft material\", \"the rectangular ottoman made of light-colored wood\". ", "option": [ "the white wooden cabinet with a minimalistic design, the rectangular white pillow made of soft material", "the white wooden cabinet with a minimalistic design, the rectangular ottoman made of light-colored wood", "the rectangular white pillow made of soft material, the rectangular ottoman made of light-colored wood" ], "answer": "the white wooden cabinet with a minimalistic design, the rectangular white pillow made of soft material", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the white wooden cabinet with a minimalistic design\", \"the rectangular white pillow made of soft material\", \"the rectangular ottoman made of light-colored wood\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the white wooden cabinet with a minimalistic design, the rectangular white pillow made of soft material', 'the white wooden cabinet with a minimalistic design, the rectangular ottoman made of light-colored wood', 'the rectangular white pillow made of soft material, the rectangular ottoman made of light-colored wood'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0256_02/00640.jpg", "scene0256_02/00680.jpg", "scene0256_02/00720.jpg", "scene0256_02/00760.jpg" ], "turn_id": 5, "scan_id": "scene0256_02", "origin_question": "Is the moderate-sized plant with white flowers and green leaves to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the moderate-sized plant with white flowers and green leaves to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0256_02/00800.jpg", "scene0256_02/00840.jpg", "scene0256_02/00880.jpg", "scene0256_02/00920.jpg" ], "turn_id": 6, "scan_id": "scene0256_02", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the rectangular white piece of paper\"; \"the bed\"; \"the rectangular, light brown, soft blanket\" ", "option": [ "the rectangular white piece of paper", "the bed", "the rectangular, light brown, soft blanket" ], "answer": "the rectangular white piece of paper", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the rectangular white piece of paper\"; \"the bed\"; \"the rectangular, light brown, soft blanket\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular white piece of paper', 'the bed', 'the rectangular, light brown, soft blanket'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0256_02/00960.jpg", "scene0256_02/01000.jpg", "scene0256_02/01040.jpg", "scene0256_02/01080.jpg" ], "turn_id": 7, "scan_id": "scene0256_02", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the rectangular, light brown, soft blanket\"; \"the person\"; \"the white wooden cabinet with a minimalistic design\" ", "option": [ "the rectangular, light brown, soft blanket", "the person", "the white wooden cabinet with a minimalistic design" ], "answer": "the person", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the rectangular, light brown, soft blanket\"; \"the person\"; \"the white wooden cabinet with a minimalistic design\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular, light brown, soft blanket', 'the person', 'the white wooden cabinet with a minimalistic design'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0277_00/00000.jpg", "scene0277_00/00040.jpg", "scene0277_00/00080.jpg", "scene0277_00/00180.jpg" ], "turn_id": 1, "scan_id": "scene0277_00", "origin_question": "Remember, have you seen any bin(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any bin(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0277_00/00220.jpg", "scene0277_00/00260.jpg", "scene0277_00/00300.jpg", "scene0277_00/00340.jpg" ], "turn_id": 2, "scan_id": "scene0277_00", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the black office chair with a mesh backrest\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the black office chair with a mesh backrest\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0277_00/00380.jpg", "scene0277_00/00420.jpg", "scene0277_00/00460.jpg", "scene0277_00/00500.jpg" ], "turn_id": 3, "scan_id": "scene0277_00", "origin_question": "Please recall \"the light rectangular wooden desk with closed drawers\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.58", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the light rectangular wooden desk with closed drawers\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0277_00/00540.jpg", "scene0277_00/00580.jpg", "scene0277_00/00620.jpg", "scene0277_00/00660.jpg" ], "turn_id": 4, "scan_id": "scene0277_00", "origin_question": "At the end of which turn were both of \"the black office chair with a mesh backrest\" and \"the rectangular single bed with a deep blue mattress\" on your right side?", "option": [], "answer": "2", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the black office chair with a mesh backrest\" and \"the rectangular single bed with a deep blue mattress\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0277_00/00700.jpg", "scene0277_00/00760.jpg", "scene0277_00/00800.jpg", "scene0277_00/00840.jpg" ], "turn_id": 5, "scan_id": "scene0277_00", "origin_question": "At the end of which turn were both of \"the black office chair with a mesh backrest\" and \"the radiator\" on your left side?", "option": [], "answer": "1", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the black office chair with a mesh backrest\" and \"the radiator\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0277_00/00880.jpg", "scene0277_00/00920.jpg", "scene0277_00/00960.jpg", "scene0277_00/01000.jpg" ], "turn_id": 6, "scan_id": "scene0277_00", "origin_question": "Using your orientation at the end of turn 4 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "106.32", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 4 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0004_region14/85fa57f01a904a4099aed1c7f381da81_i1_0.jpg", "1mp3d_0004_region14/85fa57f01a904a4099aed1c7f381da81_i2_0.jpg", "1mp3d_0004_region14/5a3f574cd3634d189563c54ac751232b_i1_5.jpg", "1mp3d_0004_region14/5a3f574cd3634d189563c54ac751232b_i2_5.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0004_region14", "origin_question": "Remember, have you seen any treadmill(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any treadmill(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "1mp3d_0004_region14/5a3f574cd3634d189563c54ac751232b_i1_0.jpg", "1mp3d_0004_region14/5a3f574cd3634d189563c54ac751232b_i2_0.jpg", "1mp3d_0004_region14/5a3f574cd3634d189563c54ac751232b_i1_1.jpg", "1mp3d_0004_region14/5a3f574cd3634d189563c54ac751232b_i2_1.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0004_region14", "origin_question": "At the end of which turn were both of \"the large rectangular window with a white frame\" and \"the door\" on your left side?", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the large rectangular window with a white frame\" and \"the door\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0004_region14/5a3f574cd3634d189563c54ac751232b_i1_2.jpg", "1mp3d_0004_region14/5a3f574cd3634d189563c54ac751232b_i2_2.jpg", "1mp3d_0004_region14/85fa57f01a904a4099aed1c7f381da81_i1_3.jpg", "1mp3d_0004_region14/85fa57f01a904a4099aed1c7f381da81_i2_3.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0004_region14", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the large rectangular window with a white frame\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the large rectangular window with a white frame\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "1mp3d_0004_region14/255411360fbb4cb88190ec9bbc7f5a45_i1_0.jpg", "1mp3d_0004_region14/255411360fbb4cb88190ec9bbc7f5a45_i2_0.jpg", "1mp3d_0004_region14/255411360fbb4cb88190ec9bbc7f5a45_i1_5.jpg", "1mp3d_0004_region14/255411360fbb4cb88190ec9bbc7f5a45_i2_5.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0004_region14", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the sleek black rectangular bag\" from your position?", "option": [], "answer": "110.05", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the sleek black rectangular bag\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "1mp3d_0004_region14/d055595a35f2464ca7b02125fd5a630d_i1_1.jpg", "1mp3d_0004_region14/d055595a35f2464ca7b02125fd5a630d_i2_1.jpg", "1mp3d_0004_region14/8bb253a3e7ac4be18262b97b052bbb5a_i1_2.jpg", "1mp3d_0004_region14/8bb253a3e7ac4be18262b97b052bbb5a_i2_2.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0004_region14", "origin_question": "Is the sleek black rectangular bag to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the sleek black rectangular bag to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "1mp3d_0004_region14/d055595a35f2464ca7b02125fd5a630d_i1_3.jpg", "1mp3d_0004_region14/d055595a35f2464ca7b02125fd5a630d_i2_3.jpg", "1mp3d_0004_region14/d055595a35f2464ca7b02125fd5a630d_i1_2.jpg", "1mp3d_0004_region14/d055595a35f2464ca7b02125fd5a630d_i2_2.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0004_region14", "origin_question": "Is the door to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the door to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0029_region35/f3e7de357f254d15bb79bc62ef0c803e_i1_4.jpg", "1mp3d_0029_region35/f3e7de357f254d15bb79bc62ef0c803e_i2_4.jpg", "1mp3d_0029_region35/0e9fdd85e24a4a35b3dc1e8cb76ebb09_i1_4.jpg", "1mp3d_0029_region35/0e9fdd85e24a4a35b3dc1e8cb76ebb09_i2_4.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0029_region35", "origin_question": "", "option": [], "answer": "", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 2 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0029_region35/0e9fdd85e24a4a35b3dc1e8cb76ebb09_i1_3.jpg", "1mp3d_0029_region35/0e9fdd85e24a4a35b3dc1e8cb76ebb09_i2_3.jpg", "1mp3d_0029_region35/f3e7de357f254d15bb79bc62ef0c803e_i2_1.jpg", "1mp3d_0029_region35/f3e7de357f254d15bb79bc62ef0c803e_i1_0.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0029_region35", "origin_question": "Remember, have you seen any toiletry(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any toiletry(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 2 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0696_00/00000.jpg", "scene0696_00/00040.jpg", "scene0696_00/00080.jpg", "scene0696_00/00120.jpg" ], "turn_id": 1, "scan_id": "scene0696_00", "origin_question": "Remember, have you seen any chair(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any chair(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0696_00/00160.jpg", "scene0696_00/00200.jpg", "scene0696_00/00240.jpg", "scene0696_00/00280.jpg" ], "turn_id": 2, "scan_id": "scene0696_00", "origin_question": "Which direction is the rectangular beige mattress made of soft materials in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the rectangular beige mattress made of soft materials in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0696_00/00320.jpg", "scene0696_00/00360.jpg", "scene0696_00/00400.jpg", "scene0696_00/00440.jpg" ], "turn_id": 3, "scan_id": "scene0696_00", "origin_question": "Please recall \"the light-colored full-size window curtain\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.5", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the light-colored full-size window curtain\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0696_00/00480.jpg", "scene0696_00/00520.jpg", "scene0696_00/00560.jpg", "scene0696_00/00600.jpg" ], "turn_id": 4, "scan_id": "scene0696_00", "origin_question": "Is the large, soft, white carpet with subtle patterns to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the large, soft, white carpet with subtle patterns to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0696_00/00640.jpg", "scene0696_00/00680.jpg", "scene0696_00/00720.jpg", "scene0696_00/00760.jpg" ], "turn_id": 5, "scan_id": "scene0696_00", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the pillow\" from your position?", "option": [], "answer": "110.2", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the pillow\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0696_00/00800.jpg", "scene0696_00/00840.jpg", "scene0696_00/00880.jpg", "scene0696_00/00920.jpg" ], "turn_id": 6, "scan_id": "scene0696_00", "origin_question": "What will be the first-time appearance order of the following categories: carpet, door, helmets?", "option": [ "helmets, carpet, door", "helmets, door, carpet", "carpet, helmets, door", "carpet, door, helmets" ], "answer": "carpet, door, helmets", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: carpet, door, helmets? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['helmets, carpet, door', 'helmets, door, carpet', 'carpet, helmets, door', 'carpet, door, helmets'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "scene0696_00/00960.jpg", "scene0696_00/01000.jpg", "scene0696_00/01040.jpg", "scene0696_00/01080.jpg" ], "turn_id": 7, "scan_id": "scene0696_00", "origin_question": "Among these three objects(\"the large, soft, white carpet with subtle patterns\",\"the rectangular beige mattress made of soft materials\",\"the door\" ), which one is the farthest from you?", "option": [ "the large, soft, white carpet with subtle patterns", "the rectangular beige mattress made of soft materials", "the door" ], "answer": "the door", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects(\"the large, soft, white carpet with subtle patterns\",\"the rectangular beige mattress made of soft materials\",\"the door\" ), which one is the farthest from you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the large, soft, white carpet with subtle patterns', 'the rectangular beige mattress made of soft materials', 'the door'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0696_00/01120.jpg", "scene0696_00/01160.jpg", "scene0696_00/01200.jpg", "scene0696_00/01240.jpg" ], "turn_id": 8, "scan_id": "scene0696_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0004_region3/508a0c321b2341fb89271997435f6842_i1_4.jpg", "1mp3d_0004_region3/508a0c321b2341fb89271997435f6842_i2_4.jpg", "1mp3d_0004_region3/9eedcaf6ec54441ba2454e247dd957db_i1_0.jpg", "1mp3d_0004_region3/9eedcaf6ec54441ba2454e247dd957db_i2_0.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0004_region3", "origin_question": "Remember, have you seen any washbasin(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any washbasin(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0004_region3/7f6aa85262424e7da75f62063d397514_i1_2.jpg", "1mp3d_0004_region3/7f6aa85262424e7da75f62063d397514_i2_2.jpg", "1mp3d_0004_region3/9eedcaf6ec54441ba2454e247dd957db_i1_5.jpg", "1mp3d_0004_region3/9eedcaf6ec54441ba2454e247dd957db_i2_5.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0004_region3", "origin_question": "Remember, have you seen any door(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any door(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "1mp3d_0004_region3/1a681f5111a94b0b8892b2f2f41ba9f9_i1_4.jpg", "1mp3d_0004_region3/1a681f5111a94b0b8892b2f2f41ba9f9_i2_4.jpg", "1mp3d_0004_region3/1a681f5111a94b0b8892b2f2f41ba9f9_i1_0.jpg", "1mp3d_0004_region3/1a681f5111a94b0b8892b2f2f41ba9f9_i2_0.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0004_region3", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0004_region3/1a681f5111a94b0b8892b2f2f41ba9f9_i1_5.jpg", "1mp3d_0004_region3/1a681f5111a94b0b8892b2f2f41ba9f9_i2_5.jpg", "1mp3d_0004_region3/ac4d3c6c11e4434fbf9cd56412176bd7_i1_1.jpg", "1mp3d_0004_region3/ac4d3c6c11e4434fbf9cd56412176bd7_i2_1.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0004_region3", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the white, soft, absorbent towel made of cotton\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the white, soft, absorbent towel made of cotton\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "1mp3d_0004_region3/ede19359f733479d82a33051431e17e5_i1_4.jpg", "1mp3d_0004_region3/ede19359f733479d82a33051431e17e5_i2_4.jpg", "1mp3d_0004_region3/ac4d3c6c11e4434fbf9cd56412176bd7_i1_2.jpg", "1mp3d_0004_region3/ac4d3c6c11e4434fbf9cd56412176bd7_i2_2.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0004_region3", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the white rectangular wooden door with chrome handles\", Y: \"the white, soft, absorbent towel made of cotton\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the white rectangular wooden door with chrome handles\", Y: \"the white, soft, absorbent towel made of cotton\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0004_region3/ede19359f733479d82a33051431e17e5_i1_5.jpg", "1mp3d_0004_region3/ede19359f733479d82a33051431e17e5_i2_5.jpg", "1mp3d_0004_region3/ede19359f733479d82a33051431e17e5_i1_0.jpg", "1mp3d_0004_region3/ede19359f733479d82a33051431e17e5_i2_0.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0004_region3", "origin_question": "", "option": [], "answer": "", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0702_01/00000.jpg", "scene0702_01/00030.jpg", "scene0702_01/00060.jpg", "scene0702_01/00090.jpg" ], "turn_id": 1, "scan_id": "scene0702_01", "origin_question": "Remember, have you seen any bin(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any bin(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0702_01/00120.jpg", "scene0702_01/00150.jpg", "scene0702_01/00180.jpg", "scene0702_01/00210.jpg" ], "turn_id": 2, "scan_id": "scene0702_01", "origin_question": "In which turn did you see both \"the curtain\" and \"the bathtub\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the curtain\" and \"the bathtub\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0702_01/00240.jpg", "scene0702_01/00420.jpg", "scene0702_01/00450.jpg", "scene0702_01/00510.jpg" ], "turn_id": 3, "scan_id": "scene0702_01", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the white ceramic toilet with a round bowl\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the white ceramic toilet with a round bowl\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0702_01/00540.jpg", "scene0702_01/00570.jpg", "scene0702_01/00600.jpg", "scene0702_01/00630.jpg" ], "turn_id": 4, "scan_id": "scene0702_01", "origin_question": "", "option": [], "answer": "", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0589_00/00000.jpg", "scene0589_00/00040.jpg", "scene0589_00/00080.jpg", "scene0589_00/00120.jpg" ], "turn_id": 1, "scan_id": "scene0589_00", "origin_question": "Remember, have you seen any case(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any case(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0589_00/00160.jpg", "scene0589_00/00200.jpg", "scene0589_00/00240.jpg", "scene0589_00/00280.jpg" ], "turn_id": 2, "scan_id": "scene0589_00", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the cone-shaped yellow lamp with a metal body\" from your position?", "option": [], "answer": "88.57", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the cone-shaped yellow lamp with a metal body\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0589_00/00320.jpg", "scene0589_00/00360.jpg", "scene0589_00/00400.jpg", "scene0589_00/00440.jpg" ], "turn_id": 3, "scan_id": "scene0589_00", "origin_question": "Which two objects are on the same side of you now? \"the tall wooden shelf with multiple panels\", \"the double bed with a bright orange-yellow bedspread\", \"the rectangular, light-colored wooden stand with a drawer\". ", "option": [ "the tall wooden shelf with multiple panels, the double bed with a bright orange-yellow bedspread", "the tall wooden shelf with multiple panels, the rectangular, light-colored wooden stand with a drawer", "the double bed with a bright orange-yellow bedspread, the rectangular, light-colored wooden stand with a drawer" ], "answer": "the tall wooden shelf with multiple panels, the double bed with a bright orange-yellow bedspread", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the tall wooden shelf with multiple panels\", \"the double bed with a bright orange-yellow bedspread\", \"the rectangular, light-colored wooden stand with a drawer\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the tall wooden shelf with multiple panels, the double bed with a bright orange-yellow bedspread', 'the tall wooden shelf with multiple panels, the rectangular, light-colored wooden stand with a drawer', 'the double bed with a bright orange-yellow bedspread, the rectangular, light-colored wooden stand with a drawer'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0589_00/00480.jpg", "scene0589_00/00520.jpg", "scene0589_00/00560.jpg", "scene0589_00/00600.jpg" ], "turn_id": 4, "scan_id": "scene0589_00", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0589_00/00640.jpg", "scene0589_00/00680.jpg", "scene0589_00/00720.jpg", "scene0589_00/00760.jpg" ], "turn_id": 5, "scan_id": "scene0589_00", "origin_question": "Is the rectangular, light-colored wooden stand with a drawer to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the rectangular, light-colored wooden stand with a drawer to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0589_00/00800.jpg", "scene0589_00/00840.jpg", "scene0589_00/00880.jpg", "scene0589_00/00920.jpg" ], "turn_id": 6, "scan_id": "scene0589_00", "origin_question": "Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "154.12", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0589_00/00960.jpg", "scene0589_00/01000.jpg", "scene0589_00/01040.jpg", "scene0589_00/01080.jpg" ], "turn_id": 7, "scan_id": "scene0589_00", "origin_question": "What will be the first-time appearance order of the following categories: case, door, picture?", "option": [ "door, case, picture", "picture, door, case", "case, door, picture", "picture, case, door" ], "answer": "case, door, picture", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: case, door, picture? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['door, case, picture', 'picture, door, case', 'case, door, picture', 'picture, case, door'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "1mp3d_0040_region31/67c20b163349417893a145ea7a3f1cd7_i1_3.jpg", "1mp3d_0040_region31/67c20b163349417893a145ea7a3f1cd7_i2_3.jpg", "1mp3d_0040_region31/e642c08656284353abdd1fe071ad1582_i1_2.jpg", "1mp3d_0040_region31/e642c08656284353abdd1fe071ad1582_i2_2.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0040_region31", "origin_question": "Remember, how many plant(s) have you seen so far? ", "option": [], "answer": "2", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many plant(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "1mp3d_0040_region31/250a0195f77f4fbd9a6e574bee058306_i1_0.jpg", "1mp3d_0040_region31/250a0195f77f4fbd9a6e574bee058306_i2_0.jpg", "1mp3d_0040_region31/250a0195f77f4fbd9a6e574bee058306_i1_1.jpg", "1mp3d_0040_region31/250a0195f77f4fbd9a6e574bee058306_i2_1.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0040_region31", "origin_question": "In which turn were you closest to \"the rectangular wooden desk with a deep brown finish\"? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the rectangular wooden desk with a deep brown finish\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "1mp3d_0040_region31/e642c08656284353abdd1fe071ad1582_i1_1.jpg", "1mp3d_0040_region31/e642c08656284353abdd1fe071ad1582_i2_1.jpg", "1mp3d_0040_region31/67c20b163349417893a145ea7a3f1cd7_i1_0.jpg", "1mp3d_0040_region31/67c20b163349417893a145ea7a3f1cd7_i2_0.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0040_region31", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "54.39", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "1mp3d_0040_region31/67c20b163349417893a145ea7a3f1cd7_i1_1.jpg", "1mp3d_0040_region31/67c20b163349417893a145ea7a3f1cd7_i2_1.jpg", "1mp3d_0040_region31/67c20b163349417893a145ea7a3f1cd7_i1_2.jpg", "1mp3d_0040_region31/67c20b163349417893a145ea7a3f1cd7_i2_2.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0040_region31", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "2.33", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "1mp3d_0040_region31/c694ace3fc364ef4990a3dd40dcc6a64_i1_2.jpg", "1mp3d_0040_region31/c694ace3fc364ef4990a3dd40dcc6a64_i2_2.jpg", "1mp3d_0040_region31/c694ace3fc364ef4990a3dd40dcc6a64_i1_3.jpg", "1mp3d_0040_region31/c694ace3fc364ef4990a3dd40dcc6a64_i2_3.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0040_region31", "origin_question": "In which turn did you see both \"the classic black leather chair with wooden armrests\" and \"the round beige woven flowerpot with dark edges\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the classic black leather chair with wooden armrests\" and \"the round beige woven flowerpot with dark edges\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "1mp3d_0040_region31/8c6f5e702e7a473987f6ae0a3dbd4442_i1_0.jpg", "1mp3d_0040_region31/8c6f5e702e7a473987f6ae0a3dbd4442_i2_0.jpg", "1mp3d_0040_region31/8c6f5e702e7a473987f6ae0a3dbd4442_i1_5.jpg", "1mp3d_0040_region31/8c6f5e702e7a473987f6ae0a3dbd4442_i2_5.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0040_region31", "origin_question": "At the end of which turn were both of \"the green cylindrical ceramic vase with a glossy surface\" and \"the large round earthy-colored pot\" on your left side?", "option": [], "answer": "2", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the green cylindrical ceramic vase with a glossy surface\" and \"the large round earthy-colored pot\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0490_00/00000.jpg", "scene0490_00/00040.jpg", "scene0490_00/00080.jpg", "scene0490_00/00120.jpg" ], "turn_id": 1, "scan_id": "scene0490_00", "origin_question": "Remember, how many blackboard(s) have you seen so far? ", "option": [], "answer": "2", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many blackboard(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0490_00/00160.jpg", "scene0490_00/00400.jpg", "scene0490_00/00440.jpg", "scene0490_00/00480.jpg" ], "turn_id": 2, "scan_id": "scene0490_00", "origin_question": "In which turn did you see both \"the table\" and \"the tv\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the table\" and \"the tv\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0490_00/00520.jpg", "scene0490_00/00560.jpg", "scene0490_00/00600.jpg", "scene0490_00/00640.jpg" ], "turn_id": 3, "scan_id": "scene0490_00", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "146.85", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0490_00/00680.jpg", "scene0490_00/00720.jpg", "scene0490_00/00760.jpg", "scene0490_00/00800.jpg" ], "turn_id": 4, "scan_id": "scene0490_00", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0490_00/00840.jpg", "scene0490_00/00880.jpg", "scene0490_00/00920.jpg", "scene0490_00/00960.jpg" ], "turn_id": 5, "scan_id": "scene0490_00", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "73.52", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0490_00/01000.jpg", "scene0490_00/01040.jpg", "scene0490_00/01080.jpg", "scene0490_00/01120.jpg" ], "turn_id": 6, "scan_id": "scene0490_00", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the table\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the table\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0553_00/00000.jpg", "scene0553_00/00040.jpg", "scene0553_00/00080.jpg", "scene0553_00/00120.jpg" ], "turn_id": 1, "scan_id": "scene0553_00", "origin_question": "Remember, have you seen any toilet paper(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any toilet paper(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0553_00/00160.jpg", "scene0553_00/00200.jpg", "scene0553_00/00240.jpg", "scene0553_00/00280.jpg" ], "turn_id": 2, "scan_id": "scene0553_00", "origin_question": "At the end of which turn were both of \"the stainless steel bar with a smooth surface\" and \"the toilet paper\" on your left side?", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the stainless steel bar with a smooth surface\" and \"the toilet paper\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0553_00/00320.jpg", "scene0553_00/00360.jpg", "scene0553_00/00400.jpg", "scene0553_00/00440.jpg" ], "turn_id": 3, "scan_id": "scene0553_00", "origin_question": "Please recall \"the stainless steel bar with a smooth surface\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.28", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the stainless steel bar with a smooth surface\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0553_00/00480.jpg", "scene0553_00/00520.jpg", "scene0553_00/00560.jpg", "scene0553_00/00600.jpg" ], "turn_id": 4, "scan_id": "scene0553_00", "origin_question": "What will be the first-time appearance order of the following categories: backpack, bin, toilet paper?", "option": [ "backpack, toilet paper, bin", "backpack, bin, toilet paper", "toilet paper, bin, backpack", "bin, backpack, toilet paper" ], "answer": "toilet paper, bin, backpack", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: backpack, bin, toilet paper? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['backpack, toilet paper, bin', 'backpack, bin, toilet paper', 'toilet paper, bin, backpack', 'bin, backpack, toilet paper'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0553_00/00640.jpg", "scene0553_00/00680.jpg", "scene0553_00/00720.jpg", "scene0553_00/00760.jpg" ], "turn_id": 5, "scan_id": "scene0553_00", "origin_question": "Please recall \"the stainless steel bar with a smooth surface\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.19", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the stainless steel bar with a smooth surface\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0553_00/00800.jpg", "scene0553_00/00840.jpg", "scene0553_00/00880.jpg", "scene0553_00/00920.jpg" ], "turn_id": 6, "scan_id": "scene0553_00", "origin_question": "Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the tall rectangular silver metal bin\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the tall rectangular silver metal bin\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0553_00/00960.jpg", "scene0553_00/01000.jpg", "scene0553_00/01040.jpg", "scene0553_00/01080.jpg" ], "turn_id": 7, "scan_id": "scene0553_00", "origin_question": "Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the rectangular black and silver plastic soap dispenser\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the rectangular black and silver plastic soap dispenser\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0553_00/01120.jpg", "scene0553_00/01160.jpg", "scene0553_00/01200.jpg", "scene0553_00/01240.jpg" ], "turn_id": 8, "scan_id": "scene0553_00", "origin_question": "Which direction is the tall rectangular silver metal bin in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the tall rectangular silver metal bin in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0553_00/01280.jpg", "scene0553_00/01320.jpg", "scene0553_00/01360.jpg", "scene0553_00/01400.jpg" ], "turn_id": 9, "scan_id": "scene0553_00", "origin_question": "Please recall \"the rectangular black and grey durable backpack\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.66", "user_message": "For the 9 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the rectangular black and grey durable backpack\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0412_00/00000.jpg", "scene0412_00/00040.jpg", "scene0412_00/00080.jpg", "scene0412_00/00120.jpg", "scene0412_00/00160.jpg" ], "turn_id": 1, "scan_id": "scene0412_00", "origin_question": "Remember, have you seen any ladder(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any ladder(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0412_00/00200.jpg", "scene0412_00/00240.jpg", "scene0412_00/00280.jpg", "scene0412_00/00320.jpg", "scene0412_00/00360.jpg" ], "turn_id": 2, "scan_id": "scene0412_00", "origin_question": "When did you first discover \"the rectangular stainless steel wall-mounted sink\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the rectangular stainless steel wall-mounted sink\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0412_00/00400.jpg", "scene0412_00/00440.jpg", "scene0412_00/00480.jpg", "scene0412_00/00520.jpg", "scene0412_00/00560.jpg" ], "turn_id": 3, "scan_id": "scene0412_00", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the rectangular stainless steel wall-mounted sink\" from your position?", "option": [], "answer": "159.46", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the rectangular stainless steel wall-mounted sink\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0412_00/00600.jpg", "scene0412_00/00640.jpg", "scene0412_00/00680.jpg", "scene0412_00/00720.jpg", "scene0412_00/00760.jpg" ], "turn_id": 4, "scan_id": "scene0412_00", "origin_question": "In which turn did you see both \"the black upright vacuum cleaner with a stylish design\" and \"the rectangular stainless steel wall-mounted sink\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the black upright vacuum cleaner with a stylish design\" and \"the rectangular stainless steel wall-mounted sink\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0412_00/00800.jpg", "scene0412_00/00840.jpg", "scene0412_00/00880.jpg", "scene0412_00/00920.jpg", "scene0412_00/00960.jpg" ], "turn_id": 5, "scan_id": "scene0412_00", "origin_question": "Is the medium-sized brown cardboard box to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the medium-sized brown cardboard box to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0412_00/01000.jpg", "scene0412_00/01040.jpg", "scene0412_00/01080.jpg", "scene0412_00/01120.jpg", "scene0412_00/01160.jpg" ], "turn_id": 6, "scan_id": "scene0412_00", "origin_question": "Please recall \"the rectangular stainless steel wall-mounted sink\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "3.35", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the rectangular stainless steel wall-mounted sink\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0412_00/01200.jpg", "scene0412_00/01240.jpg", "scene0412_00/01280.jpg", "scene0412_00/01320.jpg", "scene0412_00/01360.jpg" ], "turn_id": 7, "scan_id": "scene0412_00", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the black upright vacuum cleaner with a stylish design\" from your position?", "option": [], "answer": "121.51", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the black upright vacuum cleaner with a stylish design\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0412_00/01400.jpg", "scene0412_00/01440.jpg", "scene0412_00/01480.jpg", "scene0412_00/01520.jpg", "scene0412_00/01560.jpg" ], "turn_id": 8, "scan_id": "scene0412_00", "origin_question": "Is the rectangular stainless steel wall-mounted sink to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the rectangular stainless steel wall-mounted sink to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0412_00/01600.jpg", "scene0412_00/01640.jpg", "scene0412_00/01680.jpg", "scene0412_00/01720.jpg", "scene0412_00/01760.jpg" ], "turn_id": 9, "scan_id": "scene0412_00", "origin_question": "Which direction is the medium-sized brown cardboard box in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the medium-sized brown cardboard box in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "scene0412_00/01800.jpg", "scene0412_00/01840.jpg", "scene0412_00/01880.jpg", "scene0412_00/01920.jpg", "scene0412_00/01960.jpg" ], "turn_id": 10, "scan_id": "scene0412_00", "origin_question": "Assuming the direction you are facing at the end of the turn 9 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Right", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 9 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0509_00/00000.jpg", "scene0509_00/00040.jpg", "scene0509_00/00080.jpg", "scene0509_00/00120.jpg" ], "turn_id": 1, "scan_id": "scene0509_00", "origin_question": "Remember, have you seen any basket(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any basket(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0509_00/00160.jpg", "scene0509_00/00200.jpg", "scene0509_00/00240.jpg", "scene0509_00/00280.jpg" ], "turn_id": 2, "scan_id": "scene0509_00", "origin_question": "Remember, have you seen any bathtub(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any bathtub(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0509_00/00320.jpg", "scene0509_00/00360.jpg", "scene0509_00/00400.jpg", "scene0509_00/00440.jpg" ], "turn_id": 3, "scan_id": "scene0509_00", "origin_question": "In which turn were you closest to \"the towel\"? ", "option": [], "answer": "2", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the towel\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0509_00/00480.jpg", "scene0509_00/00520.jpg", "scene0509_00/00560.jpg", "scene0509_00/00600.jpg" ], "turn_id": 4, "scan_id": "scene0509_00", "origin_question": "In which turn did you see both \"the towel\" and \"the medium-sized polished stainless steel bar\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the towel\" and \"the medium-sized polished stainless steel bar\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0509_00/00640.jpg", "scene0509_00/00680.jpg", "scene0509_00/00720.jpg", "scene0509_00/00760.jpg" ], "turn_id": 5, "scan_id": "scene0509_00", "origin_question": "Is the medium-sized polished stainless steel bar to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the medium-sized polished stainless steel bar to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0509_00/00800.jpg", "scene0509_00/00840.jpg", "scene0509_00/00880.jpg", "scene0509_00/00920.jpg" ], "turn_id": 6, "scan_id": "scene0509_00", "origin_question": "What will be the first-time appearance order of the following categories: bin, hair dryer, sink?", "option": [ "sink, bin, hair dryer", "bin, hair dryer, sink", "bin, sink, hair dryer", "sink, hair dryer, bin" ], "answer": "bin, sink, hair dryer", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bin, hair dryer, sink? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['sink, bin, hair dryer', 'bin, hair dryer, sink', 'bin, sink, hair dryer', 'sink, hair dryer, bin'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0509_00/00960.jpg", "scene0509_00/01000.jpg", "scene0509_00/01040.jpg", "scene0509_00/01080.jpg" ], "turn_id": 7, "scan_id": "scene0509_00", "origin_question": "Is the rectangular white porcelain sink with a stylish faucet to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the rectangular white porcelain sink with a stylish faucet to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0278_01/00000.jpg", "scene0278_01/00030.jpg", "scene0278_01/00060.jpg", "scene0278_01/00090.jpg" ], "turn_id": 1, "scan_id": "scene0278_01", "origin_question": "Remember, have you seen any bin(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any bin(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0278_01/00120.jpg", "scene0278_01/00150.jpg", "scene0278_01/00180.jpg", "scene0278_01/00210.jpg" ], "turn_id": 2, "scan_id": "scene0278_01", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the small rectangular brown board\"; \"the large gray metal copier with a white top\"; \"the beige horizontal metal cabinet with four drawers\" ", "option": [ "the small rectangular brown board", "the large gray metal copier with a white top", "the beige horizontal metal cabinet with four drawers" ], "answer": "the large gray metal copier with a white top", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the small rectangular brown board\"; \"the large gray metal copier with a white top\"; \"the beige horizontal metal cabinet with four drawers\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the small rectangular brown board', 'the large gray metal copier with a white top', 'the beige horizontal metal cabinet with four drawers'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0278_01/00240.jpg", "scene0278_01/00270.jpg", "scene0278_01/00300.jpg", "scene0278_01/00330.jpg" ], "turn_id": 3, "scan_id": "scene0278_01", "origin_question": "At the end of which turn were both of \"the large gray metal copier with a white top\" and \"the medium-sized round white plastic bucket\" on your left side?", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the large gray metal copier with a white top\" and \"the medium-sized round white plastic bucket\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0278_01/00360.jpg", "scene0278_01/00390.jpg", "scene0278_01/00420.jpg", "scene0278_01/00450.jpg" ], "turn_id": 4, "scan_id": "scene0278_01", "origin_question": "Please recall \"the durable broom with a smooth wooden handle\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.14", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the durable broom with a smooth wooden handle\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0278_01/00480.jpg", "scene0278_01/00510.jpg", "scene0278_01/00540.jpg", "scene0278_01/00570.jpg" ], "turn_id": 5, "scan_id": "scene0278_01", "origin_question": "At the end of which turn were both of \"the large gray metal copier with a white top\" and \"the medium-sized round white plastic bucket\" on your left side?", "option": [], "answer": "3", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the large gray metal copier with a white top\" and \"the medium-sized round white plastic bucket\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0278_01/00600.jpg", "scene0278_01/00630.jpg", "scene0278_01/00660.jpg", "scene0278_01/00690.jpg" ], "turn_id": 6, "scan_id": "scene0278_01", "origin_question": "Is the durable broom with a smooth wooden handle to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the durable broom with a smooth wooden handle to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0222_01/00000.jpg", "scene0222_01/00040.jpg", "scene0222_01/00080.jpg", "scene0222_01/00120.jpg", "scene0222_01/00160.jpg" ], "turn_id": 1, "scan_id": "scene0222_01", "origin_question": "Remember, have you seen any door(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any door(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0222_01/00200.jpg", "scene0222_01/00240.jpg", "scene0222_01/00280.jpg", "scene0222_01/00320.jpg", "scene0222_01/00360.jpg" ], "turn_id": 2, "scan_id": "scene0222_01", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the blue rectangular plastic bin\"; \"the large white plastic cylindrical container with a green cap\"; \"the medium-sized grey cylindrical fan\" ", "option": [ "the blue rectangular plastic bin", "the large white plastic cylindrical container with a green cap", "the medium-sized grey cylindrical fan" ], "answer": "the large white plastic cylindrical container with a green cap", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the blue rectangular plastic bin\"; \"the large white plastic cylindrical container with a green cap\"; \"the medium-sized grey cylindrical fan\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the blue rectangular plastic bin', 'the large white plastic cylindrical container with a green cap', 'the medium-sized grey cylindrical fan'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0222_01/00400.jpg", "scene0222_01/00440.jpg", "scene0222_01/00480.jpg", "scene0222_01/00520.jpg", "scene0222_01/00560.jpg" ], "turn_id": 3, "scan_id": "scene0222_01", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the large white plastic cylindrical container with a green cap\" from your position?", "option": [], "answer": "42.75", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the large white plastic cylindrical container with a green cap\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0222_01/00600.jpg", "scene0222_01/00640.jpg", "scene0222_01/00680.jpg", "scene0222_01/00720.jpg", "scene0222_01/00760.jpg" ], "turn_id": 4, "scan_id": "scene0222_01", "origin_question": "At the end of which turn were both of \"the medium-sized grey cylindrical fan\" and \"the large white plastic cylindrical container with a green cap\" on your right side?", "option": [], "answer": "3", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the medium-sized grey cylindrical fan\" and \"the large white plastic cylindrical container with a green cap\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0222_01/00800.jpg", "scene0222_01/00840.jpg", "scene0222_01/00880.jpg", "scene0222_01/00920.jpg", "scene0222_01/00960.jpg" ], "turn_id": 5, "scan_id": "scene0222_01", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the sleek black lamp with an adjustable long arm\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the sleek black lamp with an adjustable long arm\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0222_01/01000.jpg", "scene0222_01/01040.jpg", "scene0222_01/01080.jpg", "scene0222_01/01120.jpg", "scene0222_01/01160.jpg" ], "turn_id": 6, "scan_id": "scene0222_01", "origin_question": "Which direction is Neatly stacked medium-sized fabric clothes in white to beige in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is Neatly stacked medium-sized fabric clothes in white to beige in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0222_01/01200.jpg", "scene0222_01/01240.jpg", "scene0222_01/01280.jpg", "scene0222_01/01320.jpg", "scene0222_01/01360.jpg" ], "turn_id": 7, "scan_id": "scene0222_01", "origin_question": "Please recall \"the cylindrical, blue soap made of plastic\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.84", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the cylindrical, blue soap made of plastic\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0222_01/01400.jpg", "scene0222_01/01440.jpg", "scene0222_01/01480.jpg", "scene0222_01/01520.jpg", "scene0222_01/01560.jpg" ], "turn_id": 8, "scan_id": "scene0222_01", "origin_question": "In which turn were you closest to \"the blue book made of paper\"? ", "option": [], "answer": "6", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the blue book made of paper\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0222_01/01600.jpg", "scene0222_01/01640.jpg", "scene0222_01/01680.jpg", "scene0222_01/01720.jpg", "scene0222_01/01760.jpg" ], "turn_id": 9, "scan_id": "scene0222_01", "origin_question": "Which two objects are on the same side of you now? \"the white rectangular metal radiator\", \"the brown cylindrical glass wine bottle\", \"the rectangular blue fabric hamper with a closed lid\". ", "option": [ "the white rectangular metal radiator, the brown cylindrical glass wine bottle", "the white rectangular metal radiator, the rectangular blue fabric hamper with a closed lid", "the brown cylindrical glass wine bottle, the rectangular blue fabric hamper with a closed lid" ], "answer": "the brown cylindrical glass wine bottle, the rectangular blue fabric hamper with a closed lid", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the white rectangular metal radiator\", \"the brown cylindrical glass wine bottle\", \"the rectangular blue fabric hamper with a closed lid\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the white rectangular metal radiator, the brown cylindrical glass wine bottle', 'the white rectangular metal radiator, the rectangular blue fabric hamper with a closed lid', 'the brown cylindrical glass wine bottle, the rectangular blue fabric hamper with a closed lid'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0222_01/01800.jpg", "scene0222_01/01840.jpg", "scene0222_01/01880.jpg", "scene0222_01/01920.jpg", "scene0222_01/01960.jpg" ], "turn_id": 10, "scan_id": "scene0222_01", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the cylindrical, blue soap made of plastic\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the cylindrical, blue soap made of plastic\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0044_region26/53a82d3ca3004b1083ffaa2f378dd0e5_i1_0.jpg", "1mp3d_0044_region26/53a82d3ca3004b1083ffaa2f378dd0e5_i2_0.jpg", "1mp3d_0044_region26/53a82d3ca3004b1083ffaa2f378dd0e5_i1_5.jpg", "1mp3d_0044_region26/53a82d3ca3004b1083ffaa2f378dd0e5_i2_5.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0044_region26", "origin_question": "Remember, have you seen any door(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any door(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 2 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "1mp3d_0044_region26/c7096f5da9c94c709477ff0d88635e14_i1_4.jpg", "1mp3d_0044_region26/c7096f5da9c94c709477ff0d88635e14_i2_4.jpg", "1mp3d_0044_region26/c7096f5da9c94c709477ff0d88635e14_i1_5.jpg", "1mp3d_0044_region26/c7096f5da9c94c709477ff0d88635e14_i2_5.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0044_region26", "origin_question": "In which turn did you see both \"the large rectangular dark brown wooden cabinet\" and \"the large dark brown square sink made of stone\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the large rectangular dark brown wooden cabinet\" and \"the large dark brown square sink made of stone\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 2 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0084_region0/b34a815938144c9dadecd6f1dc532eca_i2_3.jpg", "1mp3d_0084_region0/e0cc229442d3469893cdf81c6087f53b_i2_0.jpg", "1mp3d_0084_region0/e0cc229442d3469893cdf81c6087f53b_i2_1.jpg", "1mp3d_0084_region0/72a70368733b4efa8678994b32428fbb_i2_2.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0084_region0", "origin_question": "Remember, have you seen any shelf(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any shelf(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "1mp3d_0084_region0/72a70368733b4efa8678994b32428fbb_i1_1.jpg", "1mp3d_0084_region0/72a70368733b4efa8678994b32428fbb_i2_1.jpg", "1mp3d_0084_region0/72a70368733b4efa8678994b32428fbb_i1_0.jpg", "1mp3d_0084_region0/72a70368733b4efa8678994b32428fbb_i2_0.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0084_region0", "origin_question": "Is the small rectangular brown wooden shelf to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the small rectangular brown wooden shelf to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "1mp3d_0084_region0/72a70368733b4efa8678994b32428fbb_i2_5.jpg", "1mp3d_0084_region0/e0cc229442d3469893cdf81c6087f53b_i1_4.jpg", "1mp3d_0084_region0/e0cc229442d3469893cdf81c6087f53b_i2_4.jpg", "1mp3d_0084_region0/e0cc229442d3469893cdf81c6087f53b_i1_5.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0084_region0", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the small rectangular brown wooden shelf\" from your position?", "option": [], "answer": "86.61", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the small rectangular brown wooden shelf\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "1mp3d_0084_region0/e0cc229442d3469893cdf81c6087f53b_i2_5.jpg", "1mp3d_0084_region0/b34a815938144c9dadecd6f1dc532eca_i1_5.jpg", "1mp3d_0084_region0/b34a815938144c9dadecd6f1dc532eca_i2_5.jpg", "1mp3d_0084_region0/b34a815938144c9dadecd6f1dc532eca_i1_0.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0084_region0", "origin_question": "At the end of which turn were both of \"the small rectangular brown wooden shelf\" and \"the tall rectangular plant\" on your right side?", "option": [], "answer": "2", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the small rectangular brown wooden shelf\" and \"the tall rectangular plant\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "1mp3d_0047_region0/6ebcdb37ae1e44e2ae9dcbbff6b5a08e_i1_4.jpg", "1mp3d_0047_region0/6ebcdb37ae1e44e2ae9dcbbff6b5a08e_i2_4.jpg", "1mp3d_0047_region0/6ebcdb37ae1e44e2ae9dcbbff6b5a08e_i1_5.jpg", "1mp3d_0047_region0/6ebcdb37ae1e44e2ae9dcbbff6b5a08e_i2_5.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0047_region0", "origin_question": "Remember, how many chair(s) have you seen so far? ", "option": [], "answer": "2", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many chair(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "1mp3d_0047_region0/9e3431d5d6414ca6bd3c159a25c30999_i1_1.jpg", "1mp3d_0047_region0/9e3431d5d6414ca6bd3c159a25c30999_i2_1.jpg", "1mp3d_0047_region0/9e3431d5d6414ca6bd3c159a25c30999_i1_2.jpg", "1mp3d_0047_region0/9e3431d5d6414ca6bd3c159a25c30999_i2_2.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0047_region0", "origin_question": "In which turn did you see both \"the wardrobe\" and \"the boxy CRT TV with a silver frame and dark center\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the wardrobe\" and \"the boxy CRT TV with a silver frame and dark center\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "1mp3d_0047_region0/119f12ba94004df3b643d55195b25b21_i1_1.jpg", "1mp3d_0047_region0/119f12ba94004df3b643d55195b25b21_i2_1.jpg", "1mp3d_0047_region0/af7ed758b8c04875a5ae383252e35332_i1_1.jpg", "1mp3d_0047_region0/af7ed758b8c04875a5ae383252e35332_i2_1.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0047_region0", "origin_question": "In which turn did you see both \"the modern rectangular light grey headboard\" and \"the boxy CRT TV with a silver frame and dark center\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the modern rectangular light grey headboard\" and \"the boxy CRT TV with a silver frame and dark center\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "1mp3d_0047_region0/af7ed758b8c04875a5ae383252e35332_i1_0.jpg", "1mp3d_0047_region0/af7ed758b8c04875a5ae383252e35332_i2_0.jpg", "1mp3d_0047_region0/119f12ba94004df3b643d55195b25b21_i1_0.jpg", "1mp3d_0047_region0/119f12ba94004df3b643d55195b25b21_i2_0.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0047_region0", "origin_question": "Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Left", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "1mp3d_0047_region0/a6fea230be6d47bcbb520d23e8d55724_i1_4.jpg", "1mp3d_0047_region0/a6fea230be6d47bcbb520d23e8d55724_i2_4.jpg", "1mp3d_0047_region0/b11577607584497581c9433f5406c55b_i1_5.jpg", "1mp3d_0047_region0/b11577607584497581c9433f5406c55b_i2_5.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0047_region0", "origin_question": "How far is your current position from where you were at the end of the turn 4? (in meters)", "option": [], "answer": "2.55", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 4? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0047_region0/a6fea230be6d47bcbb520d23e8d55724_i1_2.jpg", "1mp3d_0047_region0/a6fea230be6d47bcbb520d23e8d55724_i2_2.jpg", "1mp3d_0047_region0/a6fea230be6d47bcbb520d23e8d55724_i1_1.jpg", "1mp3d_0047_region0/a6fea230be6d47bcbb520d23e8d55724_i2_1.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0047_region0", "origin_question": "Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the modern rectangular light grey headboard\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the modern rectangular light grey headboard\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0171_01/00000.jpg", "scene0171_01/00040.jpg", "scene0171_01/00080.jpg", "scene0171_01/00120.jpg" ], "turn_id": 1, "scan_id": "scene0171_01", "origin_question": "Remember, have you seen any chair(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any chair(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0171_01/00160.jpg", "scene0171_01/00200.jpg", "scene0171_01/00240.jpg", "scene0171_01/00280.jpg" ], "turn_id": 2, "scan_id": "scene0171_01", "origin_question": "When did you first discover \"the rectangular, white glass panel door\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the rectangular, white glass panel door\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0171_01/00320.jpg", "scene0171_01/00360.jpg", "scene0171_01/00400.jpg", "scene0171_01/00440.jpg" ], "turn_id": 3, "scan_id": "scene0171_01", "origin_question": "", "option": [], "answer": "", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0171_01/00480.jpg", "scene0171_01/00520.jpg", "scene0171_01/00560.jpg", "scene0171_01/00600.jpg" ], "turn_id": 4, "scan_id": "scene0171_01", "origin_question": "In which turn were you farthest from \"the smooth rectangular white glossy table\"? ", "option": [], "answer": "3", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the smooth rectangular white glossy table\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0171_01/00640.jpg", "scene0171_01/00680.jpg", "scene0171_01/00720.jpg", "scene0171_01/00760.jpg" ], "turn_id": 5, "scan_id": "scene0171_01", "origin_question": "In which turn were you closest to \"the window\"? ", "option": [], "answer": "5", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the window\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0171_01/00800.jpg", "scene0171_01/00840.jpg", "scene0171_01/00880.jpg", "scene0171_01/00920.jpg" ], "turn_id": 6, "scan_id": "scene0171_01", "origin_question": "Using your orientation at the end of turn 5 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 5 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0616_00/00000.jpg", "scene0616_00/00040.jpg", "scene0616_00/00080.jpg", "scene0616_00/00120.jpg", "scene0616_00/00160.jpg" ], "turn_id": 1, "scan_id": "scene0616_00", "origin_question": "Remember, how many table(s) have you seen so far? ", "option": [], "answer": "2", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many table(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0616_00/00200.jpg", "scene0616_00/00240.jpg", "scene0616_00/00280.jpg", "scene0616_00/00320.jpg", "scene0616_00/00360.jpg" ], "turn_id": 2, "scan_id": "scene0616_00", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the windowsill\" from your position?", "option": [], "answer": "70.0", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the windowsill\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0616_00/00400.jpg", "scene0616_00/00440.jpg", "scene0616_00/00480.jpg", "scene0616_00/00520.jpg", "scene0616_00/00560.jpg" ], "turn_id": 3, "scan_id": "scene0616_00", "origin_question": "Is the black and grey vacuum cleaner with a red hose to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the black and grey vacuum cleaner with a red hose to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0616_00/00600.jpg", "scene0616_00/00640.jpg", "scene0616_00/00680.jpg", "scene0616_00/00720.jpg", "scene0616_00/00760.jpg" ], "turn_id": 4, "scan_id": "scene0616_00", "origin_question": "At the end of which turn were both of \"the black and grey vacuum cleaner with a red hose\" and \"the round metal lamp with a cylindrical shade\" on your right side?", "option": [], "answer": "4", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the black and grey vacuum cleaner with a red hose\" and \"the round metal lamp with a cylindrical shade\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0616_00/00800.jpg", "scene0616_00/00840.jpg", "scene0616_00/00880.jpg", "scene0616_00/00920.jpg", "scene0616_00/00960.jpg" ], "turn_id": 5, "scan_id": "scene0616_00", "origin_question": "In which turn did you see both \"the black and grey vacuum cleaner with a red hose\" and \"the radiator\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the black and grey vacuum cleaner with a red hose\" and \"the radiator\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "scene0616_00/01000.jpg", "scene0616_00/01040.jpg", "scene0616_00/01080.jpg", "scene0616_00/01120.jpg", "scene0616_00/01160.jpg" ], "turn_id": 6, "scan_id": "scene0616_00", "origin_question": "Assuming the direction you are facing at the end of the turn 5 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Left", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 5 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0616_00/01200.jpg", "scene0616_00/01240.jpg", "scene0616_00/01280.jpg", "scene0616_00/01320.jpg", "scene0616_00/01360.jpg" ], "turn_id": 7, "scan_id": "scene0616_00", "origin_question": "Please recall \"the round metal lamp with a cylindrical shade\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.43", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the round metal lamp with a cylindrical shade\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0616_00/01400.jpg", "scene0616_00/01440.jpg", "scene0616_00/01480.jpg", "scene0616_00/01520.jpg", "scene0616_00/01560.jpg" ], "turn_id": 8, "scan_id": "scene0616_00", "origin_question": "At the end of which turn were both of \"the window\" and \"the round metal lamp with a cylindrical shade\" on your left side?", "option": [], "answer": "5", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the window\" and \"the round metal lamp with a cylindrical shade\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0616_00/01600.jpg", "scene0616_00/01640.jpg", "scene0616_00/01680.jpg", "scene0616_00/01720.jpg", "scene0616_00/01760.jpg" ], "turn_id": 9, "scan_id": "scene0616_00", "origin_question": "Using your orientation at the end of turn 8 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 8 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0616_00/01800.jpg", "scene0616_00/01840.jpg", "scene0616_00/01880.jpg", "scene0616_00/01920.jpg", "scene0616_00/01960.jpg" ], "turn_id": 10, "scan_id": "scene0616_00", "origin_question": "In which turn were you closest to \"the round metal lamp with a cylindrical shade\"? ", "option": [], "answer": "9", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the round metal lamp with a cylindrical shade\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0334_00/00000.jpg", "scene0334_00/00040.jpg", "scene0334_00/00080.jpg", "scene0334_00/00120.jpg" ], "turn_id": 1, "scan_id": "scene0334_00", "origin_question": "Remember, have you seen any couch(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any couch(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0334_00/00160.jpg", "scene0334_00/00200.jpg", "scene0334_00/00240.jpg", "scene0334_00/00280.jpg" ], "turn_id": 2, "scan_id": "scene0334_00", "origin_question": "In which turn were you closest to \"the round coffee table with a light brown wooden veneer\"? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the round coffee table with a light brown wooden veneer\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0334_00/00320.jpg", "scene0334_00/00360.jpg", "scene0334_00/00400.jpg", "scene0334_00/00440.jpg" ], "turn_id": 3, "scan_id": "scene0334_00", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "22.32", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0334_00/00480.jpg", "scene0334_00/00520.jpg", "scene0334_00/00560.jpg", "scene0334_00/00600.jpg" ], "turn_id": 4, "scan_id": "scene0334_00", "origin_question": "Remember, have you seen any bench(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any bench(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "scene0334_00/00640.jpg", "scene0334_00/00680.jpg", "scene0334_00/00720.jpg", "scene0334_00/00760.jpg" ], "turn_id": 5, "scan_id": "scene0334_00", "origin_question": "Assuming the direction you are facing at the end of the turn 4 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Right", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 4 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0334_00/00800.jpg", "scene0334_00/00840.jpg", "scene0334_00/00880.jpg", "scene0334_00/00950.jpg" ], "turn_id": 6, "scan_id": "scene0334_00", "origin_question": "How far is your current position from where you were at the end of the turn 1? (in meters)", "option": [], "answer": "4.63", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 1? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0334_00/00990.jpg", "scene0334_00/01030.jpg", "scene0334_00/01070.jpg", "scene0334_00/01110.jpg" ], "turn_id": 7, "scan_id": "scene0334_00", "origin_question": "Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the round coffee table with a light brown wooden veneer\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the round coffee table with a light brown wooden veneer\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0040_region8/a7c413edfeba4661930406b501a30131_i1_4.jpg", "1mp3d_0040_region8/a7c413edfeba4661930406b501a30131_i2_4.jpg", "1mp3d_0040_region8/cfa7edd3cdea4c428cd1b2e2bdb4aa31_i1_5.jpg", "1mp3d_0040_region8/cfa7edd3cdea4c428cd1b2e2bdb4aa31_i2_5.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0040_region8", "origin_question": "Remember, have you seen any switch(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any switch(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 2 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0040_region8/a7c413edfeba4661930406b501a30131_i1_2.jpg", "1mp3d_0040_region8/a7c413edfeba4661930406b501a30131_i2_2.jpg", "1mp3d_0040_region8/cfa7edd3cdea4c428cd1b2e2bdb4aa31_i2_2.jpg", "1mp3d_0040_region8/cfa7edd3cdea4c428cd1b2e2bdb4aa31_i1_3.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0040_region8", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 2 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "1mp3d_0074_region13/1d383b2f6a454cdbb4d942b127a12320_i1_1.jpg", "1mp3d_0074_region13/1d383b2f6a454cdbb4d942b127a12320_i2_1.jpg", "1mp3d_0074_region13/20cebd99bcf84b1c83bc9b6abbd78102_i1_0.jpg", "1mp3d_0074_region13/20cebd99bcf84b1c83bc9b6abbd78102_i2_0.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0074_region13", "origin_question": "Remember, how many window(s) have you seen so far? ", "option": [], "answer": "2", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many window(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "1mp3d_0074_region13/71d49ea7d95d406d944f73ad62031b79_i1_5.jpg", "1mp3d_0074_region13/71d49ea7d95d406d944f73ad62031b79_i2_5.jpg", "1mp3d_0074_region13/6b2856b4c9874823859788e1c35b9e5d_i1_5.jpg", "1mp3d_0074_region13/6b2856b4c9874823859788e1c35b9e5d_i2_5.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0074_region13", "origin_question": "At the end of which turn were both of \"the small, sturdy black shelf with a simple design\" and \"the black and white polka dot circular mat\" on your right side?", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the small, sturdy black shelf with a simple design\" and \"the black and white polka dot circular mat\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "1mp3d_0074_region13/6b2856b4c9874823859788e1c35b9e5d_i1_0.jpg", "1mp3d_0074_region13/6b2856b4c9874823859788e1c35b9e5d_i2_0.jpg", "1mp3d_0074_region13/6b2856b4c9874823859788e1c35b9e5d_i1_1.jpg", "1mp3d_0074_region13/6b2856b4c9874823859788e1c35b9e5d_i2_1.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0074_region13", "origin_question": "At the end of which turn were both of \"the black and white polka dot circular mat\" and \"the tall tripod stand made of dark wood\" on your right side?", "option": [], "answer": "2", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the black and white polka dot circular mat\" and \"the tall tripod stand made of dark wood\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "1mp3d_0074_region13/71d49ea7d95d406d944f73ad62031b79_i1_1.jpg", "1mp3d_0074_region13/71d49ea7d95d406d944f73ad62031b79_i2_1.jpg", "1mp3d_0074_region13/20cebd99bcf84b1c83bc9b6abbd78102_i1_4.jpg", "1mp3d_0074_region13/20cebd99bcf84b1c83bc9b6abbd78102_i2_4.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0074_region13", "origin_question": "By the end of the 3 turn, were you getting closer to or farther away from these two objects: X: \"the small, sturdy black shelf with a simple design\", Y: \"the long, gray, semi-transparent curtain with a black ribbon\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to X and farther away from Y", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 3 turn, were you getting closer to or farther away from these two objects: X: \"the small, sturdy black shelf with a simple design\", Y: \"the long, gray, semi-transparent curtain with a black ribbon\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "1mp3d_0074_region13/20cebd99bcf84b1c83bc9b6abbd78102_i1_5.jpg", "1mp3d_0074_region13/20cebd99bcf84b1c83bc9b6abbd78102_i2_5.jpg", "1mp3d_0074_region13/2bc3dc4bf7cc4489a8cbcbed9d7d6b25_i1_3.jpg", "1mp3d_0074_region13/2bc3dc4bf7cc4489a8cbcbed9d7d6b25_i2_3.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0074_region13", "origin_question": "In which turn were you closest to \"the black and white polka dot circular mat\"? ", "option": [], "answer": "5", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the black and white polka dot circular mat\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "1mp3d_0074_region13/100d11cb989b46078ef7edb1b4dd34b9_i2_3.jpg", "1mp3d_0074_region13/100d11cb989b46078ef7edb1b4dd34b9_i1_2.jpg", "1mp3d_0074_region13/100d11cb989b46078ef7edb1b4dd34b9_i2_2.jpg", "1mp3d_0074_region13/2bc3dc4bf7cc4489a8cbcbed9d7d6b25_i1_4.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0074_region13", "origin_question": "Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "23.78", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "1mp3d_0074_region13/2bc3dc4bf7cc4489a8cbcbed9d7d6b25_i2_4.jpg", "1mp3d_0074_region13/2bc3dc4bf7cc4489a8cbcbed9d7d6b25_i1_5.jpg", "1mp3d_0074_region13/2bc3dc4bf7cc4489a8cbcbed9d7d6b25_i2_5.jpg", "1mp3d_0074_region13/2bc3dc4bf7cc4489a8cbcbed9d7d6b25_i1_0.jpg" ], "turn_id": 7, "scan_id": "1mp3d_0074_region13", "origin_question": "Which direction is the tall tripod stand made of dark wood in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the tall tripod stand made of dark wood in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0057_region0/788770d79fc940cfa0b70fdd7c4f58d0_i1_3.jpg", "1mp3d_0057_region0/788770d79fc940cfa0b70fdd7c4f58d0_i2_3.jpg", "1mp3d_0057_region0/479ed85f86aa4a8bab37194b5eeaa051_i1_2.jpg", "1mp3d_0057_region0/479ed85f86aa4a8bab37194b5eeaa051_i2_2.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0057_region0", "origin_question": "Remember, have you seen any drawer(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any drawer(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "1mp3d_0057_region0/f3c0b6536ff04efaae476fd547aea850_i1_0.jpg", "1mp3d_0057_region0/f3c0b6536ff04efaae476fd547aea850_i2_0.jpg", "1mp3d_0057_region0/479ed85f86aa4a8bab37194b5eeaa051_i1_0.jpg", "1mp3d_0057_region0/479ed85f86aa4a8bab37194b5eeaa051_i2_0.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0057_region0", "origin_question": "In which turn were you closest to \"the rectangular light brown wooden drawer with a handle\"? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the rectangular light brown wooden drawer with a handle\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "1mp3d_0057_region0/479ed85f86aa4a8bab37194b5eeaa051_i1_1.jpg", "1mp3d_0057_region0/479ed85f86aa4a8bab37194b5eeaa051_i2_1.jpg", "1mp3d_0057_region0/788770d79fc940cfa0b70fdd7c4f58d0_i1_5.jpg", "1mp3d_0057_region0/788770d79fc940cfa0b70fdd7c4f58d0_i2_5.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0057_region0", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the rectangular light brown wooden drawer with a handle\"; \"the small, light brown wooden stand\"; \"the rectangular, striped cotton pillow in warm colors\" ", "option": [ "the rectangular light brown wooden drawer with a handle", "the small, light brown wooden stand", "the rectangular, striped cotton pillow in warm colors" ], "answer": "the rectangular, striped cotton pillow in warm colors", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the rectangular light brown wooden drawer with a handle\"; \"the small, light brown wooden stand\"; \"the rectangular, striped cotton pillow in warm colors\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular light brown wooden drawer with a handle', 'the small, light brown wooden stand', 'the rectangular, striped cotton pillow in warm colors'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "1mp3d_0057_region0/788770d79fc940cfa0b70fdd7c4f58d0_i1_1.jpg", "1mp3d_0057_region0/788770d79fc940cfa0b70fdd7c4f58d0_i2_1.jpg", "1mp3d_0057_region0/788770d79fc940cfa0b70fdd7c4f58d0_i1_0.jpg", "1mp3d_0057_region0/788770d79fc940cfa0b70fdd7c4f58d0_i2_0.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0057_region0", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the rectangular light brown wooden drawer with a handle\" from your position?", "option": [], "answer": "178.39", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the rectangular light brown wooden drawer with a handle\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "1mp3d_0057_region0/cd7869a1f80046b5adeeb20d8108a409_i1_4.jpg", "1mp3d_0057_region0/cd7869a1f80046b5adeeb20d8108a409_i2_4.jpg", "1mp3d_0057_region0/a8794beea1a240dfa0d97e84d6ea00bb_i1_5.jpg", "1mp3d_0057_region0/a8794beea1a240dfa0d97e84d6ea00bb_i2_5.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0057_region0", "origin_question": "At the end of which turn were both of \"the black flat-screen television of medium size\" and \"the rectangular, striped cotton pillow in warm colors\" on your left side?", "option": [], "answer": "5", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the black flat-screen television of medium size\" and \"the rectangular, striped cotton pillow in warm colors\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "1mp3d_0057_region0/cd7869a1f80046b5adeeb20d8108a409_i1_5.jpg", "1mp3d_0057_region0/cd7869a1f80046b5adeeb20d8108a409_i2_5.jpg", "1mp3d_0057_region0/a8794beea1a240dfa0d97e84d6ea00bb_i1_3.jpg", "1mp3d_0057_region0/a8794beea1a240dfa0d97e84d6ea00bb_i2_3.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0057_region0", "origin_question": "What will be the first-time appearance order of the following categories: pillow, stand, tv?", "option": [ "pillow, stand, tv", "tv, pillow, stand", "pillow, tv, stand", "tv, stand, pillow" ], "answer": "tv, stand, pillow", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: pillow, stand, tv? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['pillow, stand, tv', 'tv, pillow, stand', 'pillow, tv, stand', 'tv, stand, pillow'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0387_00/00000.jpg", "scene0387_00/00030.jpg", "scene0387_00/00060.jpg", "scene0387_00/00090.jpg" ], "turn_id": 1, "scan_id": "scene0387_00", "origin_question": "Remember, have you seen any cap(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any cap(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0387_00/00120.jpg", "scene0387_00/00150.jpg", "scene0387_00/00180.jpg", "scene0387_00/00210.jpg" ], "turn_id": 2, "scan_id": "scene0387_00", "origin_question": "When did you first discover \"the long blue rectangular plastic bin with a lid\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the long blue rectangular plastic bin with a lid\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0387_00/00240.jpg", "scene0387_00/00270.jpg", "scene0387_00/00300.jpg", "scene0387_00/00330.jpg" ], "turn_id": 3, "scan_id": "scene0387_00", "origin_question": "In which turn did you see both \"the window\" and \"the long blue rectangular plastic bin with a lid\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the window\" and \"the long blue rectangular plastic bin with a lid\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0387_00/00360.jpg", "scene0387_00/00390.jpg", "scene0387_00/00420.jpg", "scene0387_00/00450.jpg" ], "turn_id": 4, "scan_id": "scene0387_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "scene0387_00/00480.jpg", "scene0387_00/00510.jpg", "scene0387_00/00540.jpg", "scene0387_00/00570.jpg" ], "turn_id": 5, "scan_id": "scene0387_00", "origin_question": "Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Right", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0387_00/00600.jpg", "scene0387_00/00630.jpg", "scene0387_00/00660.jpg", "scene0387_00/00690.jpg" ], "turn_id": 6, "scan_id": "scene0387_00", "origin_question": "How far is your current position from where you were at the end of the turn 5? (in meters)", "option": [], "answer": "1.93", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 5? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0387_00/00720.jpg", "scene0387_00/00750.jpg", "scene0387_00/00780.jpg", "scene0387_00/00810.jpg" ], "turn_id": 7, "scan_id": "scene0387_00", "origin_question": "Using your orientation at the end of turn 5 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "70.59", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 5 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0029_region4/0662f8da68b94d83a47af1e38f28f4e5_i1_3.jpg", "1mp3d_0029_region4/0662f8da68b94d83a47af1e38f28f4e5_i2_3.jpg", "1mp3d_0029_region4/9a503b8a42274ecc9c1fde72a893b75f_i2_4.jpg", "1mp3d_0029_region4/9a503b8a42274ecc9c1fde72a893b75f_i2_5.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0029_region4", "origin_question": "Remember, have you seen any stairs(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any stairs(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 3 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0029_region4/9a503b8a42274ecc9c1fde72a893b75f_i1_3.jpg", "1mp3d_0029_region4/9a503b8a42274ecc9c1fde72a893b75f_i2_3.jpg", "1mp3d_0029_region4/9a503b8a42274ecc9c1fde72a893b75f_i1_0.jpg", "1mp3d_0029_region4/9a503b8a42274ecc9c1fde72a893b75f_i2_0.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0029_region4", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 3 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0029_region4/2f211467f33544ff97c624d91f37e5fd_i1_0.jpg", "1mp3d_0029_region4/2f211467f33544ff97c624d91f37e5fd_i2_0.jpg", "1mp3d_0029_region4/2f211467f33544ff97c624d91f37e5fd_i1_5.jpg", "1mp3d_0029_region4/2f211467f33544ff97c624d91f37e5fd_i2_5.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0029_region4", "origin_question": "", "option": [], "answer": "", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 3 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0574_00/00000.jpg", "scene0574_00/00040.jpg", "scene0574_00/00080.jpg", "scene0574_00/00120.jpg" ], "turn_id": 1, "scan_id": "scene0574_00", "origin_question": "Remember, have you seen any shoe(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any shoe(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0574_00/00160.jpg", "scene0574_00/00200.jpg", "scene0574_00/00240.jpg", "scene0574_00/00280.jpg" ], "turn_id": 2, "scan_id": "scene0574_00", "origin_question": "When did you first discover \"the light gray rectangular board\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the light gray rectangular board\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0574_00/00320.jpg", "scene0574_00/00360.jpg", "scene0574_00/00400.jpg", "scene0574_00/00440.jpg" ], "turn_id": 3, "scan_id": "scene0574_00", "origin_question": "Remember, have you seen any bench(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any bench(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0574_00/00480.jpg", "scene0574_00/00520.jpg", "scene0574_00/00560.jpg", "scene0574_00/00600.jpg" ], "turn_id": 4, "scan_id": "scene0574_00", "origin_question": "Which two objects are on the same side of you now? \"the light gray rectangular board\", \"the bin\", \"the faucet\". ", "option": [ "the light gray rectangular board, the bin", "the light gray rectangular board, the faucet", "the bin, the faucet" ], "answer": "the light gray rectangular board, the bin", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the light gray rectangular board\", \"the bin\", \"the faucet\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the light gray rectangular board, the bin', 'the light gray rectangular board, the faucet', 'the bin, the faucet'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0574_00/00640.jpg", "scene0574_00/00680.jpg", "scene0574_00/00720.jpg", "scene0574_00/00760.jpg" ], "turn_id": 5, "scan_id": "scene0574_00", "origin_question": "What will be the first-time appearance order of the following categories: bench, bin, faucet?", "option": [ "bin, bench, faucet", "bench, bin, faucet", "bench, faucet, bin", "bin, faucet, bench" ], "answer": "bench, bin, faucet", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bench, bin, faucet? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['bin, bench, faucet', 'bench, bin, faucet', 'bench, faucet, bin', 'bin, faucet, bench'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "scene0574_00/00800.jpg", "scene0574_00/00840.jpg", "scene0574_00/00880.jpg", "scene0574_00/00920.jpg" ], "turn_id": 6, "scan_id": "scene0574_00", "origin_question": "Among these three objects(\"the bench\",\"the bin\",\"the sink\" ), which one is the closest to you?", "option": [ "the bench", "the bin", "the sink" ], "answer": "the sink", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects(\"the bench\",\"the bin\",\"the sink\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the bench', 'the bin', 'the sink'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0614_00/00000.jpg", "scene0614_00/00030.jpg", "scene0614_00/00060.jpg", "scene0614_00/00090.jpg" ], "turn_id": 1, "scan_id": "scene0614_00", "origin_question": "Remember, how many chair(s) have you seen so far? ", "option": [], "answer": "4", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many chair(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0614_00/00120.jpg", "scene0614_00/00150.jpg", "scene0614_00/00180.jpg", "scene0614_00/00210.jpg" ], "turn_id": 2, "scan_id": "scene0614_00", "origin_question": "In which turn did you see both \"the radiator\" and \"the small round orange plastic bowl\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the radiator\" and \"the small round orange plastic bowl\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0614_00/00240.jpg", "scene0614_00/00270.jpg", "scene0614_00/00300.jpg", "scene0614_00/00330.jpg" ], "turn_id": 3, "scan_id": "scene0614_00", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the clothes\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the clothes\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0614_00/00360.jpg", "scene0614_00/00390.jpg", "scene0614_00/00420.jpg", "scene0614_00/00450.jpg" ], "turn_id": 4, "scan_id": "scene0614_00", "origin_question": "Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0614_00/00480.jpg", "scene0614_00/00510.jpg", "scene0614_00/00540.jpg", "scene0614_00/00570.jpg" ], "turn_id": 5, "scan_id": "scene0614_00", "origin_question": "Which direction is the bright rectangular canvas picture in red and blue in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the bright rectangular canvas picture in red and blue in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "scene0614_00/00600.jpg", "scene0614_00/00630.jpg", "scene0614_00/00660.jpg", "scene0614_00/00690.jpg" ], "turn_id": 6, "scan_id": "scene0614_00", "origin_question": "Among these three objects(\"the small round orange plastic bowl\",\"the stylish black laptop with a silver keyboard\",\"the round wall clock with a black frame and white face\" ), which one is the farthest from you?", "option": [ "the small round orange plastic bowl", "the stylish black laptop with a silver keyboard", "the round wall clock with a black frame and white face" ], "answer": "the round wall clock with a black frame and white face", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects(\"the small round orange plastic bowl\",\"the stylish black laptop with a silver keyboard\",\"the round wall clock with a black frame and white face\" ), which one is the farthest from you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the small round orange plastic bowl', 'the stylish black laptop with a silver keyboard', 'the round wall clock with a black frame and white face'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0614_00/00720.jpg", "scene0614_00/00750.jpg", "scene0614_00/00780.jpg", "scene0614_00/00810.jpg" ], "turn_id": 7, "scan_id": "scene0614_00", "origin_question": "Which two objects are on the same side of you now? \"the clothes\", \"the light brown rectangular wooden door\", \"the round wall clock with a black frame and white face\". ", "option": [ "the clothes, the light brown rectangular wooden door", "the clothes, the round wall clock with a black frame and white face", "the light brown rectangular wooden door, the round wall clock with a black frame and white face" ], "answer": "the clothes, the light brown rectangular wooden door", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the clothes\", \"the light brown rectangular wooden door\", \"the round wall clock with a black frame and white face\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the clothes, the light brown rectangular wooden door', 'the clothes, the round wall clock with a black frame and white face', 'the light brown rectangular wooden door, the round wall clock with a black frame and white face'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0067_region26/7b14552fe83b4e76974f711b8fcfc46f_i1_3.jpg", "1mp3d_0067_region26/7b14552fe83b4e76974f711b8fcfc46f_i2_3.jpg", "1mp3d_0067_region26/7b14552fe83b4e76974f711b8fcfc46f_i1_4.jpg", "1mp3d_0067_region26/7b14552fe83b4e76974f711b8fcfc46f_i2_4.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0067_region26", "origin_question": "Remember, have you seen any fireplace(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any fireplace(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0067_region26/7e3ee6d9948547619b1ef64c8bf3b350_i1_3.jpg", "1mp3d_0067_region26/7e3ee6d9948547619b1ef64c8bf3b350_i2_3.jpg", "1mp3d_0067_region26/7b14552fe83b4e76974f711b8fcfc46f_i1_2.jpg", "1mp3d_0067_region26/7b14552fe83b4e76974f711b8fcfc46f_i2_2.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0067_region26", "origin_question": "Remember, have you seen any shelf(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any shelf(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "1mp3d_0067_region26/7b14552fe83b4e76974f711b8fcfc46f_i1_1.jpg", "1mp3d_0067_region26/7b14552fe83b4e76974f711b8fcfc46f_i2_1.jpg", "1mp3d_0067_region26/7b14552fe83b4e76974f711b8fcfc46f_i1_0.jpg", "1mp3d_0067_region26/7b14552fe83b4e76974f711b8fcfc46f_i2_0.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0067_region26", "origin_question": "In which turn did you see both \"the metallic grey rectangular washing machine made of metal\" and \"the smooth rectangular granite countertop in light gray and white\" simultaneously? ", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the metallic grey rectangular washing machine made of metal\" and \"the smooth rectangular granite countertop in light gray and white\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0067_region26/7b14552fe83b4e76974f711b8fcfc46f_i1_5.jpg", "1mp3d_0067_region26/7b14552fe83b4e76974f711b8fcfc46f_i2_5.jpg", "1mp3d_0067_region26/7e3ee6d9948547619b1ef64c8bf3b350_i1_1.jpg", "1mp3d_0067_region26/7e3ee6d9948547619b1ef64c8bf3b350_i2_1.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0067_region26", "origin_question": "", "option": [], "answer": "", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "1mp3d_0067_region26/7e3ee6d9948547619b1ef64c8bf3b350_i1_0.jpg", "1mp3d_0067_region26/7e3ee6d9948547619b1ef64c8bf3b350_i2_0.jpg", "1mp3d_0067_region26/7e3ee6d9948547619b1ef64c8bf3b350_i1_2.jpg", "1mp3d_0067_region26/7e3ee6d9948547619b1ef64c8bf3b350_i2_2.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0067_region26", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "2.26", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "1mp3d_0067_region26/7e3ee6d9948547619b1ef64c8bf3b350_i1_4.jpg", "1mp3d_0067_region26/7e3ee6d9948547619b1ef64c8bf3b350_i2_4.jpg", "1mp3d_0067_region26/7e3ee6d9948547619b1ef64c8bf3b350_i1_5.jpg", "1mp3d_0067_region26/7e3ee6d9948547619b1ef64c8bf3b350_i2_5.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0067_region26", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the chrome-plated metal faucet with a shiny silver color\" from your position?", "option": [], "answer": "119.35", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the chrome-plated metal faucet with a shiny silver color\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0631_00/00000.jpg", "scene0631_00/00030.jpg", "scene0631_00/00060.jpg", "scene0631_00/00090.jpg" ], "turn_id": 1, "scan_id": "scene0631_00", "origin_question": "Remember, have you seen any sink(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any sink(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0631_00/00120.jpg", "scene0631_00/00150.jpg", "scene0631_00/00180.jpg", "scene0631_00/00210.jpg" ], "turn_id": 2, "scan_id": "scene0631_00", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the rectangular ceramic sink with a light brown coating\"; \"the rectangular light beige ceramic vanity countertop\"; \"the white plastic bin on a silver base\" ", "option": [ "the rectangular ceramic sink with a light brown coating", "the rectangular light beige ceramic vanity countertop", "the white plastic bin on a silver base" ], "answer": "the white plastic bin on a silver base", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the rectangular ceramic sink with a light brown coating\"; \"the rectangular light beige ceramic vanity countertop\"; \"the white plastic bin on a silver base\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular ceramic sink with a light brown coating', 'the rectangular light beige ceramic vanity countertop', 'the white plastic bin on a silver base'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0631_00/00240.jpg", "scene0631_00/00270.jpg", "scene0631_00/00300.jpg", "scene0631_00/00330.jpg" ], "turn_id": 3, "scan_id": "scene0631_00", "origin_question": "Which direction is the small blue cylindrical bottle with a slender neck in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the small blue cylindrical bottle with a slender neck in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0631_00/00360.jpg", "scene0631_00/00390.jpg", "scene0631_00/00420.jpg", "scene0631_00/00450.jpg" ], "turn_id": 4, "scan_id": "scene0631_00", "origin_question": "Which two objects are on the same side of you now? \"the circular green plastic bucket with a handle\", \"the window\", \"the white plastic bin on a silver base\". ", "option": [ "the circular green plastic bucket with a handle, the window", "the circular green plastic bucket with a handle, the white plastic bin on a silver base", "the window, the white plastic bin on a silver base" ], "answer": "the window, the white plastic bin on a silver base", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the circular green plastic bucket with a handle\", \"the window\", \"the white plastic bin on a silver base\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the circular green plastic bucket with a handle, the window', 'the circular green plastic bucket with a handle, the white plastic bin on a silver base', 'the window, the white plastic bin on a silver base'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0631_00/00480.jpg", "scene0631_00/00510.jpg", "scene0631_00/00540.jpg", "scene0631_00/00570.jpg" ], "turn_id": 5, "scan_id": "scene0631_00", "origin_question": "Which two objects are on the same side of you now? \"the mirror\", \"the window\", \"the white ceramic toilet with a round bowl\". ", "option": [ "the mirror, the window", "the mirror, the white ceramic toilet with a round bowl", "the window, the white ceramic toilet with a round bowl" ], "answer": "the window, the white ceramic toilet with a round bowl", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the mirror\", \"the window\", \"the white ceramic toilet with a round bowl\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the mirror, the window', 'the mirror, the white ceramic toilet with a round bowl', 'the window, the white ceramic toilet with a round bowl'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0631_00/00600.jpg", "scene0631_00/00630.jpg", "scene0631_00/00660.jpg", "scene0631_00/00690.jpg" ], "turn_id": 6, "scan_id": "scene0631_00", "origin_question": "In which turn were you closest to \"the round dial scale with a black plastic platform\"? ", "option": [], "answer": "4", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the round dial scale with a black plastic platform\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0631_00/00720.jpg", "scene0631_00/00750.jpg", "scene0631_00/00780.jpg", "scene0631_00/00810.jpg" ], "turn_id": 7, "scan_id": "scene0631_00", "origin_question": "In which turn were you closest to \"the white plastic bin on a silver base\"? ", "option": [], "answer": "7", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the white plastic bin on a silver base\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0134_02/00000.jpg", "scene0134_02/00030.jpg", "scene0134_02/00060.jpg", "scene0134_02/00090.jpg" ], "turn_id": 1, "scan_id": "scene0134_02", "origin_question": "Remember, how many chair(s) have you seen so far? ", "option": [], "answer": "4", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many chair(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0134_02/00120.jpg", "scene0134_02/00150.jpg", "scene0134_02/00180.jpg", "scene0134_02/00210.jpg" ], "turn_id": 2, "scan_id": "scene0134_02", "origin_question": "When did you first discover \"the rectangular wooden cabinet in rich brown color\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the rectangular wooden cabinet in rich brown color\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0134_02/00240.jpg", "scene0134_02/00270.jpg", "scene0134_02/00300.jpg", "scene0134_02/00330.jpg" ], "turn_id": 3, "scan_id": "scene0134_02", "origin_question": "Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "81.69", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0134_02/00360.jpg", "scene0134_02/00390.jpg", "scene0134_02/00420.jpg", "scene0134_02/00450.jpg" ], "turn_id": 4, "scan_id": "scene0134_02", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0134_02/00480.jpg", "scene0134_02/00510.jpg", "scene0134_02/00540.jpg", "scene0134_02/00570.jpg" ], "turn_id": 5, "scan_id": "scene0134_02", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the rectangular wooden cabinet in rich brown color\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the rectangular wooden cabinet in rich brown color\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0134_02/00600.jpg", "scene0134_02/00630.jpg", "scene0134_02/00660.jpg", "scene0134_02/00690.jpg" ], "turn_id": 6, "scan_id": "scene0134_02", "origin_question": "Using your orientation at the end of turn 4 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "61.82", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 4 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0040_region45/ce025bf7dc1048f890d587c1f9435ee4_i1_2.jpg", "1mp3d_0040_region45/ce025bf7dc1048f890d587c1f9435ee4_i2_2.jpg", "1mp3d_0040_region45/ce025bf7dc1048f890d587c1f9435ee4_i1_1.jpg", "1mp3d_0040_region45/ce025bf7dc1048f890d587c1f9435ee4_i2_1.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0040_region45", "origin_question": "Remember, have you seen any chandelier(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any chandelier(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "1mp3d_0040_region45/466ef42243d641888f8044a19d263a56_i1_5.jpg", "1mp3d_0040_region45/466ef42243d641888f8044a19d263a56_i2_5.jpg", "1mp3d_0040_region45/466ef42243d641888f8044a19d263a56_i1_0.jpg", "1mp3d_0040_region45/466ef42243d641888f8044a19d263a56_i2_0.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0040_region45", "origin_question": "In which turn did you see both \"the stainless steel modern refrigerator with two doors\" and \"the tall, slender dark bottle with a shiny surface\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the stainless steel modern refrigerator with two doors\" and \"the tall, slender dark bottle with a shiny surface\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "1mp3d_0040_region45/a2899329906b4c77a07ab17d76d501b5_i1_0.jpg", "1mp3d_0040_region45/a2899329906b4c77a07ab17d76d501b5_i2_0.jpg", "1mp3d_0040_region45/618c554cbed548a1a6ad278f3e50fbc7_i1_1.jpg", "1mp3d_0040_region45/618c554cbed548a1a6ad278f3e50fbc7_i2_1.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0040_region45", "origin_question": "Among these three objects(\"then elegant chandelier with five arms and beige lampshades\",\"the black rectangular gas stove with stainless steel accents\",\"the medium-sized brown wooden container with a round top\" ), which one is the closest to you?", "option": [ "then elegant chandelier with five arms and beige lampshades", "the black rectangular gas stove with stainless steel accents", "the medium-sized brown wooden container with a round top" ], "answer": "the black rectangular gas stove with stainless steel accents", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects(\"then elegant chandelier with five arms and beige lampshades\",\"the black rectangular gas stove with stainless steel accents\",\"the medium-sized brown wooden container with a round top\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['then elegant chandelier with five arms and beige lampshades', 'the black rectangular gas stove with stainless steel accents', 'the medium-sized brown wooden container with a round top'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0040_region45/618c554cbed548a1a6ad278f3e50fbc7_i1_0.jpg", "1mp3d_0040_region45/618c554cbed548a1a6ad278f3e50fbc7_i2_0.jpg", "1mp3d_0040_region45/618c554cbed548a1a6ad278f3e50fbc7_i1_5.jpg", "1mp3d_0040_region45/618c554cbed548a1a6ad278f3e50fbc7_i2_5.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0040_region45", "origin_question": "", "option": [], "answer": "", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "1mp3d_0040_region45/618c554cbed548a1a6ad278f3e50fbc7_i1_3.jpg", "1mp3d_0040_region45/618c554cbed548a1a6ad278f3e50fbc7_i2_3.jpg", "1mp3d_0040_region45/a2899329906b4c77a07ab17d76d501b5_i1_2.jpg", "1mp3d_0040_region45/a2899329906b4c77a07ab17d76d501b5_i2_2.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0040_region45", "origin_question": "What will be the first-time appearance order of the following categories: bottle, table, window?", "option": [ "table, bottle, window", "bottle, window, table", "table, window, bottle", "window, bottle, table" ], "answer": "table, window, bottle", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bottle, table, window? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['table, bottle, window', 'bottle, window, table', 'table, window, bottle', 'window, bottle, table'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0040_region45/a2899329906b4c77a07ab17d76d501b5_i1_4.jpg", "1mp3d_0040_region45/a2899329906b4c77a07ab17d76d501b5_i2_4.jpg", "1mp3d_0040_region45/a2899329906b4c77a07ab17d76d501b5_i1_5.jpg", "1mp3d_0040_region45/a2899329906b4c77a07ab17d76d501b5_i2_5.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0040_region45", "origin_question": "", "option": [], "answer": "", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0050_02/00000.jpg", "scene0050_02/00040.jpg", "scene0050_02/00080.jpg", "scene0050_02/00120.jpg", "scene0050_02/00160.jpg" ], "turn_id": 1, "scan_id": "scene0050_02", "origin_question": "Remember, have you seen any printer(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any printer(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0050_02/00200.jpg", "scene0050_02/00240.jpg", "scene0050_02/00280.jpg", "scene0050_02/00320.jpg", "scene0050_02/00360.jpg" ], "turn_id": 2, "scan_id": "scene0050_02", "origin_question": "Which two objects are on the same side of you now? \"the rectangular light-colored book with a glossy surface\", \"the rectangular wooden office desk in good condition\", \"the medium-sized beige printer with a metallic head\". ", "option": [ "the rectangular light-colored book with a glossy surface, the rectangular wooden office desk in good condition", "the rectangular light-colored book with a glossy surface, the medium-sized beige printer with a metallic head", "the rectangular wooden office desk in good condition, the medium-sized beige printer with a metallic head" ], "answer": "the rectangular wooden office desk in good condition, the medium-sized beige printer with a metallic head", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the rectangular light-colored book with a glossy surface\", \"the rectangular wooden office desk in good condition\", \"the medium-sized beige printer with a metallic head\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular light-colored book with a glossy surface, the rectangular wooden office desk in good condition', 'the rectangular light-colored book with a glossy surface, the medium-sized beige printer with a metallic head', 'the rectangular wooden office desk in good condition, the medium-sized beige printer with a metallic head'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0050_02/00400.jpg", "scene0050_02/00440.jpg", "scene0050_02/00480.jpg", "scene0050_02/00520.jpg", "scene0050_02/00560.jpg" ], "turn_id": 3, "scan_id": "scene0050_02", "origin_question": "Please recall \"the rectangular light-colored book with a glossy surface\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.73", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the rectangular light-colored book with a glossy surface\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0050_02/00600.jpg", "scene0050_02/00640.jpg", "scene0050_02/00680.jpg", "scene0050_02/00720.jpg", "scene0050_02/00760.jpg" ], "turn_id": 4, "scan_id": "scene0050_02", "origin_question": "Which direction is the medium-sized green cup with a handle in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the medium-sized green cup with a handle in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0050_02/00800.jpg", "scene0050_02/00840.jpg", "scene0050_02/00880.jpg", "scene0050_02/00920.jpg", "scene0050_02/00960.jpg" ], "turn_id": 5, "scan_id": "scene0050_02", "origin_question": "Please recall \"the round wall clock with a dark clock face\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.37", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the round wall clock with a dark clock face\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0050_02/01000.jpg", "scene0050_02/01040.jpg", "scene0050_02/01080.jpg", "scene0050_02/01120.jpg", "scene0050_02/01160.jpg" ], "turn_id": 6, "scan_id": "scene0050_02", "origin_question": "What will be the first-time appearance order of the following categories: shelf, table, tissue box?", "option": [ "shelf, tissue box, table", "shelf, table, tissue box", "tissue box, shelf, table", "table, tissue box, shelf" ], "answer": "shelf, table, tissue box", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: shelf, table, tissue box? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['shelf, tissue box, table', 'shelf, table, tissue box', 'tissue box, shelf, table', 'table, tissue box, shelf'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0050_02/01200.jpg", "scene0050_02/01240.jpg", "scene0050_02/01280.jpg", "scene0050_02/01320.jpg", "scene0050_02/01360.jpg" ], "turn_id": 7, "scan_id": "scene0050_02", "origin_question": "Compared to your position at the end of the 6 turn, are you now closer or farther away from \"the medium-sized beige printer with a metallic head\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 6 turn, are you now closer or farther away from \"the medium-sized beige printer with a metallic head\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0050_02/01400.jpg", "scene0050_02/01440.jpg", "scene0050_02/01480.jpg", "scene0050_02/01520.jpg", "scene0050_02/01560.jpg" ], "turn_id": 8, "scan_id": "scene0050_02", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the classic upright piano in deep brown wood\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the classic upright piano in deep brown wood\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0050_02/01600.jpg", "scene0050_02/01640.jpg", "scene0050_02/01680.jpg", "scene0050_02/01720.jpg", "scene0050_02/01760.jpg" ], "turn_id": 9, "scan_id": "scene0050_02", "origin_question": "Please recall \"the rectangular light-colored book with a glossy surface\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.91", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the rectangular light-colored book with a glossy surface\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0050_02/01800.jpg", "scene0050_02/01840.jpg", "scene0050_02/01880.jpg", "scene0050_02/01920.jpg", "scene0050_02/01960.jpg" ], "turn_id": 10, "scan_id": "scene0050_02", "origin_question": "Compared to your position at the end of the 7 turn, are you now closer or farther away from \"the medium-sized green cup with a handle\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 7 turn, are you now closer or farther away from \"the medium-sized green cup with a handle\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0541_02/00000.jpg", "scene0541_02/00040.jpg", "scene0541_02/00080.jpg", "scene0541_02/00120.jpg" ], "turn_id": 1, "scan_id": "scene0541_02", "origin_question": "Remember, have you seen any jacket(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any jacket(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0541_02/00160.jpg", "scene0541_02/00200.jpg", "scene0541_02/00240.jpg", "scene0541_02/00280.jpg" ], "turn_id": 2, "scan_id": "scene0541_02", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the rectangular wooden bed with a plaid blanket\"; \"the curtain\"; \"the telephone\" ", "option": [ "the rectangular wooden bed with a plaid blanket", "the curtain", "the telephone" ], "answer": "the telephone", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the rectangular wooden bed with a plaid blanket\"; \"the curtain\"; \"the telephone\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular wooden bed with a plaid blanket', 'the curtain', 'the telephone'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0541_02/00320.jpg", "scene0541_02/00360.jpg", "scene0541_02/00400.jpg", "scene0541_02/00440.jpg" ], "turn_id": 3, "scan_id": "scene0541_02", "origin_question": "Please recall \"the telephone\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.47", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the telephone\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0541_02/00480.jpg", "scene0541_02/00520.jpg", "scene0541_02/00560.jpg", "scene0541_02/00600.jpg" ], "turn_id": 4, "scan_id": "scene0541_02", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the telephone\" from your position?", "option": [], "answer": "86.06", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the telephone\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0541_02/00640.jpg", "scene0541_02/00680.jpg", "scene0541_02/00720.jpg", "scene0541_02/00760.jpg" ], "turn_id": 5, "scan_id": "scene0541_02", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the light brown wooden wardrobe with a simple design\" from your position?", "option": [], "answer": "119.1", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the light brown wooden wardrobe with a simple design\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0541_02/00800.jpg", "scene0541_02/00840.jpg", "scene0541_02/00880.jpg", "scene0541_02/00920.jpg" ], "turn_id": 6, "scan_id": "scene0541_02", "origin_question": "How far is your current position from where you were at the end of the turn 1? (in meters)", "option": [], "answer": "2.04", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 1? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0541_02/00960.jpg", "scene0541_02/01000.jpg", "scene0541_02/01040.jpg", "scene0541_02/01080.jpg" ], "turn_id": 7, "scan_id": "scene0541_02", "origin_question": "Compared to your position at the end of the 5 turn, are you now closer or farther away from \"the light brown wooden wardrobe with a simple design\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 5 turn, are you now closer or farther away from \"the light brown wooden wardrobe with a simple design\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0541_02/01120.jpg", "scene0541_02/01160.jpg", "scene0541_02/01200.jpg", "scene0541_02/01240.jpg" ], "turn_id": 8, "scan_id": "scene0541_02", "origin_question": "", "option": [], "answer": "", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0541_02/01280.jpg", "scene0541_02/01320.jpg", "scene0541_02/01360.jpg", "scene0541_02/01400.jpg" ], "turn_id": 9, "scan_id": "scene0541_02", "origin_question": "", "option": [], "answer": "", "user_message": "For the 9 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0486_00/00000.jpg", "scene0486_00/00040.jpg", "scene0486_00/00080.jpg", "scene0486_00/00120.jpg", "scene0486_00/00160.jpg" ], "turn_id": 1, "scan_id": "scene0486_00", "origin_question": "Remember, have you seen any bowl(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any bowl(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0486_00/00200.jpg", "scene0486_00/00240.jpg", "scene0486_00/00280.jpg", "scene0486_00/00320.jpg", "scene0486_00/00360.jpg" ], "turn_id": 2, "scan_id": "scene0486_00", "origin_question": "When did you first discover \"the sturdy, rectangular matte black counter\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the sturdy, rectangular matte black counter\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0486_00/00400.jpg", "scene0486_00/00440.jpg", "scene0486_00/00480.jpg", "scene0486_00/00520.jpg", "scene0486_00/00560.jpg" ], "turn_id": 3, "scan_id": "scene0486_00", "origin_question": "Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "34.73", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0486_00/00600.jpg", "scene0486_00/00640.jpg", "scene0486_00/00680.jpg", "scene0486_00/00720.jpg", "scene0486_00/00760.jpg" ], "turn_id": 4, "scan_id": "scene0486_00", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the modern rectangular black stainless steel dishwasher\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the modern rectangular black stainless steel dishwasher\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0486_00/00800.jpg", "scene0486_00/00840.jpg", "scene0486_00/00880.jpg", "scene0486_00/00920.jpg", "scene0486_00/00960.jpg" ], "turn_id": 5, "scan_id": "scene0486_00", "origin_question": "In which turn did you see both \"the sleek black stainless steel refrigerator with two doors\" and \"the sturdy, rectangular matte black counter\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the sleek black stainless steel refrigerator with two doors\" and \"the sturdy, rectangular matte black counter\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0486_00/01000.jpg", "scene0486_00/01040.jpg", "scene0486_00/01080.jpg", "scene0486_00/01120.jpg", "scene0486_00/01160.jpg" ], "turn_id": 6, "scan_id": "scene0486_00", "origin_question": "Please recall \"the stylish stainless steel double basin sink\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.44", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the stylish stainless steel double basin sink\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0486_00/01200.jpg", "scene0486_00/01240.jpg", "scene0486_00/01280.jpg", "scene0486_00/01320.jpg", "scene0486_00/01360.jpg" ], "turn_id": 7, "scan_id": "scene0486_00", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the sturdy, rectangular matte black counter\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the sturdy, rectangular matte black counter\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0486_00/01400.jpg", "scene0486_00/01440.jpg", "scene0486_00/01480.jpg", "scene0486_00/01520.jpg", "scene0486_00/01560.jpg" ], "turn_id": 8, "scan_id": "scene0486_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0486_00/01600.jpg", "scene0486_00/01640.jpg", "scene0486_00/01680.jpg", "scene0486_00/01720.jpg", "scene0486_00/01760.jpg" ], "turn_id": 9, "scan_id": "scene0486_00", "origin_question": "What will be the first-time appearance order of the following categories: bin, cabinet, cup?", "option": [ "cabinet, bin, cup", "cup, cabinet, bin", "bin, cabinet, cup", "cabinet, cup, bin" ], "answer": "cabinet, bin, cup", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bin, cabinet, cup? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['cabinet, bin, cup', 'cup, cabinet, bin', 'bin, cabinet, cup', 'cabinet, cup, bin'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0486_00/01800.jpg", "scene0486_00/01840.jpg", "scene0486_00/01880.jpg", "scene0486_00/01920.jpg", "scene0486_00/01960.jpg" ], "turn_id": 10, "scan_id": "scene0486_00", "origin_question": "Which two objects are on the same side of you now? \"the modern rectangular black stainless steel dishwasher\", \"the ceramic cylindrical cup with red and white colors\", \"the cylindrical, white tissue roll made of soft paper\". ", "option": [ "the modern rectangular black stainless steel dishwasher, the ceramic cylindrical cup with red and white colors", "the modern rectangular black stainless steel dishwasher, the cylindrical, white tissue roll made of soft paper", "the ceramic cylindrical cup with red and white colors, the cylindrical, white tissue roll made of soft paper" ], "answer": "the ceramic cylindrical cup with red and white colors, the cylindrical, white tissue roll made of soft paper", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the modern rectangular black stainless steel dishwasher\", \"the ceramic cylindrical cup with red and white colors\", \"the cylindrical, white tissue roll made of soft paper\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the modern rectangular black stainless steel dishwasher, the ceramic cylindrical cup with red and white colors', 'the modern rectangular black stainless steel dishwasher, the cylindrical, white tissue roll made of soft paper', 'the ceramic cylindrical cup with red and white colors, the cylindrical, white tissue roll made of soft paper'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0057_region13/4955496ce75d4347bab8ed9dfa24780b_i1_3.jpg", "1mp3d_0057_region13/4955496ce75d4347bab8ed9dfa24780b_i2_3.jpg", "1mp3d_0057_region13/b3af4a4e99914332ac39c7a0935246ea_i2_1.jpg", "1mp3d_0057_region13/b3af4a4e99914332ac39c7a0935246ea_i2_2.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0057_region13", "origin_question": "Remember, have you seen any kitchenware(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any kitchenware(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "1mp3d_0057_region13/1012e5dfb0c9403da80b9323f3aea94f_i1_5.jpg", "1mp3d_0057_region13/1012e5dfb0c9403da80b9323f3aea94f_i2_5.jpg", "1mp3d_0057_region13/b3af4a4e99914332ac39c7a0935246ea_i1_0.jpg", "1mp3d_0057_region13/b3af4a4e99914332ac39c7a0935246ea_i2_0.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0057_region13", "origin_question": "At the end of which turn were both of \"the large rectangular mirror with a brown frame\" and \"the medium gray circular flowerpot with a smooth surface\" on your left side?", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the large rectangular mirror with a brown frame\" and \"the medium gray circular flowerpot with a smooth surface\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "1mp3d_0057_region13/1012e5dfb0c9403da80b9323f3aea94f_i1_4.jpg", "1mp3d_0057_region13/1012e5dfb0c9403da80b9323f3aea94f_i2_4.jpg", "1mp3d_0057_region13/88f88c66ebd34a318219f68c92407cc0_i1_4.jpg", "1mp3d_0057_region13/88f88c66ebd34a318219f68c92407cc0_i2_4.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0057_region13", "origin_question": "Among these three objects(\"the standard white toilet paper roll\",\"the medium gray circular flowerpot with a smooth surface\",\"the towel\" ), which one is the closest to you?", "option": [ "the standard white toilet paper roll", "the medium gray circular flowerpot with a smooth surface", "the towel" ], "answer": "the towel", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects(\"the standard white toilet paper roll\",\"the medium gray circular flowerpot with a smooth surface\",\"the towel\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the standard white toilet paper roll', 'the medium gray circular flowerpot with a smooth surface', 'the towel'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0057_region13/1012e5dfb0c9403da80b9323f3aea94f_i2_3.jpg", "1mp3d_0057_region13/1012e5dfb0c9403da80b9323f3aea94f_i1_1.jpg", "1mp3d_0057_region13/1012e5dfb0c9403da80b9323f3aea94f_i2_1.jpg", "1mp3d_0057_region13/88f88c66ebd34a318219f68c92407cc0_i1_1.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0057_region13", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the towel\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the towel\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "1mp3d_0057_region13/88f88c66ebd34a318219f68c92407cc0_i2_1.jpg", "1mp3d_0057_region13/88f88c66ebd34a318219f68c92407cc0_i1_0.jpg", "1mp3d_0057_region13/88f88c66ebd34a318219f68c92407cc0_i2_0.jpg", "1mp3d_0057_region13/88f88c66ebd34a318219f68c92407cc0_i1_5.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0057_region13", "origin_question": "At the end of which turn were both of \"the standard white toilet paper roll\" and \"the large rectangular mirror with a brown frame\" on your left side?", "option": [], "answer": "4", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the standard white toilet paper roll\" and \"the large rectangular mirror with a brown frame\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0057_region9/500527577f0b4e0db97d0a3e4de565dd_i2_5.jpg", "1mp3d_0057_region9/54f491ad92e347c892e35a0d48d497d2_i1_0.jpg", "1mp3d_0057_region9/54f491ad92e347c892e35a0d48d497d2_i2_0.jpg", "1mp3d_0057_region9/500527577f0b4e0db97d0a3e4de565dd_i1_1.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0057_region9", "origin_question": "Remember, have you seen any faucet(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any faucet(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 2 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "1mp3d_0057_region9/500527577f0b4e0db97d0a3e4de565dd_i2_1.jpg", "1mp3d_0057_region9/500527577f0b4e0db97d0a3e4de565dd_i1_2.jpg", "1mp3d_0057_region9/500527577f0b4e0db97d0a3e4de565dd_i2_2.jpg", "1mp3d_0057_region9/54f491ad92e347c892e35a0d48d497d2_i1_1.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0057_region9", "origin_question": "When did you first discover \"the faucet\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the faucet\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 2 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0229_02/00000.jpg", "scene0229_02/00030.jpg", "scene0229_02/00060.jpg", "scene0229_02/00090.jpg" ], "turn_id": 1, "scan_id": "scene0229_02", "origin_question": "Remember, have you seen any cap(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any cap(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0229_02/00120.jpg", "scene0229_02/00150.jpg", "scene0229_02/00180.jpg", "scene0229_02/00210.jpg" ], "turn_id": 2, "scan_id": "scene0229_02", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the small rectangular black metal shelf with compartments\"; \"the medium-sized rectangular brown wooden table\"; \"the white and brown microwave oven in good condition\" ", "option": [ "the small rectangular black metal shelf with compartments", "the medium-sized rectangular brown wooden table", "the white and brown microwave oven in good condition" ], "answer": "the small rectangular black metal shelf with compartments", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the small rectangular black metal shelf with compartments\"; \"the medium-sized rectangular brown wooden table\"; \"the white and brown microwave oven in good condition\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the small rectangular black metal shelf with compartments', 'the medium-sized rectangular brown wooden table', 'the white and brown microwave oven in good condition'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0229_02/00240.jpg", "scene0229_02/00270.jpg", "scene0229_02/00300.jpg", "scene0229_02/00330.jpg" ], "turn_id": 3, "scan_id": "scene0229_02", "origin_question": "By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the white and brown microwave oven in good condition\", Y: \"the medium-sized rectangular brown wooden table\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the white and brown microwave oven in good condition\", Y: \"the medium-sized rectangular brown wooden table\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0229_02/00360.jpg", "scene0229_02/00390.jpg", "scene0229_02/00420.jpg", "scene0229_02/00450.jpg" ], "turn_id": 4, "scan_id": "scene0229_02", "origin_question": "When did you first discover \"the small rectangular black metal shelf with compartments\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the small rectangular black metal shelf with compartments\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0229_02/00480.jpg", "scene0229_02/00510.jpg", "scene0229_02/00540.jpg", "scene0229_02/00580.jpg" ], "turn_id": 5, "scan_id": "scene0229_02", "origin_question": "Which direction is the black and white sports shoe with a white sole in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the black and white sports shoe with a white sole in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0229_02/00610.jpg", "scene0229_02/00640.jpg", "scene0229_02/00670.jpg", "scene0229_02/00700.jpg" ], "turn_id": 6, "scan_id": "scene0229_02", "origin_question": "Please recall \"the gray rectangular folding board with a black handle\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.05", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the gray rectangular folding board with a black handle\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0421_00/00000.jpg", "scene0421_00/00040.jpg", "scene0421_00/00080.jpg", "scene0421_00/00120.jpg", "scene0421_00/00160.jpg" ], "turn_id": 1, "scan_id": "scene0421_00", "origin_question": "Remember, have you seen any broom(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any broom(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0421_00/00200.jpg", "scene0421_00/00240.jpg", "scene0421_00/00280.jpg", "scene0421_00/00320.jpg", "scene0421_00/00360.jpg" ], "turn_id": 2, "scan_id": "scene0421_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0421_00/00400.jpg", "scene0421_00/00440.jpg", "scene0421_00/00480.jpg", "scene0421_00/00520.jpg", "scene0421_00/00560.jpg" ], "turn_id": 3, "scan_id": "scene0421_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0421_00/00600.jpg", "scene0421_00/00640.jpg", "scene0421_00/00680.jpg", "scene0421_00/00720.jpg", "scene0421_00/00760.jpg" ], "turn_id": 4, "scan_id": "scene0421_00", "origin_question": "What will be the first-time appearance order of the following categories: basket, clothes dryer, door?", "option": [ "door, basket, clothes dryer", "door, clothes dryer, basket", "basket, door, clothes dryer", "clothes dryer, door, basket" ], "answer": "door, basket, clothes dryer", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: basket, clothes dryer, door? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['door, basket, clothes dryer', 'door, clothes dryer, basket', 'basket, door, clothes dryer', 'clothes dryer, door, basket'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0421_00/00800.jpg", "scene0421_00/00840.jpg", "scene0421_00/00880.jpg", "scene0421_00/00920.jpg", "scene0421_00/00960.jpg" ], "turn_id": 5, "scan_id": "scene0421_00", "origin_question": "At the end of which turn were both of \"the medium-sized rectangular beige cardboard box\" and \"the stylish white stainless steel clothes dryer with buttons\" on your left side?", "option": [], "answer": "1", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the medium-sized rectangular beige cardboard box\" and \"the stylish white stainless steel clothes dryer with buttons\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0421_00/01000.jpg", "scene0421_00/01040.jpg", "scene0421_00/01080.jpg", "scene0421_00/01120.jpg", "scene0421_00/01160.jpg" ], "turn_id": 6, "scan_id": "scene0421_00", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the yellow-handled broom with a black head\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the yellow-handled broom with a black head\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0421_00/01200.jpg", "scene0421_00/01240.jpg", "scene0421_00/01280.jpg", "scene0421_00/01320.jpg", "scene0421_00/01360.jpg" ], "turn_id": 7, "scan_id": "scene0421_00", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the rectangular blue and white cotton blanket\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the rectangular blue and white cotton blanket\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0421_00/01400.jpg", "scene0421_00/01440.jpg", "scene0421_00/01480.jpg", "scene0421_00/01520.jpg", "scene0421_00/01560.jpg" ], "turn_id": 8, "scan_id": "scene0421_00", "origin_question": "What will be the first-time appearance order of the following categories: door, shelf, washing machine?", "option": [ "door, washing machine, shelf", "shelf, washing machine, door", "shelf, door, washing machine", "door, shelf, washing machine" ], "answer": "door, shelf, washing machine", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: door, shelf, washing machine? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['door, washing machine, shelf', 'shelf, washing machine, door', 'shelf, door, washing machine', 'door, shelf, washing machine'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0421_00/01600.jpg", "scene0421_00/01640.jpg", "scene0421_00/01680.jpg", "scene0421_00/01720.jpg", "scene0421_00/01760.jpg" ], "turn_id": 9, "scan_id": "scene0421_00", "origin_question": "What will be the first-time appearance order of the following categories: basket, broom, washing machine?", "option": [ "broom, basket, washing machine", "broom, washing machine, basket", "basket, washing machine, broom", "washing machine, broom, basket" ], "answer": "broom, basket, washing machine", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: basket, broom, washing machine? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['broom, basket, washing machine', 'broom, washing machine, basket', 'basket, washing machine, broom', 'washing machine, broom, basket'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0421_00/01800.jpg", "scene0421_00/01840.jpg", "scene0421_00/01880.jpg", "scene0421_00/01920.jpg", "scene0421_00/01960.jpg" ], "turn_id": 10, "scan_id": "scene0421_00", "origin_question": "What will be the first-time appearance order of the following categories: basket, blanket, broom?", "option": [ "basket, broom, blanket", "blanket, broom, basket", "broom, blanket, basket", "broom, basket, blanket" ], "answer": "broom, basket, blanket", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: basket, blanket, broom? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['basket, broom, blanket', 'blanket, broom, basket', 'broom, blanket, basket', 'broom, basket, blanket'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0599_02/00000.jpg", "scene0599_02/00040.jpg", "scene0599_02/00080.jpg", "scene0599_02/00120.jpg", "scene0599_02/00160.jpg" ], "turn_id": 1, "scan_id": "scene0599_02", "origin_question": "Remember, have you seen any window(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any window(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0599_02/00200.jpg", "scene0599_02/00240.jpg", "scene0599_02/00280.jpg", "scene0599_02/00320.jpg", "scene0599_02/00360.jpg" ], "turn_id": 2, "scan_id": "scene0599_02", "origin_question": "At the end of which turn were both of \"the standard-sized rectangular gray plastic bin\" and \"the vintage white telephone with a rectangular shape\" on your left side?", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the standard-sized rectangular gray plastic bin\" and \"the vintage white telephone with a rectangular shape\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0599_02/00400.jpg", "scene0599_02/00440.jpg", "scene0599_02/00480.jpg", "scene0599_02/00520.jpg", "scene0599_02/00560.jpg" ], "turn_id": 3, "scan_id": "scene0599_02", "origin_question": "When did you first discover \"the light brown rectangular wooden table\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the light brown rectangular wooden table\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "scene0599_02/00600.jpg", "scene0599_02/00640.jpg", "scene0599_02/00680.jpg", "scene0599_02/00720.jpg", "scene0599_02/00760.jpg" ], "turn_id": 4, "scan_id": "scene0599_02", "origin_question": "Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Right", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0599_02/00800.jpg", "scene0599_02/00840.jpg", "scene0599_02/00880.jpg", "scene0599_02/00920.jpg", "scene0599_02/00960.jpg" ], "turn_id": 5, "scan_id": "scene0599_02", "origin_question": "At the end of which turn were both of \"the light gray rectangular plastic dispenser\" and \"the deep purple and black backpack with a white stripe\" on your right side?", "option": [], "answer": "2", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the light gray rectangular plastic dispenser\" and \"the deep purple and black backpack with a white stripe\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0599_02/01000.jpg", "scene0599_02/01040.jpg", "scene0599_02/01080.jpg", "scene0599_02/01120.jpg", "scene0599_02/01160.jpg" ], "turn_id": 6, "scan_id": "scene0599_02", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the vintage white telephone with a rectangular shape\" from your position?", "option": [], "answer": "177.18", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the vintage white telephone with a rectangular shape\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0599_02/01200.jpg", "scene0599_02/01240.jpg", "scene0599_02/01280.jpg", "scene0599_02/01320.jpg", "scene0599_02/01360.jpg" ], "turn_id": 7, "scan_id": "scene0599_02", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the deep purple and black backpack with a white stripe\" from your position?", "option": [], "answer": "177.83", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the deep purple and black backpack with a white stripe\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0599_02/01400.jpg", "scene0599_02/01440.jpg", "scene0599_02/01480.jpg", "scene0599_02/01520.jpg", "scene0599_02/01560.jpg" ], "turn_id": 8, "scan_id": "scene0599_02", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the standard-sized rectangular gray plastic bin\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the standard-sized rectangular gray plastic bin\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0599_02/01600.jpg", "scene0599_02/01640.jpg", "scene0599_02/01680.jpg", "scene0599_02/01720.jpg", "scene0599_02/01760.jpg" ], "turn_id": 9, "scan_id": "scene0599_02", "origin_question": "By the end of the 5 turn, were you getting closer to or farther away from these two objects: X: \"the light gray rectangular plastic dispenser\", Y: \"the deep purple and black backpack with a white stripe\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to X and farther away from Y", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 5 turn, were you getting closer to or farther away from these two objects: X: \"the light gray rectangular plastic dispenser\", Y: \"the deep purple and black backpack with a white stripe\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0599_02/01800.jpg", "scene0599_02/01840.jpg", "scene0599_02/01880.jpg", "scene0599_02/01920.jpg", "scene0599_02/01960.jpg" ], "turn_id": 10, "scan_id": "scene0599_02", "origin_question": "Please recall \"the vintage white telephone with a rectangular shape\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.91", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the vintage white telephone with a rectangular shape\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42897599/Validation_42897599_42897599_476302.229.png", "Validation_42897599/Validation_42897599_42897599_476304.211.png", "Validation_42897599/Validation_42897599_42897599_476306.210.png", "Validation_42897599/Validation_42897599_42897599_476308.209.png", "Validation_42897599/Validation_42897599_42897599_476310.209.png" ], "turn_id": 1, "scan_id": "Validation_42897599", "origin_question": "Remember, have you seen any plant(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any plant(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_42897599/Validation_42897599_42897599_476312.208.png", "Validation_42897599/Validation_42897599_42897599_476314.207.png", "Validation_42897599/Validation_42897599_42897599_476316.206.png", "Validation_42897599/Validation_42897599_42897599_476318.205.png", "Validation_42897599/Validation_42897599_42897599_476320.204.png" ], "turn_id": 2, "scan_id": "Validation_42897599", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the medium-brown wooden door with glass panels\", Y: \"the glossy rectangular table with the clear surface\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the medium-brown wooden door with glass panels\", Y: \"the glossy rectangular table with the clear surface\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "Validation_42897599/Validation_42897599_42897599_476322.204.png", "Validation_42897599/Validation_42897599_42897599_476324.203.png", "Validation_42897599/Validation_42897599_42897599_476326.202.png", "Validation_42897599/Validation_42897599_42897599_476328.201.png", "Validation_42897599/Validation_42897599_42897599_476330.200.png" ], "turn_id": 3, "scan_id": "Validation_42897599", "origin_question": "Among these three objects(\"the medium-brown wooden door with glass panels\",\"the glossy rectangular table with the clear surface\",\"the small rectangular window with the white frame\" ), which one is the closest to you?", "option": [ "the medium-brown wooden door with glass panels", "the glossy rectangular table with the clear surface", "the small rectangular window with the white frame" ], "answer": "the glossy rectangular table with the clear surface", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the medium-brown wooden door with glass panels\",\"the glossy rectangular table with the clear surface\",\"the small rectangular window with the white frame\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the medium-brown wooden door with glass panels', 'the glossy rectangular table with the clear surface', 'the small rectangular window with the white frame'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_42897599/Validation_42897599_42897599_476332.200.png", "Validation_42897599/Validation_42897599_42897599_476334.199.png", "Validation_42897599/Validation_42897599_42897599_476336.198.png", "Validation_42897599/Validation_42897599_42897599_476338.214.png", "Validation_42897599/Validation_42897599_42897599_476340.213.png" ], "turn_id": 4, "scan_id": "Validation_42897599", "origin_question": "Which two objects are on the same side of you now? \"the medium-brown wooden door with glass panels\", \"the glossy rectangular table with the clear surface\", \"the sleek modern stainless steel and black glass oven\". ", "option": [ "the medium-brown wooden door with glass panels, the glossy rectangular table with the clear surface", "the medium-brown wooden door with glass panels, the sleek modern stainless steel and black glass oven", "the glossy rectangular table with the clear surface, the sleek modern stainless steel and black glass oven" ], "answer": "the glossy rectangular table with the clear surface, the sleek modern stainless steel and black glass oven", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the medium-brown wooden door with glass panels\", \"the glossy rectangular table with the clear surface\", \"the sleek modern stainless steel and black glass oven\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the medium-brown wooden door with glass panels, the glossy rectangular table with the clear surface', 'the medium-brown wooden door with glass panels, the sleek modern stainless steel and black glass oven', 'the glossy rectangular table with the clear surface, the sleek modern stainless steel and black glass oven'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_42897599/Validation_42897599_42897599_476342.212.png", "Validation_42897599/Validation_42897599_42897599_476344.211.png", "Validation_42897599/Validation_42897599_42897599_476346.211.png", "Validation_42897599/Validation_42897599_42897599_476348.210.png", "Validation_42897599/Validation_42897599_42897599_476350.209.png" ], "turn_id": 5, "scan_id": "Validation_42897599", "origin_question": "Which two objects are on the same side of you now? \"the deep red horizontal curtain made of smooth material\", \"the sleek stainless steel refrigerator\", \"the silver stainless steel bin with the black lid\". ", "option": [ "the deep red horizontal curtain made of smooth material, the sleek stainless steel refrigerator", "the deep red horizontal curtain made of smooth material, the silver stainless steel bin with the black lid", "the sleek stainless steel refrigerator, the silver stainless steel bin with the black lid" ], "answer": "the sleek stainless steel refrigerator, the silver stainless steel bin with the black lid", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the deep red horizontal curtain made of smooth material\", \"the sleek stainless steel refrigerator\", \"the silver stainless steel bin with the black lid\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the deep red horizontal curtain made of smooth material, the sleek stainless steel refrigerator', 'the deep red horizontal curtain made of smooth material, the silver stainless steel bin with the black lid', 'the sleek stainless steel refrigerator, the silver stainless steel bin with the black lid'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_42897599/Validation_42897599_42897599_476352.208.png", "Validation_42897599/Validation_42897599_42897599_476354.207.png", "Validation_42897599/Validation_42897599_42897599_476356.206.png", "Validation_42897599/Validation_42897599_42897599_476358.206.png", "Validation_42897599/Validation_42897599_42897599_476360.205.png" ], "turn_id": 6, "scan_id": "Validation_42897599", "origin_question": "When did you first discover \"the silver stainless steel bin with the black lid\"? (index of the turn)", "option": [], "answer": "3", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the silver stainless steel bin with the black lid\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_42897599/Validation_42897599_42897599_476362.204.png", "Validation_42897599/Validation_42897599_42897599_476364.203.png", "Validation_42897599/Validation_42897599_42897599_476366.202.png", "Validation_42897599/Validation_42897599_42897599_476368.202.png", "Validation_42897599/Validation_42897599_42897599_476370.201.png" ], "turn_id": 7, "scan_id": "Validation_42897599", "origin_question": "What will be the first-time appearance order of the following categories: picture, plant, refrigerator?", "option": [ "picture, refrigerator, plant", "picture, plant, refrigerator", "plant, refrigerator, picture", "plant, picture, refrigerator" ], "answer": "picture, plant, refrigerator", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: picture, plant, refrigerator? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['picture, refrigerator, plant', 'picture, plant, refrigerator', 'plant, refrigerator, picture', 'plant, picture, refrigerator'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_42897599/Validation_42897599_42897599_476372.200.png", "Validation_42897599/Validation_42897599_42897599_476374.199.png", "Validation_42897599/Validation_42897599_42897599_476376.198.png", "Validation_42897599/Validation_42897599_42897599_476378.198.png", "Validation_42897599/Validation_42897599_42897599_476380.197.png" ], "turn_id": 8, "scan_id": "Validation_42897599", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the deep red horizontal curtain made of smooth material\", Y: \"the glossy rectangular table with the clear surface\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the deep red horizontal curtain made of smooth material\", Y: \"the glossy rectangular table with the clear surface\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_42897599/Validation_42897599_42897599_476382.213.png", "Validation_42897599/Validation_42897599_42897599_476384.212.png", "Validation_42897599/Validation_42897599_42897599_476386.211.png", "Validation_42897599/Validation_42897599_42897599_476388.210.png", "Validation_42897599/Validation_42897599_42897599_476390.209.png" ], "turn_id": 9, "scan_id": "Validation_42897599", "origin_question": "Which direction is the small rectangular window with the white frame in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the small rectangular window with the white frame in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_42897599/Validation_42897599_42897599_476392.209.png", "Validation_42897599/Validation_42897599_42897599_476394.208.png", "Validation_42897599/Validation_42897599_42897599_476396.207.png", "Validation_42897599/Validation_42897599_42897599_476398.206.png", "Validation_42897599/Validation_42897599_42897599_476400.205.png" ], "turn_id": 10, "scan_id": "Validation_42897599", "origin_question": "Is the sleek stainless steel refrigerator to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the sleek stainless steel refrigerator to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47429971/Validation_47429971_47429971_16279.772.png", "Validation_47429971/Validation_47429971_47429971_16281.771.png", "Validation_47429971/Validation_47429971_47429971_16283.771.png", "Validation_47429971/Validation_47429971_47429971_16285.770.png", "Validation_47429971/Validation_47429971_47429971_16287.769.png" ], "turn_id": 1, "scan_id": "Validation_47429971", "origin_question": "Remember, have you seen any candle(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any candle(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47429971/Validation_47429971_47429971_16289.768.png", "Validation_47429971/Validation_47429971_47429971_16291.767.png", "Validation_47429971/Validation_47429971_47429971_16293.766.png", "Validation_47429971/Validation_47429971_47429971_16295.765.png", "Validation_47429971/Validation_47429971_47429971_16297.765.png" ], "turn_id": 2, "scan_id": "Validation_47429971", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47429971/Validation_47429971_47429971_16299.764.png", "Validation_47429971/Validation_47429971_47429971_16301.763.png", "Validation_47429971/Validation_47429971_47429971_16303.762.png", "Validation_47429971/Validation_47429971_47429971_16305.761.png", "Validation_47429971/Validation_47429971_47429971_16307.761.png" ], "turn_id": 3, "scan_id": "Validation_47429971", "origin_question": "In which turn were you farthest from \"the rectangular stainless steel sink\"? ", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the rectangular stainless steel sink\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_47429971/Validation_47429971_47429971_16309.760.png", "Validation_47429971/Validation_47429971_47429971_16311.759.png", "Validation_47429971/Validation_47429971_47429971_16313.758.png", "Validation_47429971/Validation_47429971_47429971_16315.774.png", "Validation_47429971/Validation_47429971_47429971_16317.773.png" ], "turn_id": 4, "scan_id": "Validation_47429971", "origin_question": "Please recall \"the rectangular stainless steel sink\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.36", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the rectangular stainless steel sink\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_47429971/Validation_47429971_47429971_16319.772.png", "Validation_47429971/Validation_47429971_47429971_16321.771.png", "Validation_47429971/Validation_47429971_47429971_16323.771.png", "Validation_47429971/Validation_47429971_47429971_16325.770.png", "Validation_47429971/Validation_47429971_47429971_16327.769.png" ], "turn_id": 5, "scan_id": "Validation_47429971", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the rectangular stainless steel sink\" from your position?", "option": [], "answer": "115.13", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the rectangular stainless steel sink\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_47429971/Validation_47429971_47429971_16329.768.png", "Validation_47429971/Validation_47429971_47429971_16331.767.png", "Validation_47429971/Validation_47429971_47429971_16333.766.png", "Validation_47429971/Validation_47429971_47429971_16335.766.png", "Validation_47429971/Validation_47429971_47429971_16337.765.png" ], "turn_id": 6, "scan_id": "Validation_47429971", "origin_question": "Is the rectangular stainless steel sink to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the rectangular stainless steel sink to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_47429971/Validation_47429971_47429971_16339.764.png", "Validation_47429971/Validation_47429971_47429971_16341.763.png", "Validation_47429971/Validation_47429971_47429971_16343.762.png", "Validation_47429971/Validation_47429971_47429971_16345.762.png", "Validation_47429971/Validation_47429971_47429971_16347.761.png" ], "turn_id": 7, "scan_id": "Validation_47429971", "origin_question": "In which turn did you see both \"the rectangular stainless steel sink\" and \"the refrigerator\" simultaneously? ", "option": [], "answer": "3", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the rectangular stainless steel sink\" and \"the refrigerator\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_47429971/Validation_47429971_47429971_16349.760.png", "Validation_47429971/Validation_47429971_47429971_16351.759.png", "Validation_47429971/Validation_47429971_47429971_16353.758.png", "Validation_47429971/Validation_47429971_47429971_16355.757.png", "Validation_47429971/Validation_47429971_47429971_16357.773.png" ], "turn_id": 8, "scan_id": "Validation_47429971", "origin_question": "Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "77.13", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47332895/Validation_47332895_47332895_9046.813.png", "Validation_47332895/Validation_47332895_47332895_9048.795.png", "Validation_47332895/Validation_47332895_47332895_9050.794.png", "Validation_47332895/Validation_47332895_47332895_9052.794.png" ], "turn_id": 1, "scan_id": "Validation_47332895", "origin_question": "Remember, have you seen any window(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any window(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47332895/Validation_47332895_47332895_9054.793.png", "Validation_47332895/Validation_47332895_47332895_9056.792.png", "Validation_47332895/Validation_47332895_47332895_9058.791.png", "Validation_47332895/Validation_47332895_47332895_9060.790.png" ], "turn_id": 2, "scan_id": "Validation_47332895", "origin_question": "In which turn were you closest to \"the bold lime green modern bed with the headboard\"? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the bold lime green modern bed with the headboard\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47332895/Validation_47332895_47332895_9062.790.png", "Validation_47332895/Validation_47332895_47332895_9064.789.png", "Validation_47332895/Validation_47332895_47332895_9066.788.png", "Validation_47332895/Validation_47332895_47332895_9068.787.png" ], "turn_id": 3, "scan_id": "Validation_47332895", "origin_question": "In which turn were you closest to \"the compact cube-shaped brown wooden stool\"? ", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the compact cube-shaped brown wooden stool\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47332895/Validation_47332895_47332895_9070.786.png", "Validation_47332895/Validation_47332895_47332895_9072.785.png", "Validation_47332895/Validation_47332895_47332895_9074.785.png", "Validation_47332895/Validation_47332895_47332895_9076.784.png" ], "turn_id": 4, "scan_id": "Validation_47332895", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the sleek rectangular white steel radiator\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the sleek rectangular white steel radiator\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47332895/Validation_47332895_47332895_9078.800.png", "Validation_47332895/Validation_47332895_47332895_9080.799.png", "Validation_47332895/Validation_47332895_47332895_9082.798.png", "Validation_47332895/Validation_47332895_47332895_9084.797.png" ], "turn_id": 5, "scan_id": "Validation_47332895", "origin_question": "Which direction is the compact cube-shaped brown wooden stool in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the compact cube-shaped brown wooden stool in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_47332895/Validation_47332895_47332895_9086.797.png", "Validation_47332895/Validation_47332895_47332895_9088.796.png", "Validation_47332895/Validation_47332895_47332895_9090.795.png", "Validation_47332895/Validation_47332895_47332895_9092.794.png" ], "turn_id": 6, "scan_id": "Validation_47332895", "origin_question": "Which two objects are on the same side of you now? \"the bold lime green modern bed with the headboard\", \"the modern two-pane sliding window with the white frame\", \"the compact cube-shaped brown wooden stool\". ", "option": [ "the bold lime green modern bed with the headboard, the modern two-pane sliding window with the white frame", "the bold lime green modern bed with the headboard, the compact cube-shaped brown wooden stool", "the modern two-pane sliding window with the white frame, the compact cube-shaped brown wooden stool" ], "answer": "the modern two-pane sliding window with the white frame, the compact cube-shaped brown wooden stool", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the bold lime green modern bed with the headboard\", \"the modern two-pane sliding window with the white frame\", \"the compact cube-shaped brown wooden stool\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the bold lime green modern bed with the headboard, the modern two-pane sliding window with the white frame', 'the bold lime green modern bed with the headboard, the compact cube-shaped brown wooden stool', 'the modern two-pane sliding window with the white frame, the compact cube-shaped brown wooden stool'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_47332895/Validation_47332895_47332895_9094.793.png", "Validation_47332895/Validation_47332895_47332895_9096.792.png", "Validation_47332895/Validation_47332895_47332895_9098.792.png", "Validation_47332895/Validation_47332895_47332895_9100.791.png" ], "turn_id": 7, "scan_id": "Validation_47332895", "origin_question": "How far is your current position from where you were at the end of the turn 1? (in meters)", "option": [], "answer": "1.89", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 1? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_47332895/Validation_47332895_47332895_9102.790.png", "Validation_47332895/Validation_47332895_47332895_9104.789.png", "Validation_47332895/Validation_47332895_47332895_9106.788.png", "Validation_47332895/Validation_47332895_47332895_9108.788.png" ], "turn_id": 8, "scan_id": "Validation_47332895", "origin_question": "How far is your current position from where you were at the end of the turn 1? (in meters)", "option": [], "answer": "3.02", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 1? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47895743/Validation_47895743_47895743_385144.771.png", "Validation_47895743/Validation_47895743_47895743_385146.253.png", "Validation_47895743/Validation_47895743_47895743_385147.753.png", "Validation_47895743/Validation_47895743_47895743_385149.252.png" ], "turn_id": 1, "scan_id": "Validation_47895743", "origin_question": "Remember, have you seen any bed(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any bed(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_47895743/Validation_47895743_47895743_385150.752.png", "Validation_47895743/Validation_47895743_47895743_385152.251.png", "Validation_47895743/Validation_47895743_47895743_385153.750.png", "Validation_47895743/Validation_47895743_47895743_385155.250.png" ], "turn_id": 2, "scan_id": "Validation_47895743", "origin_question": "In which turn did you see both \"the clothes\" and \"the shelf\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the clothes\" and \"the shelf\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_47895743/Validation_47895743_47895743_385156.749.png", "Validation_47895743/Validation_47895743_47895743_385158.249.png", "Validation_47895743/Validation_47895743_47895743_385159.748.png", "Validation_47895743/Validation_47895743_47895743_385161.247.png" ], "turn_id": 3, "scan_id": "Validation_47895743", "origin_question": "Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance forward or backward from that position?", "option": [ "Forward", "Backward" ], "answer": "Forward", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance forward or backward from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Forward', 'Backward'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_47895743/Validation_47895743_47895743_385162.747.png", "Validation_47895743/Validation_47895743_47895743_385164.246.png", "Validation_47895743/Validation_47895743_47895743_385165.746.png", "Validation_47895743/Validation_47895743_47895743_385167.245.png" ], "turn_id": 4, "scan_id": "Validation_47895743", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "1.07", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42444968/Validation_42444968_42444968_40245.400.png", "Validation_42444968/Validation_42444968_42444968_40247.383.png", "Validation_42444968/Validation_42444968_42444968_40249.382.png", "Validation_42444968/Validation_42444968_42444968_40251.381.png" ], "turn_id": 1, "scan_id": "Validation_42444968", "origin_question": "Remember, have you seen any toothpaste(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any toothpaste(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_42444968/Validation_42444968_42444968_40253.380.png", "Validation_42444968/Validation_42444968_42444968_40255.396.png", "Validation_42444968/Validation_42444968_42444968_40257.395.png", "Validation_42444968/Validation_42444968_42444968_40259.395.png" ], "turn_id": 2, "scan_id": "Validation_42444968", "origin_question": "In which turn were you farthest from \"the modern stainless steel range hood with the sleek design\"? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the modern stainless steel range hood with the sleek design\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_42444968/Validation_42444968_42444968_40261.394.png", "Validation_42444968/Validation_42444968_42444968_40263.393.png", "Validation_42444968/Validation_42444968_42444968_40265.392.png", "Validation_42444968/Validation_42444968_42444968_40267.391.png" ], "turn_id": 3, "scan_id": "Validation_42444968", "origin_question": "Which direction is the compact rectangular plastic bin in the neutral color in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the compact rectangular plastic bin in the neutral color in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "Validation_42444968/Validation_42444968_42444968_40269.391.png", "Validation_42444968/Validation_42444968_42444968_40271.390.png", "Validation_42444968/Validation_42444968_42444968_40273.389.png", "Validation_42444968/Validation_42444968_42444968_40275.388.png" ], "turn_id": 4, "scan_id": "Validation_42444968", "origin_question": "Among these three objects(\"the compact, built-in white dishwasher\",\"the rectangular metal-framed window in the neutral tone\",\"the modern stainless steel range hood with the sleek design\" ), which one is the farthest from you?", "option": [ "the compact, built-in white dishwasher", "the rectangular metal-framed window in the neutral tone", "the modern stainless steel range hood with the sleek design" ], "answer": "the rectangular metal-framed window in the neutral tone", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects(\"the compact, built-in white dishwasher\",\"the rectangular metal-framed window in the neutral tone\",\"the modern stainless steel range hood with the sleek design\" ), which one is the farthest from you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the compact, built-in white dishwasher', 'the rectangular metal-framed window in the neutral tone', 'the modern stainless steel range hood with the sleek design'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_42444968/Validation_42444968_42444968_40277.387.png", "Validation_42444968/Validation_42444968_42444968_40279.386.png", "Validation_42444968/Validation_42444968_42444968_40281.385.png", "Validation_42444968/Validation_42444968_42444968_40283.385.png" ], "turn_id": 5, "scan_id": "Validation_42444968", "origin_question": "Is the modern stainless steel range hood with the sleek design to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the modern stainless steel range hood with the sleek design to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_42444968/Validation_42444968_42444968_40285.384.png", "Validation_42444968/Validation_42444968_42444968_40287.383.png", "Validation_42444968/Validation_42444968_42444968_40289.382.png", "Validation_42444968/Validation_42444968_42444968_40291.381.png" ], "turn_id": 6, "scan_id": "Validation_42444968", "origin_question": "Which direction is the rectangular carpet with dark base and light patterns in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the rectangular carpet with dark base and light patterns in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_42444968/Validation_42444968_42444968_40293.380.png", "Validation_42444968/Validation_42444968_42444968_40295.396.png", "Validation_42444968/Validation_42444968_42444968_40297.395.png", "Validation_42444968/Validation_42444968_42444968_40299.395.png" ], "turn_id": 7, "scan_id": "Validation_42444968", "origin_question": "Which two objects are on the same side of you now? \"the rectangular metal-framed window in the neutral tone\", \"the modern stainless steel range hood with the sleek design\", \"the light wood multi-level shelf\". ", "option": [ "the rectangular metal-framed window in the neutral tone, the modern stainless steel range hood with the sleek design", "the rectangular metal-framed window in the neutral tone, the light wood multi-level shelf", "the modern stainless steel range hood with the sleek design, the light wood multi-level shelf" ], "answer": "the rectangular metal-framed window in the neutral tone, the modern stainless steel range hood with the sleek design", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the rectangular metal-framed window in the neutral tone\", \"the modern stainless steel range hood with the sleek design\", \"the light wood multi-level shelf\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular metal-framed window in the neutral tone, the modern stainless steel range hood with the sleek design', 'the rectangular metal-framed window in the neutral tone, the light wood multi-level shelf', 'the modern stainless steel range hood with the sleek design, the light wood multi-level shelf'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_42444968/Validation_42444968_42444968_40301.394.png", "Validation_42444968/Validation_42444968_42444968_40303.393.png", "Validation_42444968/Validation_42444968_42444968_40305.392.png", "Validation_42444968/Validation_42444968_42444968_40307.391.png" ], "turn_id": 8, "scan_id": "Validation_42444968", "origin_question": "Is the compact, sleek white front-loading washing machine to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the compact, sleek white front-loading washing machine to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_42444968/Validation_42444968_42444968_40309.391.png", "Validation_42444968/Validation_42444968_42444968_40311.390.png", "Validation_42444968/Validation_42444968_42444968_40313.389.png", "Validation_42444968/Validation_42444968_42444968_40315.388.png" ], "turn_id": 9, "scan_id": "Validation_42444968", "origin_question": "Which two objects are on the same side of you now? \"the compact, built-in white dishwasher\", \"the compact rectangular plastic bin in the neutral color\", \"the modern circular glass table with the chrome base\". ", "option": [ "the compact, built-in white dishwasher, the compact rectangular plastic bin in the neutral color", "the compact, built-in white dishwasher, the modern circular glass table with the chrome base", "the compact rectangular plastic bin in the neutral color, the modern circular glass table with the chrome base" ], "answer": "the compact, built-in white dishwasher, the compact rectangular plastic bin in the neutral color", "user_message": "For the 9 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the compact, built-in white dishwasher\", \"the compact rectangular plastic bin in the neutral color\", \"the modern circular glass table with the chrome base\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the compact, built-in white dishwasher, the compact rectangular plastic bin in the neutral color', 'the compact, built-in white dishwasher, the modern circular glass table with the chrome base', 'the compact rectangular plastic bin in the neutral color, the modern circular glass table with the chrome base'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47895750/Validation_47895750_47895750_385202.364.png", "Validation_47895750/Validation_47895750_47895750_385203.846.png", "Validation_47895750/Validation_47895750_47895750_385205.346.png", "Validation_47895750/Validation_47895750_47895750_385206.845.png" ], "turn_id": 1, "scan_id": "Validation_47895750", "origin_question": "", "option": [], "answer": "", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_47895750/Validation_47895750_47895750_385208.345.png", "Validation_47895750/Validation_47895750_47895750_385209.844.png", "Validation_47895750/Validation_47895750_47895750_385211.343.png", "Validation_47895750/Validation_47895750_47895750_385212.843.png" ], "turn_id": 2, "scan_id": "Validation_47895750", "origin_question": "When did you first discover \"the floral coverlet with the green and white design on the bed\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the floral coverlet with the green and white design on the bed\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47895750/Validation_47895750_47895750_385214.342.png", "Validation_47895750/Validation_47895750_47895750_385215.858.png", "Validation_47895750/Validation_47895750_47895750_385217.357.png", "Validation_47895750/Validation_47895750_47895750_385218.857.png" ], "turn_id": 3, "scan_id": "Validation_47895750", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the rectangular pillow with the green leaf pattern\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the rectangular pillow with the green leaf pattern\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_47895750/Validation_47895750_47895750_385220.356.png", "Validation_47895750/Validation_47895750_47895750_385221.856.png", "Validation_47895750/Validation_47895750_47895750_385223.355.png", "Validation_47895750/Validation_47895750_47895750_385224.854.png" ], "turn_id": 4, "scan_id": "Validation_47895750", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the cloud-shaped white light fixture with the modern design\" from your position?", "option": [], "answer": "138.2", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the cloud-shaped white light fixture with the modern design\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_47895750/Validation_47895750_47895750_385226.354.png", "Validation_47895750/Validation_47895750_47895750_385227.853.png", "Validation_47895750/Validation_47895750_47895750_385229.352.png", "Validation_47895750/Validation_47895750_47895750_385230.852.png" ], "turn_id": 5, "scan_id": "Validation_47895750", "origin_question": "What will be the first-time appearance order of the following categories: clothes, light, tv?", "option": [ "tv, light, clothes", "clothes, light, tv", "tv, clothes, light", "light, tv, clothes" ], "answer": "clothes, light, tv", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: clothes, light, tv? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['tv, light, clothes', 'clothes, light, tv', 'tv, clothes, light', 'light, tv, clothes'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47430490/Validation_47430490_47430490_313673.285.png", "Validation_47430490/Validation_47430490_47430490_313675.284.png", "Validation_47430490/Validation_47430490_47430490_313677.283.png", "Validation_47430490/Validation_47430490_47430490_313679.282.png", "Validation_47430490/Validation_47430490_47430490_313681.281.png" ], "turn_id": 1, "scan_id": "Validation_47430490", "origin_question": "Remember, have you seen any towel(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any towel(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "Validation_47430490/Validation_47430490_47430490_313683.281.png", "Validation_47430490/Validation_47430490_47430490_313685.280.png", "Validation_47430490/Validation_47430490_47430490_313687.279.png", "Validation_47430490/Validation_47430490_47430490_313689.278.png", "Validation_47430490/Validation_47430490_47430490_313691.277.png" ], "turn_id": 2, "scan_id": "Validation_47430490", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the vibrant orange heart-shaped pillow with plush fabric\"; \"the compact white table with the sleek design\"; \"the vibrant blue flexible cushion\" ", "option": [ "the vibrant orange heart-shaped pillow with plush fabric", "the compact white table with the sleek design", "the vibrant blue flexible cushion" ], "answer": "the vibrant orange heart-shaped pillow with plush fabric", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the vibrant orange heart-shaped pillow with plush fabric\"; \"the compact white table with the sleek design\"; \"the vibrant blue flexible cushion\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the vibrant orange heart-shaped pillow with plush fabric', 'the compact white table with the sleek design', 'the vibrant blue flexible cushion'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_47430490/Validation_47430490_47430490_313693.277.png", "Validation_47430490/Validation_47430490_47430490_313695.276.png", "Validation_47430490/Validation_47430490_47430490_313697.275.png", "Validation_47430490/Validation_47430490_47430490_313699.274.png", "Validation_47430490/Validation_47430490_47430490_313701.273.png" ], "turn_id": 3, "scan_id": "Validation_47430490", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_47430490/Validation_47430490_47430490_313703.273.png", "Validation_47430490/Validation_47430490_47430490_313705.272.png", "Validation_47430490/Validation_47430490_47430490_313707.271.png", "Validation_47430490/Validation_47430490_47430490_313709.270.png", "Validation_47430490/Validation_47430490_47430490_313711.269.png" ], "turn_id": 4, "scan_id": "Validation_47430490", "origin_question": "At the end of which turn were both of \"the light, flowing white curtain with the minimalist design\" and \"the sleek black chair with smooth faux leather\" on your left side?", "option": [], "answer": "3", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the light, flowing white curtain with the minimalist design\" and \"the sleek black chair with smooth faux leather\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47430490/Validation_47430490_47430490_313713.269.png", "Validation_47430490/Validation_47430490_47430490_313715.268.png", "Validation_47430490/Validation_47430490_47430490_313717.284.png", "Validation_47430490/Validation_47430490_47430490_313719.283.png", "Validation_47430490/Validation_47430490_47430490_313721.282.png" ], "turn_id": 5, "scan_id": "Validation_47430490", "origin_question": "Which direction is the vibrant blue flexible cushion in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the vibrant blue flexible cushion in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_47430490/Validation_47430490_47430490_313723.281.png", "Validation_47430490/Validation_47430490_47430490_313725.280.png", "Validation_47430490/Validation_47430490_47430490_313727.279.png", "Validation_47430490/Validation_47430490_47430490_313729.279.png", "Validation_47430490/Validation_47430490_47430490_313731.278.png" ], "turn_id": 6, "scan_id": "Validation_47430490", "origin_question": "At the end of which turn were both of \"the vibrant blue flexible cushion\" and \"the rectangular wooden window with the brown frame\" on your left side?", "option": [], "answer": "2", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the vibrant blue flexible cushion\" and \"the rectangular wooden window with the brown frame\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_47430490/Validation_47430490_47430490_313733.277.png", "Validation_47430490/Validation_47430490_47430490_313735.276.png", "Validation_47430490/Validation_47430490_47430490_313737.275.png", "Validation_47430490/Validation_47430490_47430490_313739.275.png", "Validation_47430490/Validation_47430490_47430490_313741.274.png" ], "turn_id": 7, "scan_id": "Validation_47430490", "origin_question": "What will be the first-time appearance order of the following categories: book, curtain, toy?", "option": [ "toy, curtain, book", "toy, book, curtain", "book, curtain, toy", "book, toy, curtain" ], "answer": "toy, curtain, book", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: book, curtain, toy? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['toy, curtain, book', 'toy, book, curtain', 'book, curtain, toy', 'book, toy, curtain'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_47430490/Validation_47430490_47430490_313743.273.png", "Validation_47430490/Validation_47430490_47430490_313745.272.png", "Validation_47430490/Validation_47430490_47430490_313747.271.png", "Validation_47430490/Validation_47430490_47430490_313749.270.png", "Validation_47430490/Validation_47430490_47430490_313751.270.png" ], "turn_id": 8, "scan_id": "Validation_47430490", "origin_question": "Is the sleek black chair with smooth faux leather to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the sleek black chair with smooth faux leather to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47333898/Validation_47333898_47333898_81319.993.png", "Validation_47333898/Validation_47333898_47333898_81321.976.png", "Validation_47333898/Validation_47333898_47333898_81323.975.png", "Validation_47333898/Validation_47333898_47333898_81325.974.png", "Validation_47333898/Validation_47333898_47333898_81327.973.png" ], "turn_id": 1, "scan_id": "Validation_47333898", "origin_question": "Remember, have you seen any light(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any light(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47333898/Validation_47333898_47333898_81329.973.png", "Validation_47333898/Validation_47333898_47333898_81331.972.png", "Validation_47333898/Validation_47333898_47333898_81333.971.png", "Validation_47333898/Validation_47333898_47333898_81335.970.png", "Validation_47333898/Validation_47333898_47333898_81337.969.png" ], "turn_id": 2, "scan_id": "Validation_47333898", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_47333898/Validation_47333898_47333898_81339.968.png", "Validation_47333898/Validation_47333898_47333898_81341.968.png", "Validation_47333898/Validation_47333898_47333898_81343.967.png", "Validation_47333898/Validation_47333898_47333898_81345.966.png", "Validation_47333898/Validation_47333898_47333898_81347.965.png" ], "turn_id": 3, "scan_id": "Validation_47333898", "origin_question": "In which turn did you see both \"the light\" and \"the window\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the light\" and \"the window\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_47333898/Validation_47333898_47333898_81349.964.png", "Validation_47333898/Validation_47333898_47333898_81351.963.png", "Validation_47333898/Validation_47333898_47333898_81353.963.png", "Validation_47333898/Validation_47333898_47333898_81355.979.png", "Validation_47333898/Validation_47333898_47333898_81357.978.png" ], "turn_id": 4, "scan_id": "Validation_47333898", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_47333898/Validation_47333898_47333898_81359.977.png", "Validation_47333898/Validation_47333898_47333898_81361.976.png", "Validation_47333898/Validation_47333898_47333898_81363.975.png", "Validation_47333898/Validation_47333898_47333898_81365.974.png", "Validation_47333898/Validation_47333898_47333898_81367.974.png" ], "turn_id": 5, "scan_id": "Validation_47333898", "origin_question": "Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_47333898/Validation_47333898_47333898_81369.973.png", "Validation_47333898/Validation_47333898_47333898_81371.972.png", "Validation_47333898/Validation_47333898_47333898_81373.971.png", "Validation_47333898/Validation_47333898_47333898_81375.970.png", "Validation_47333898/Validation_47333898_47333898_81377.970.png" ], "turn_id": 6, "scan_id": "Validation_47333898", "origin_question": "Which two objects are on the same side of you now? \"the large round blue and white ceramic plate\", \"the window\", \"the sleek, white washing machine with the modern design\". ", "option": [ "the large round blue and white ceramic plate, the window", "the large round blue and white ceramic plate, the sleek, white washing machine with the modern design", "the window, the sleek, white washing machine with the modern design" ], "answer": "the large round blue and white ceramic plate, the window", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the large round blue and white ceramic plate\", \"the window\", \"the sleek, white washing machine with the modern design\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the large round blue and white ceramic plate, the window', 'the large round blue and white ceramic plate, the sleek, white washing machine with the modern design', 'the window, the sleek, white washing machine with the modern design'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "Validation_47333898/Validation_47333898_47333898_81379.969.png", "Validation_47333898/Validation_47333898_47333898_81381.968.png", "Validation_47333898/Validation_47333898_47333898_81383.967.png", "Validation_47333898/Validation_47333898_47333898_81385.966.png", "Validation_47333898/Validation_47333898_47333898_81387.966.png" ], "turn_id": 7, "scan_id": "Validation_47333898", "origin_question": "Among these three objects(\"the window\",\"the oven\",\"the sleek, white washing machine with the modern design\" ), which one is the farthest from you?", "option": [ "the window", "the oven", "the sleek, white washing machine with the modern design" ], "answer": "the sleek, white washing machine with the modern design", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the window\",\"the oven\",\"the sleek, white washing machine with the modern design\" ), which one is the farthest from you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the window', 'the oven', 'the sleek, white washing machine with the modern design'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "Validation_47333898/Validation_47333898_47333898_81389.965.png", "Validation_47333898/Validation_47333898_47333898_81391.964.png", "Validation_47333898/Validation_47333898_47333898_81393.963.png", "Validation_47333898/Validation_47333898_47333898_81395.962.png", "Validation_47333898/Validation_47333898_47333898_81397.978.png" ], "turn_id": 8, "scan_id": "Validation_47333898", "origin_question": "Among these three objects(\"the large round blue and white ceramic plate\",\"the stainless steel rectangular sink with the polished finish\",\"the large, white metal refrigerator with ample storage\" ), which one is the closest to you?", "option": [ "the large round blue and white ceramic plate", "the stainless steel rectangular sink with the polished finish", "the large, white metal refrigerator with ample storage" ], "answer": "the large, white metal refrigerator with ample storage", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the large round blue and white ceramic plate\",\"the stainless steel rectangular sink with the polished finish\",\"the large, white metal refrigerator with ample storage\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the large round blue and white ceramic plate', 'the stainless steel rectangular sink with the polished finish', 'the large, white metal refrigerator with ample storage'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_41254246/Validation_41254246_41254246_101384.032.png", "Validation_41254246/Validation_41254246_41254246_101385.998.png", "Validation_41254246/Validation_41254246_41254246_101387.997.png", "Validation_41254246/Validation_41254246_41254246_101389.997.png", "Validation_41254246/Validation_41254246_41254246_101392.012.png" ], "turn_id": 1, "scan_id": "Validation_41254246", "origin_question": "Remember, have you seen any sink(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any sink(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_41254246/Validation_41254246_41254246_101394.012.png", "Validation_41254246/Validation_41254246_41254246_101396.011.png", "Validation_41254246/Validation_41254246_41254246_101398.010.png", "Validation_41254246/Validation_41254246_41254246_101400.009.png", "Validation_41254246/Validation_41254246_41254246_101402.008.png" ], "turn_id": 2, "scan_id": "Validation_41254246", "origin_question": "When did you first discover \"the rectangular wooden door with the smooth surface\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the rectangular wooden door with the smooth surface\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_41254246/Validation_41254246_41254246_101404.008.png", "Validation_41254246/Validation_41254246_41254246_101406.007.png", "Validation_41254246/Validation_41254246_41254246_101408.006.png", "Validation_41254246/Validation_41254246_41254246_101410.005.png", "Validation_41254246/Validation_41254246_41254246_101412.004.png" ], "turn_id": 3, "scan_id": "Validation_41254246", "origin_question": "When did you first discover \"the bed\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the bed\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_41254246/Validation_41254246_41254246_101414.003.png", "Validation_41254246/Validation_41254246_41254246_101416.003.png", "Validation_41254246/Validation_41254246_41254246_101418.002.png", "Validation_41254246/Validation_41254246_41254246_101420.001.png", "Validation_41254246/Validation_41254246_41254246_101422.000.png" ], "turn_id": 4, "scan_id": "Validation_41254246", "origin_question": "In which turn did you see both \"the bed\" and \"the rectangular wooden door with the smooth surface\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the bed\" and \"the rectangular wooden door with the smooth surface\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_41254246/Validation_41254246_41254246_101423.999.png", "Validation_41254246/Validation_41254246_41254246_101425.998.png", "Validation_41254246/Validation_41254246_41254246_101427.997.png", "Validation_41254246/Validation_41254246_41254246_101429.997.png", "Validation_41254246/Validation_41254246_41254246_101432.012.png" ], "turn_id": 5, "scan_id": "Validation_41254246", "origin_question": "Is the plush, rectangular light-colored cotton towel to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the plush, rectangular light-colored cotton towel to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_41254246/Validation_41254246_41254246_101434.012.png", "Validation_41254246/Validation_41254246_41254246_101436.011.png", "Validation_41254246/Validation_41254246_41254246_101438.010.png", "Validation_41254246/Validation_41254246_41254246_101440.009.png", "Validation_41254246/Validation_41254246_41254246_101442.008.png" ], "turn_id": 6, "scan_id": "Validation_41254246", "origin_question": "Is the rectangular wooden door with the smooth surface to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the rectangular wooden door with the smooth surface to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_41254246/Validation_41254246_41254246_101444.008.png", "Validation_41254246/Validation_41254246_41254246_101446.007.png", "Validation_41254246/Validation_41254246_41254246_101448.006.png", "Validation_41254246/Validation_41254246_41254246_101450.005.png", "Validation_41254246/Validation_41254246_41254246_101452.004.png" ], "turn_id": 7, "scan_id": "Validation_41254246", "origin_question": "Please recall \"the rectangular wooden door with the smooth surface\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.74", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the rectangular wooden door with the smooth surface\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_41254246/Validation_41254246_41254246_101454.003.png", "Validation_41254246/Validation_41254246_41254246_101456.003.png", "Validation_41254246/Validation_41254246_41254246_101458.002.png", "Validation_41254246/Validation_41254246_41254246_101460.001.png", "Validation_41254246/Validation_41254246_41254246_101462.000.png" ], "turn_id": 8, "scan_id": "Validation_41254246", "origin_question": "Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the rectangular window with the neutral-colored frame\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the rectangular window with the neutral-colored frame\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_41254246/Validation_41254246_41254246_101463.999.png", "Validation_41254246/Validation_41254246_41254246_101465.999.png", "Validation_41254246/Validation_41254246_41254246_101467.998.png", "Validation_41254246/Validation_41254246_41254246_101469.997.png", "Validation_41254246/Validation_41254246_41254246_101471.996.png" ], "turn_id": 9, "scan_id": "Validation_41254246", "origin_question": "Which two objects are on the same side of you now? \"the rectangular wooden door with the smooth surface\", \"the plush, rectangular light-colored cotton towel\", \"the rectangular window with the neutral-colored frame\". ", "option": [ "the rectangular wooden door with the smooth surface, the plush, rectangular light-colored cotton towel", "the rectangular wooden door with the smooth surface, the rectangular window with the neutral-colored frame", "the plush, rectangular light-colored cotton towel, the rectangular window with the neutral-colored frame" ], "answer": "the plush, rectangular light-colored cotton towel, the rectangular window with the neutral-colored frame", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the rectangular wooden door with the smooth surface\", \"the plush, rectangular light-colored cotton towel\", \"the rectangular window with the neutral-colored frame\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular wooden door with the smooth surface, the plush, rectangular light-colored cotton towel', 'the rectangular wooden door with the smooth surface, the rectangular window with the neutral-colored frame', 'the plush, rectangular light-colored cotton towel, the rectangular window with the neutral-colored frame'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47115548/Validation_47115548_47115548_15696.136.png", "Validation_47115548/Validation_47115548_47115548_15698.118.png", "Validation_47115548/Validation_47115548_47115548_15700.117.png", "Validation_47115548/Validation_47115548_47115548_15702.117.png" ], "turn_id": 1, "scan_id": "Validation_47115548", "origin_question": "Remember, have you seen any chair(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any chair(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47115548/Validation_47115548_47115548_15704.116.png", "Validation_47115548/Validation_47115548_47115548_15706.115.png", "Validation_47115548/Validation_47115548_47115548_15708.114.png", "Validation_47115548/Validation_47115548_47115548_15710.113.png" ], "turn_id": 2, "scan_id": "Validation_47115548", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the compact rectangular wooden table with the light brown finish\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the compact rectangular wooden table with the light brown finish\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_47115548/Validation_47115548_47115548_15712.113.png", "Validation_47115548/Validation_47115548_47115548_15714.112.png", "Validation_47115548/Validation_47115548_47115548_15716.111.png", "Validation_47115548/Validation_47115548_47115548_15718.110.png" ], "turn_id": 3, "scan_id": "Validation_47115548", "origin_question": "When did you first discover \"the sleek white vertical radiator with metal fins\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the sleek white vertical radiator with metal fins\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_47115548/Validation_47115548_47115548_15720.126.png", "Validation_47115548/Validation_47115548_47115548_15722.125.png", "Validation_47115548/Validation_47115548_47115548_15724.125.png", "Validation_47115548/Validation_47115548_47115548_15726.124.png" ], "turn_id": 4, "scan_id": "Validation_47115548", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "24.67", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_47115548/Validation_47115548_47115548_15728.123.png", "Validation_47115548/Validation_47115548_47115548_15730.122.png", "Validation_47115548/Validation_47115548_47115548_15732.121.png", "Validation_47115548/Validation_47115548_47115548_15734.121.png" ], "turn_id": 5, "scan_id": "Validation_47115548", "origin_question": "Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "76.82", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_47115548/Validation_47115548_47115548_15736.120.png", "Validation_47115548/Validation_47115548_47115548_15738.119.png", "Validation_47115548/Validation_47115548_47115548_15740.118.png", "Validation_47115548/Validation_47115548_47115548_15742.118.png" ], "turn_id": 6, "scan_id": "Validation_47115548", "origin_question": "By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the light wood paneled door with the natural finish\", Y: \"the classic white rectangular window with multiple panes\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to Y and farther away from X", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the light wood paneled door with the natural finish\", Y: \"the classic white rectangular window with multiple panes\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_47115548/Validation_47115548_47115548_15744.117.png", "Validation_47115548/Validation_47115548_47115548_15746.116.png", "Validation_47115548/Validation_47115548_47115548_15748.115.png", "Validation_47115548/Validation_47115548_47115548_15750.114.png" ], "turn_id": 7, "scan_id": "Validation_47115548", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "121.77", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_47115548/Validation_47115548_47115548_15752.113.png", "Validation_47115548/Validation_47115548_47115548_15754.113.png", "Validation_47115548/Validation_47115548_47115548_15756.112.png", "Validation_47115548/Validation_47115548_47115548_15758.111.png" ], "turn_id": 8, "scan_id": "Validation_47115548", "origin_question": "In which turn did you see both \"the classic white rectangular window with multiple panes\" and \"the weathered light-colored wooden box\" simultaneously? ", "option": [], "answer": "8", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the classic white rectangular window with multiple panes\" and \"the weathered light-colored wooden box\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_47115548/Validation_47115548_47115548_15760.127.png", "Validation_47115548/Validation_47115548_47115548_15762.126.png", "Validation_47115548/Validation_47115548_47115548_15764.125.png", "Validation_47115548/Validation_47115548_47115548_15766.125.png" ], "turn_id": 9, "scan_id": "Validation_47115548", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the light wood paneled door with the natural finish\", Y: \"the compact rectangular wooden table with the light brown finish\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to Y and farther away from X", "user_message": "For the 9 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the light wood paneled door with the natural finish\", Y: \"the compact rectangular wooden table with the light brown finish\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "Validation_45662987/Validation_45662987_45662987_27867.211.png", "Validation_45662987/Validation_45662987_45662987_27869.193.png", "Validation_45662987/Validation_45662987_45662987_27871.192.png", "Validation_45662987/Validation_45662987_45662987_27873.192.png" ], "turn_id": 1, "scan_id": "Validation_45662987", "origin_question": "Remember, how many mailbox(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many mailbox(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_45662987/Validation_45662987_45662987_27875.191.png", "Validation_45662987/Validation_45662987_45662987_27877.190.png", "Validation_45662987/Validation_45662987_45662987_27879.189.png", "Validation_45662987/Validation_45662987_45662987_27881.188.png" ], "turn_id": 2, "scan_id": "Validation_45662987", "origin_question": "At the end of which turn were both of \"the fireplace\" and \"the sleek rectangular metal mirror with the silver-gray frame\" on your left side?", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the fireplace\" and \"the sleek rectangular metal mirror with the silver-gray frame\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "Validation_45662987/Validation_45662987_45662987_27883.188.png", "Validation_45662987/Validation_45662987_45662987_27885.187.png", "Validation_45662987/Validation_45662987_45662987_27887.186.png", "Validation_45662987/Validation_45662987_45662987_27889.185.png" ], "turn_id": 3, "scan_id": "Validation_45662987", "origin_question": "Among these three objects(\"the fireplace\",\"the carpet\",\"the glossy green plant in the cylindrical pot\" ), which one is the closest to you?", "option": [ "the fireplace", "the carpet", "the glossy green plant in the cylindrical pot" ], "answer": "the carpet", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects(\"the fireplace\",\"the carpet\",\"the glossy green plant in the cylindrical pot\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the fireplace', 'the carpet', 'the glossy green plant in the cylindrical pot'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "Validation_45662987/Validation_45662987_45662987_27891.184.png", "Validation_45662987/Validation_45662987_45662987_27893.184.png", "Validation_45662987/Validation_45662987_45662987_27895.183.png", "Validation_45662987/Validation_45662987_45662987_27897.199.png" ], "turn_id": 4, "scan_id": "Validation_45662987", "origin_question": "Among these three objects(\"the large modern black monitor with the sleek design\",\"the large rectangular window with the white frame\",\"the carpet\" ), which one is the closest to you?", "option": [ "the large modern black monitor with the sleek design", "the large rectangular window with the white frame", "the carpet" ], "answer": "the carpet", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects(\"the large modern black monitor with the sleek design\",\"the large rectangular window with the white frame\",\"the carpet\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the large modern black monitor with the sleek design', 'the large rectangular window with the white frame', 'the carpet'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_45662987/Validation_45662987_45662987_27899.198.png", "Validation_45662987/Validation_45662987_45662987_27901.197.png", "Validation_45662987/Validation_45662987_45662987_27903.196.png", "Validation_45662987/Validation_45662987_45662987_27905.196.png" ], "turn_id": 5, "scan_id": "Validation_45662987", "origin_question": "Is the carpet to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the carpet to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_45662987/Validation_45662987_45662987_27907.195.png", "Validation_45662987/Validation_45662987_45662987_27909.194.png", "Validation_45662987/Validation_45662987_45662987_27911.193.png", "Validation_45662987/Validation_45662987_45662987_27913.192.png" ], "turn_id": 6, "scan_id": "Validation_45662987", "origin_question": "Is the rectangular white door with the paneled design to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the rectangular white door with the paneled design to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_45662987/Validation_45662987_45662987_27915.192.png", "Validation_45662987/Validation_45662987_45662987_27917.191.png", "Validation_45662987/Validation_45662987_45662987_27919.190.png", "Validation_45662987/Validation_45662987_45662987_27921.189.png" ], "turn_id": 7, "scan_id": "Validation_45662987", "origin_question": "At the end of which turn were both of \"the fireplace\" and \"the carpet\" on your left side?", "option": [], "answer": "1", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the fireplace\" and \"the carpet\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_45662987/Validation_45662987_45662987_27923.189.png", "Validation_45662987/Validation_45662987_45662987_27925.188.png", "Validation_45662987/Validation_45662987_45662987_27927.187.png", "Validation_45662987/Validation_45662987_45662987_27929.186.png" ], "turn_id": 8, "scan_id": "Validation_45662987", "origin_question": "Which two objects are on the same side of you now? \"the rectangular white door with the paneled design\", \"the large modern black monitor with the sleek design\", \"the large rectangular window with the white frame\". ", "option": [ "the rectangular white door with the paneled design, the large modern black monitor with the sleek design", "the rectangular white door with the paneled design, the large rectangular window with the white frame", "the large modern black monitor with the sleek design, the large rectangular window with the white frame" ], "answer": "the large modern black monitor with the sleek design, the large rectangular window with the white frame", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the rectangular white door with the paneled design\", \"the large modern black monitor with the sleek design\", \"the large rectangular window with the white frame\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular white door with the paneled design, the large modern black monitor with the sleek design', 'the rectangular white door with the paneled design, the large rectangular window with the white frame', 'the large modern black monitor with the sleek design, the large rectangular window with the white frame'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47332004/Validation_47332004_47332004_171922.455.png", "Validation_47332004/Validation_47332004_47332004_171924.437.png", "Validation_47332004/Validation_47332004_47332004_171926.437.png", "Validation_47332004/Validation_47332004_47332004_171928.436.png" ], "turn_id": 1, "scan_id": "Validation_47332004", "origin_question": "", "option": [], "answer": "", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47332004/Validation_47332004_47332004_171930.435.png", "Validation_47332004/Validation_47332004_47332004_171932.434.png", "Validation_47332004/Validation_47332004_47332004_171934.433.png", "Validation_47332004/Validation_47332004_47332004_171936.432.png" ], "turn_id": 2, "scan_id": "Validation_47332004", "origin_question": "In which turn were you closest to \"the soft, checkered black and white towel\"? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the soft, checkered black and white towel\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_47332004/Validation_47332004_47332004_171938.432.png", "Validation_47332004/Validation_47332004_47332004_171940.431.png", "Validation_47332004/Validation_47332004_47332004_171942.430.png", "Validation_47332004/Validation_47332004_47332004_171944.429.png" ], "turn_id": 3, "scan_id": "Validation_47332004", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular white ceramic sink with the glossy finish\", Y: \"the small, smooth cylindrical ceramic flowerpot\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular white ceramic sink with the glossy finish\", Y: \"the small, smooth cylindrical ceramic flowerpot\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_47332004/Validation_47332004_47332004_171946.429.png", "Validation_47332004/Validation_47332004_47332004_171948.444.png", "Validation_47332004/Validation_47332004_47332004_171950.443.png", "Validation_47332004/Validation_47332004_47332004_171952.443.png" ], "turn_id": 4, "scan_id": "Validation_47332004", "origin_question": "Is the rectangular white ceramic sink with the glossy finish to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the rectangular white ceramic sink with the glossy finish to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47332004/Validation_47332004_47332004_171954.442.png", "Validation_47332004/Validation_47332004_47332004_171956.441.png", "Validation_47332004/Validation_47332004_47332004_171958.440.png", "Validation_47332004/Validation_47332004_47332004_171960.439.png" ], "turn_id": 5, "scan_id": "Validation_47332004", "origin_question": "Which direction is the soft, checkered black and white towel in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the soft, checkered black and white towel in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47332004/Validation_47332004_47332004_171962.439.png", "Validation_47332004/Validation_47332004_47332004_171964.438.png", "Validation_47332004/Validation_47332004_47332004_171966.437.png", "Validation_47332004/Validation_47332004_47332004_171968.436.png" ], "turn_id": 6, "scan_id": "Validation_47332004", "origin_question": "Which direction is the small, smooth cylindrical ceramic flowerpot in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the small, smooth cylindrical ceramic flowerpot in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_47332004/Validation_47332004_47332004_171970.435.png", "Validation_47332004/Validation_47332004_47332004_171972.435.png", "Validation_47332004/Validation_47332004_47332004_171974.434.png", "Validation_47332004/Validation_47332004_47332004_171976.433.png" ], "turn_id": 7, "scan_id": "Validation_47332004", "origin_question": "Please recall \"the small, smooth cylindrical ceramic flowerpot\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.91", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the small, smooth cylindrical ceramic flowerpot\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_48458657/Validation_48458657_48458657_7584.704.png", "Validation_48458657/Validation_48458657_48458657_7586.187.png", "Validation_48458657/Validation_48458657_48458657_7587.687.png", "Validation_48458657/Validation_48458657_48458657_7589.186.png" ], "turn_id": 1, "scan_id": "Validation_48458657", "origin_question": "Remember, have you seen any sink(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any sink(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_48458657/Validation_48458657_48458657_7590.685.png", "Validation_48458657/Validation_48458657_48458657_7592.185.png", "Validation_48458657/Validation_48458657_48458657_7593.684.png", "Validation_48458657/Validation_48458657_48458657_7595.184.png" ], "turn_id": 2, "scan_id": "Validation_48458657", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_48458657/Validation_48458657_48458657_7596.683.png", "Validation_48458657/Validation_48458657_48458657_7598.182.png", "Validation_48458657/Validation_48458657_48458657_7599.682.png", "Validation_48458657/Validation_48458657_48458657_7601.198.png" ], "turn_id": 3, "scan_id": "Validation_48458657", "origin_question": "When did you first discover \"the sleek wooden door with the light wood finish\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the sleek wooden door with the light wood finish\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_48458657/Validation_48458657_48458657_7602.697.png", "Validation_48458657/Validation_48458657_48458657_7604.197.png", "Validation_48458657/Validation_48458657_48458657_7605.696.png", "Validation_48458657/Validation_48458657_48458657_7607.195.png" ], "turn_id": 4, "scan_id": "Validation_48458657", "origin_question": "Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the compact white cabinet with the sleek design\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the compact white cabinet with the sleek design\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "Validation_48458657/Validation_48458657_48458657_7608.695.png", "Validation_48458657/Validation_48458657_48458657_7610.194.png", "Validation_48458657/Validation_48458657_48458657_7611.693.png", "Validation_48458657/Validation_48458657_48458657_7613.193.png" ], "turn_id": 5, "scan_id": "Validation_48458657", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the sleek wooden door with the light wood finish\"; \"the towel\"; \"the sink\" ", "option": [ "the sleek wooden door with the light wood finish", "the towel", "the sink" ], "answer": "the towel", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the sleek wooden door with the light wood finish\"; \"the towel\"; \"the sink\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the sleek wooden door with the light wood finish', 'the towel', 'the sink'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47331651/Validation_47331651_47331651_22850.252.png", "Validation_47331651/Validation_47331651_47331651_22852.234.png", "Validation_47331651/Validation_47331651_47331651_22854.234.png", "Validation_47331651/Validation_47331651_47331651_22856.233.png", "Validation_47331651/Validation_47331651_47331651_22858.232.png" ], "turn_id": 1, "scan_id": "Validation_47331651", "origin_question": "Remember, have you seen any mop(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any mop(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_47331651/Validation_47331651_47331651_22860.231.png", "Validation_47331651/Validation_47331651_47331651_22862.231.png", "Validation_47331651/Validation_47331651_47331651_22864.230.png", "Validation_47331651/Validation_47331651_47331651_22866.229.png", "Validation_47331651/Validation_47331651_47331651_22868.228.png" ], "turn_id": 2, "scan_id": "Validation_47331651", "origin_question": "In which turn did you see both \"the rectangular white door with gold-toned finishes\" and \"the compact, white microwave with metal and plastic construction\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the rectangular white door with gold-toned finishes\" and \"the compact, white microwave with metal and plastic construction\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_47331651/Validation_47331651_47331651_22870.227.png", "Validation_47331651/Validation_47331651_47331651_22872.226.png", "Validation_47331651/Validation_47331651_47331651_22874.226.png", "Validation_47331651/Validation_47331651_47331651_22876.225.png", "Validation_47331651/Validation_47331651_47331651_22878.224.png" ], "turn_id": 3, "scan_id": "Validation_47331651", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the vibrant, multicolored cup\", Y: \"the white, average-sized top-freezer refrigerator\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the vibrant, multicolored cup\", Y: \"the white, average-sized top-freezer refrigerator\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_47331651/Validation_47331651_47331651_22880.223.png", "Validation_47331651/Validation_47331651_47331651_22882.222.png", "Validation_47331651/Validation_47331651_47331651_22884.238.png", "Validation_47331651/Validation_47331651_47331651_22886.237.png", "Validation_47331651/Validation_47331651_47331651_22888.236.png" ], "turn_id": 4, "scan_id": "Validation_47331651", "origin_question": "Which two objects are on the same side of you now? \"the rectangular white door with gold-toned finishes\", \"the compact, white microwave with metal and plastic construction\", \"the vibrant, multicolored cup\". ", "option": [ "the rectangular white door with gold-toned finishes, the compact, white microwave with metal and plastic construction", "the rectangular white door with gold-toned finishes, the vibrant, multicolored cup", "the compact, white microwave with metal and plastic construction, the vibrant, multicolored cup" ], "answer": "the rectangular white door with gold-toned finishes, the compact, white microwave with metal and plastic construction", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the rectangular white door with gold-toned finishes\", \"the compact, white microwave with metal and plastic construction\", \"the vibrant, multicolored cup\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular white door with gold-toned finishes, the compact, white microwave with metal and plastic construction', 'the rectangular white door with gold-toned finishes, the vibrant, multicolored cup', 'the compact, white microwave with metal and plastic construction, the vibrant, multicolored cup'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47331651/Validation_47331651_47331651_22890.236.png", "Validation_47331651/Validation_47331651_47331651_22892.235.png", "Validation_47331651/Validation_47331651_47331651_22894.234.png", "Validation_47331651/Validation_47331651_47331651_22896.233.png", "Validation_47331651/Validation_47331651_47331651_22898.232.png" ], "turn_id": 5, "scan_id": "Validation_47331651", "origin_question": "", "option": [], "answer": "", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_47331651/Validation_47331651_47331651_22900.232.png", "Validation_47331651/Validation_47331651_47331651_22902.231.png", "Validation_47331651/Validation_47331651_47331651_22904.230.png", "Validation_47331651/Validation_47331651_47331651_22906.229.png", "Validation_47331651/Validation_47331651_47331651_22908.228.png" ], "turn_id": 6, "scan_id": "Validation_47331651", "origin_question": "What will be the first-time appearance order of the following categories: door, microwave, towel?", "option": [ "microwave, towel, door", "door, microwave, towel", "door, towel, microwave", "microwave, door, towel" ], "answer": "door, microwave, towel", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: door, microwave, towel? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['microwave, towel, door', 'door, microwave, towel', 'door, towel, microwave', 'microwave, door, towel'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47331651/Validation_47331651_47331651_22910.228.png", "Validation_47331651/Validation_47331651_47331651_22912.227.png", "Validation_47331651/Validation_47331651_47331651_22914.226.png", "Validation_47331651/Validation_47331651_47331651_22916.225.png", "Validation_47331651/Validation_47331651_47331651_22918.224.png" ], "turn_id": 7, "scan_id": "Validation_47331651", "origin_question": "In which turn were you closest to \"the white, average-sized top-freezer refrigerator\"? ", "option": [], "answer": "5", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the white, average-sized top-freezer refrigerator\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_47331651/Validation_47331651_47331651_22920.224.png", "Validation_47331651/Validation_47331651_47331651_22922.223.png", "Validation_47331651/Validation_47331651_47331651_22924.238.png", "Validation_47331651/Validation_47331651_47331651_22926.238.png", "Validation_47331651/Validation_47331651_47331651_22928.237.png" ], "turn_id": 8, "scan_id": "Validation_47331651", "origin_question": "Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Left", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_47331651/Validation_47331651_47331651_22930.236.png", "Validation_47331651/Validation_47331651_47331651_22932.235.png", "Validation_47331651/Validation_47331651_47331651_22934.234.png", "Validation_47331651/Validation_47331651_47331651_22936.234.png", "Validation_47331651/Validation_47331651_47331651_22938.233.png" ], "turn_id": 9, "scan_id": "Validation_47331651", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the slender dark bottle with the classic shape\" from your position?", "option": [], "answer": "179.33", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the slender dark bottle with the classic shape\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_47331651/Validation_47331651_47331651_22940.232.png", "Validation_47331651/Validation_47331651_47331651_22942.231.png", "Validation_47331651/Validation_47331651_47331651_22944.230.png", "Validation_47331651/Validation_47331651_47331651_22946.230.png", "Validation_47331651/Validation_47331651_47331651_22948.229.png" ], "turn_id": 10, "scan_id": "Validation_47331651", "origin_question": "By the end of the 4 turn, were you getting closer to or farther away from these two objects: X: \"the compact, white microwave with metal and plastic construction\", Y: \"the sleek white front-loading washing machine\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to Y and farther away from X", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 4 turn, were you getting closer to or farther away from these two objects: X: \"the compact, white microwave with metal and plastic construction\", Y: \"the sleek white front-loading washing machine\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47331988/Validation_47331988_47331988_170595.573.png", "Validation_47331988/Validation_47331988_47331988_170597.556.png", "Validation_47331988/Validation_47331988_47331988_170599.555.png", "Validation_47331988/Validation_47331988_47331988_170601.554.png", "Validation_47331988/Validation_47331988_47331988_170603.554.png" ], "turn_id": 1, "scan_id": "Validation_47331988", "origin_question": "Remember, have you seen any cabinet(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any cabinet(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47331988/Validation_47331988_47331988_170605.553.png", "Validation_47331988/Validation_47331988_47331988_170607.552.png", "Validation_47331988/Validation_47331988_47331988_170609.551.png", "Validation_47331988/Validation_47331988_47331988_170611.550.png", "Validation_47331988/Validation_47331988_47331988_170613.549.png" ], "turn_id": 2, "scan_id": "Validation_47331988", "origin_question": "In which turn were you closest to \"the rectangular modern white door with glass panels\"? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the rectangular modern white door with glass panels\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47331988/Validation_47331988_47331988_170615.549.png", "Validation_47331988/Validation_47331988_47331988_170617.565.png", "Validation_47331988/Validation_47331988_47331988_170619.564.png", "Validation_47331988/Validation_47331988_47331988_170621.563.png", "Validation_47331988/Validation_47331988_47331988_170623.562.png" ], "turn_id": 3, "scan_id": "Validation_47331988", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the sleek, dark wood cabinet with adjustable shelving\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the sleek, dark wood cabinet with adjustable shelving\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47331988/Validation_47331988_47331988_170625.561.png", "Validation_47331988/Validation_47331988_47331988_170627.560.png", "Validation_47331988/Validation_47331988_47331988_170629.560.png", "Validation_47331988/Validation_47331988_47331988_170631.559.png", "Validation_47331988/Validation_47331988_47331988_170633.558.png" ], "turn_id": 4, "scan_id": "Validation_47331988", "origin_question": "In which turn were you farthest from \"the sleek, dark wood cabinet with adjustable shelving\"? ", "option": [], "answer": "1", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the sleek, dark wood cabinet with adjustable shelving\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_47331988/Validation_47331988_47331988_170635.557.png", "Validation_47331988/Validation_47331988_47331988_170637.557.png", "Validation_47331988/Validation_47331988_47331988_170639.556.png", "Validation_47331988/Validation_47331988_47331988_170641.555.png", "Validation_47331988/Validation_47331988_47331988_170643.554.png" ], "turn_id": 5, "scan_id": "Validation_47331988", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "24.89", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47331988/Validation_47331988_47331988_170645.553.png", "Validation_47331988/Validation_47331988_47331988_170647.552.png", "Validation_47331988/Validation_47331988_47331988_170649.552.png", "Validation_47331988/Validation_47331988_47331988_170651.551.png", "Validation_47331988/Validation_47331988_47331988_170653.550.png" ], "turn_id": 6, "scan_id": "Validation_47331988", "origin_question": "Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the sleek, dark wood cabinet with adjustable shelving\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the sleek, dark wood cabinet with adjustable shelving\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47331988/Validation_47331988_47331988_170655.549.png", "Validation_47331988/Validation_47331988_47331988_170657.548.png", "Validation_47331988/Validation_47331988_47331988_170659.548.png", "Validation_47331988/Validation_47331988_47331988_170661.563.png", "Validation_47331988/Validation_47331988_47331988_170663.563.png" ], "turn_id": 7, "scan_id": "Validation_47331988", "origin_question": "", "option": [], "answer": "", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_47331988/Validation_47331988_47331988_170665.562.png", "Validation_47331988/Validation_47331988_47331988_170667.561.png", "Validation_47331988/Validation_47331988_47331988_170669.560.png", "Validation_47331988/Validation_47331988_47331988_170671.559.png", "Validation_47331988/Validation_47331988_47331988_170673.559.png" ], "turn_id": 8, "scan_id": "Validation_47331988", "origin_question": "In which turn did you see both \"the sleek, dark wood cabinet with adjustable shelving\" and \"the modern rectangular bed with the brown duvet\" simultaneously? ", "option": [], "answer": "6", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the sleek, dark wood cabinet with adjustable shelving\" and \"the modern rectangular bed with the brown duvet\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_47331988/Validation_47331988_47331988_170675.558.png", "Validation_47331988/Validation_47331988_47331988_170677.557.png", "Validation_47331988/Validation_47331988_47331988_170679.556.png", "Validation_47331988/Validation_47331988_47331988_170681.555.png", "Validation_47331988/Validation_47331988_47331988_170683.554.png" ], "turn_id": 9, "scan_id": "Validation_47331988", "origin_question": "Is the rectangular modern white door with glass panels to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the rectangular modern white door with glass panels to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_47331988/Validation_47331988_47331988_170685.554.png", "Validation_47331988/Validation_47331988_47331988_170687.553.png", "Validation_47331988/Validation_47331988_47331988_170689.552.png", "Validation_47331988/Validation_47331988_47331988_170691.551.png", "Validation_47331988/Validation_47331988_47331988_170693.551.png" ], "turn_id": 10, "scan_id": "Validation_47331988", "origin_question": "Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Left", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42899729/Validation_42899729_42899729_800856.978.png", "Validation_42899729/Validation_42899729_42899729_800858.978.png", "Validation_42899729/Validation_42899729_42899729_800860.977.png", "Validation_42899729/Validation_42899729_42899729_800862.976.png" ], "turn_id": 1, "scan_id": "Validation_42899729", "origin_question": "Remember, have you seen any bed(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any bed(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_42899729/Validation_42899729_42899729_800864.975.png", "Validation_42899729/Validation_42899729_42899729_800866.974.png", "Validation_42899729/Validation_42899729_42899729_800868.973.png", "Validation_42899729/Validation_42899729_42899729_800870.973.png" ], "turn_id": 2, "scan_id": "Validation_42899729", "origin_question": "In which turn were you closest to \"the bed\"? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the bed\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_42899729/Validation_42899729_42899729_800872.972.png", "Validation_42899729/Validation_42899729_42899729_800874.971.png", "Validation_42899729/Validation_42899729_42899729_800876.970.png", "Validation_42899729/Validation_42899729_42899729_800878.969.png" ], "turn_id": 3, "scan_id": "Validation_42899729", "origin_question": "When did you first discover \"the compact, white plastic bin without the lid\"? (index of the turn)", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the compact, white plastic bin without the lid\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_42899729/Validation_42899729_42899729_800880.968.png", "Validation_42899729/Validation_42899729_42899729_800882.968.png", "Validation_42899729/Validation_42899729_42899729_800884.967.png", "Validation_42899729/Validation_42899729_42899729_800886.966.png" ], "turn_id": 4, "scan_id": "Validation_42899729", "origin_question": "Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_42899729/Validation_42899729_42899729_800888.965.png", "Validation_42899729/Validation_42899729_42899729_800890.964.png", "Validation_42899729/Validation_42899729_42899729_800892.980.png", "Validation_42899729/Validation_42899729_42899729_800894.979.png" ], "turn_id": 5, "scan_id": "Validation_42899729", "origin_question": "Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance forward or backward from that position?", "option": [ "Forward", "Backward" ], "answer": "Backward", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance forward or backward from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Forward', 'Backward'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_42899729/Validation_42899729_42899729_800896.978.png", "Validation_42899729/Validation_42899729_42899729_800898.978.png", "Validation_42899729/Validation_42899729_42899729_800900.977.png", "Validation_42899729/Validation_42899729_42899729_800902.976.png" ], "turn_id": 6, "scan_id": "Validation_42899729", "origin_question": "How far is your current position from where you were at the end of the turn 3? (in meters)", "option": [], "answer": "3.13", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 3? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_42899729/Validation_42899729_42899729_800904.975.png", "Validation_42899729/Validation_42899729_42899729_800906.974.png", "Validation_42899729/Validation_42899729_42899729_800908.974.png", "Validation_42899729/Validation_42899729_42899729_800910.973.png" ], "turn_id": 7, "scan_id": "Validation_42899729", "origin_question": "Which direction is the compact, white plastic bin without the lid in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the compact, white plastic bin without the lid in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_42899729/Validation_42899729_42899729_800912.972.png", "Validation_42899729/Validation_42899729_42899729_800914.971.png", "Validation_42899729/Validation_42899729_42899729_800916.970.png", "Validation_42899729/Validation_42899729_42899729_800918.969.png" ], "turn_id": 8, "scan_id": "Validation_42899729", "origin_question": "", "option": [], "answer": "", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42899690/Validation_42899690_42899690_789384.928.png", "Validation_42899690/Validation_42899690_42899690_789386.411.png", "Validation_42899690/Validation_42899690_42899690_789387.910.png", "Validation_42899690/Validation_42899690_42899690_789389.409.png" ], "turn_id": 1, "scan_id": "Validation_42899690", "origin_question": "Remember, have you seen any arch(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any arch(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_42899690/Validation_42899690_42899690_789390.909.png", "Validation_42899690/Validation_42899690_42899690_789392.408.png", "Validation_42899690/Validation_42899690_42899690_789393.907.png", "Validation_42899690/Validation_42899690_42899690_789395.407.png" ], "turn_id": 2, "scan_id": "Validation_42899690", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_42899690/Validation_42899690_42899690_789396.906.png", "Validation_42899690/Validation_42899690_42899690_789398.406.png", "Validation_42899690/Validation_42899690_42899690_789399.905.png", "Validation_42899690/Validation_42899690_42899690_789401.404.png" ], "turn_id": 3, "scan_id": "Validation_42899690", "origin_question": "Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Left", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_42899690/Validation_42899690_42899690_789402.904.png", "Validation_42899690/Validation_42899690_42899690_789404.403.png", "Validation_42899690/Validation_42899690_42899690_789405.902.png", "Validation_42899690/Validation_42899690_42899690_789407.402.png" ], "turn_id": 4, "scan_id": "Validation_42899690", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_42899690/Validation_42899690_42899690_789408.901.png", "Validation_42899690/Validation_42899690_42899690_789410.401.png", "Validation_42899690/Validation_42899690_42899690_789411.900.png", "Validation_42899690/Validation_42899690_42899690_789413.400.png" ], "turn_id": 5, "scan_id": "Validation_42899690", "origin_question": "Using your orientation at the end of turn 4 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "140.68", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 4 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_42899690/Validation_42899690_42899690_789414.899.png", "Validation_42899690/Validation_42899690_42899690_789416.398.png", "Validation_42899690/Validation_42899690_42899690_789417.898.png", "Validation_42899690/Validation_42899690_42899690_789419.397.png" ], "turn_id": 6, "scan_id": "Validation_42899690", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47333932/Validation_47333932_47333932_84063.222.png", "Validation_47333932/Validation_47333932_47333932_84065.204.png", "Validation_47333932/Validation_47333932_47333932_84067.203.png", "Validation_47333932/Validation_47333932_47333932_84069.202.png", "Validation_47333932/Validation_47333932_47333932_84071.202.png" ], "turn_id": 1, "scan_id": "Validation_47333932", "origin_question": "Remember, have you seen any bed(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any bed(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_47333932/Validation_47333932_47333932_84073.201.png", "Validation_47333932/Validation_47333932_47333932_84075.200.png", "Validation_47333932/Validation_47333932_47333932_84077.199.png", "Validation_47333932/Validation_47333932_47333932_84079.198.png", "Validation_47333932/Validation_47333932_47333932_84081.198.png" ], "turn_id": 2, "scan_id": "Validation_47333932", "origin_question": "In which turn did you see both \"the sleek white door with the metal handle\" and \"the window\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the sleek white door with the metal handle\" and \"the window\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_47333932/Validation_47333932_47333932_84083.197.png", "Validation_47333932/Validation_47333932_47333932_84085.213.png", "Validation_47333932/Validation_47333932_47333932_84087.212.png", "Validation_47333932/Validation_47333932_47333932_84089.211.png", "Validation_47333932/Validation_47333932_47333932_84091.210.png" ], "turn_id": 3, "scan_id": "Validation_47333932", "origin_question": "In which turn did you see both \"the warm brown wooden table with the smooth surface\" and \"the sleek white door with the metal handle\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the warm brown wooden table with the smooth surface\" and \"the sleek white door with the metal handle\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_47333932/Validation_47333932_47333932_84093.209.png", "Validation_47333932/Validation_47333932_47333932_84095.209.png", "Validation_47333932/Validation_47333932_47333932_84097.208.png", "Validation_47333932/Validation_47333932_47333932_84099.207.png", "Validation_47333932/Validation_47333932_47333932_84101.206.png" ], "turn_id": 4, "scan_id": "Validation_47333932", "origin_question": "Is the medium-sized cylindrical ceramic flowerpot to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the medium-sized cylindrical ceramic flowerpot to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_47333932/Validation_47333932_47333932_84103.205.png", "Validation_47333932/Validation_47333932_47333932_84105.205.png", "Validation_47333932/Validation_47333932_47333932_84107.204.png", "Validation_47333932/Validation_47333932_47333932_84109.203.png", "Validation_47333932/Validation_47333932_47333932_84111.202.png" ], "turn_id": 5, "scan_id": "Validation_47333932", "origin_question": "Please recall \"the medium-sized rectangular brown cardboard box\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.46", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the medium-sized rectangular brown cardboard box\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "Validation_47333932/Validation_47333932_47333932_84113.201.png", "Validation_47333932/Validation_47333932_47333932_84115.200.png", "Validation_47333932/Validation_47333932_47333932_84117.200.png", "Validation_47333932/Validation_47333932_47333932_84119.199.png", "Validation_47333932/Validation_47333932_47333932_84121.198.png" ], "turn_id": 6, "scan_id": "Validation_47333932", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the window\"; \"the tall, light-colored shelf\"; \"the sleek, minimalistic stool with the sturdy frame\" ", "option": [ "the window", "the tall, light-colored shelf", "the sleek, minimalistic stool with the sturdy frame" ], "answer": "the tall, light-colored shelf", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the window\"; \"the tall, light-colored shelf\"; \"the sleek, minimalistic stool with the sturdy frame\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the window', 'the tall, light-colored shelf', 'the sleek, minimalistic stool with the sturdy frame'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_47333932/Validation_47333932_47333932_84123.197.png", "Validation_47333932/Validation_47333932_47333932_84125.196.png", "Validation_47333932/Validation_47333932_47333932_84127.212.png", "Validation_47333932/Validation_47333932_47333932_84129.211.png", "Validation_47333932/Validation_47333932_47333932_84131.211.png" ], "turn_id": 7, "scan_id": "Validation_47333932", "origin_question": "Is the bed to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the bed to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_47333932/Validation_47333932_47333932_84133.210.png", "Validation_47333932/Validation_47333932_47333932_84135.209.png", "Validation_47333932/Validation_47333932_47333932_84137.208.png", "Validation_47333932/Validation_47333932_47333932_84139.207.png", "Validation_47333932/Validation_47333932_47333932_84141.206.png" ], "turn_id": 8, "scan_id": "Validation_47333932", "origin_question": "Which two objects are on the same side of you now? \"the bed\", \"the tall, light-colored shelf\", \"the medium-sized cylindrical ceramic flowerpot\". ", "option": [ "the bed, the tall, light-colored shelf", "the bed, the medium-sized cylindrical ceramic flowerpot", "the tall, light-colored shelf, the medium-sized cylindrical ceramic flowerpot" ], "answer": "the bed, the tall, light-colored shelf", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the bed\", \"the tall, light-colored shelf\", \"the medium-sized cylindrical ceramic flowerpot\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the bed, the tall, light-colored shelf', 'the bed, the medium-sized cylindrical ceramic flowerpot', 'the tall, light-colored shelf, the medium-sized cylindrical ceramic flowerpot'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_47333932/Validation_47333932_47333932_84143.206.png", "Validation_47333932/Validation_47333932_47333932_84145.205.png", "Validation_47333932/Validation_47333932_47333932_84147.204.png", "Validation_47333932/Validation_47333932_47333932_84149.203.png", "Validation_47333932/Validation_47333932_47333932_84151.202.png" ], "turn_id": 9, "scan_id": "Validation_47333932", "origin_question": "When did you first discover \"the bed\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the bed\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47333932/Validation_47333932_47333932_84153.201.png", "Validation_47333932/Validation_47333932_47333932_84155.201.png", "Validation_47333932/Validation_47333932_47333932_84157.200.png", "Validation_47333932/Validation_47333932_47333932_84159.199.png", "Validation_47333932/Validation_47333932_47333932_84161.198.png" ], "turn_id": 10, "scan_id": "Validation_47333932", "origin_question": "Which direction is the warm brown wooden table with the smooth surface in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the warm brown wooden table with the smooth surface in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_45260857/Validation_45260857_45260857_497.701.png", "Validation_45260857/Validation_45260857_45260857_499.683.png", "Validation_45260857/Validation_45260857_45260857_501.682.png", "Validation_45260857/Validation_45260857_45260857_503.682.png" ], "turn_id": 1, "scan_id": "Validation_45260857", "origin_question": "Remember, have you seen any chair(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any chair(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_45260857/Validation_45260857_45260857_505.681.png", "Validation_45260857/Validation_45260857_45260857_507.680.png", "Validation_45260857/Validation_45260857_45260857_509.679.png", "Validation_45260857/Validation_45260857_45260857_511.678.png" ], "turn_id": 2, "scan_id": "Validation_45260857", "origin_question": "In which turn were you farthest from \"the vintage metal-framed bed with the cream color scheme\"? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the vintage metal-framed bed with the cream color scheme\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_45260857/Validation_45260857_45260857_513.694.png", "Validation_45260857/Validation_45260857_45260857_515.693.png", "Validation_45260857/Validation_45260857_45260857_517.692.png", "Validation_45260857/Validation_45260857_45260857_519.692.png" ], "turn_id": 3, "scan_id": "Validation_45260857", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_45260857/Validation_45260857_45260857_521.691.png", "Validation_45260857/Validation_45260857_45260857_523.690.png", "Validation_45260857/Validation_45260857_45260857_525.689.png", "Validation_45260857/Validation_45260857_45260857_527.688.png" ], "turn_id": 4, "scan_id": "Validation_45260857", "origin_question": "Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "157.98", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_45260857/Validation_45260857_45260857_529.688.png", "Validation_45260857/Validation_45260857_45260857_531.687.png", "Validation_45260857/Validation_45260857_45260857_533.686.png", "Validation_45260857/Validation_45260857_45260857_535.685.png" ], "turn_id": 5, "scan_id": "Validation_45260857", "origin_question": "Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "139.02", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_45260857/Validation_45260857_45260857_537.684.png", "Validation_45260857/Validation_45260857_45260857_539.683.png", "Validation_45260857/Validation_45260857_45260857_541.683.png", "Validation_45260857/Validation_45260857_45260857_543.682.png" ], "turn_id": 6, "scan_id": "Validation_45260857", "origin_question": "Which direction is the fireplace in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the fireplace in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_41125722/Validation_41125722_41125722_4501.534.png", "Validation_41125722/Validation_41125722_41125722_4503.517.png", "Validation_41125722/Validation_41125722_41125722_4505.533.png", "Validation_41125722/Validation_41125722_41125722_4507.532.png", "Validation_41125722/Validation_41125722_41125722_4509.531.png" ], "turn_id": 1, "scan_id": "Validation_41125722", "origin_question": "Remember, have you seen any basket(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any basket(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_41125722/Validation_41125722_41125722_4511.530.png", "Validation_41125722/Validation_41125722_41125722_4513.529.png", "Validation_41125722/Validation_41125722_41125722_4515.529.png", "Validation_41125722/Validation_41125722_41125722_4517.528.png", "Validation_41125722/Validation_41125722_41125722_4519.527.png" ], "turn_id": 2, "scan_id": "Validation_41125722", "origin_question": "When did you first discover \"an irregularly shaped picture in dark blue and cream\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"an irregularly shaped picture in dark blue and cream\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_41125722/Validation_41125722_41125722_4521.526.png", "Validation_41125722/Validation_41125722_41125722_4523.525.png", "Validation_41125722/Validation_41125722_41125722_4525.525.png", "Validation_41125722/Validation_41125722_41125722_4527.524.png", "Validation_41125722/Validation_41125722_41125722_4529.523.png" ], "turn_id": 3, "scan_id": "Validation_41125722", "origin_question": "In which turn did you see both \"an irregularly shaped picture in dark blue and cream\" and \"the rectangular, soft white pillow\" simultaneously? ", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"an irregularly shaped picture in dark blue and cream\" and \"the rectangular, soft white pillow\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_41125722/Validation_41125722_41125722_4531.522.png", "Validation_41125722/Validation_41125722_41125722_4533.521.png", "Validation_41125722/Validation_41125722_41125722_4535.520.png", "Validation_41125722/Validation_41125722_41125722_4537.520.png", "Validation_41125722/Validation_41125722_41125722_4539.519.png" ], "turn_id": 4, "scan_id": "Validation_41125722", "origin_question": "In which turn did you see both \"an irregularly shaped picture in dark blue and cream\" and \"the rectangular bed with soft, neutral-colored fabric\" simultaneously? ", "option": [], "answer": "3", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"an irregularly shaped picture in dark blue and cream\" and \"the rectangular bed with soft, neutral-colored fabric\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_41125722/Validation_41125722_41125722_4541.518.png", "Validation_41125722/Validation_41125722_41125722_4543.517.png", "Validation_41125722/Validation_41125722_41125722_4545.533.png", "Validation_41125722/Validation_41125722_41125722_4547.532.png", "Validation_41125722/Validation_41125722_41125722_4549.531.png" ], "turn_id": 5, "scan_id": "Validation_41125722", "origin_question": "Compared to your position at the end of the 3 turn, are you now closer or farther away from \"an irregularly shaped picture in dark blue and cream\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 3 turn, are you now closer or farther away from \"an irregularly shaped picture in dark blue and cream\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_41125722/Validation_41125722_41125722_4551.531.png", "Validation_41125722/Validation_41125722_41125722_4553.530.png", "Validation_41125722/Validation_41125722_41125722_4555.529.png", "Validation_41125722/Validation_41125722_41125722_4557.528.png", "Validation_41125722/Validation_41125722_41125722_4559.527.png" ], "turn_id": 6, "scan_id": "Validation_41125722", "origin_question": "In which turn were you closest to \"an irregularly shaped picture in dark blue and cream\"? ", "option": [], "answer": "3", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"an irregularly shaped picture in dark blue and cream\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_41125722/Validation_41125722_41125722_4561.527.png", "Validation_41125722/Validation_41125722_41125722_4563.526.png", "Validation_41125722/Validation_41125722_41125722_4565.525.png", "Validation_41125722/Validation_41125722_41125722_4567.524.png", "Validation_41125722/Validation_41125722_41125722_4569.523.png" ], "turn_id": 7, "scan_id": "Validation_41125722", "origin_question": "What will be the first-time appearance order of the following categories: bed, cabinet, towel?", "option": [ "towel, cabinet, bed", "bed, cabinet, towel", "cabinet, towel, bed", "cabinet, bed, towel" ], "answer": "cabinet, bed, towel", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bed, cabinet, towel? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['towel, cabinet, bed', 'bed, cabinet, towel', 'cabinet, towel, bed', 'cabinet, bed, towel'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_41125722/Validation_41125722_41125722_4571.523.png", "Validation_41125722/Validation_41125722_41125722_4573.522.png", "Validation_41125722/Validation_41125722_41125722_4575.521.png", "Validation_41125722/Validation_41125722_41125722_4577.520.png", "Validation_41125722/Validation_41125722_41125722_4579.519.png" ], "turn_id": 8, "scan_id": "Validation_41125722", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "132.33", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_41125722/Validation_41125722_41125722_4581.518.png", "Validation_41125722/Validation_41125722_41125722_4583.518.png", "Validation_41125722/Validation_41125722_41125722_4585.517.png", "Validation_41125722/Validation_41125722_41125722_4587.533.png", "Validation_41125722/Validation_41125722_41125722_4589.532.png" ], "turn_id": 9, "scan_id": "Validation_41125722", "origin_question": "In which turn were you closest to \"the sleek, glossy white door made of composite or wood\"? ", "option": [], "answer": "1", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the sleek, glossy white door made of composite or wood\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_41125722/Validation_41125722_41125722_4591.531.png", "Validation_41125722/Validation_41125722_41125722_4593.530.png", "Validation_41125722/Validation_41125722_41125722_4595.529.png", "Validation_41125722/Validation_41125722_41125722_4597.529.png", "Validation_41125722/Validation_41125722_41125722_4599.528.png" ], "turn_id": 10, "scan_id": "Validation_41125722", "origin_question": "When did you first discover \"the circular silver-gray clock\"? (index of the turn)", "option": [], "answer": "8", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the circular silver-gray clock\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_41254432/Validation_41254432_41254432_96416.333.png", "Validation_41254432/Validation_41254432_41254432_96418.316.png", "Validation_41254432/Validation_41254432_41254432_96420.315.png", "Validation_41254432/Validation_41254432_41254432_96422.314.png", "Validation_41254432/Validation_41254432_41254432_96424.314.png" ], "turn_id": 1, "scan_id": "Validation_41254432", "origin_question": "Remember, have you seen any mailbox(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any mailbox(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_41254432/Validation_41254432_41254432_96426.313.png", "Validation_41254432/Validation_41254432_41254432_96428.312.png", "Validation_41254432/Validation_41254432_41254432_96430.311.png", "Validation_41254432/Validation_41254432_41254432_96432.310.png", "Validation_41254432/Validation_41254432_41254432_96434.310.png" ], "turn_id": 2, "scan_id": "Validation_41254432", "origin_question": "When did you first discover \"the black rectangular speaker with the lighter frame\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the black rectangular speaker with the lighter frame\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_41254432/Validation_41254432_41254432_96436.309.png", "Validation_41254432/Validation_41254432_41254432_96438.308.png", "Validation_41254432/Validation_41254432_41254432_96440.307.png", "Validation_41254432/Validation_41254432_41254432_96442.306.png", "Validation_41254432/Validation_41254432_41254432_96444.306.png" ], "turn_id": 3, "scan_id": "Validation_41254432", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_41254432/Validation_41254432_41254432_96446.305.png", "Validation_41254432/Validation_41254432_41254432_96448.304.png", "Validation_41254432/Validation_41254432_41254432_96450.303.png", "Validation_41254432/Validation_41254432_41254432_96452.302.png", "Validation_41254432/Validation_41254432_41254432_96454.302.png" ], "turn_id": 4, "scan_id": "Validation_41254432", "origin_question": "In which turn were you closest to \"the compact metallic silver microwave with the sleek design\"? ", "option": [], "answer": "4", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the compact metallic silver microwave with the sleek design\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_41254432/Validation_41254432_41254432_96456.318.png", "Validation_41254432/Validation_41254432_41254432_96458.317.png", "Validation_41254432/Validation_41254432_41254432_96460.316.png", "Validation_41254432/Validation_41254432_41254432_96462.315.png", "Validation_41254432/Validation_41254432_41254432_96464.314.png" ], "turn_id": 5, "scan_id": "Validation_41254432", "origin_question": "At the end of which turn were both of \"the rectangular gray plastic bin with the closed lid\" and \"the stainless steel rectangular stove with multiple burners\" on your left side?", "option": [], "answer": "2", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the rectangular gray plastic bin with the closed lid\" and \"the stainless steel rectangular stove with multiple burners\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_41254432/Validation_41254432_41254432_96466.314.png", "Validation_41254432/Validation_41254432_41254432_96468.313.png", "Validation_41254432/Validation_41254432_41254432_96470.312.png", "Validation_41254432/Validation_41254432_41254432_96472.311.png", "Validation_41254432/Validation_41254432_41254432_96474.310.png" ], "turn_id": 6, "scan_id": "Validation_41254432", "origin_question": "Please recall \"the compact stainless steel built-in oven\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.63", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the compact stainless steel built-in oven\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_41254432/Validation_41254432_41254432_96476.310.png", "Validation_41254432/Validation_41254432_41254432_96478.309.png", "Validation_41254432/Validation_41254432_41254432_96480.308.png", "Validation_41254432/Validation_41254432_41254432_96482.307.png", "Validation_41254432/Validation_41254432_41254432_96484.306.png" ], "turn_id": 7, "scan_id": "Validation_41254432", "origin_question": "When did you first discover \"the compact metallic silver microwave with the sleek design\"? (index of the turn)", "option": [], "answer": "4", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the compact metallic silver microwave with the sleek design\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_41254432/Validation_41254432_41254432_96486.306.png", "Validation_41254432/Validation_41254432_41254432_96488.305.png", "Validation_41254432/Validation_41254432_41254432_96490.304.png", "Validation_41254432/Validation_41254432_41254432_96492.303.png", "Validation_41254432/Validation_41254432_41254432_96494.303.png" ], "turn_id": 8, "scan_id": "Validation_41254432", "origin_question": "Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Right", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_41254432/Validation_41254432_41254432_96496.302.png", "Validation_41254432/Validation_41254432_41254432_96498.318.png", "Validation_41254432/Validation_41254432_41254432_96500.317.png", "Validation_41254432/Validation_41254432_41254432_96502.316.png", "Validation_41254432/Validation_41254432_41254432_96504.315.png" ], "turn_id": 9, "scan_id": "Validation_41254432", "origin_question": "Assuming the direction you are facing at the end of the turn 8 is forward, did you move a certain distance forward or backward from that position?", "option": [ "Forward", "Backward" ], "answer": "Forward", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 8 is forward, did you move a certain distance forward or backward from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Forward', 'Backward'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_41254432/Validation_41254432_41254432_96506.314.png", "Validation_41254432/Validation_41254432_41254432_96508.314.png", "Validation_41254432/Validation_41254432_41254432_96510.313.png", "Validation_41254432/Validation_41254432_41254432_96512.312.png", "Validation_41254432/Validation_41254432_41254432_96514.311.png" ], "turn_id": 10, "scan_id": "Validation_41254432", "origin_question": "At the end of which turn were both of \"the sleek black flat-screen monitor with the thin bezel\" and \"the sleek, light-colored wooden door with the metallic handle\" on your left side?", "option": [], "answer": "5", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the sleek black flat-screen monitor with the thin bezel\" and \"the sleek, light-colored wooden door with the metallic handle\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42898538/Validation_42898538_42898538_751950.159.png", "Validation_42898538/Validation_42898538_42898538_751952.142.png", "Validation_42898538/Validation_42898538_42898538_751954.158.png", "Validation_42898538/Validation_42898538_42898538_751956.157.png", "Validation_42898538/Validation_42898538_42898538_751958.156.png" ], "turn_id": 1, "scan_id": "Validation_42898538", "origin_question": "Remember, have you seen any lamp(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any lamp(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_42898538/Validation_42898538_42898538_751960.155.png", "Validation_42898538/Validation_42898538_42898538_751962.154.png", "Validation_42898538/Validation_42898538_42898538_751964.154.png", "Validation_42898538/Validation_42898538_42898538_751966.153.png", "Validation_42898538/Validation_42898538_42898538_751968.152.png" ], "turn_id": 2, "scan_id": "Validation_42898538", "origin_question": "In which turn were you closest to \"the medium-brown wooden door with grain patterns\"? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the medium-brown wooden door with grain patterns\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_42898538/Validation_42898538_42898538_751970.151.png", "Validation_42898538/Validation_42898538_42898538_751972.150.png", "Validation_42898538/Validation_42898538_42898538_751974.150.png", "Validation_42898538/Validation_42898538_42898538_751976.149.png", "Validation_42898538/Validation_42898538_42898538_751978.148.png" ], "turn_id": 3, "scan_id": "Validation_42898538", "origin_question": "Please recall \"the medium-brown wooden door with grain patterns\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.67", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the medium-brown wooden door with grain patterns\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_42898538/Validation_42898538_42898538_751980.147.png", "Validation_42898538/Validation_42898538_42898538_751982.146.png", "Validation_42898538/Validation_42898538_42898538_751984.145.png", "Validation_42898538/Validation_42898538_42898538_751986.145.png", "Validation_42898538/Validation_42898538_42898538_751988.144.png" ], "turn_id": 4, "scan_id": "Validation_42898538", "origin_question": "Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_42898538/Validation_42898538_42898538_751990.143.png", "Validation_42898538/Validation_42898538_42898538_751992.142.png", "Validation_42898538/Validation_42898538_42898538_751994.158.png", "Validation_42898538/Validation_42898538_42898538_751996.157.png", "Validation_42898538/Validation_42898538_42898538_751998.156.png" ], "turn_id": 5, "scan_id": "Validation_42898538", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the dark wood bed with the light, textured comforter\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the dark wood bed with the light, textured comforter\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_42898538/Validation_42898538_42898538_752000.155.png", "Validation_42898538/Validation_42898538_42898538_752002.155.png", "Validation_42898538/Validation_42898538_42898538_752004.154.png", "Validation_42898538/Validation_42898538_42898538_752006.153.png", "Validation_42898538/Validation_42898538_42898538_752008.152.png" ], "turn_id": 6, "scan_id": "Validation_42898538", "origin_question": "Using your orientation at the end of turn 5 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 5 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_42898538/Validation_42898538_42898538_752010.151.png", "Validation_42898538/Validation_42898538_42898538_752012.150.png", "Validation_42898538/Validation_42898538_42898538_752014.150.png", "Validation_42898538/Validation_42898538_42898538_752016.149.png", "Validation_42898538/Validation_42898538_42898538_752018.148.png" ], "turn_id": 7, "scan_id": "Validation_42898538", "origin_question": "In which turn were you closest to \"the large rectangular window with the white frame\"? ", "option": [], "answer": "7", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the large rectangular window with the white frame\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_42898538/Validation_42898538_42898538_752020.147.png", "Validation_42898538/Validation_42898538_42898538_752022.146.png", "Validation_42898538/Validation_42898538_42898538_752024.145.png", "Validation_42898538/Validation_42898538_42898538_752026.145.png", "Validation_42898538/Validation_42898538_42898538_752028.144.png" ], "turn_id": 8, "scan_id": "Validation_42898538", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the dark wood bed with the light, textured comforter\" from your position?", "option": [], "answer": "117.87", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the dark wood bed with the light, textured comforter\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_42898538/Validation_42898538_42898538_752030.143.png", "Validation_42898538/Validation_42898538_42898538_752032.142.png", "Validation_42898538/Validation_42898538_42898538_752034.141.png", "Validation_42898538/Validation_42898538_42898538_752036.141.png", "Validation_42898538/Validation_42898538_42898538_752038.156.png" ], "turn_id": 9, "scan_id": "Validation_42898538", "origin_question": "Which two objects are on the same side of you now? \"the medium-brown wooden door with grain patterns\", \"the warm brown rectangular varnished table\", \"the compact, polished wood stool in light beige\". ", "option": [ "the medium-brown wooden door with grain patterns, the warm brown rectangular varnished table", "the medium-brown wooden door with grain patterns, the compact, polished wood stool in light beige", "the warm brown rectangular varnished table, the compact, polished wood stool in light beige" ], "answer": "the warm brown rectangular varnished table, the compact, polished wood stool in light beige", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the medium-brown wooden door with grain patterns\", \"the warm brown rectangular varnished table\", \"the compact, polished wood stool in light beige\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the medium-brown wooden door with grain patterns, the warm brown rectangular varnished table', 'the medium-brown wooden door with grain patterns, the compact, polished wood stool in light beige', 'the warm brown rectangular varnished table, the compact, polished wood stool in light beige'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_42898538/Validation_42898538_42898538_752040.156.png", "Validation_42898538/Validation_42898538_42898538_752042.155.png", "Validation_42898538/Validation_42898538_42898538_752044.154.png", "Validation_42898538/Validation_42898538_42898538_752046.153.png", "Validation_42898538/Validation_42898538_42898538_752048.152.png" ], "turn_id": 10, "scan_id": "Validation_42898538", "origin_question": "Which direction is the dark wood bed with the light, textured comforter in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the dark wood bed with the light, textured comforter in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47331333/Validation_47331333_47331333_180638.634.png", "Validation_47331333/Validation_47331333_47331333_180640.600.png", "Validation_47331333/Validation_47331333_47331333_180642.599.png", "Validation_47331333/Validation_47331333_47331333_180644.615.png" ], "turn_id": 1, "scan_id": "Validation_47331333", "origin_question": "Remember, have you seen any pool table(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any pool table(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47331333/Validation_47331333_47331333_180646.614.png", "Validation_47331333/Validation_47331333_47331333_180648.614.png", "Validation_47331333/Validation_47331333_47331333_180650.613.png", "Validation_47331333/Validation_47331333_47331333_180652.612.png" ], "turn_id": 2, "scan_id": "Validation_47331333", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the classic rectangular wooden door with the medium-brown finish\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the classic rectangular wooden door with the medium-brown finish\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_47331333/Validation_47331333_47331333_180654.611.png", "Validation_47331333/Validation_47331333_47331333_180656.610.png", "Validation_47331333/Validation_47331333_47331333_180658.609.png", "Validation_47331333/Validation_47331333_47331333_180660.609.png" ], "turn_id": 3, "scan_id": "Validation_47331333", "origin_question": "In which turn did you see both \"the muted purple bed with the contrasting throw blanket\" and \"the soft, light yellow rectangular pillow\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the muted purple bed with the contrasting throw blanket\" and \"the soft, light yellow rectangular pillow\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47331333/Validation_47331333_47331333_180662.608.png", "Validation_47331333/Validation_47331333_47331333_180664.607.png", "Validation_47331333/Validation_47331333_47331333_180666.606.png", "Validation_47331333/Validation_47331333_47331333_180668.605.png" ], "turn_id": 4, "scan_id": "Validation_47331333", "origin_question": "", "option": [], "answer": "", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_47331333/Validation_47331333_47331333_180670.605.png", "Validation_47331333/Validation_47331333_47331333_180672.604.png", "Validation_47331333/Validation_47331333_47331333_180674.603.png", "Validation_47331333/Validation_47331333_47331333_180676.602.png" ], "turn_id": 5, "scan_id": "Validation_47331333", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "1.73", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47331333/Validation_47331333_47331333_180678.601.png", "Validation_47331333/Validation_47331333_47331333_180680.601.png", "Validation_47331333/Validation_47331333_47331333_180682.600.png", "Validation_47331333/Validation_47331333_47331333_180684.616.png" ], "turn_id": 6, "scan_id": "Validation_47331333", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the printer\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the printer\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_47331333/Validation_47331333_47331333_180686.615.png", "Validation_47331333/Validation_47331333_47331333_180688.614.png", "Validation_47331333/Validation_47331333_47331333_180690.613.png", "Validation_47331333/Validation_47331333_47331333_180692.612.png" ], "turn_id": 7, "scan_id": "Validation_47331333", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the muted purple bed with the contrasting throw blanket\" from your position?", "option": [], "answer": "172.89", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the muted purple bed with the contrasting throw blanket\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42446137/Validation_42446137_42446137_79049.624.png", "Validation_42446137/Validation_42446137_42446137_79051.623.png", "Validation_42446137/Validation_42446137_42446137_79053.622.png", "Validation_42446137/Validation_42446137_42446137_79055.621.png" ], "turn_id": 1, "scan_id": "Validation_42446137", "origin_question": "Remember, have you seen any mirror(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any mirror(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_42446137/Validation_42446137_42446137_79057.620.png", "Validation_42446137/Validation_42446137_42446137_79059.620.png", "Validation_42446137/Validation_42446137_42446137_79061.619.png", "Validation_42446137/Validation_42446137_42446137_79063.618.png" ], "turn_id": 2, "scan_id": "Validation_42446137", "origin_question": "At the end of which turn were both of \"the modern rectangular light wood table\" and \"the sleek, black rectangular monitor with the thin bezel\" on your right side?", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the modern rectangular light wood table\" and \"the sleek, black rectangular monitor with the thin bezel\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_42446137/Validation_42446137_42446137_79065.617.png", "Validation_42446137/Validation_42446137_42446137_79067.616.png", "Validation_42446137/Validation_42446137_42446137_79069.616.png", "Validation_42446137/Validation_42446137_42446137_79071.615.png" ], "turn_id": 3, "scan_id": "Validation_42446137", "origin_question": "", "option": [], "answer": "", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_42446137/Validation_42446137_42446137_79073.614.png", "Validation_42446137/Validation_42446137_42446137_79075.613.png", "Validation_42446137/Validation_42446137_42446137_79077.612.png", "Validation_42446137/Validation_42446137_42446137_79079.612.png" ], "turn_id": 4, "scan_id": "Validation_42446137", "origin_question": "Is the sleek, black rectangular monitor with the thin bezel to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the sleek, black rectangular monitor with the thin bezel to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_42446137/Validation_42446137_42446137_79081.611.png", "Validation_42446137/Validation_42446137_42446137_79083.610.png", "Validation_42446137/Validation_42446137_42446137_79085.609.png", "Validation_42446137/Validation_42446137_42446137_79087.609.png" ], "turn_id": 5, "scan_id": "Validation_42446137", "origin_question": "Please recall \"the rectangular beige pillow with vintage print\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.97", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the rectangular beige pillow with vintage print\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_42446137/Validation_42446137_42446137_79089.608.png", "Validation_42446137/Validation_42446137_42446137_79091.624.png", "Validation_42446137/Validation_42446137_42446137_79093.623.png", "Validation_42446137/Validation_42446137_42446137_79095.622.png" ], "turn_id": 6, "scan_id": "Validation_42446137", "origin_question": "When did you first discover \"the rectangular beige pillow with vintage print\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the rectangular beige pillow with vintage print\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42446137/Validation_42446137_42446137_79097.621.png", "Validation_42446137/Validation_42446137_42446137_79099.620.png", "Validation_42446137/Validation_42446137_42446137_79101.620.png", "Validation_42446137/Validation_42446137_42446137_79103.619.png" ], "turn_id": 7, "scan_id": "Validation_42446137", "origin_question": "Remember, have you seen any bottle(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any bottle(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "Validation_42446137/Validation_42446137_42446137_79105.618.png", "Validation_42446137/Validation_42446137_42446137_79107.617.png", "Validation_42446137/Validation_42446137_42446137_79109.616.png", "Validation_42446137/Validation_42446137_42446137_79111.616.png" ], "turn_id": 8, "scan_id": "Validation_42446137", "origin_question": "Among these three objects(\"the medium-sized rectangular mirror with the colored frame\",\"an elegant dark brown wooden cabinet with the rectangular shape\",\"the classic wooden shelf in rich brown color\" ), which one is the farthest from you?", "option": [ "the medium-sized rectangular mirror with the colored frame", "an elegant dark brown wooden cabinet with the rectangular shape", "the classic wooden shelf in rich brown color" ], "answer": "the medium-sized rectangular mirror with the colored frame", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects(\"the medium-sized rectangular mirror with the colored frame\",\"an elegant dark brown wooden cabinet with the rectangular shape\",\"the classic wooden shelf in rich brown color\" ), which one is the farthest from you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the medium-sized rectangular mirror with the colored frame', 'an elegant dark brown wooden cabinet with the rectangular shape', 'the classic wooden shelf in rich brown color'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_41125760/Validation_41125760_41125760_5344.916.png", "Validation_41125760/Validation_41125760_41125760_5346.899.png", "Validation_41125760/Validation_41125760_41125760_5348.898.png", "Validation_41125760/Validation_41125760_41125760_5350.897.png", "Validation_41125760/Validation_41125760_41125760_5352.896.png" ], "turn_id": 1, "scan_id": "Validation_41125760", "origin_question": "Remember, have you seen any salt(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any salt(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_41125760/Validation_41125760_41125760_5354.896.png", "Validation_41125760/Validation_41125760_41125760_5356.911.png", "Validation_41125760/Validation_41125760_41125760_5358.911.png", "Validation_41125760/Validation_41125760_41125760_5360.910.png", "Validation_41125760/Validation_41125760_41125760_5362.909.png" ], "turn_id": 2, "scan_id": "Validation_41125760", "origin_question": "When did you first discover \"the modern wall-mounted shelf with vertical bars\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the modern wall-mounted shelf with vertical bars\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_41125760/Validation_41125760_41125760_5364.908.png", "Validation_41125760/Validation_41125760_41125760_5366.907.png", "Validation_41125760/Validation_41125760_41125760_5368.907.png", "Validation_41125760/Validation_41125760_41125760_5370.906.png", "Validation_41125760/Validation_41125760_41125760_5372.905.png" ], "turn_id": 3, "scan_id": "Validation_41125760", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "33.43", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_41125760/Validation_41125760_41125760_5374.904.png", "Validation_41125760/Validation_41125760_41125760_5376.903.png", "Validation_41125760/Validation_41125760_41125760_5378.902.png", "Validation_41125760/Validation_41125760_41125760_5380.902.png", "Validation_41125760/Validation_41125760_41125760_5382.901.png" ], "turn_id": 4, "scan_id": "Validation_41125760", "origin_question": "Which direction is the modern wall-mounted shelf with vertical bars in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the modern wall-mounted shelf with vertical bars in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_41125760/Validation_41125760_41125760_5384.900.png", "Validation_41125760/Validation_41125760_41125760_5386.899.png", "Validation_41125760/Validation_41125760_41125760_5388.898.png", "Validation_41125760/Validation_41125760_41125760_5390.898.png", "Validation_41125760/Validation_41125760_41125760_5392.897.png" ], "turn_id": 5, "scan_id": "Validation_41125760", "origin_question": "Please recall \"the large, soft white cotton towel\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.59", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the large, soft white cotton towel\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_41125760/Validation_41125760_41125760_5394.896.png", "Validation_41125760/Validation_41125760_41125760_5396.912.png", "Validation_41125760/Validation_41125760_41125760_5398.911.png", "Validation_41125760/Validation_41125760_41125760_5400.910.png", "Validation_41125760/Validation_41125760_41125760_5402.909.png" ], "turn_id": 6, "scan_id": "Validation_41125760", "origin_question": "Please recall \"the modern wall-mounted shelf with vertical bars\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.75", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the modern wall-mounted shelf with vertical bars\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_41125760/Validation_41125760_41125760_5404.908.png", "Validation_41125760/Validation_41125760_41125760_5406.908.png", "Validation_41125760/Validation_41125760_41125760_5408.907.png", "Validation_41125760/Validation_41125760_41125760_5410.906.png", "Validation_41125760/Validation_41125760_41125760_5412.905.png" ], "turn_id": 7, "scan_id": "Validation_41125760", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_41125760/Validation_41125760_41125760_5414.904.png", "Validation_41125760/Validation_41125760_41125760_5416.904.png", "Validation_41125760/Validation_41125760_41125760_5418.903.png", "Validation_41125760/Validation_41125760_41125760_5420.902.png", "Validation_41125760/Validation_41125760_41125760_5422.901.png" ], "turn_id": 8, "scan_id": "Validation_41125760", "origin_question": "Is the rectangular sleek silver-gray mirror to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the rectangular sleek silver-gray mirror to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47333904/Validation_47333904_47333904_81402.393.png", "Validation_47333904/Validation_47333904_47333904_81404.375.png", "Validation_47333904/Validation_47333904_47333904_81406.375.png", "Validation_47333904/Validation_47333904_47333904_81408.374.png", "Validation_47333904/Validation_47333904_47333904_81410.373.png" ], "turn_id": 1, "scan_id": "Validation_47333904", "origin_question": "Remember, have you seen any window(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any window(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47333904/Validation_47333904_47333904_81412.372.png", "Validation_47333904/Validation_47333904_47333904_81414.371.png", "Validation_47333904/Validation_47333904_47333904_81416.371.png", "Validation_47333904/Validation_47333904_47333904_81418.370.png", "Validation_47333904/Validation_47333904_47333904_81420.369.png" ], "turn_id": 2, "scan_id": "Validation_47333904", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the medium-sized white ceramic flowerpot with the glossy finish\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the medium-sized white ceramic flowerpot with the glossy finish\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_47333904/Validation_47333904_47333904_81422.368.png", "Validation_47333904/Validation_47333904_47333904_81424.367.png", "Validation_47333904/Validation_47333904_47333904_81426.367.png", "Validation_47333904/Validation_47333904_47333904_81428.366.png", "Validation_47333904/Validation_47333904_47333904_81430.365.png" ], "turn_id": 3, "scan_id": "Validation_47333904", "origin_question": "At the end of which turn were both of \"the sturdy rectangular white table with the clean finish\" and \"the sink\" on your right side?", "option": [], "answer": "2", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the sturdy rectangular white table with the clean finish\" and \"the sink\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47333904/Validation_47333904_47333904_81432.364.png", "Validation_47333904/Validation_47333904_47333904_81434.363.png", "Validation_47333904/Validation_47333904_47333904_81436.363.png", "Validation_47333904/Validation_47333904_47333904_81438.378.png", "Validation_47333904/Validation_47333904_47333904_81440.378.png" ], "turn_id": 4, "scan_id": "Validation_47333904", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the large rectangular window with the white frame\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the large rectangular window with the white frame\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "Validation_47333904/Validation_47333904_47333904_81442.377.png", "Validation_47333904/Validation_47333904_47333904_81444.376.png", "Validation_47333904/Validation_47333904_47333904_81446.375.png", "Validation_47333904/Validation_47333904_47333904_81448.374.png", "Validation_47333904/Validation_47333904_47333904_81450.373.png" ], "turn_id": 5, "scan_id": "Validation_47333904", "origin_question": "Among these three objects(\"the sturdy rectangular white table with the clean finish\",\"the large rectangular window with the white frame\",\"the modern light with the cone-shaped light fixture\" ), which one is the closest to you?", "option": [ "the sturdy rectangular white table with the clean finish", "the large rectangular window with the white frame", "the modern light with the cone-shaped light fixture" ], "answer": "the large rectangular window with the white frame", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the sturdy rectangular white table with the clean finish\",\"the large rectangular window with the white frame\",\"the modern light with the cone-shaped light fixture\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the sturdy rectangular white table with the clean finish', 'the large rectangular window with the white frame', 'the modern light with the cone-shaped light fixture'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_47333904/Validation_47333904_47333904_81452.373.png", "Validation_47333904/Validation_47333904_47333904_81454.372.png", "Validation_47333904/Validation_47333904_47333904_81456.371.png", "Validation_47333904/Validation_47333904_47333904_81458.370.png", "Validation_47333904/Validation_47333904_47333904_81460.369.png" ], "turn_id": 6, "scan_id": "Validation_47333904", "origin_question": "Is the sturdy rectangular white table with the clean finish to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the sturdy rectangular white table with the clean finish to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_47333904/Validation_47333904_47333904_81462.368.png", "Validation_47333904/Validation_47333904_47333904_81464.368.png", "Validation_47333904/Validation_47333904_47333904_81466.367.png", "Validation_47333904/Validation_47333904_47333904_81468.366.png", "Validation_47333904/Validation_47333904_47333904_81470.365.png" ], "turn_id": 7, "scan_id": "Validation_47333904", "origin_question": "Using your orientation at the end of turn 6 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 6 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_47333904/Validation_47333904_47333904_81472.364.png", "Validation_47333904/Validation_47333904_47333904_81474.364.png", "Validation_47333904/Validation_47333904_47333904_81476.363.png", "Validation_47333904/Validation_47333904_47333904_81478.362.png", "Validation_47333904/Validation_47333904_47333904_81480.361.png" ], "turn_id": 8, "scan_id": "Validation_47333904", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47331963/Validation_47331963_47331963_167609.759.png", "Validation_47331963/Validation_47331963_47331963_167611.742.png", "Validation_47331963/Validation_47331963_47331963_167613.741.png", "Validation_47331963/Validation_47331963_47331963_167615.740.png", "Validation_47331963/Validation_47331963_47331963_167617.739.png" ], "turn_id": 1, "scan_id": "Validation_47331963", "origin_question": "Remember, have you seen any cap(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any cap(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47331963/Validation_47331963_47331963_167619.739.png", "Validation_47331963/Validation_47331963_47331963_167621.738.png", "Validation_47331963/Validation_47331963_47331963_167623.737.png", "Validation_47331963/Validation_47331963_47331963_167625.736.png", "Validation_47331963/Validation_47331963_47331963_167627.736.png" ], "turn_id": 2, "scan_id": "Validation_47331963", "origin_question": "Remember, have you seen any sack(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any sack(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_47331963/Validation_47331963_47331963_167629.751.png", "Validation_47331963/Validation_47331963_47331963_167631.750.png", "Validation_47331963/Validation_47331963_47331963_167633.750.png", "Validation_47331963/Validation_47331963_47331963_167635.749.png", "Validation_47331963/Validation_47331963_47331963_167637.748.png" ], "turn_id": 3, "scan_id": "Validation_47331963", "origin_question": "When did you first discover \"the stainless steel vertical refrigerator with adjustable shelves\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the stainless steel vertical refrigerator with adjustable shelves\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_47331963/Validation_47331963_47331963_167639.747.png", "Validation_47331963/Validation_47331963_47331963_167641.746.png", "Validation_47331963/Validation_47331963_47331963_167643.746.png", "Validation_47331963/Validation_47331963_47331963_167645.745.png", "Validation_47331963/Validation_47331963_47331963_167647.744.png" ], "turn_id": 4, "scan_id": "Validation_47331963", "origin_question": "By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the stainless steel vertical refrigerator with adjustable shelves\", Y: \"the oven\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "farther away from both X and Y", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the stainless steel vertical refrigerator with adjustable shelves\", Y: \"the oven\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_47331963/Validation_47331963_47331963_167649.743.png", "Validation_47331963/Validation_47331963_47331963_167651.742.png", "Validation_47331963/Validation_47331963_47331963_167653.742.png", "Validation_47331963/Validation_47331963_47331963_167655.741.png", "Validation_47331963/Validation_47331963_47331963_167657.740.png" ], "turn_id": 5, "scan_id": "Validation_47331963", "origin_question": "Please recall \"the stainless steel vertical refrigerator with adjustable shelves\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.73", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the stainless steel vertical refrigerator with adjustable shelves\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_47331963/Validation_47331963_47331963_167659.739.png", "Validation_47331963/Validation_47331963_47331963_167661.738.png", "Validation_47331963/Validation_47331963_47331963_167663.737.png", "Validation_47331963/Validation_47331963_47331963_167665.737.png", "Validation_47331963/Validation_47331963_47331963_167667.736.png" ], "turn_id": 6, "scan_id": "Validation_47331963", "origin_question": "What will be the first-time appearance order of the following categories: oven, refrigerator, window?", "option": [ "refrigerator, oven, window", "window, oven, refrigerator", "oven, refrigerator, window", "refrigerator, window, oven" ], "answer": "refrigerator, oven, window", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: oven, refrigerator, window? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['refrigerator, oven, window', 'window, oven, refrigerator', 'oven, refrigerator, window', 'refrigerator, window, oven'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_47331963/Validation_47331963_47331963_167669.752.png", "Validation_47331963/Validation_47331963_47331963_167671.751.png", "Validation_47331963/Validation_47331963_47331963_167673.750.png", "Validation_47331963/Validation_47331963_47331963_167675.749.png", "Validation_47331963/Validation_47331963_47331963_167677.748.png" ], "turn_id": 7, "scan_id": "Validation_47331963", "origin_question": "Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance forward or backward from that position?", "option": [ "Forward", "Backward" ], "answer": "Backward", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance forward or backward from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Forward', 'Backward'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_47331963/Validation_47331963_47331963_167679.748.png", "Validation_47331963/Validation_47331963_47331963_167681.747.png", "Validation_47331963/Validation_47331963_47331963_167683.746.png", "Validation_47331963/Validation_47331963_47331963_167685.745.png", "Validation_47331963/Validation_47331963_47331963_167687.745.png" ], "turn_id": 8, "scan_id": "Validation_47331963", "origin_question": "In which turn did you see both \"the sleek stainless steel dishwasher with modern design\" and \"the sleek rectangular gray shelf\" simultaneously? ", "option": [], "answer": "8", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the sleek stainless steel dishwasher with modern design\" and \"the sleek rectangular gray shelf\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_47331963/Validation_47331963_47331963_167689.744.png", "Validation_47331963/Validation_47331963_47331963_167691.743.png", "Validation_47331963/Validation_47331963_47331963_167693.742.png", "Validation_47331963/Validation_47331963_47331963_167695.741.png", "Validation_47331963/Validation_47331963_47331963_167697.741.png" ], "turn_id": 9, "scan_id": "Validation_47331963", "origin_question": "Assuming the direction you are facing at the end of the turn 7 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Right", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 7 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_47331963/Validation_47331963_47331963_167699.740.png", "Validation_47331963/Validation_47331963_47331963_167701.739.png", "Validation_47331963/Validation_47331963_47331963_167703.738.png", "Validation_47331963/Validation_47331963_47331963_167705.737.png", "Validation_47331963/Validation_47331963_47331963_167707.736.png" ], "turn_id": 10, "scan_id": "Validation_47331963", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the polished stainless steel rectangular sink\" from your position?", "option": [], "answer": "168.77", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the polished stainless steel rectangular sink\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_45260920/Validation_45260920_45260920_2467.435.png", "Validation_45260920/Validation_45260920_45260920_2469.418.png", "Validation_45260920/Validation_45260920_45260920_2471.417.png", "Validation_45260920/Validation_45260920_45260920_2473.416.png", "Validation_45260920/Validation_45260920_45260920_2475.432.png" ], "turn_id": 1, "scan_id": "Validation_45260920", "origin_question": "Remember, have you seen any fireplace(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any fireplace(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_45260920/Validation_45260920_45260920_2477.431.png", "Validation_45260920/Validation_45260920_45260920_2479.430.png", "Validation_45260920/Validation_45260920_45260920_2481.430.png", "Validation_45260920/Validation_45260920_45260920_2483.429.png", "Validation_45260920/Validation_45260920_45260920_2485.428.png" ], "turn_id": 2, "scan_id": "Validation_45260920", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_45260920/Validation_45260920_45260920_2487.427.png", "Validation_45260920/Validation_45260920_45260920_2489.427.png", "Validation_45260920/Validation_45260920_45260920_2491.426.png", "Validation_45260920/Validation_45260920_45260920_2493.425.png", "Validation_45260920/Validation_45260920_45260920_2495.424.png" ], "turn_id": 3, "scan_id": "Validation_45260920", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "1.03", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_45260920/Validation_45260920_45260920_2497.423.png", "Validation_45260920/Validation_45260920_45260920_2499.423.png", "Validation_45260920/Validation_45260920_45260920_2501.422.png", "Validation_45260920/Validation_45260920_45260920_2503.421.png", "Validation_45260920/Validation_45260920_45260920_2505.420.png" ], "turn_id": 4, "scan_id": "Validation_45260920", "origin_question": "When did you first discover \"the rectangular dark cast iron fireplace with floral tiles\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the rectangular dark cast iron fireplace with floral tiles\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_45260920/Validation_45260920_45260920_2507.419.png", "Validation_45260920/Validation_45260920_45260920_2509.419.png", "Validation_45260920/Validation_45260920_45260920_2511.418.png", "Validation_45260920/Validation_45260920_45260920_2513.417.png", "Validation_45260920/Validation_45260920_45260920_2515.416.png" ], "turn_id": 5, "scan_id": "Validation_45260920", "origin_question": "What will be the first-time appearance order of the following categories: cabinet, couch, fireplace?", "option": [ "fireplace, cabinet, couch", "couch, cabinet, fireplace", "fireplace, couch, cabinet", "cabinet, couch, fireplace" ], "answer": "fireplace, cabinet, couch", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: cabinet, couch, fireplace? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['fireplace, cabinet, couch', 'couch, cabinet, fireplace', 'fireplace, couch, cabinet', 'cabinet, couch, fireplace'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_45260920/Validation_45260920_45260920_2517.432.png", "Validation_45260920/Validation_45260920_45260920_2519.431.png", "Validation_45260920/Validation_45260920_45260920_2521.431.png", "Validation_45260920/Validation_45260920_45260920_2523.430.png", "Validation_45260920/Validation_45260920_45260920_2525.429.png" ], "turn_id": 6, "scan_id": "Validation_45260920", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the small, sleek monitor with the dark plastic frame\" from your position?", "option": [], "answer": "66.71", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the small, sleek monitor with the dark plastic frame\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_45260920/Validation_45260920_45260920_2527.428.png", "Validation_45260920/Validation_45260920_45260920_2529.427.png", "Validation_45260920/Validation_45260920_45260920_2531.427.png", "Validation_45260920/Validation_45260920_45260920_2533.426.png", "Validation_45260920/Validation_45260920_45260920_2535.425.png" ], "turn_id": 7, "scan_id": "Validation_45260920", "origin_question": "In which turn did you see both \"the rectangular dark cast iron fireplace with floral tiles\" and \"the classic couch with the high, curved backrest\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the rectangular dark cast iron fireplace with floral tiles\" and \"the classic couch with the high, curved backrest\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_45260920/Validation_45260920_45260920_2537.424.png", "Validation_45260920/Validation_45260920_45260920_2539.423.png", "Validation_45260920/Validation_45260920_45260920_2541.423.png", "Validation_45260920/Validation_45260920_45260920_2543.422.png", "Validation_45260920/Validation_45260920_45260920_2545.421.png" ], "turn_id": 8, "scan_id": "Validation_45260920", "origin_question": "How far is your current position from where you were at the end of the turn 4? (in meters)", "option": [], "answer": "1.4", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 4? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "Validation_42446522/Validation_42446522_42446522_197315.890.png", "Validation_42446522/Validation_42446522_42446522_197317.873.png", "Validation_42446522/Validation_42446522_42446522_197319.872.png", "Validation_42446522/Validation_42446522_42446522_197321.871.png", "Validation_42446522/Validation_42446522_42446522_197323.870.png" ], "turn_id": 1, "scan_id": "Validation_42446522", "origin_question": "Remember, how many chair(s) have you seen so far? ", "option": [], "answer": "2", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many chair(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_42446522/Validation_42446522_42446522_197325.870.png", "Validation_42446522/Validation_42446522_42446522_197327.885.png", "Validation_42446522/Validation_42446522_42446522_197329.884.png", "Validation_42446522/Validation_42446522_42446522_197331.884.png", "Validation_42446522/Validation_42446522_42446522_197333.883.png" ], "turn_id": 2, "scan_id": "Validation_42446522", "origin_question": "Which direction is the sleek, rectangular black monitor with the thin bezel in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the sleek, rectangular black monitor with the thin bezel in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_42446522/Validation_42446522_42446522_197335.882.png", "Validation_42446522/Validation_42446522_42446522_197337.881.png", "Validation_42446522/Validation_42446522_42446522_197339.880.png", "Validation_42446522/Validation_42446522_42446522_197341.880.png", "Validation_42446522/Validation_42446522_42446522_197343.879.png" ], "turn_id": 3, "scan_id": "Validation_42446522", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "83.23", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_42446522/Validation_42446522_42446522_197345.878.png", "Validation_42446522/Validation_42446522_42446522_197347.877.png", "Validation_42446522/Validation_42446522_42446522_197349.876.png", "Validation_42446522/Validation_42446522_42446522_197351.875.png", "Validation_42446522/Validation_42446522_42446522_197353.875.png" ], "turn_id": 4, "scan_id": "Validation_42446522", "origin_question": "Please recall \"the rectangular white pvc window with the minimalist design\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "4.46", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the rectangular white pvc window with the minimalist design\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_42446522/Validation_42446522_42446522_197355.874.png", "Validation_42446522/Validation_42446522_42446522_197357.873.png", "Validation_42446522/Validation_42446522_42446522_197359.872.png", "Validation_42446522/Validation_42446522_42446522_197361.871.png", "Validation_42446522/Validation_42446522_42446522_197363.871.png" ], "turn_id": 5, "scan_id": "Validation_42446522", "origin_question": "Which two objects are on the same side of you now? \"the tall rectangular deep red shelf\", \"the simple, neutral-toned vertical door\", \"the rectangular white box made of robust material\". ", "option": [ "the tall rectangular deep red shelf, the simple, neutral-toned vertical door", "the tall rectangular deep red shelf, the rectangular white box made of robust material", "the simple, neutral-toned vertical door, the rectangular white box made of robust material" ], "answer": "the simple, neutral-toned vertical door, the rectangular white box made of robust material", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the tall rectangular deep red shelf\", \"the simple, neutral-toned vertical door\", \"the rectangular white box made of robust material\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the tall rectangular deep red shelf, the simple, neutral-toned vertical door', 'the tall rectangular deep red shelf, the rectangular white box made of robust material', 'the simple, neutral-toned vertical door, the rectangular white box made of robust material'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_42446522/Validation_42446522_42446522_197365.870.png", "Validation_42446522/Validation_42446522_42446522_197367.886.png", "Validation_42446522/Validation_42446522_42446522_197369.885.png", "Validation_42446522/Validation_42446522_42446522_197371.884.png", "Validation_42446522/Validation_42446522_42446522_197373.883.png" ], "turn_id": 6, "scan_id": "Validation_42446522", "origin_question": "Is the cylindrical gray plastic bin with the secure lid to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the cylindrical gray plastic bin with the secure lid to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "Validation_42446522/Validation_42446522_42446522_197375.882.png", "Validation_42446522/Validation_42446522_42446522_197377.881.png", "Validation_42446522/Validation_42446522_42446522_197379.881.png", "Validation_42446522/Validation_42446522_42446522_197381.880.png", "Validation_42446522/Validation_42446522_42446522_197383.879.png" ], "turn_id": 7, "scan_id": "Validation_42446522", "origin_question": "Among these three objects(\"the sleek, rectangular black monitor with the thin bezel\",\"the simple, neutral-toned vertical door\",\"the rectangular white box made of robust material\" ), which one is the closest to you?", "option": [ "the sleek, rectangular black monitor with the thin bezel", "the simple, neutral-toned vertical door", "the rectangular white box made of robust material" ], "answer": "the sleek, rectangular black monitor with the thin bezel", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the sleek, rectangular black monitor with the thin bezel\",\"the simple, neutral-toned vertical door\",\"the rectangular white box made of robust material\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the sleek, rectangular black monitor with the thin bezel', 'the simple, neutral-toned vertical door', 'the rectangular white box made of robust material'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_42446522/Validation_42446522_42446522_197385.878.png", "Validation_42446522/Validation_42446522_42446522_197387.877.png", "Validation_42446522/Validation_42446522_42446522_197389.877.png", "Validation_42446522/Validation_42446522_42446522_197391.876.png", "Validation_42446522/Validation_42446522_42446522_197393.875.png" ], "turn_id": 8, "scan_id": "Validation_42446522", "origin_question": "Which two objects are on the same side of you now? \"the sleek, rectangular black monitor with the thin bezel\", \"the medium-sized, bright green nylon bag\", \"the tall rectangular deep red shelf\". ", "option": [ "the sleek, rectangular black monitor with the thin bezel, the medium-sized, bright green nylon bag", "the sleek, rectangular black monitor with the thin bezel, the tall rectangular deep red shelf", "the medium-sized, bright green nylon bag, the tall rectangular deep red shelf" ], "answer": "the sleek, rectangular black monitor with the thin bezel, the medium-sized, bright green nylon bag", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the sleek, rectangular black monitor with the thin bezel\", \"the medium-sized, bright green nylon bag\", \"the tall rectangular deep red shelf\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the sleek, rectangular black monitor with the thin bezel, the medium-sized, bright green nylon bag', 'the sleek, rectangular black monitor with the thin bezel, the tall rectangular deep red shelf', 'the medium-sized, bright green nylon bag, the tall rectangular deep red shelf'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_42446522/Validation_42446522_42446522_197395.874.png", "Validation_42446522/Validation_42446522_42446522_197397.873.png", "Validation_42446522/Validation_42446522_42446522_197399.872.png", "Validation_42446522/Validation_42446522_42446522_197401.872.png", "Validation_42446522/Validation_42446522_42446522_197403.871.png" ], "turn_id": 9, "scan_id": "Validation_42446522", "origin_question": "How far is your current position from where you were at the end of the turn 8? (in meters)", "option": [], "answer": "1.76", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 8? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47115474/Validation_47115474_47115474_13961.697.png", "Validation_47115474/Validation_47115474_47115474_13963.180.png", "Validation_47115474/Validation_47115474_47115474_13964.679.png", "Validation_47115474/Validation_47115474_47115474_13966.179.png" ], "turn_id": 1, "scan_id": "Validation_47115474", "origin_question": "Remember, have you seen any sack(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any sack(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47115474/Validation_47115474_47115474_13967.678.png", "Validation_47115474/Validation_47115474_47115474_13969.194.png", "Validation_47115474/Validation_47115474_47115474_13970.693.png", "Validation_47115474/Validation_47115474_47115474_13972.193.png" ], "turn_id": 2, "scan_id": "Validation_47115474", "origin_question": "In which turn were you farthest from \"the sleek, oval, glossy white acrylic bathtub\"? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the sleek, oval, glossy white acrylic bathtub\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_47115474/Validation_47115474_47115474_13973.692.png", "Validation_47115474/Validation_47115474_47115474_13975.192.png", "Validation_47115474/Validation_47115474_47115474_13976.691.png", "Validation_47115474/Validation_47115474_47115474_13978.190.png" ], "turn_id": 3, "scan_id": "Validation_47115474", "origin_question": "Please recall \"the white shelf with horizontal bars\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.23", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the white shelf with horizontal bars\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_47115474/Validation_47115474_47115474_13979.690.png", "Validation_47115474/Validation_47115474_47115474_13981.189.png", "Validation_47115474/Validation_47115474_47115474_13982.689.png", "Validation_47115474/Validation_47115474_47115474_13984.188.png" ], "turn_id": 4, "scan_id": "Validation_47115474", "origin_question": "Which two objects are on the same side of you now? \"the sleek, oval, glossy white acrylic bathtub\", \"the sleek, oval white ceramic sink\", \"the small white cylindrical plastic bin\". ", "option": [ "the sleek, oval, glossy white acrylic bathtub, the sleek, oval white ceramic sink", "the sleek, oval, glossy white acrylic bathtub, the small white cylindrical plastic bin", "the sleek, oval white ceramic sink, the small white cylindrical plastic bin" ], "answer": "the sleek, oval white ceramic sink, the small white cylindrical plastic bin", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the sleek, oval, glossy white acrylic bathtub\", \"the sleek, oval white ceramic sink\", \"the small white cylindrical plastic bin\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the sleek, oval, glossy white acrylic bathtub, the sleek, oval white ceramic sink', 'the sleek, oval, glossy white acrylic bathtub, the small white cylindrical plastic bin', 'the sleek, oval white ceramic sink, the small white cylindrical plastic bin'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47115474/Validation_47115474_47115474_13985.687.png", "Validation_47115474/Validation_47115474_47115474_13987.187.png", "Validation_47115474/Validation_47115474_47115474_13988.686.png", "Validation_47115474/Validation_47115474_47115474_13990.185.png" ], "turn_id": 5, "scan_id": "Validation_47115474", "origin_question": "Which direction is the rectangular mirror with the sleek, neutral-colored frame in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the rectangular mirror with the sleek, neutral-colored frame in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_47115474/Validation_47115474_47115474_13991.685.png", "Validation_47115474/Validation_47115474_47115474_13993.184.png", "Validation_47115474/Validation_47115474_47115474_13994.683.png", "Validation_47115474/Validation_47115474_47115474_13996.183.png" ], "turn_id": 6, "scan_id": "Validation_47115474", "origin_question": "Please recall \"the sleek, rectangular white toilet made of porcelain\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.79", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the sleek, rectangular white toilet made of porcelain\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47115474/Validation_47115474_47115474_13997.682.png", "Validation_47115474/Validation_47115474_47115474_13999.182.png", "Validation_47115474/Validation_47115474_47115474_14000.681.png", "Validation_47115474/Validation_47115474_47115474_14002.180.png" ], "turn_id": 7, "scan_id": "Validation_47115474", "origin_question": "Which direction is the sleek, oval white ceramic sink in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the sleek, oval white ceramic sink in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_41254398/Validation_41254398_41254398_94959.571.png", "Validation_41254398/Validation_41254398_41254398_94961.571.png", "Validation_41254398/Validation_41254398_41254398_94963.570.png", "Validation_41254398/Validation_41254398_41254398_94965.569.png" ], "turn_id": 1, "scan_id": "Validation_41254398", "origin_question": "Remember, have you seen any frame(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any frame(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_41254398/Validation_41254398_41254398_94967.568.png", "Validation_41254398/Validation_41254398_41254398_94969.568.png", "Validation_41254398/Validation_41254398_41254398_94971.567.png", "Validation_41254398/Validation_41254398_41254398_94973.566.png" ], "turn_id": 2, "scan_id": "Validation_41254398", "origin_question": "In which turn did you see both \"the classic white porcelain toilet with the rounded design\" and \"the white oval porcelain sink on the pedestal\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the classic white porcelain toilet with the rounded design\" and \"the white oval porcelain sink on the pedestal\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_41254398/Validation_41254398_41254398_94975.565.png", "Validation_41254398/Validation_41254398_41254398_94977.564.png", "Validation_41254398/Validation_41254398_41254398_94979.564.png", "Validation_41254398/Validation_41254398_41254398_94981.563.png" ], "turn_id": 3, "scan_id": "Validation_41254398", "origin_question": "Which two objects are on the same side of you now? \"the sleek, rectangular white fiberglass bathtub\", \"the muted light-colored rectangular cotton towel\", \"an elegant off-white wooden door with brass hardware\". ", "option": [ "the sleek, rectangular white fiberglass bathtub, the muted light-colored rectangular cotton towel", "the sleek, rectangular white fiberglass bathtub, an elegant off-white wooden door with brass hardware", "the muted light-colored rectangular cotton towel, an elegant off-white wooden door with brass hardware" ], "answer": "the sleek, rectangular white fiberglass bathtub, the muted light-colored rectangular cotton towel", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the sleek, rectangular white fiberglass bathtub\", \"the muted light-colored rectangular cotton towel\", \"an elegant off-white wooden door with brass hardware\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the sleek, rectangular white fiberglass bathtub, the muted light-colored rectangular cotton towel', 'the sleek, rectangular white fiberglass bathtub, an elegant off-white wooden door with brass hardware', 'the muted light-colored rectangular cotton towel, an elegant off-white wooden door with brass hardware'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_41254398/Validation_41254398_41254398_94983.562.png", "Validation_41254398/Validation_41254398_41254398_94985.561.png", "Validation_41254398/Validation_41254398_41254398_94987.560.png", "Validation_41254398/Validation_41254398_41254398_94989.560.png" ], "turn_id": 4, "scan_id": "Validation_41254398", "origin_question": "At the end of which turn were both of \"an elegant off-white wooden door with brass hardware\" and \"the window\" on your left side?", "option": [], "answer": "2", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"an elegant off-white wooden door with brass hardware\" and \"the window\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_41254398/Validation_41254398_41254398_94991.575.png", "Validation_41254398/Validation_41254398_41254398_94993.575.png", "Validation_41254398/Validation_41254398_41254398_94995.574.png", "Validation_41254398/Validation_41254398_41254398_94997.573.png" ], "turn_id": 5, "scan_id": "Validation_41254398", "origin_question": "Please recall \"the muted light-colored rectangular cotton towel\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.74", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the muted light-colored rectangular cotton towel\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_41254398/Validation_41254398_41254398_94999.572.png", "Validation_41254398/Validation_41254398_41254398_95001.572.png", "Validation_41254398/Validation_41254398_41254398_95003.571.png", "Validation_41254398/Validation_41254398_41254398_95005.570.png" ], "turn_id": 6, "scan_id": "Validation_41254398", "origin_question": "Which direction is the muted light-colored rectangular cotton towel in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the muted light-colored rectangular cotton towel in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_41254398/Validation_41254398_41254398_95007.569.png", "Validation_41254398/Validation_41254398_41254398_95009.568.png", "Validation_41254398/Validation_41254398_41254398_95011.568.png", "Validation_41254398/Validation_41254398_41254398_95013.567.png" ], "turn_id": 7, "scan_id": "Validation_41254398", "origin_question": "Using your orientation at the end of turn 5 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 5 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_41254398/Validation_41254398_41254398_95015.566.png", "Validation_41254398/Validation_41254398_41254398_95017.565.png", "Validation_41254398/Validation_41254398_41254398_95019.564.png", "Validation_41254398/Validation_41254398_41254398_95021.564.png" ], "turn_id": 8, "scan_id": "Validation_41254398", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the window\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the window\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42899699/Validation_42899699_42899699_793077.365.png", "Validation_42899699/Validation_42899699_42899699_793079.348.png", "Validation_42899699/Validation_42899699_42899699_793081.364.png", "Validation_42899699/Validation_42899699_42899699_793083.363.png" ], "turn_id": 1, "scan_id": "Validation_42899699", "origin_question": "Remember, have you seen any statue(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any statue(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_42899699/Validation_42899699_42899699_793085.362.png", "Validation_42899699/Validation_42899699_42899699_793087.361.png", "Validation_42899699/Validation_42899699_42899699_793089.360.png", "Validation_42899699/Validation_42899699_42899699_793091.359.png" ], "turn_id": 2, "scan_id": "Validation_42899699", "origin_question": "Please recall \"an off-white modern rectangular door with the metallic doorknob\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.97", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"an off-white modern rectangular door with the metallic doorknob\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_42899699/Validation_42899699_42899699_793093.359.png", "Validation_42899699/Validation_42899699_42899699_793095.358.png", "Validation_42899699/Validation_42899699_42899699_793097.357.png", "Validation_42899699/Validation_42899699_42899699_793099.356.png" ], "turn_id": 3, "scan_id": "Validation_42899699", "origin_question": "Which direction is the compact metal and wood table in neutral tones in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the compact metal and wood table in neutral tones in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_42899699/Validation_42899699_42899699_793101.355.png", "Validation_42899699/Validation_42899699_42899699_793103.355.png", "Validation_42899699/Validation_42899699_42899699_793105.354.png", "Validation_42899699/Validation_42899699_42899699_793107.353.png" ], "turn_id": 4, "scan_id": "Validation_42899699", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the sleek rectangular white bathtub made of acrylic\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the sleek rectangular white bathtub made of acrylic\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_42899699/Validation_42899699_42899699_793109.352.png", "Validation_42899699/Validation_42899699_42899699_793111.351.png", "Validation_42899699/Validation_42899699_42899699_793113.350.png", "Validation_42899699/Validation_42899699_42899699_793115.350.png" ], "turn_id": 5, "scan_id": "Validation_42899699", "origin_question": "Which two objects are on the same side of you now? \"the cylindrical dark basket made of durable material\", \"the compact metal and wood table in neutral tones\", \"an off-white modern rectangular door with the metallic doorknob\". ", "option": [ "the cylindrical dark basket made of durable material, the compact metal and wood table in neutral tones", "the cylindrical dark basket made of durable material, an off-white modern rectangular door with the metallic doorknob", "the compact metal and wood table in neutral tones, an off-white modern rectangular door with the metallic doorknob" ], "answer": "the compact metal and wood table in neutral tones, an off-white modern rectangular door with the metallic doorknob", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the cylindrical dark basket made of durable material\", \"the compact metal and wood table in neutral tones\", \"an off-white modern rectangular door with the metallic doorknob\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the cylindrical dark basket made of durable material, the compact metal and wood table in neutral tones', 'the cylindrical dark basket made of durable material, an off-white modern rectangular door with the metallic doorknob', 'the compact metal and wood table in neutral tones, an off-white modern rectangular door with the metallic doorknob'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_42899699/Validation_42899699_42899699_793117.349.png", "Validation_42899699/Validation_42899699_42899699_793119.348.png", "Validation_42899699/Validation_42899699_42899699_793121.347.png", "Validation_42899699/Validation_42899699_42899699_793123.363.png" ], "turn_id": 6, "scan_id": "Validation_42899699", "origin_question": "What will be the first-time appearance order of the following categories: basket, cabinet, toilet?", "option": [ "toilet, cabinet, basket", "cabinet, basket, toilet", "cabinet, toilet, basket", "toilet, basket, cabinet" ], "answer": "cabinet, basket, toilet", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: basket, cabinet, toilet? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['toilet, cabinet, basket', 'cabinet, basket, toilet', 'cabinet, toilet, basket', 'toilet, basket, cabinet'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_42899699/Validation_42899699_42899699_793125.362.png", "Validation_42899699/Validation_42899699_42899699_793127.361.png", "Validation_42899699/Validation_42899699_42899699_793129.360.png", "Validation_42899699/Validation_42899699_42899699_793131.360.png" ], "turn_id": 7, "scan_id": "Validation_42899699", "origin_question": "Assuming the direction you are facing at the end of the turn 3 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Right", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 3 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_42899699/Validation_42899699_42899699_793133.359.png", "Validation_42899699/Validation_42899699_42899699_793135.358.png", "Validation_42899699/Validation_42899699_42899699_793137.357.png", "Validation_42899699/Validation_42899699_42899699_793139.356.png" ], "turn_id": 8, "scan_id": "Validation_42899699", "origin_question": "How far is your current position from where you were at the end of the turn 4? (in meters)", "option": [], "answer": "3.6", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 4? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42898862/Validation_42898862_42898862_196312.721.png", "Validation_42898862/Validation_42898862_42898862_196314.720.png", "Validation_42898862/Validation_42898862_42898862_196316.720.png", "Validation_42898862/Validation_42898862_42898862_196318.719.png", "Validation_42898862/Validation_42898862_42898862_196320.718.png" ], "turn_id": 1, "scan_id": "Validation_42898862", "origin_question": "Remember, have you seen any blanket(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any blanket(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_42898862/Validation_42898862_42898862_196322.717.png", "Validation_42898862/Validation_42898862_42898862_196324.716.png", "Validation_42898862/Validation_42898862_42898862_196326.716.png", "Validation_42898862/Validation_42898862_42898862_196328.715.png", "Validation_42898862/Validation_42898862_42898862_196330.714.png" ], "turn_id": 2, "scan_id": "Validation_42898862", "origin_question": "At the end of which turn were both of \"an elegant white ceramic round flowerpot\" and \"the cylindrical lamp with the neutral fabric shade\" on your right side?", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"an elegant white ceramic round flowerpot\" and \"the cylindrical lamp with the neutral fabric shade\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_42898862/Validation_42898862_42898862_196332.713.png", "Validation_42898862/Validation_42898862_42898862_196334.712.png", "Validation_42898862/Validation_42898862_42898862_196336.712.png", "Validation_42898862/Validation_42898862_42898862_196338.728.png", "Validation_42898862/Validation_42898862_42898862_196340.727.png" ], "turn_id": 3, "scan_id": "Validation_42898862", "origin_question": "At the end of which turn were both of \"the light blue rectangular plush cotton pillow\" and \"an elegant white ceramic round flowerpot\" on your right side?", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the light blue rectangular plush cotton pillow\" and \"an elegant white ceramic round flowerpot\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_42898862/Validation_42898862_42898862_196342.726.png", "Validation_42898862/Validation_42898862_42898862_196344.725.png", "Validation_42898862/Validation_42898862_42898862_196346.724.png", "Validation_42898862/Validation_42898862_42898862_196348.724.png", "Validation_42898862/Validation_42898862_42898862_196350.723.png" ], "turn_id": 4, "scan_id": "Validation_42898862", "origin_question": "Please recall \"the cylindrical lamp with the neutral fabric shade\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "5.8", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the cylindrical lamp with the neutral fabric shade\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "Validation_42898862/Validation_42898862_42898862_196352.722.png", "Validation_42898862/Validation_42898862_42898862_196354.721.png", "Validation_42898862/Validation_42898862_42898862_196356.720.png", "Validation_42898862/Validation_42898862_42898862_196358.720.png", "Validation_42898862/Validation_42898862_42898862_196360.719.png" ], "turn_id": 5, "scan_id": "Validation_42898862", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"an elegant white ceramic round flowerpot\"; \"the white ceramic cup in the cylindrical shape\"; \"the rectangular fireplace with the neutral-toned frame\" ", "option": [ "an elegant white ceramic round flowerpot", "the white ceramic cup in the cylindrical shape", "the rectangular fireplace with the neutral-toned frame" ], "answer": "the white ceramic cup in the cylindrical shape", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"an elegant white ceramic round flowerpot\"; \"the white ceramic cup in the cylindrical shape\"; \"the rectangular fireplace with the neutral-toned frame\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['an elegant white ceramic round flowerpot', 'the white ceramic cup in the cylindrical shape', 'the rectangular fireplace with the neutral-toned frame'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_42898862/Validation_42898862_42898862_196362.718.png", "Validation_42898862/Validation_42898862_42898862_196364.717.png", "Validation_42898862/Validation_42898862_42898862_196366.716.png", "Validation_42898862/Validation_42898862_42898862_196368.716.png", "Validation_42898862/Validation_42898862_42898862_196370.715.png" ], "turn_id": 6, "scan_id": "Validation_42898862", "origin_question": "Which two objects are on the same side of you now? \"the rectangular fireplace with the neutral-toned frame\", \"the sleek rectangular monitor with the black bezel\", \"an elegant white ceramic round flowerpot\". ", "option": [ "the rectangular fireplace with the neutral-toned frame, the sleek rectangular monitor with the black bezel", "the rectangular fireplace with the neutral-toned frame, an elegant white ceramic round flowerpot", "the sleek rectangular monitor with the black bezel, an elegant white ceramic round flowerpot" ], "answer": "the rectangular fireplace with the neutral-toned frame, the sleek rectangular monitor with the black bezel", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the rectangular fireplace with the neutral-toned frame\", \"the sleek rectangular monitor with the black bezel\", \"an elegant white ceramic round flowerpot\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular fireplace with the neutral-toned frame, the sleek rectangular monitor with the black bezel', 'the rectangular fireplace with the neutral-toned frame, an elegant white ceramic round flowerpot', 'the sleek rectangular monitor with the black bezel, an elegant white ceramic round flowerpot'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42898862/Validation_42898862_42898862_196372.714.png", "Validation_42898862/Validation_42898862_42898862_196374.713.png", "Validation_42898862/Validation_42898862_42898862_196376.712.png", "Validation_42898862/Validation_42898862_42898862_196378.712.png", "Validation_42898862/Validation_42898862_42898862_196380.728.png" ], "turn_id": 7, "scan_id": "Validation_42898862", "origin_question": "Remember, have you seen any guitar(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any guitar(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_42898862/Validation_42898862_42898862_196382.727.png", "Validation_42898862/Validation_42898862_42898862_196384.726.png", "Validation_42898862/Validation_42898862_42898862_196386.725.png", "Validation_42898862/Validation_42898862_42898862_196388.724.png", "Validation_42898862/Validation_42898862_42898862_196390.724.png" ], "turn_id": 8, "scan_id": "Validation_42898862", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"an elegant white ceramic round flowerpot\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"an elegant white ceramic round flowerpot\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_42898862/Validation_42898862_42898862_196392.723.png", "Validation_42898862/Validation_42898862_42898862_196394.722.png", "Validation_42898862/Validation_42898862_42898862_196396.721.png", "Validation_42898862/Validation_42898862_42898862_196398.720.png", "Validation_42898862/Validation_42898862_42898862_196400.720.png" ], "turn_id": 9, "scan_id": "Validation_42898862", "origin_question": "Which direction is the rectangular fireplace with the neutral-toned frame in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the rectangular fireplace with the neutral-toned frame in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_42898862/Validation_42898862_42898862_196402.719.png", "Validation_42898862/Validation_42898862_42898862_196404.718.png", "Validation_42898862/Validation_42898862_42898862_196406.717.png", "Validation_42898862/Validation_42898862_42898862_196408.717.png", "Validation_42898862/Validation_42898862_42898862_196410.716.png" ], "turn_id": 10, "scan_id": "Validation_42898862", "origin_question": "Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "119.34", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47430468/Validation_47430468_47430468_311454.963.png", "Validation_47430468/Validation_47430468_47430468_311456.946.png", "Validation_47430468/Validation_47430468_47430468_311458.945.png", "Validation_47430468/Validation_47430468_47430468_311460.944.png" ], "turn_id": 1, "scan_id": "Validation_47430468", "origin_question": "Remember, have you seen any teapot(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any teapot(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_47430468/Validation_47430468_47430468_311462.943.png", "Validation_47430468/Validation_47430468_47430468_311464.943.png", "Validation_47430468/Validation_47430468_47430468_311466.942.png", "Validation_47430468/Validation_47430468_47430468_311468.941.png" ], "turn_id": 2, "scan_id": "Validation_47430468", "origin_question": "In which turn did you see both \"the large rectangular window with the dark brown frame\" and \"the deep purple sleek curtain with the minimalist style\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the large rectangular window with the dark brown frame\" and \"the deep purple sleek curtain with the minimalist style\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47430468/Validation_47430468_47430468_311470.940.png", "Validation_47430468/Validation_47430468_47430468_311472.939.png", "Validation_47430468/Validation_47430468_47430468_311474.939.png", "Validation_47430468/Validation_47430468_47430468_311476.938.png" ], "turn_id": 3, "scan_id": "Validation_47430468", "origin_question": "", "option": [], "answer": "", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47430468/Validation_47430468_47430468_311478.937.png", "Validation_47430468/Validation_47430468_47430468_311480.936.png", "Validation_47430468/Validation_47430468_47430468_311482.935.png", "Validation_47430468/Validation_47430468_47430468_311484.934.png" ], "turn_id": 4, "scan_id": "Validation_47430468", "origin_question": "", "option": [], "answer": "", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_47430468/Validation_47430468_47430468_311486.934.png", "Validation_47430468/Validation_47430468_47430468_311488.950.png", "Validation_47430468/Validation_47430468_47430468_311490.949.png", "Validation_47430468/Validation_47430468_47430468_311492.948.png" ], "turn_id": 5, "scan_id": "Validation_47430468", "origin_question": "Please recall \"the large rectangular window with the dark brown frame\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.97", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the large rectangular window with the dark brown frame\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_47430468/Validation_47430468_47430468_311494.947.png", "Validation_47430468/Validation_47430468_47430468_311496.946.png", "Validation_47430468/Validation_47430468_47430468_311498.945.png", "Validation_47430468/Validation_47430468_47430468_311500.945.png" ], "turn_id": 6, "scan_id": "Validation_47430468", "origin_question": "Is the mix of medium-sized kitchenware glass and plastic bottles to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the mix of medium-sized kitchenware glass and plastic bottles to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_47430468/Validation_47430468_47430468_311502.944.png", "Validation_47430468/Validation_47430468_47430468_311504.943.png", "Validation_47430468/Validation_47430468_47430468_311506.942.png", "Validation_47430468/Validation_47430468_47430468_311508.941.png" ], "turn_id": 7, "scan_id": "Validation_47430468", "origin_question": "What will be the first-time appearance order of the following categories: couch, flowerpot, kitchenware?", "option": [ "flowerpot, couch, kitchenware", "kitchenware, flowerpot, couch", "couch, flowerpot, kitchenware", "flowerpot, kitchenware, couch" ], "answer": "flowerpot, couch, kitchenware", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: couch, flowerpot, kitchenware? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['flowerpot, couch, kitchenware', 'kitchenware, flowerpot, couch', 'couch, flowerpot, kitchenware', 'flowerpot, kitchenware, couch'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_47430468/Validation_47430468_47430468_311510.941.png", "Validation_47430468/Validation_47430468_47430468_311512.940.png", "Validation_47430468/Validation_47430468_47430468_311514.939.png", "Validation_47430468/Validation_47430468_47430468_311516.938.png" ], "turn_id": 8, "scan_id": "Validation_47430468", "origin_question": "At the end of which turn were both of \"the large rectangular window with the dark brown frame\" and \"the modern white plastic stool with the textured top\" on your left side?", "option": [], "answer": "1", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the large rectangular window with the dark brown frame\" and \"the modern white plastic stool with the textured top\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_41125756/Validation_41125756_41125756_5286.923.png", "Validation_41125756/Validation_41125756_41125756_5288.906.png", "Validation_41125756/Validation_41125756_41125756_5290.905.png", "Validation_41125756/Validation_41125756_41125756_5292.904.png" ], "turn_id": 1, "scan_id": "Validation_41125756", "origin_question": "Remember, have you seen any mailbox(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any mailbox(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_41125756/Validation_41125756_41125756_5294.903.png", "Validation_41125756/Validation_41125756_41125756_5296.903.png", "Validation_41125756/Validation_41125756_41125756_5298.902.png", "Validation_41125756/Validation_41125756_41125756_5300.901.png" ], "turn_id": 2, "scan_id": "Validation_41125756", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_41125756/Validation_41125756_41125756_5302.900.png", "Validation_41125756/Validation_41125756_41125756_5304.899.png", "Validation_41125756/Validation_41125756_41125756_5306.898.png", "Validation_41125756/Validation_41125756_41125756_5308.898.png" ], "turn_id": 3, "scan_id": "Validation_41125756", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the compact, rounded rectangular white toilet\", Y: \"the sink\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the compact, rounded rectangular white toilet\", Y: \"the sink\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_41125756/Validation_41125756_41125756_5310.897.png", "Validation_41125756/Validation_41125756_41125756_5312.913.png", "Validation_41125756/Validation_41125756_41125756_5314.912.png", "Validation_41125756/Validation_41125756_41125756_5316.911.png" ], "turn_id": 4, "scan_id": "Validation_41125756", "origin_question": "When did you first discover \"the sleek modern door with smooth white surfaces\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the sleek modern door with smooth white surfaces\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_41125756/Validation_41125756_41125756_5318.910.png", "Validation_41125756/Validation_41125756_41125756_5320.909.png", "Validation_41125756/Validation_41125756_41125756_5322.909.png", "Validation_41125756/Validation_41125756_41125756_5324.908.png" ], "turn_id": 5, "scan_id": "Validation_41125756", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the compact, rounded rectangular white toilet\" from your position?", "option": [], "answer": "124.47", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the compact, rounded rectangular white toilet\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_41125756/Validation_41125756_41125756_5326.907.png", "Validation_41125756/Validation_41125756_41125756_5328.906.png", "Validation_41125756/Validation_41125756_41125756_5330.905.png", "Validation_41125756/Validation_41125756_41125756_5332.904.png" ], "turn_id": 6, "scan_id": "Validation_41125756", "origin_question": "What will be the first-time appearance order of the following categories: bathtub, door, mirror?", "option": [ "mirror, bathtub, door", "door, mirror, bathtub", "door, bathtub, mirror", "bathtub, mirror, door" ], "answer": "door, mirror, bathtub", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bathtub, door, mirror? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['mirror, bathtub, door', 'door, mirror, bathtub', 'door, bathtub, mirror', 'bathtub, mirror, door'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_41125756/Validation_41125756_41125756_5334.904.png", "Validation_41125756/Validation_41125756_41125756_5336.903.png", "Validation_41125756/Validation_41125756_41125756_5338.902.png", "Validation_41125756/Validation_41125756_41125756_5340.901.png" ], "turn_id": 7, "scan_id": "Validation_41125756", "origin_question": "What will be the first-time appearance order of the following categories: bathtub, door, sink?", "option": [ "door, bathtub, sink", "bathtub, sink, door", "door, sink, bathtub", "sink, bathtub, door" ], "answer": "door, sink, bathtub", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bathtub, door, sink? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['door, bathtub, sink', 'bathtub, sink, door', 'door, sink, bathtub', 'sink, bathtub, door'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_44358518/Validation_44358518_44358518_52704.414.png", "Validation_44358518/Validation_44358518_44358518_52706.413.png", "Validation_44358518/Validation_44358518_44358518_52708.412.png", "Validation_44358518/Validation_44358518_44358518_52710.411.png", "Validation_44358518/Validation_44358518_44358518_52712.410.png" ], "turn_id": 1, "scan_id": "Validation_44358518", "origin_question": "Remember, have you seen any chair(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any chair(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "Validation_44358518/Validation_44358518_44358518_52714.410.png", "Validation_44358518/Validation_44358518_44358518_52716.409.png", "Validation_44358518/Validation_44358518_44358518_52718.408.png", "Validation_44358518/Validation_44358518_44358518_52720.407.png", "Validation_44358518/Validation_44358518_44358518_52722.406.png" ], "turn_id": 2, "scan_id": "Validation_44358518", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the sleek rectangular table with light wood top and gray legs\"; \"the sleek black monitor with the thin bezel\"; \"the tall indoor potted plant with long green leaves\" ", "option": [ "the sleek rectangular table with light wood top and gray legs", "the sleek black monitor with the thin bezel", "the tall indoor potted plant with long green leaves" ], "answer": "the tall indoor potted plant with long green leaves", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the sleek rectangular table with light wood top and gray legs\"; \"the sleek black monitor with the thin bezel\"; \"the tall indoor potted plant with long green leaves\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the sleek rectangular table with light wood top and gray legs', 'the sleek black monitor with the thin bezel', 'the tall indoor potted plant with long green leaves'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_44358518/Validation_44358518_44358518_52724.405.png", "Validation_44358518/Validation_44358518_44358518_52726.404.png", "Validation_44358518/Validation_44358518_44358518_52728.404.png", "Validation_44358518/Validation_44358518_44358518_52730.403.png", "Validation_44358518/Validation_44358518_44358518_52732.402.png" ], "turn_id": 3, "scan_id": "Validation_44358518", "origin_question": "When did you first discover \"the sleek rectangular table with light wood top and gray legs\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the sleek rectangular table with light wood top and gray legs\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_44358518/Validation_44358518_44358518_52734.401.png", "Validation_44358518/Validation_44358518_44358518_52736.400.png", "Validation_44358518/Validation_44358518_44358518_52738.416.png", "Validation_44358518/Validation_44358518_44358518_52740.415.png", "Validation_44358518/Validation_44358518_44358518_52742.415.png" ], "turn_id": 4, "scan_id": "Validation_44358518", "origin_question": "Remember, have you seen any chair(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any chair(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_44358518/Validation_44358518_44358518_52744.414.png", "Validation_44358518/Validation_44358518_44358518_52746.413.png", "Validation_44358518/Validation_44358518_44358518_52748.412.png", "Validation_44358518/Validation_44358518_44358518_52750.411.png", "Validation_44358518/Validation_44358518_44358518_52752.410.png" ], "turn_id": 5, "scan_id": "Validation_44358518", "origin_question": "How far is your current position from where you were at the end of the turn 3? (in meters)", "option": [], "answer": "1.75", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 3? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_44358518/Validation_44358518_44358518_52754.410.png", "Validation_44358518/Validation_44358518_44358518_52756.409.png", "Validation_44358518/Validation_44358518_44358518_52758.408.png", "Validation_44358518/Validation_44358518_44358518_52760.407.png", "Validation_44358518/Validation_44358518_44358518_52762.406.png" ], "turn_id": 6, "scan_id": "Validation_44358518", "origin_question": "By the end of the 3 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular white wooden fireplace\", Y: \"the tall indoor potted plant with long green leaves\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "farther away from both X and Y", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 3 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular white wooden fireplace\", Y: \"the tall indoor potted plant with long green leaves\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_44358518/Validation_44358518_44358518_52764.405.png", "Validation_44358518/Validation_44358518_44358518_52766.405.png", "Validation_44358518/Validation_44358518_44358518_52768.404.png", "Validation_44358518/Validation_44358518_44358518_52770.403.png", "Validation_44358518/Validation_44358518_44358518_52772.402.png" ], "turn_id": 7, "scan_id": "Validation_44358518", "origin_question": "Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Right", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_44358518/Validation_44358518_44358518_52774.401.png", "Validation_44358518/Validation_44358518_44358518_52776.400.png", "Validation_44358518/Validation_44358518_44358518_52778.416.png", "Validation_44358518/Validation_44358518_44358518_52780.415.png", "Validation_44358518/Validation_44358518_44358518_52782.415.png" ], "turn_id": 8, "scan_id": "Validation_44358518", "origin_question": "Which direction is the rectangular white wooden fireplace in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the rectangular white wooden fireplace in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_44358518/Validation_44358518_44358518_52784.414.png", "Validation_44358518/Validation_44358518_44358518_52786.413.png", "Validation_44358518/Validation_44358518_44358518_52788.412.png", "Validation_44358518/Validation_44358518_44358518_52790.411.png", "Validation_44358518/Validation_44358518_44358518_52792.410.png" ], "turn_id": 9, "scan_id": "Validation_44358518", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the tall indoor potted plant with long green leaves\" from your position?", "option": [], "answer": "155.81", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the tall indoor potted plant with long green leaves\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_44358518/Validation_44358518_44358518_52794.410.png", "Validation_44358518/Validation_44358518_44358518_52796.409.png", "Validation_44358518/Validation_44358518_44358518_52798.408.png", "Validation_44358518/Validation_44358518_44358518_52800.407.png", "Validation_44358518/Validation_44358518_44358518_52802.406.png" ], "turn_id": 10, "scan_id": "Validation_44358518", "origin_question": "What will be the first-time appearance order of the following categories: couch, fireplace, plant?", "option": [ "fireplace, plant, couch", "plant, couch, fireplace", "plant, fireplace, couch", "couch, fireplace, plant" ], "answer": "fireplace, plant, couch", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: couch, fireplace, plant? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['fireplace, plant, couch', 'plant, couch, fireplace', 'plant, fireplace, couch', 'couch, fireplace, plant'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47333457/Validation_47333457_47333457_54892.931.png", "Validation_47333457/Validation_47333457_47333457_54894.913.png", "Validation_47333457/Validation_47333457_47333457_54896.912.png", "Validation_47333457/Validation_47333457_47333457_54898.912.png" ], "turn_id": 1, "scan_id": "Validation_47333457", "origin_question": "Remember, have you seen any statue(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any statue(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47333457/Validation_47333457_47333457_54900.911.png", "Validation_47333457/Validation_47333457_47333457_54902.910.png", "Validation_47333457/Validation_47333457_47333457_54904.909.png", "Validation_47333457/Validation_47333457_47333457_54906.908.png" ], "turn_id": 2, "scan_id": "Validation_47333457", "origin_question": "Which direction is the rectangular stainless steel sink in the kitchen in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the rectangular stainless steel sink in the kitchen in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_47333457/Validation_47333457_47333457_54908.924.png", "Validation_47333457/Validation_47333457_47333457_54910.923.png", "Validation_47333457/Validation_47333457_47333457_54912.923.png", "Validation_47333457/Validation_47333457_47333457_54914.922.png" ], "turn_id": 3, "scan_id": "Validation_47333457", "origin_question": "Is the compact, transparent plastic case with the pink lid to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the compact, transparent plastic case with the pink lid to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47333457/Validation_47333457_47333457_54916.921.png", "Validation_47333457/Validation_47333457_47333457_54918.920.png", "Validation_47333457/Validation_47333457_47333457_54920.919.png", "Validation_47333457/Validation_47333457_47333457_54922.918.png" ], "turn_id": 4, "scan_id": "Validation_47333457", "origin_question": "Which direction is the compact, transparent plastic case with the pink lid in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the compact, transparent plastic case with the pink lid in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_47333457/Validation_47333457_47333457_54924.918.png", "Validation_47333457/Validation_47333457_47333457_54926.917.png", "Validation_47333457/Validation_47333457_47333457_54928.916.png", "Validation_47333457/Validation_47333457_47333457_54930.915.png" ], "turn_id": 5, "scan_id": "Validation_47333457", "origin_question": "Please recall \"the compact, transparent plastic case with the pink lid\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.21", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the compact, transparent plastic case with the pink lid\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47333457/Validation_47333457_47333457_54932.914.png", "Validation_47333457/Validation_47333457_47333457_54934.913.png", "Validation_47333457/Validation_47333457_47333457_54936.913.png", "Validation_47333457/Validation_47333457_47333457_54938.912.png" ], "turn_id": 6, "scan_id": "Validation_47333457", "origin_question": "Which direction is the bed in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the bed in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_41125731/Validation_41125731_41125731_4608.224.png", "Validation_41125731/Validation_41125731_41125731_4610.223.png", "Validation_41125731/Validation_41125731_41125731_4612.223.png", "Validation_41125731/Validation_41125731_41125731_4614.222.png", "Validation_41125731/Validation_41125731_41125731_4617.420.png" ], "turn_id": 1, "scan_id": "Validation_41125731", "origin_question": "Remember, have you seen any door(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any door(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_41125731/Validation_41125731_41125731_4619.420.png", "Validation_41125731/Validation_41125731_41125731_4621.419.png", "Validation_41125731/Validation_41125731_41125731_4623.418.png", "Validation_41125731/Validation_41125731_41125731_4625.417.png", "Validation_41125731/Validation_41125731_41125731_4627.416.png" ], "turn_id": 2, "scan_id": "Validation_41125731", "origin_question": "When did you first discover \"the sleek, white wooden door with the modern design\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the sleek, white wooden door with the modern design\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_41125731/Validation_41125731_41125731_4629.432.png", "Validation_41125731/Validation_41125731_41125731_4631.431.png", "Validation_41125731/Validation_41125731_41125731_4633.431.png", "Validation_41125731/Validation_41125731_41125731_4635.430.png", "Validation_41125731/Validation_41125731_41125731_4637.429.png" ], "turn_id": 3, "scan_id": "Validation_41125731", "origin_question": "Which direction is the sleek clock with the neutral or metallic face in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the sleek clock with the neutral or metallic face in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_41125731/Validation_41125731_41125731_4639.428.png", "Validation_41125731/Validation_41125731_41125731_4641.427.png", "Validation_41125731/Validation_41125731_41125731_4643.426.png", "Validation_41125731/Validation_41125731_41125731_4645.426.png", "Validation_41125731/Validation_41125731_41125731_4647.425.png" ], "turn_id": 4, "scan_id": "Validation_41125731", "origin_question": "In which turn did you see both \"the sleek clock with the neutral or metallic face\" and \"the cozy bed with soft pastel pink bedding\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the sleek clock with the neutral or metallic face\" and \"the cozy bed with soft pastel pink bedding\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_41125731/Validation_41125731_41125731_4649.424.png", "Validation_41125731/Validation_41125731_41125731_4651.423.png", "Validation_41125731/Validation_41125731_41125731_4653.422.png", "Validation_41125731/Validation_41125731_41125731_4655.422.png", "Validation_41125731/Validation_41125731_41125731_4657.421.png" ], "turn_id": 5, "scan_id": "Validation_41125731", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the sleek, white wooden door with the modern design\" from your position?", "option": [], "answer": "120.85", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the sleek, white wooden door with the modern design\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_41125731/Validation_41125731_41125731_4659.420.png", "Validation_41125731/Validation_41125731_41125731_4661.419.png", "Validation_41125731/Validation_41125731_41125731_4663.418.png", "Validation_41125731/Validation_41125731_41125731_4665.417.png", "Validation_41125731/Validation_41125731_41125731_4667.417.png" ], "turn_id": 6, "scan_id": "Validation_41125731", "origin_question": "", "option": [], "answer": "", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_41125731/Validation_41125731_41125731_4669.416.png", "Validation_41125731/Validation_41125731_41125731_4671.432.png", "Validation_41125731/Validation_41125731_41125731_4673.431.png", "Validation_41125731/Validation_41125731_41125731_4675.430.png", "Validation_41125731/Validation_41125731_41125731_4677.429.png" ], "turn_id": 7, "scan_id": "Validation_41125731", "origin_question": "Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_41125731/Validation_41125731_41125731_4679.428.png", "Validation_41125731/Validation_41125731_41125731_4681.428.png", "Validation_41125731/Validation_41125731_41125731_4683.427.png", "Validation_41125731/Validation_41125731_41125731_4685.426.png", "Validation_41125731/Validation_41125731_41125731_4687.425.png" ], "turn_id": 8, "scan_id": "Validation_41125731", "origin_question": "At the end of which turn were both of \"the sleek clock with the neutral or metallic face\" and \"the rectangular window with the sleek white frame\" on your right side?", "option": [], "answer": "1", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the sleek clock with the neutral or metallic face\" and \"the rectangular window with the sleek white frame\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "Validation_45663150/Validation_45663150_45663150_63507.970.png", "Validation_45663150/Validation_45663150_45663150_63509.953.png", "Validation_45663150/Validation_45663150_45663150_63511.952.png", "Validation_45663150/Validation_45663150_45663150_63513.951.png", "Validation_45663150/Validation_45663150_45663150_63515.950.png" ], "turn_id": 1, "scan_id": "Validation_45663150", "origin_question": "Remember, how many kitchen island(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many kitchen island(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "Validation_45663150/Validation_45663150_45663150_63517.950.png", "Validation_45663150/Validation_45663150_45663150_63519.949.png", "Validation_45663150/Validation_45663150_45663150_63521.965.png", "Validation_45663150/Validation_45663150_45663150_63523.964.png", "Validation_45663150/Validation_45663150_45663150_63525.963.png" ], "turn_id": 2, "scan_id": "Validation_45663150", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the sleek lamp with the transparent glass base\"; \"the matte black cast iron fireplace\"; \"the sleek white metal radiator with the compact design\" ", "option": [ "the sleek lamp with the transparent glass base", "the matte black cast iron fireplace", "the sleek white metal radiator with the compact design" ], "answer": "the sleek white metal radiator with the compact design", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the sleek lamp with the transparent glass base\"; \"the matte black cast iron fireplace\"; \"the sleek white metal radiator with the compact design\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the sleek lamp with the transparent glass base', 'the matte black cast iron fireplace', 'the sleek white metal radiator with the compact design'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_45663150/Validation_45663150_45663150_63527.962.png", "Validation_45663150/Validation_45663150_45663150_63529.961.png", "Validation_45663150/Validation_45663150_45663150_63531.961.png", "Validation_45663150/Validation_45663150_45663150_63533.960.png", "Validation_45663150/Validation_45663150_45663150_63535.959.png" ], "turn_id": 3, "scan_id": "Validation_45663150", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "1.03", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_45663150/Validation_45663150_45663150_63537.958.png", "Validation_45663150/Validation_45663150_45663150_63539.958.png", "Validation_45663150/Validation_45663150_45663150_63541.957.png", "Validation_45663150/Validation_45663150_45663150_63543.956.png", "Validation_45663150/Validation_45663150_45663150_63545.955.png" ], "turn_id": 4, "scan_id": "Validation_45663150", "origin_question": "When did you first discover \"the matte black cast iron fireplace\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the matte black cast iron fireplace\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_45663150/Validation_45663150_45663150_63547.954.png", "Validation_45663150/Validation_45663150_45663150_63549.954.png", "Validation_45663150/Validation_45663150_45663150_63551.953.png", "Validation_45663150/Validation_45663150_45663150_63553.952.png", "Validation_45663150/Validation_45663150_45663150_63555.951.png" ], "turn_id": 5, "scan_id": "Validation_45663150", "origin_question": "What will be the first-time appearance order of the following categories: carpet, fireplace, radiator?", "option": [ "radiator, carpet, fireplace", "carpet, fireplace, radiator", "fireplace, carpet, radiator", "fireplace, radiator, carpet" ], "answer": "fireplace, carpet, radiator", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: carpet, fireplace, radiator? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['radiator, carpet, fireplace', 'carpet, fireplace, radiator', 'fireplace, carpet, radiator', 'fireplace, radiator, carpet'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_45663150/Validation_45663150_45663150_63557.950.png", "Validation_45663150/Validation_45663150_45663150_63559.950.png", "Validation_45663150/Validation_45663150_45663150_63561.949.png", "Validation_45663150/Validation_45663150_45663150_63563.965.png", "Validation_45663150/Validation_45663150_45663150_63565.964.png" ], "turn_id": 6, "scan_id": "Validation_45663150", "origin_question": "Is the sleek black monitor with the matte finish to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the sleek black monitor with the matte finish to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_45663150/Validation_45663150_45663150_63567.963.png", "Validation_45663150/Validation_45663150_45663150_63569.962.png", "Validation_45663150/Validation_45663150_45663150_63571.962.png", "Validation_45663150/Validation_45663150_45663150_63573.961.png", "Validation_45663150/Validation_45663150_45663150_63575.960.png" ], "turn_id": 7, "scan_id": "Validation_45663150", "origin_question": "By the end of the 5 turn, were you getting closer to or farther away from these two objects: X: \"the warm, earthy-colored plush carpet\", Y: \"the sleek black monitor with the matte finish\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "farther away from both X and Y", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 5 turn, were you getting closer to or farther away from these two objects: X: \"the warm, earthy-colored plush carpet\", Y: \"the sleek black monitor with the matte finish\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_45663150/Validation_45663150_45663150_63577.959.png", "Validation_45663150/Validation_45663150_45663150_63579.958.png", "Validation_45663150/Validation_45663150_45663150_63581.958.png", "Validation_45663150/Validation_45663150_45663150_63583.957.png", "Validation_45663150/Validation_45663150_45663150_63585.956.png" ], "turn_id": 8, "scan_id": "Validation_45663150", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the matte black cast iron fireplace\", Y: \"the rustic brown square table with straight legs\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the matte black cast iron fireplace\", Y: \"the rustic brown square table with straight legs\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_45663150/Validation_45663150_45663150_63587.955.png", "Validation_45663150/Validation_45663150_45663150_63589.954.png", "Validation_45663150/Validation_45663150_45663150_63591.954.png", "Validation_45663150/Validation_45663150_45663150_63593.953.png", "Validation_45663150/Validation_45663150_45663150_63595.952.png" ], "turn_id": 9, "scan_id": "Validation_45663150", "origin_question": "How far is your current position from where you were at the end of the turn 7? (in meters)", "option": [], "answer": "2.81", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 7? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_45663150/Validation_45663150_45663150_63597.951.png", "Validation_45663150/Validation_45663150_45663150_63599.950.png", "Validation_45663150/Validation_45663150_45663150_63601.950.png", "Validation_45663150/Validation_45663150_45663150_63603.949.png", "Validation_45663150/Validation_45663150_45663150_63605.965.png" ], "turn_id": 10, "scan_id": "Validation_45663150", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the medium circular clock with the light face\" from your position?", "option": [], "answer": "157.62", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the medium circular clock with the light face\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47895556/Validation_47895556_47895556_457799.158.png", "Validation_47895556/Validation_47895556_47895556_457801.141.png", "Validation_47895556/Validation_47895556_47895556_457803.157.png", "Validation_47895556/Validation_47895556_47895556_457805.156.png" ], "turn_id": 1, "scan_id": "Validation_47895556", "origin_question": "Remember, have you seen any glove(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any glove(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47895556/Validation_47895556_47895556_457807.155.png", "Validation_47895556/Validation_47895556_47895556_457809.154.png", "Validation_47895556/Validation_47895556_47895556_457811.154.png", "Validation_47895556/Validation_47895556_47895556_457813.153.png" ], "turn_id": 2, "scan_id": "Validation_47895556", "origin_question": "In which turn were you closest to \"the washing machine\"? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the washing machine\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_47895556/Validation_47895556_47895556_457815.152.png", "Validation_47895556/Validation_47895556_47895556_457817.151.png", "Validation_47895556/Validation_47895556_47895556_457819.150.png", "Validation_47895556/Validation_47895556_47895556_457821.149.png" ], "turn_id": 3, "scan_id": "Validation_47895556", "origin_question": "In which turn did you see both \"the modern, oval-shaped white ceramic toilet\" and \"the washing machine\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the modern, oval-shaped white ceramic toilet\" and \"the washing machine\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_47895556/Validation_47895556_47895556_457823.149.png", "Validation_47895556/Validation_47895556_47895556_457825.148.png", "Validation_47895556/Validation_47895556_47895556_457827.147.png", "Validation_47895556/Validation_47895556_47895556_457829.146.png" ], "turn_id": 4, "scan_id": "Validation_47895556", "origin_question": "In which turn did you see both \"the modern, oval-shaped white ceramic toilet\" and \"the medium-sized red plastic basin\" simultaneously? ", "option": [], "answer": "3", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the modern, oval-shaped white ceramic toilet\" and \"the medium-sized red plastic basin\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_47895556/Validation_47895556_47895556_457831.145.png", "Validation_47895556/Validation_47895556_47895556_457833.145.png", "Validation_47895556/Validation_47895556_47895556_457835.144.png", "Validation_47895556/Validation_47895556_47895556_457837.143.png" ], "turn_id": 5, "scan_id": "Validation_47895556", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_47895556/Validation_47895556_47895556_457839.142.png", "Validation_47895556/Validation_47895556_47895556_457841.141.png", "Validation_47895556/Validation_47895556_47895556_457843.157.png", "Validation_47895556/Validation_47895556_47895556_457845.156.png" ], "turn_id": 6, "scan_id": "Validation_47895556", "origin_question": "Please recall \"the modern, oval-shaped white ceramic toilet\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.82", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the modern, oval-shaped white ceramic toilet\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_45260899/Validation_45260899_45260899_848.808.png", "Validation_45260899/Validation_45260899_45260899_850.791.png", "Validation_45260899/Validation_45260899_45260899_852.790.png", "Validation_45260899/Validation_45260899_45260899_854.789.png", "Validation_45260899/Validation_45260899_45260899_856.788.png" ], "turn_id": 1, "scan_id": "Validation_45260899", "origin_question": "Remember, have you seen any sink(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any sink(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_45260899/Validation_45260899_45260899_858.788.png", "Validation_45260899/Validation_45260899_45260899_860.787.png", "Validation_45260899/Validation_45260899_45260899_862.786.png", "Validation_45260899/Validation_45260899_45260899_864.785.png", "Validation_45260899/Validation_45260899_45260899_866.784.png" ], "turn_id": 2, "scan_id": "Validation_45260899", "origin_question": "In which turn were you farthest from \"the sleek rectangular beige bathtub made of acrylic\"? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the sleek rectangular beige bathtub made of acrylic\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_45260899/Validation_45260899_45260899_868.784.png", "Validation_45260899/Validation_45260899_45260899_870.783.png", "Validation_45260899/Validation_45260899_45260899_872.782.png", "Validation_45260899/Validation_45260899_45260899_874.781.png", "Validation_45260899/Validation_45260899_45260899_876.781.png" ], "turn_id": 3, "scan_id": "Validation_45260899", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "55.66", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_45260899/Validation_45260899_45260899_878.780.png", "Validation_45260899/Validation_45260899_45260899_880.779.png", "Validation_45260899/Validation_45260899_45260899_882.778.png", "Validation_45260899/Validation_45260899_45260899_884.777.png", "Validation_45260899/Validation_45260899_45260899_886.793.png" ], "turn_id": 4, "scan_id": "Validation_45260899", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the cylindrical olive green bin with the stainless steel lid\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the cylindrical olive green bin with the stainless steel lid\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_45260899/Validation_45260899_45260899_888.793.png", "Validation_45260899/Validation_45260899_45260899_890.792.png", "Validation_45260899/Validation_45260899_45260899_892.791.png", "Validation_45260899/Validation_45260899_45260899_894.790.png", "Validation_45260899/Validation_45260899_45260899_896.789.png" ], "turn_id": 5, "scan_id": "Validation_45260899", "origin_question": "In which turn were you closest to \"the glossy white ceramic sink with polished metallic fixtures\"? ", "option": [], "answer": "5", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the glossy white ceramic sink with polished metallic fixtures\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_45260899/Validation_45260899_45260899_898.789.png", "Validation_45260899/Validation_45260899_45260899_900.788.png", "Validation_45260899/Validation_45260899_45260899_902.787.png", "Validation_45260899/Validation_45260899_45260899_904.786.png", "Validation_45260899/Validation_45260899_45260899_906.785.png" ], "turn_id": 6, "scan_id": "Validation_45260899", "origin_question": "In which turn did you see both \"the sleek, modern white toilet with the rounded bowl\" and \"the cylindrical olive green bin with the stainless steel lid\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the sleek, modern white toilet with the rounded bowl\" and \"the cylindrical olive green bin with the stainless steel lid\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_45260899/Validation_45260899_45260899_908.785.png", "Validation_45260899/Validation_45260899_45260899_910.784.png", "Validation_45260899/Validation_45260899_45260899_912.783.png", "Validation_45260899/Validation_45260899_45260899_914.782.png", "Validation_45260899/Validation_45260899_45260899_916.781.png" ], "turn_id": 7, "scan_id": "Validation_45260899", "origin_question": "When did you first discover \"the cylindrical olive green bin with the stainless steel lid\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the cylindrical olive green bin with the stainless steel lid\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_45260899/Validation_45260899_45260899_918.781.png", "Validation_45260899/Validation_45260899_45260899_920.780.png", "Validation_45260899/Validation_45260899_45260899_922.779.png", "Validation_45260899/Validation_45260899_45260899_924.778.png", "Validation_45260899/Validation_45260899_45260899_926.794.png" ], "turn_id": 8, "scan_id": "Validation_45260899", "origin_question": "Which direction is the cylindrical olive green bin with the stainless steel lid in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the cylindrical olive green bin with the stainless steel lid in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_45260899/Validation_45260899_45260899_928.793.png", "Validation_45260899/Validation_45260899_45260899_930.793.png", "Validation_45260899/Validation_45260899_45260899_932.792.png", "Validation_45260899/Validation_45260899_45260899_934.791.png", "Validation_45260899/Validation_45260899_45260899_936.790.png" ], "turn_id": 9, "scan_id": "Validation_45260899", "origin_question": "Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "117.28", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_45260899/Validation_45260899_45260899_938.789.png", "Validation_45260899/Validation_45260899_45260899_940.789.png", "Validation_45260899/Validation_45260899_45260899_942.788.png", "Validation_45260899/Validation_45260899_45260899_944.787.png", "Validation_45260899/Validation_45260899_45260899_946.786.png" ], "turn_id": 10, "scan_id": "Validation_45260899", "origin_question": "How far is your current position from where you were at the end of the turn 5? (in meters)", "option": [], "answer": "1.52", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 5? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47430051/Validation_47430051_47430051_21132.855.png", "Validation_47430051/Validation_47430051_47430051_21134.838.png", "Validation_47430051/Validation_47430051_47430051_21136.837.png", "Validation_47430051/Validation_47430051_47430051_21138.836.png" ], "turn_id": 1, "scan_id": "Validation_47430051", "origin_question": "", "option": [], "answer": "", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47430051/Validation_47430051_47430051_21140.835.png", "Validation_47430051/Validation_47430051_47430051_21142.834.png", "Validation_47430051/Validation_47430051_47430051_21144.833.png", "Validation_47430051/Validation_47430051_47430051_21146.833.png" ], "turn_id": 2, "scan_id": "Validation_47430051", "origin_question": "Remember, have you seen any bucket(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any bucket(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_47430051/Validation_47430051_47430051_21148.832.png", "Validation_47430051/Validation_47430051_47430051_21150.831.png", "Validation_47430051/Validation_47430051_47430051_21152.830.png", "Validation_47430051/Validation_47430051_47430051_21154.829.png" ], "turn_id": 3, "scan_id": "Validation_47430051", "origin_question": "In which turn did you see both \"the bathtub\" and \"the dark gray plastic bucket with the square shape\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the bathtub\" and \"the dark gray plastic bucket with the square shape\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_47430051/Validation_47430051_47430051_21156.828.png", "Validation_47430051/Validation_47430051_47430051_21158.828.png", "Validation_47430051/Validation_47430051_47430051_21160.827.png", "Validation_47430051/Validation_47430051_47430051_21162.826.png" ], "turn_id": 4, "scan_id": "Validation_47430051", "origin_question": "By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the small white woven box\", Y: \"the white ceramic round flowerpot\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the small white woven box\", Y: \"the white ceramic round flowerpot\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47430051/Validation_47430051_47430051_21164.825.png", "Validation_47430051/Validation_47430051_47430051_21166.841.png", "Validation_47430051/Validation_47430051_47430051_21168.840.png", "Validation_47430051/Validation_47430051_47430051_21170.839.png" ], "turn_id": 5, "scan_id": "Validation_47430051", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the compact grey metal cabinet with the sleek rectangle shape\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the compact grey metal cabinet with the sleek rectangle shape\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_47430051/Validation_47430051_47430051_21172.839.png", "Validation_47430051/Validation_47430051_47430051_21174.838.png", "Validation_47430051/Validation_47430051_47430051_21176.837.png", "Validation_47430051/Validation_47430051_47430051_21178.836.png" ], "turn_id": 6, "scan_id": "Validation_47430051", "origin_question": "In which turn did you see both \"the white ceramic round flowerpot\" and \"the sleek white front-loading washing machine with the silver rim\" simultaneously? ", "option": [], "answer": "5", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the white ceramic round flowerpot\" and \"the sleek white front-loading washing machine with the silver rim\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_47430051/Validation_47430051_47430051_21180.835.png", "Validation_47430051/Validation_47430051_47430051_21182.835.png", "Validation_47430051/Validation_47430051_47430051_21184.834.png", "Validation_47430051/Validation_47430051_47430051_21186.833.png" ], "turn_id": 7, "scan_id": "Validation_47430051", "origin_question": "What will be the first-time appearance order of the following categories: bathtub, flowerpot, towel?", "option": [ "flowerpot, bathtub, towel", "bathtub, flowerpot, towel", "flowerpot, towel, bathtub", "towel, flowerpot, bathtub" ], "answer": "towel, flowerpot, bathtub", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bathtub, flowerpot, towel? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['flowerpot, bathtub, towel', 'bathtub, flowerpot, towel', 'flowerpot, towel, bathtub', 'towel, flowerpot, bathtub'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_45261190/Validation_45261190_45261190_3686.914.png", "Validation_45261190/Validation_45261190_45261190_3688.880.png", "Validation_45261190/Validation_45261190_45261190_3690.879.png", "Validation_45261190/Validation_45261190_45261190_3692.878.png" ], "turn_id": 1, "scan_id": "Validation_45261190", "origin_question": "Remember, have you seen any kettle(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any kettle(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_45261190/Validation_45261190_45261190_3694.894.png", "Validation_45261190/Validation_45261190_45261190_3696.894.png", "Validation_45261190/Validation_45261190_45261190_3698.893.png", "Validation_45261190/Validation_45261190_45261190_3700.892.png" ], "turn_id": 2, "scan_id": "Validation_45261190", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_45261190/Validation_45261190_45261190_3702.891.png", "Validation_45261190/Validation_45261190_45261190_3704.890.png", "Validation_45261190/Validation_45261190_45261190_3706.890.png", "Validation_45261190/Validation_45261190_45261190_3708.889.png" ], "turn_id": 3, "scan_id": "Validation_45261190", "origin_question": "Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "37.17", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_45261190/Validation_45261190_45261190_3710.888.png", "Validation_45261190/Validation_45261190_45261190_3712.887.png", "Validation_45261190/Validation_45261190_45261190_3714.886.png", "Validation_45261190/Validation_45261190_45261190_3716.886.png" ], "turn_id": 4, "scan_id": "Validation_45261190", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_45261190/Validation_45261190_45261190_3718.885.png", "Validation_45261190/Validation_45261190_45261190_3720.884.png", "Validation_45261190/Validation_45261190_45261190_3722.883.png", "Validation_45261190/Validation_45261190_45261190_3724.883.png" ], "turn_id": 5, "scan_id": "Validation_45261190", "origin_question": "", "option": [], "answer": "", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_45261190/Validation_45261190_45261190_3726.882.png", "Validation_45261190/Validation_45261190_45261190_3728.881.png", "Validation_45261190/Validation_45261190_45261190_3730.880.png", "Validation_45261190/Validation_45261190_45261190_3732.879.png" ], "turn_id": 6, "scan_id": "Validation_45261190", "origin_question": "", "option": [], "answer": "", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_45261190/Validation_45261190_45261190_3734.878.png", "Validation_45261190/Validation_45261190_45261190_3736.894.png", "Validation_45261190/Validation_45261190_45261190_3738.894.png", "Validation_45261190/Validation_45261190_45261190_3740.893.png" ], "turn_id": 7, "scan_id": "Validation_45261190", "origin_question": "How far is your current position from where you were at the end of the turn 3? (in meters)", "option": [], "answer": "2.78", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 3? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_45261190/Validation_45261190_45261190_3742.892.png", "Validation_45261190/Validation_45261190_45261190_3744.891.png", "Validation_45261190/Validation_45261190_45261190_3746.890.png", "Validation_45261190/Validation_45261190_45261190_3748.890.png" ], "turn_id": 8, "scan_id": "Validation_45261190", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the bed\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the bed\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_45261190/Validation_45261190_45261190_3750.889.png", "Validation_45261190/Validation_45261190_45261190_3752.888.png", "Validation_45261190/Validation_45261190_45261190_3754.887.png", "Validation_45261190/Validation_45261190_45261190_3756.887.png" ], "turn_id": 9, "scan_id": "Validation_45261190", "origin_question": "Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "143.22", "user_message": "For the 9 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47334105/Validation_47334105_47334105_92883.753.png", "Validation_47334105/Validation_47334105_47334105_92885.752.png", "Validation_47334105/Validation_47334105_47334105_92887.751.png", "Validation_47334105/Validation_47334105_47334105_92889.750.png", "Validation_47334105/Validation_47334105_47334105_92891.749.png" ], "turn_id": 1, "scan_id": "Validation_47334105", "origin_question": "Remember, have you seen any monitor(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any monitor(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_47334105/Validation_47334105_47334105_92893.749.png", "Validation_47334105/Validation_47334105_47334105_92895.748.png", "Validation_47334105/Validation_47334105_47334105_92897.747.png", "Validation_47334105/Validation_47334105_47334105_92899.746.png", "Validation_47334105/Validation_47334105_47334105_92901.745.png" ], "turn_id": 2, "scan_id": "Validation_47334105", "origin_question": "At the end of which turn were both of \"the couch\" and \"the large rectangular cream-colored flowerpot\" on your left side?", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the couch\" and \"the large rectangular cream-colored flowerpot\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_47334105/Validation_47334105_47334105_92903.745.png", "Validation_47334105/Validation_47334105_47334105_92905.744.png", "Validation_47334105/Validation_47334105_47334105_92907.760.png", "Validation_47334105/Validation_47334105_47334105_92909.759.png", "Validation_47334105/Validation_47334105_47334105_92911.758.png" ], "turn_id": 3, "scan_id": "Validation_47334105", "origin_question": "Which two objects are on the same side of you now? \"the couch\", \"the large rectangular cream-colored flowerpot\", \"the dark oval chair with soft cushions\". ", "option": [ "the couch, the large rectangular cream-colored flowerpot", "the couch, the dark oval chair with soft cushions", "the large rectangular cream-colored flowerpot, the dark oval chair with soft cushions" ], "answer": "the large rectangular cream-colored flowerpot, the dark oval chair with soft cushions", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the couch\", \"the large rectangular cream-colored flowerpot\", \"the dark oval chair with soft cushions\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the couch, the large rectangular cream-colored flowerpot', 'the couch, the dark oval chair with soft cushions', 'the large rectangular cream-colored flowerpot, the dark oval chair with soft cushions'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47334105/Validation_47334105_47334105_92913.757.png", "Validation_47334105/Validation_47334105_47334105_92915.756.png", "Validation_47334105/Validation_47334105_47334105_92917.755.png", "Validation_47334105/Validation_47334105_47334105_92919.755.png", "Validation_47334105/Validation_47334105_47334105_92921.754.png" ], "turn_id": 4, "scan_id": "Validation_47334105", "origin_question": "Which direction is the large rectangular cream-colored flowerpot in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the large rectangular cream-colored flowerpot in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47334105/Validation_47334105_47334105_92923.753.png", "Validation_47334105/Validation_47334105_47334105_92925.752.png", "Validation_47334105/Validation_47334105_47334105_92927.751.png", "Validation_47334105/Validation_47334105_47334105_92929.751.png", "Validation_47334105/Validation_47334105_47334105_92931.750.png" ], "turn_id": 5, "scan_id": "Validation_47334105", "origin_question": "Which direction is the couch in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the couch in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_47334105/Validation_47334105_47334105_92933.749.png", "Validation_47334105/Validation_47334105_47334105_92935.748.png", "Validation_47334105/Validation_47334105_47334105_92937.747.png", "Validation_47334105/Validation_47334105_47334105_92939.747.png", "Validation_47334105/Validation_47334105_47334105_92941.746.png" ], "turn_id": 6, "scan_id": "Validation_47334105", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "2.45", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47334105/Validation_47334105_47334105_92943.745.png", "Validation_47334105/Validation_47334105_47334105_92945.744.png", "Validation_47334105/Validation_47334105_47334105_92947.743.png", "Validation_47334105/Validation_47334105_47334105_92949.759.png", "Validation_47334105/Validation_47334105_47334105_92951.758.png" ], "turn_id": 7, "scan_id": "Validation_47334105", "origin_question": "Which direction is the couch in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the couch in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_47334105/Validation_47334105_47334105_92953.758.png", "Validation_47334105/Validation_47334105_47334105_92955.757.png", "Validation_47334105/Validation_47334105_47334105_92957.756.png", "Validation_47334105/Validation_47334105_47334105_92959.755.png", "Validation_47334105/Validation_47334105_47334105_92961.754.png" ], "turn_id": 8, "scan_id": "Validation_47334105", "origin_question": "In which turn did you see both \"the couch\" and \"the sleek rectangular neutral-colored table made of wood\" simultaneously? ", "option": [], "answer": "4", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the couch\" and \"the sleek rectangular neutral-colored table made of wood\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_47334105/Validation_47334105_47334105_92963.754.png", "Validation_47334105/Validation_47334105_47334105_92965.753.png", "Validation_47334105/Validation_47334105_47334105_92967.752.png", "Validation_47334105/Validation_47334105_47334105_92969.751.png", "Validation_47334105/Validation_47334105_47334105_92971.750.png" ], "turn_id": 9, "scan_id": "Validation_47334105", "origin_question": "Please recall \"the large rectangular window with the light-colored frame\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.11", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the large rectangular window with the light-colored frame\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "Validation_47334105/Validation_47334105_47334105_92973.750.png", "Validation_47334105/Validation_47334105_47334105_92975.749.png", "Validation_47334105/Validation_47334105_47334105_92977.748.png", "Validation_47334105/Validation_47334105_47334105_92979.747.png", "Validation_47334105/Validation_47334105_47334105_92981.746.png" ], "turn_id": 10, "scan_id": "Validation_47334105", "origin_question": "Among these three objects(\"the couch\",\"the sleek rectangular neutral-colored table made of wood\",\"the microwave\" ), which one is the farthest from you?", "option": [ "the couch", "the sleek rectangular neutral-colored table made of wood", "the microwave" ], "answer": "the microwave", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the couch\",\"the sleek rectangular neutral-colored table made of wood\",\"the microwave\" ), which one is the farthest from you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the couch', 'the sleek rectangular neutral-colored table made of wood', 'the microwave'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47331970/Validation_47331970_47331970_167361.160.png", "Validation_47331970/Validation_47331970_47331970_167363.142.png", "Validation_47331970/Validation_47331970_47331970_167365.141.png", "Validation_47331970/Validation_47331970_47331970_167367.141.png", "Validation_47331970/Validation_47331970_47331970_167369.140.png" ], "turn_id": 1, "scan_id": "Validation_47331970", "origin_question": "Remember, have you seen any air conditioner(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any air conditioner(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47331970/Validation_47331970_47331970_167371.139.png", "Validation_47331970/Validation_47331970_47331970_167373.138.png", "Validation_47331970/Validation_47331970_47331970_167375.138.png", "Validation_47331970/Validation_47331970_47331970_167377.153.png", "Validation_47331970/Validation_47331970_47331970_167379.152.png" ], "turn_id": 2, "scan_id": "Validation_47331970", "origin_question": "Remember, have you seen any camera(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any camera(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47331970/Validation_47331970_47331970_167381.152.png", "Validation_47331970/Validation_47331970_47331970_167383.151.png", "Validation_47331970/Validation_47331970_47331970_167385.150.png", "Validation_47331970/Validation_47331970_47331970_167387.149.png", "Validation_47331970/Validation_47331970_47331970_167389.149.png" ], "turn_id": 3, "scan_id": "Validation_47331970", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the tall, black stainless steel refrigerator\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the tall, black stainless steel refrigerator\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_47331970/Validation_47331970_47331970_167391.148.png", "Validation_47331970/Validation_47331970_47331970_167393.147.png", "Validation_47331970/Validation_47331970_47331970_167395.146.png", "Validation_47331970/Validation_47331970_47331970_167397.145.png", "Validation_47331970/Validation_47331970_47331970_167399.144.png" ], "turn_id": 4, "scan_id": "Validation_47331970", "origin_question": "Please recall \"the oven\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.88", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the oven\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_47331970/Validation_47331970_47331970_167401.144.png", "Validation_47331970/Validation_47331970_47331970_167403.143.png", "Validation_47331970/Validation_47331970_47331970_167405.142.png", "Validation_47331970/Validation_47331970_47331970_167407.141.png", "Validation_47331970/Validation_47331970_47331970_167409.140.png" ], "turn_id": 5, "scan_id": "Validation_47331970", "origin_question": "At the end of which turn were both of \"the medium-sized terracottthe flowerpot with the rounded shape\" and \"the medium-sized, white metal radiator with horizontal slats\" on your left side?", "option": [], "answer": "5", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the medium-sized terracottthe flowerpot with the rounded shape\" and \"the medium-sized, white metal radiator with horizontal slats\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_47331970/Validation_47331970_47331970_167411.140.png", "Validation_47331970/Validation_47331970_47331970_167413.139.png", "Validation_47331970/Validation_47331970_47331970_167415.138.png", "Validation_47331970/Validation_47331970_47331970_167417.137.png", "Validation_47331970/Validation_47331970_47331970_167419.153.png" ], "turn_id": 6, "scan_id": "Validation_47331970", "origin_question": "What will be the first-time appearance order of the following categories: oven, radiator, refrigerator?", "option": [ "oven, radiator, refrigerator", "refrigerator, oven, radiator", "radiator, oven, refrigerator", "refrigerator, radiator, oven" ], "answer": "refrigerator, oven, radiator", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: oven, radiator, refrigerator? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['oven, radiator, refrigerator', 'refrigerator, oven, radiator', 'radiator, oven, refrigerator', 'refrigerator, radiator, oven'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47331970/Validation_47331970_47331970_167421.152.png", "Validation_47331970/Validation_47331970_47331970_167423.152.png", "Validation_47331970/Validation_47331970_47331970_167425.151.png", "Validation_47331970/Validation_47331970_47331970_167427.150.png", "Validation_47331970/Validation_47331970_47331970_167429.149.png" ], "turn_id": 7, "scan_id": "Validation_47331970", "origin_question": "Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the matte grey and white wooden shelf\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the matte grey and white wooden shelf\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_47331970/Validation_47331970_47331970_167431.148.png", "Validation_47331970/Validation_47331970_47331970_167433.147.png", "Validation_47331970/Validation_47331970_47331970_167435.147.png", "Validation_47331970/Validation_47331970_47331970_167437.146.png", "Validation_47331970/Validation_47331970_47331970_167439.145.png" ], "turn_id": 8, "scan_id": "Validation_47331970", "origin_question": "Which two objects are on the same side of you now? \"the tall, black stainless steel refrigerator\", \"the oven\", \"the medium-sized, white metal radiator with horizontal slats\". ", "option": [ "the tall, black stainless steel refrigerator, the oven", "the tall, black stainless steel refrigerator, the medium-sized, white metal radiator with horizontal slats", "the oven, the medium-sized, white metal radiator with horizontal slats" ], "answer": "the tall, black stainless steel refrigerator, the oven", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the tall, black stainless steel refrigerator\", \"the oven\", \"the medium-sized, white metal radiator with horizontal slats\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the tall, black stainless steel refrigerator, the oven', 'the tall, black stainless steel refrigerator, the medium-sized, white metal radiator with horizontal slats', 'the oven, the medium-sized, white metal radiator with horizontal slats'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_47331970/Validation_47331970_47331970_167441.144.png", "Validation_47331970/Validation_47331970_47331970_167443.143.png", "Validation_47331970/Validation_47331970_47331970_167445.143.png", "Validation_47331970/Validation_47331970_47331970_167447.142.png", "Validation_47331970/Validation_47331970_47331970_167449.141.png" ], "turn_id": 9, "scan_id": "Validation_47331970", "origin_question": "What will be the first-time appearance order of the following categories: dishwasher, oven, window?", "option": [ "oven, dishwasher, window", "window, dishwasher, oven", "dishwasher, window, oven", "dishwasher, oven, window" ], "answer": "oven, dishwasher, window", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: dishwasher, oven, window? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['oven, dishwasher, window', 'window, dishwasher, oven', 'dishwasher, window, oven', 'dishwasher, oven, window'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_47331970/Validation_47331970_47331970_167451.140.png", "Validation_47331970/Validation_47331970_47331970_167453.139.png", "Validation_47331970/Validation_47331970_47331970_167455.139.png", "Validation_47331970/Validation_47331970_47331970_167457.138.png", "Validation_47331970/Validation_47331970_47331970_167459.137.png" ], "turn_id": 10, "scan_id": "Validation_47331970", "origin_question": "What will be the first-time appearance order of the following categories: couch, oven, refrigerator?", "option": [ "refrigerator, couch, oven", "couch, oven, refrigerator", "refrigerator, oven, couch", "oven, couch, refrigerator" ], "answer": "refrigerator, oven, couch", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: couch, oven, refrigerator? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['refrigerator, couch, oven', 'couch, oven, refrigerator', 'refrigerator, oven, couch', 'oven, couch, refrigerator'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47332886/Validation_47332886_47332886_7542.275.png", "Validation_47332886/Validation_47332886_47332886_7544.258.png", "Validation_47332886/Validation_47332886_47332886_7546.257.png", "Validation_47332886/Validation_47332886_47332886_7548.256.png" ], "turn_id": 1, "scan_id": "Validation_47332886", "origin_question": "Remember, have you seen any stapler(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any stapler(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47332886/Validation_47332886_47332886_7550.256.png", "Validation_47332886/Validation_47332886_47332886_7552.271.png", "Validation_47332886/Validation_47332886_47332886_7554.271.png", "Validation_47332886/Validation_47332886_47332886_7556.270.png" ], "turn_id": 2, "scan_id": "Validation_47332886", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the muted red rectangular backpack made of synthetic fabric\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the muted red rectangular backpack made of synthetic fabric\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_47332886/Validation_47332886_47332886_7558.269.png", "Validation_47332886/Validation_47332886_47332886_7560.268.png", "Validation_47332886/Validation_47332886_47332886_7562.267.png", "Validation_47332886/Validation_47332886_47332886_7564.266.png" ], "turn_id": 3, "scan_id": "Validation_47332886", "origin_question": "Which two objects are on the same side of you now? \"the light-toned wooden cabinet with closed doors\", \"the sleek black monitor with the rectangular display\", \"the sleek, red round chair with the metallic base\". ", "option": [ "the light-toned wooden cabinet with closed doors, the sleek black monitor with the rectangular display", "the light-toned wooden cabinet with closed doors, the sleek, red round chair with the metallic base", "the sleek black monitor with the rectangular display, the sleek, red round chair with the metallic base" ], "answer": "the sleek black monitor with the rectangular display, the sleek, red round chair with the metallic base", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the light-toned wooden cabinet with closed doors\", \"the sleek black monitor with the rectangular display\", \"the sleek, red round chair with the metallic base\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the light-toned wooden cabinet with closed doors, the sleek black monitor with the rectangular display', 'the light-toned wooden cabinet with closed doors, the sleek, red round chair with the metallic base', 'the sleek black monitor with the rectangular display, the sleek, red round chair with the metallic base'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_47332886/Validation_47332886_47332886_7566.266.png", "Validation_47332886/Validation_47332886_47332886_7568.265.png", "Validation_47332886/Validation_47332886_47332886_7570.264.png", "Validation_47332886/Validation_47332886_47332886_7572.263.png" ], "turn_id": 4, "scan_id": "Validation_47332886", "origin_question": "Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "22.45", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47332886/Validation_47332886_47332886_7574.262.png", "Validation_47332886/Validation_47332886_47332886_7576.262.png", "Validation_47332886/Validation_47332886_47332886_7578.261.png", "Validation_47332886/Validation_47332886_47332886_7580.260.png" ], "turn_id": 5, "scan_id": "Validation_47332886", "origin_question": "", "option": [], "answer": "", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_47332886/Validation_47332886_47332886_7582.259.png", "Validation_47332886/Validation_47332886_47332886_7584.258.png", "Validation_47332886/Validation_47332886_47332886_7586.258.png", "Validation_47332886/Validation_47332886_47332886_7588.257.png" ], "turn_id": 6, "scan_id": "Validation_47332886", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "1.18", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42899720/Validation_42899720_42899720_797614.190.png", "Validation_42899720/Validation_42899720_42899720_797616.189.png", "Validation_42899720/Validation_42899720_42899720_797618.188.png", "Validation_42899720/Validation_42899720_42899720_797620.187.png" ], "turn_id": 1, "scan_id": "Validation_42899720", "origin_question": "Remember, have you seen any toilet(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any toilet(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_42899720/Validation_42899720_42899720_797622.187.png", "Validation_42899720/Validation_42899720_42899720_797624.186.png", "Validation_42899720/Validation_42899720_42899720_797626.185.png", "Validation_42899720/Validation_42899720_42899720_797628.184.png" ], "turn_id": 2, "scan_id": "Validation_42899720", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the sleek rectangular white ceramic sink\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the sleek rectangular white ceramic sink\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_42899720/Validation_42899720_42899720_797630.183.png", "Validation_42899720/Validation_42899720_42899720_797632.182.png", "Validation_42899720/Validation_42899720_42899720_797634.182.png", "Validation_42899720/Validation_42899720_42899720_797636.181.png" ], "turn_id": 3, "scan_id": "Validation_42899720", "origin_question": "Which direction is the small, silver cylindrical stainless steel bin in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the small, silver cylindrical stainless steel bin in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_42899720/Validation_42899720_42899720_797638.180.png", "Validation_42899720/Validation_42899720_42899720_797640.179.png", "Validation_42899720/Validation_42899720_42899720_797642.178.png", "Validation_42899720/Validation_42899720_42899720_797644.178.png" ], "turn_id": 4, "scan_id": "Validation_42899720", "origin_question": "In which turn were you closest to \"the compact oval-shaped white porcelain toilet\"? ", "option": [], "answer": "2", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the compact oval-shaped white porcelain toilet\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_42899720/Validation_42899720_42899720_797646.177.png", "Validation_42899720/Validation_42899720_42899720_797648.176.png", "Validation_42899720/Validation_42899720_42899720_797650.192.png", "Validation_42899720/Validation_42899720_42899720_797652.191.png" ], "turn_id": 5, "scan_id": "Validation_42899720", "origin_question": "Which two objects are on the same side of you now? \"the sleek rectangular white ceramic sink\", \"the small, silver cylindrical stainless steel bin\", \"the sleek, white rectangular door with the metallic handle\". ", "option": [ "the sleek rectangular white ceramic sink, the small, silver cylindrical stainless steel bin", "the sleek rectangular white ceramic sink, the sleek, white rectangular door with the metallic handle", "the small, silver cylindrical stainless steel bin, the sleek, white rectangular door with the metallic handle" ], "answer": "the sleek rectangular white ceramic sink, the small, silver cylindrical stainless steel bin", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the sleek rectangular white ceramic sink\", \"the small, silver cylindrical stainless steel bin\", \"the sleek, white rectangular door with the metallic handle\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the sleek rectangular white ceramic sink, the small, silver cylindrical stainless steel bin', 'the sleek rectangular white ceramic sink, the sleek, white rectangular door with the metallic handle', 'the small, silver cylindrical stainless steel bin, the sleek, white rectangular door with the metallic handle'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_42899720/Validation_42899720_42899720_797654.190.png", "Validation_42899720/Validation_42899720_42899720_797656.189.png", "Validation_42899720/Validation_42899720_42899720_797658.188.png", "Validation_42899720/Validation_42899720_42899720_797660.188.png" ], "turn_id": 6, "scan_id": "Validation_42899720", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the sleek rectangular white ceramic sink\", Y: \"the sleek, white rectangular door with the metallic handle\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the sleek rectangular white ceramic sink\", Y: \"the sleek, white rectangular door with the metallic handle\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_45261620/Validation_45261620_45261620_54932.305.png", "Validation_45261620/Validation_45261620_45261620_54934.288.png", "Validation_45261620/Validation_45261620_45261620_54936.287.png", "Validation_45261620/Validation_45261620_45261620_54938.286.png" ], "turn_id": 1, "scan_id": "Validation_45261620", "origin_question": "Remember, have you seen any laptop(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any laptop(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "Validation_45261620/Validation_45261620_45261620_54940.285.png", "Validation_45261620/Validation_45261620_45261620_54942.285.png", "Validation_45261620/Validation_45261620_45261620_54944.284.png", "Validation_45261620/Validation_45261620_45261620_54946.283.png" ], "turn_id": 2, "scan_id": "Validation_45261620", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"an elongated rectangular mirror with the metallic frame\"; \"the small rectangular picture with muted blue and yellow tones\"; \"the plant with vibrant yellow blossoms and green leaves\" ", "option": [ "an elongated rectangular mirror with the metallic frame", "the small rectangular picture with muted blue and yellow tones", "the plant with vibrant yellow blossoms and green leaves" ], "answer": "an elongated rectangular mirror with the metallic frame", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"an elongated rectangular mirror with the metallic frame\"; \"the small rectangular picture with muted blue and yellow tones\"; \"the plant with vibrant yellow blossoms and green leaves\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['an elongated rectangular mirror with the metallic frame', 'the small rectangular picture with muted blue and yellow tones', 'the plant with vibrant yellow blossoms and green leaves'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_45261620/Validation_45261620_45261620_54948.282.png", "Validation_45261620/Validation_45261620_45261620_54950.281.png", "Validation_45261620/Validation_45261620_45261620_54952.281.png", "Validation_45261620/Validation_45261620_45261620_54954.280.png" ], "turn_id": 3, "scan_id": "Validation_45261620", "origin_question": "In which turn were you closest to \"the small rectangular picture with muted blue and yellow tones\"? ", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the small rectangular picture with muted blue and yellow tones\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_45261620/Validation_45261620_45261620_54956.279.png", "Validation_45261620/Validation_45261620_45261620_54958.278.png", "Validation_45261620/Validation_45261620_45261620_54960.277.png", "Validation_45261620/Validation_45261620_45261620_54962.293.png" ], "turn_id": 4, "scan_id": "Validation_45261620", "origin_question": "At the end of which turn were both of \"the plant with vibrant yellow blossoms and green leaves\" and \"the sleek rectangular metal and glass chandelier\" on your right side?", "option": [], "answer": "3", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the plant with vibrant yellow blossoms and green leaves\" and \"the sleek rectangular metal and glass chandelier\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_45261620/Validation_45261620_45261620_54964.293.png", "Validation_45261620/Validation_45261620_45261620_54966.292.png", "Validation_45261620/Validation_45261620_45261620_54968.291.png", "Validation_45261620/Validation_45261620_45261620_54970.290.png" ], "turn_id": 5, "scan_id": "Validation_45261620", "origin_question": "Please recall \"the black monitor with the thin frame and matte screen\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.9", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the black monitor with the thin frame and matte screen\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_45261620/Validation_45261620_45261620_54972.289.png", "Validation_45261620/Validation_45261620_45261620_54974.289.png", "Validation_45261620/Validation_45261620_45261620_54976.288.png", "Validation_45261620/Validation_45261620_45261620_54978.287.png" ], "turn_id": 6, "scan_id": "Validation_45261620", "origin_question": "How far is your current position from where you were at the end of the turn 5? (in meters)", "option": [], "answer": "1.0", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 5? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_45261620/Validation_45261620_45261620_54980.286.png", "Validation_45261620/Validation_45261620_45261620_54982.285.png", "Validation_45261620/Validation_45261620_45261620_54984.285.png", "Validation_45261620/Validation_45261620_45261620_54986.284.png" ], "turn_id": 7, "scan_id": "Validation_45261620", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the sleek rectangular metal and glass chandelier\" from your position?", "option": [], "answer": "176.64", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the sleek rectangular metal and glass chandelier\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "Validation_42898854/Validation_42898854_42898854_196183.722.png", "Validation_42898854/Validation_42898854_42898854_196185.721.png", "Validation_42898854/Validation_42898854_42898854_196187.721.png", "Validation_42898854/Validation_42898854_42898854_196189.720.png", "Validation_42898854/Validation_42898854_42898854_196191.719.png" ], "turn_id": 1, "scan_id": "Validation_42898854", "origin_question": "Remember, how many door(s) have you seen so far? ", "option": [], "answer": "3", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many door(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_42898854/Validation_42898854_42898854_196193.718.png", "Validation_42898854/Validation_42898854_42898854_196195.717.png", "Validation_42898854/Validation_42898854_42898854_196197.717.png", "Validation_42898854/Validation_42898854_42898854_196199.716.png", "Validation_42898854/Validation_42898854_42898854_196201.715.png" ], "turn_id": 2, "scan_id": "Validation_42898854", "origin_question": "In which turn were you closest to \"the tall green plant with colorful ornaments\"? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the tall green plant with colorful ornaments\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_42898854/Validation_42898854_42898854_196203.714.png", "Validation_42898854/Validation_42898854_42898854_196205.714.png", "Validation_42898854/Validation_42898854_42898854_196207.713.png", "Validation_42898854/Validation_42898854_42898854_196209.712.png", "Validation_42898854/Validation_42898854_42898854_196211.711.png" ], "turn_id": 3, "scan_id": "Validation_42898854", "origin_question": "Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Right", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_42898854/Validation_42898854_42898854_196213.710.png", "Validation_42898854/Validation_42898854_42898854_196215.726.png", "Validation_42898854/Validation_42898854_42898854_196217.725.png", "Validation_42898854/Validation_42898854_42898854_196219.725.png", "Validation_42898854/Validation_42898854_42898854_196221.724.png" ], "turn_id": 4, "scan_id": "Validation_42898854", "origin_question": "Which direction is the glossy white ceramic vase with the rounded shape in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the glossy white ceramic vase with the rounded shape in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_42898854/Validation_42898854_42898854_196223.723.png", "Validation_42898854/Validation_42898854_42898854_196225.722.png", "Validation_42898854/Validation_42898854_42898854_196227.721.png", "Validation_42898854/Validation_42898854_42898854_196229.721.png", "Validation_42898854/Validation_42898854_42898854_196231.720.png" ], "turn_id": 5, "scan_id": "Validation_42898854", "origin_question": "What will be the first-time appearance order of the following categories: basket, couch, shelf?", "option": [ "shelf, basket, couch", "shelf, couch, basket", "basket, couch, shelf", "couch, shelf, basket" ], "answer": "couch, shelf, basket", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: basket, couch, shelf? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['shelf, basket, couch', 'shelf, couch, basket', 'basket, couch, shelf', 'couch, shelf, basket'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_42898854/Validation_42898854_42898854_196233.719.png", "Validation_42898854/Validation_42898854_42898854_196235.718.png", "Validation_42898854/Validation_42898854_42898854_196237.717.png", "Validation_42898854/Validation_42898854_42898854_196239.717.png", "Validation_42898854/Validation_42898854_42898854_196241.716.png" ], "turn_id": 6, "scan_id": "Validation_42898854", "origin_question": "What will be the first-time appearance order of the following categories: light, monitor, window?", "option": [ "monitor, window, light", "light, window, monitor", "window, light, monitor", "light, monitor, window" ], "answer": "light, window, monitor", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: light, monitor, window? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['monitor, window, light', 'light, window, monitor', 'window, light, monitor', 'light, monitor, window'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_42898854/Validation_42898854_42898854_196243.715.png", "Validation_42898854/Validation_42898854_42898854_196245.714.png", "Validation_42898854/Validation_42898854_42898854_196247.714.png", "Validation_42898854/Validation_42898854_42898854_196249.713.png", "Validation_42898854/Validation_42898854_42898854_196251.712.png" ], "turn_id": 7, "scan_id": "Validation_42898854", "origin_question": "Assuming the direction you are facing at the end of the turn 3 is forward, did you move a certain distance forward or backward from that position?", "option": [ "Forward", "Backward" ], "answer": "Backward", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 3 is forward, did you move a certain distance forward or backward from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Forward', 'Backward'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_42898854/Validation_42898854_42898854_196253.711.png", "Validation_42898854/Validation_42898854_42898854_196255.727.png", "Validation_42898854/Validation_42898854_42898854_196257.726.png", "Validation_42898854/Validation_42898854_42898854_196259.725.png", "Validation_42898854/Validation_42898854_42898854_196261.725.png" ], "turn_id": 8, "scan_id": "Validation_42898854", "origin_question": "What will be the first-time appearance order of the following categories: couch, fireplace, oven?", "option": [ "fireplace, oven, couch", "oven, fireplace, couch", "fireplace, couch, oven", "couch, fireplace, oven" ], "answer": "couch, fireplace, oven", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: couch, fireplace, oven? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['fireplace, oven, couch', 'oven, fireplace, couch', 'fireplace, couch, oven', 'couch, fireplace, oven'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_42898854/Validation_42898854_42898854_196263.724.png", "Validation_42898854/Validation_42898854_42898854_196265.723.png", "Validation_42898854/Validation_42898854_42898854_196267.722.png", "Validation_42898854/Validation_42898854_42898854_196269.722.png", "Validation_42898854/Validation_42898854_42898854_196271.721.png" ], "turn_id": 9, "scan_id": "Validation_42898854", "origin_question": "In which turn were you closest to \"the tall green plant with colorful ornaments\"? ", "option": [], "answer": "9", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the tall green plant with colorful ornaments\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_42898854/Validation_42898854_42898854_196273.720.png", "Validation_42898854/Validation_42898854_42898854_196275.719.png", "Validation_42898854/Validation_42898854_42898854_196277.718.png", "Validation_42898854/Validation_42898854_42898854_196279.717.png", "Validation_42898854/Validation_42898854_42898854_196281.717.png" ], "turn_id": 10, "scan_id": "Validation_42898854", "origin_question": "By the end of the 4 turn, were you getting closer to or farther away from these two objects: X: \"the modern rectangular sink with the glossy surface\", Y: \"the dishwasher\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 4 turn, were you getting closer to or farther away from these two objects: X: \"the modern rectangular sink with the glossy surface\", Y: \"the dishwasher\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_48458747/Validation_48458747_48458747_139029.899.png", "Validation_48458747/Validation_48458747_48458747_139031.899.png", "Validation_48458747/Validation_48458747_48458747_139033.898.png", "Validation_48458747/Validation_48458747_48458747_139035.897.png" ], "turn_id": 1, "scan_id": "Validation_48458747", "origin_question": "Remember, have you seen any monitor(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any monitor(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_48458747/Validation_48458747_48458747_139037.896.png", "Validation_48458747/Validation_48458747_48458747_139039.895.png", "Validation_48458747/Validation_48458747_48458747_139041.894.png", "Validation_48458747/Validation_48458747_48458747_139043.894.png" ], "turn_id": 2, "scan_id": "Validation_48458747", "origin_question": "In which turn did you see both \"the sleek rectangular stainless steel sink\" and \"the compact white microwave made of metal and plastic\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the sleek rectangular stainless steel sink\" and \"the compact white microwave made of metal and plastic\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_48458747/Validation_48458747_48458747_139045.893.png", "Validation_48458747/Validation_48458747_48458747_139047.892.png", "Validation_48458747/Validation_48458747_48458747_139049.891.png", "Validation_48458747/Validation_48458747_48458747_139051.891.png" ], "turn_id": 3, "scan_id": "Validation_48458747", "origin_question": "Is the sleek black rectangular monitor made of plastic to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the sleek black rectangular monitor made of plastic to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_48458747/Validation_48458747_48458747_139053.890.png", "Validation_48458747/Validation_48458747_48458747_139055.889.png", "Validation_48458747/Validation_48458747_48458747_139057.888.png", "Validation_48458747/Validation_48458747_48458747_139059.887.png" ], "turn_id": 4, "scan_id": "Validation_48458747", "origin_question": "In which turn did you see both \"the sleek black rectangular monitor made of plastic\" and \"the sleek rectangular stainless steel sink\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the sleek black rectangular monitor made of plastic\" and \"the sleek rectangular stainless steel sink\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_48458747/Validation_48458747_48458747_139061.887.png", "Validation_48458747/Validation_48458747_48458747_139063.886.png", "Validation_48458747/Validation_48458747_48458747_139065.885.png", "Validation_48458747/Validation_48458747_48458747_139067.884.png" ], "turn_id": 5, "scan_id": "Validation_48458747", "origin_question": "", "option": [], "answer": "", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_48458747/Validation_48458747_48458747_139069.883.png", "Validation_48458747/Validation_48458747_48458747_139071.883.png", "Validation_48458747/Validation_48458747_48458747_139073.899.png", "Validation_48458747/Validation_48458747_48458747_139075.898.png" ], "turn_id": 6, "scan_id": "Validation_48458747", "origin_question": "At the end of which turn were both of \"the compact white microwave made of metal and plastic\" and \"the modern black kettle with the glossy finish\" on your right side?", "option": [], "answer": "2", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the compact white microwave made of metal and plastic\" and \"the modern black kettle with the glossy finish\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "Validation_48458747/Validation_48458747_48458747_139077.897.png", "Validation_48458747/Validation_48458747_48458747_139079.896.png", "Validation_48458747/Validation_48458747_48458747_139081.895.png", "Validation_48458747/Validation_48458747_48458747_139083.895.png" ], "turn_id": 7, "scan_id": "Validation_48458747", "origin_question": "Among these three objects(\"the stainless steel built-in oven with the glass door\",\"the sleek, modern black glass stove\",\"the sleek rectangular stainless steel sink\" ), which one is the farthest from you?", "option": [ "the stainless steel built-in oven with the glass door", "the sleek, modern black glass stove", "the sleek rectangular stainless steel sink" ], "answer": "the sleek rectangular stainless steel sink", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects(\"the stainless steel built-in oven with the glass door\",\"the sleek, modern black glass stove\",\"the sleek rectangular stainless steel sink\" ), which one is the farthest from you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the stainless steel built-in oven with the glass door', 'the sleek, modern black glass stove', 'the sleek rectangular stainless steel sink'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_48018559/Validation_48018559_48018559_3341.866.png", "Validation_48018559/Validation_48018559_48018559_3343.349.png", "Validation_48018559/Validation_48018559_48018559_3344.848.png", "Validation_48018559/Validation_48018559_48018559_3346.348.png" ], "turn_id": 1, "scan_id": "Validation_48018559", "origin_question": "Remember, have you seen any window(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any window(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "Validation_48018559/Validation_48018559_48018559_3347.847.png", "Validation_48018559/Validation_48018559_48018559_3349.347.png", "Validation_48018559/Validation_48018559_48018559_3350.846.png", "Validation_48018559/Validation_48018559_48018559_3352.345.png" ], "turn_id": 2, "scan_id": "Validation_48018559", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the frosted glass dome-shaped light\"; \"the refrigerator\"; \"the compact wooden square stool in the natural wood tone\" ", "option": [ "the frosted glass dome-shaped light", "the refrigerator", "the compact wooden square stool in the natural wood tone" ], "answer": "the frosted glass dome-shaped light", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the frosted glass dome-shaped light\"; \"the refrigerator\"; \"the compact wooden square stool in the natural wood tone\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the frosted glass dome-shaped light', 'the refrigerator', 'the compact wooden square stool in the natural wood tone'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_48018559/Validation_48018559_48018559_3353.845.png", "Validation_48018559/Validation_48018559_48018559_3355.344.png", "Validation_48018559/Validation_48018559_48018559_3356.843.png", "Validation_48018559/Validation_48018559_48018559_3358.343.png" ], "turn_id": 3, "scan_id": "Validation_48018559", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_48018559/Validation_48018559_48018559_3359.842.png", "Validation_48018559/Validation_48018559_48018559_3361.342.png", "Validation_48018559/Validation_48018559_48018559_3362.841.png", "Validation_48018559/Validation_48018559_48018559_3364.340.png" ], "turn_id": 4, "scan_id": "Validation_48018559", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_48018559/Validation_48018559_48018559_3365.840.png", "Validation_48018559/Validation_48018559_48018559_3367.339.png", "Validation_48018559/Validation_48018559_48018559_3368.855.png", "Validation_48018559/Validation_48018559_48018559_3370.355.png" ], "turn_id": 5, "scan_id": "Validation_48018559", "origin_question": "When did you first discover \"the frosted glass dome-shaped light\"? (index of the turn)", "option": [], "answer": "5", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the frosted glass dome-shaped light\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_48018559/Validation_48018559_48018559_3371.854.png", "Validation_48018559/Validation_48018559_48018559_3373.354.png", "Validation_48018559/Validation_48018559_48018559_3374.853.png", "Validation_48018559/Validation_48018559_48018559_3376.352.png" ], "turn_id": 6, "scan_id": "Validation_48018559", "origin_question": "In which turn were you closest to \"the compact wooden square stool in the natural wood tone\"? ", "option": [], "answer": "4", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the compact wooden square stool in the natural wood tone\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47430483/Validation_47430483_47430483_310492.870.png", "Validation_47430483/Validation_47430483_47430483_310494.869.png", "Validation_47430483/Validation_47430483_47430483_310496.869.png", "Validation_47430483/Validation_47430483_47430483_310498.868.png", "Validation_47430483/Validation_47430483_47430483_310500.867.png" ], "turn_id": 1, "scan_id": "Validation_47430483", "origin_question": "Remember, have you seen any door(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any door(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47430483/Validation_47430483_47430483_310502.866.png", "Validation_47430483/Validation_47430483_47430483_310504.865.png", "Validation_47430483/Validation_47430483_47430483_310506.865.png", "Validation_47430483/Validation_47430483_47430483_310508.864.png", "Validation_47430483/Validation_47430483_47430483_310510.863.png" ], "turn_id": 2, "scan_id": "Validation_47430483", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47430483/Validation_47430483_47430483_310512.862.png", "Validation_47430483/Validation_47430483_47430483_310514.861.png", "Validation_47430483/Validation_47430483_47430483_310516.861.png", "Validation_47430483/Validation_47430483_47430483_310518.860.png", "Validation_47430483/Validation_47430483_47430483_310520.859.png" ], "turn_id": 3, "scan_id": "Validation_47430483", "origin_question": "", "option": [], "answer": "", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_47430483/Validation_47430483_47430483_310522.875.png", "Validation_47430483/Validation_47430483_47430483_310524.874.png", "Validation_47430483/Validation_47430483_47430483_310526.873.png", "Validation_47430483/Validation_47430483_47430483_310528.872.png", "Validation_47430483/Validation_47430483_47430483_310530.872.png" ], "turn_id": 4, "scan_id": "Validation_47430483", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "1.67", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47430483/Validation_47430483_47430483_310532.871.png", "Validation_47430483/Validation_47430483_47430483_310534.870.png", "Validation_47430483/Validation_47430483_47430483_310536.869.png", "Validation_47430483/Validation_47430483_47430483_310538.868.png", "Validation_47430483/Validation_47430483_47430483_310540.867.png" ], "turn_id": 5, "scan_id": "Validation_47430483", "origin_question": "In which turn were you closest to \"the large wooden window with the brown frame\"? ", "option": [], "answer": "5", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the large wooden window with the brown frame\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47430483/Validation_47430483_47430483_310542.867.png", "Validation_47430483/Validation_47430483_47430483_310544.866.png", "Validation_47430483/Validation_47430483_47430483_310546.865.png", "Validation_47430483/Validation_47430483_47430483_310548.864.png", "Validation_47430483/Validation_47430483_47430483_310550.863.png" ], "turn_id": 6, "scan_id": "Validation_47430483", "origin_question": "", "option": [], "answer": "", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47430483/Validation_47430483_47430483_310552.863.png", "Validation_47430483/Validation_47430483_47430483_310554.862.png", "Validation_47430483/Validation_47430483_47430483_310556.861.png", "Validation_47430483/Validation_47430483_47430483_310558.860.png", "Validation_47430483/Validation_47430483_47430483_310560.859.png" ], "turn_id": 7, "scan_id": "Validation_47430483", "origin_question": "", "option": [], "answer": "", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_47430483/Validation_47430483_47430483_310562.859.png", "Validation_47430483/Validation_47430483_47430483_310564.858.png", "Validation_47430483/Validation_47430483_47430483_310566.874.png", "Validation_47430483/Validation_47430483_47430483_310568.873.png", "Validation_47430483/Validation_47430483_47430483_310570.872.png" ], "turn_id": 8, "scan_id": "Validation_47430483", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the large wooden window with the brown frame\", Y: \"the plum-colored rectangular soft pillow\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the large wooden window with the brown frame\", Y: \"the plum-colored rectangular soft pillow\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47430483/Validation_47430483_47430483_310572.871.png", "Validation_47430483/Validation_47430483_47430483_310574.870.png", "Validation_47430483/Validation_47430483_47430483_310576.870.png", "Validation_47430483/Validation_47430483_47430483_310578.869.png", "Validation_47430483/Validation_47430483_47430483_310580.868.png" ], "turn_id": 9, "scan_id": "Validation_47430483", "origin_question": "Which direction is the large wooden window with the brown frame in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the large wooden window with the brown frame in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_41159529/Validation_41159529_41159529_39091.836.png", "Validation_41159529/Validation_41159529_41159529_39093.818.png", "Validation_41159529/Validation_41159529_41159529_39095.817.png", "Validation_41159529/Validation_41159529_41159529_39097.817.png" ], "turn_id": 1, "scan_id": "Validation_41159529", "origin_question": "Remember, have you seen any car(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any car(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_41159529/Validation_41159529_41159529_39099.816.png", "Validation_41159529/Validation_41159529_41159529_39101.815.png", "Validation_41159529/Validation_41159529_41159529_39103.814.png", "Validation_41159529/Validation_41159529_41159529_39105.813.png" ], "turn_id": 2, "scan_id": "Validation_41159529", "origin_question": "When did you first discover \"the light\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the light\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_41159529/Validation_41159529_41159529_39107.812.png", "Validation_41159529/Validation_41159529_41159529_39109.812.png", "Validation_41159529/Validation_41159529_41159529_39111.811.png", "Validation_41159529/Validation_41159529_41159529_39113.810.png" ], "turn_id": 3, "scan_id": "Validation_41159529", "origin_question": "When did you first discover \"the light wood door with the soft tan finish\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the light wood door with the soft tan finish\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_41159529/Validation_41159529_41159529_39115.809.png", "Validation_41159529/Validation_41159529_41159529_39117.808.png", "Validation_41159529/Validation_41159529_41159529_39119.808.png", "Validation_41159529/Validation_41159529_41159529_39121.807.png" ], "turn_id": 4, "scan_id": "Validation_41159529", "origin_question": "", "option": [], "answer": "", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_41159529/Validation_41159529_41159529_39123.806.png", "Validation_41159529/Validation_41159529_41159529_39125.805.png", "Validation_41159529/Validation_41159529_41159529_39127.804.png", "Validation_41159529/Validation_41159529_41159529_39129.820.png" ], "turn_id": 5, "scan_id": "Validation_41159529", "origin_question": "Which direction is the table in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the table in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_41159529/Validation_41159529_41159529_39131.819.png", "Validation_41159529/Validation_41159529_41159529_39133.818.png", "Validation_41159529/Validation_41159529_41159529_39135.817.png", "Validation_41159529/Validation_41159529_41159529_39137.817.png" ], "turn_id": 6, "scan_id": "Validation_41159529", "origin_question": "What will be the first-time appearance order of the following categories: door, light, table?", "option": [ "table, door, light", "door, table, light", "table, light, door", "door, light, table" ], "answer": "door, table, light", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: door, light, table? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['table, door, light', 'door, table, light', 'table, light, door', 'door, light, table'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_41159529/Validation_41159529_41159529_39139.816.png", "Validation_41159529/Validation_41159529_41159529_39141.815.png", "Validation_41159529/Validation_41159529_41159529_39143.814.png", "Validation_41159529/Validation_41159529_41159529_39145.813.png" ], "turn_id": 7, "scan_id": "Validation_41159529", "origin_question": "Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the stainless steel kettle with the classic rounded shape\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the stainless steel kettle with the classic rounded shape\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_45260905/Validation_45260905_45260905_1436.613.png", "Validation_45260905/Validation_45260905_45260905_1438.595.png", "Validation_45260905/Validation_45260905_45260905_1440.595.png", "Validation_45260905/Validation_45260905_45260905_1442.594.png" ], "turn_id": 1, "scan_id": "Validation_45260905", "origin_question": "Remember, have you seen any purse(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any purse(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_45260905/Validation_45260905_45260905_1444.593.png", "Validation_45260905/Validation_45260905_45260905_1446.592.png", "Validation_45260905/Validation_45260905_45260905_1448.592.png", "Validation_45260905/Validation_45260905_45260905_1450.591.png" ], "turn_id": 2, "scan_id": "Validation_45260905", "origin_question": "In which turn were you closest to \"the rectangular blue and white cotton pillow\"? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the rectangular blue and white cotton pillow\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_45260905/Validation_45260905_45260905_1452.590.png", "Validation_45260905/Validation_45260905_45260905_1454.589.png", "Validation_45260905/Validation_45260905_45260905_1456.588.png", "Validation_45260905/Validation_45260905_45260905_1458.588.png" ], "turn_id": 3, "scan_id": "Validation_45260905", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_45260905/Validation_45260905_45260905_1460.587.png", "Validation_45260905/Validation_45260905_45260905_1462.586.png", "Validation_45260905/Validation_45260905_45260905_1464.602.png", "Validation_45260905/Validation_45260905_45260905_1466.601.png" ], "turn_id": 4, "scan_id": "Validation_45260905", "origin_question": "Is the sleek black monitor with the matte finish to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the sleek black monitor with the matte finish to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_45260905/Validation_45260905_45260905_1468.600.png", "Validation_45260905/Validation_45260905_45260905_1470.600.png", "Validation_45260905/Validation_45260905_45260905_1472.599.png", "Validation_45260905/Validation_45260905_45260905_1474.598.png" ], "turn_id": 5, "scan_id": "Validation_45260905", "origin_question": "Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "21.92", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_45260905/Validation_45260905_45260905_1476.597.png", "Validation_45260905/Validation_45260905_45260905_1478.596.png", "Validation_45260905/Validation_45260905_45260905_1480.596.png", "Validation_45260905/Validation_45260905_45260905_1482.595.png" ], "turn_id": 6, "scan_id": "Validation_45260905", "origin_question": "In which turn did you see both \"the fireplace\" and \"the sleek black monitor with the matte finish\" simultaneously? ", "option": [], "answer": "3", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the fireplace\" and \"the sleek black monitor with the matte finish\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_45260905/Validation_45260905_45260905_1484.594.png", "Validation_45260905/Validation_45260905_45260905_1486.593.png", "Validation_45260905/Validation_45260905_45260905_1488.592.png", "Validation_45260905/Validation_45260905_45260905_1490.592.png" ], "turn_id": 7, "scan_id": "Validation_45260905", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "126.92", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_45260905/Validation_45260905_45260905_1492.591.png", "Validation_45260905/Validation_45260905_45260905_1494.590.png", "Validation_45260905/Validation_45260905_45260905_1496.589.png", "Validation_45260905/Validation_45260905_45260905_1498.589.png" ], "turn_id": 8, "scan_id": "Validation_45260905", "origin_question": "Please recall \"the lush indoor plant with dense green ovate leaves\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "3.47", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the lush indoor plant with dense green ovate leaves\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_45260905/Validation_45260905_45260905_1500.588.png", "Validation_45260905/Validation_45260905_45260905_1502.603.png", "Validation_45260905/Validation_45260905_45260905_1504.603.png", "Validation_45260905/Validation_45260905_45260905_1506.602.png" ], "turn_id": 9, "scan_id": "Validation_45260905", "origin_question": "Using your orientation at the end of turn 5 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 9 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 5 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_48018375/Validation_48018375_48018375_390050.848.png", "Validation_48018375/Validation_48018375_48018375_390052.830.png", "Validation_48018375/Validation_48018375_48018375_390054.830.png", "Validation_48018375/Validation_48018375_48018375_390056.829.png", "Validation_48018375/Validation_48018375_48018375_390058.828.png" ], "turn_id": 1, "scan_id": "Validation_48018375", "origin_question": "Remember, have you seen any light(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any light(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_48018375/Validation_48018375_48018375_390060.827.png", "Validation_48018375/Validation_48018375_48018375_390062.843.png", "Validation_48018375/Validation_48018375_48018375_390064.842.png", "Validation_48018375/Validation_48018375_48018375_390066.841.png", "Validation_48018375/Validation_48018375_48018375_390068.840.png" ], "turn_id": 2, "scan_id": "Validation_48018375", "origin_question": "In which turn did you see both \"the medium-sized wooden shelf with the warm brown finish\" and \"the sleek rectangular monitor with the thin black bezel\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the medium-sized wooden shelf with the warm brown finish\" and \"the sleek rectangular monitor with the thin black bezel\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_48018375/Validation_48018375_48018375_390070.840.png", "Validation_48018375/Validation_48018375_48018375_390072.839.png", "Validation_48018375/Validation_48018375_48018375_390074.838.png", "Validation_48018375/Validation_48018375_48018375_390076.837.png", "Validation_48018375/Validation_48018375_48018375_390078.836.png" ], "turn_id": 3, "scan_id": "Validation_48018375", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the rich brown wooden cuboidal box\" from your position?", "option": [], "answer": "179.51", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the rich brown wooden cuboidal box\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_48018375/Validation_48018375_48018375_390080.835.png", "Validation_48018375/Validation_48018375_48018375_390082.835.png", "Validation_48018375/Validation_48018375_48018375_390084.834.png", "Validation_48018375/Validation_48018375_48018375_390086.833.png", "Validation_48018375/Validation_48018375_48018375_390088.832.png" ], "turn_id": 4, "scan_id": "Validation_48018375", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the vintage double-globe light with frosted glass orbs\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the vintage double-globe light with frosted glass orbs\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "Validation_48018375/Validation_48018375_48018375_390090.831.png", "Validation_48018375/Validation_48018375_48018375_390092.831.png", "Validation_48018375/Validation_48018375_48018375_390094.830.png", "Validation_48018375/Validation_48018375_48018375_390096.829.png", "Validation_48018375/Validation_48018375_48018375_390098.828.png" ], "turn_id": 5, "scan_id": "Validation_48018375", "origin_question": "Among these three objects(\"the vintage double-globe light with frosted glass orbs\",\"the sink\",\"the sleek rectangular light-colored wood table\" ), which one is the farthest from you?", "option": [ "the vintage double-globe light with frosted glass orbs", "the sink", "the sleek rectangular light-colored wood table" ], "answer": "the sink", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the vintage double-globe light with frosted glass orbs\",\"the sink\",\"the sleek rectangular light-colored wood table\" ), which one is the farthest from you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the vintage double-globe light with frosted glass orbs', 'the sink', 'the sleek rectangular light-colored wood table'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_48018375/Validation_48018375_48018375_390100.827.png", "Validation_48018375/Validation_48018375_48018375_390102.826.png", "Validation_48018375/Validation_48018375_48018375_390104.842.png", "Validation_48018375/Validation_48018375_48018375_390106.841.png", "Validation_48018375/Validation_48018375_48018375_390108.841.png" ], "turn_id": 6, "scan_id": "Validation_48018375", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the rich brown wooden cuboidal box\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the rich brown wooden cuboidal box\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_48018375/Validation_48018375_48018375_390110.840.png", "Validation_48018375/Validation_48018375_48018375_390112.839.png", "Validation_48018375/Validation_48018375_48018375_390114.838.png", "Validation_48018375/Validation_48018375_48018375_390116.837.png", "Validation_48018375/Validation_48018375_48018375_390118.836.png" ], "turn_id": 7, "scan_id": "Validation_48018375", "origin_question": "At the end of which turn were both of \"the rich brown wooden cuboidal box\" and \"the white cast iron radiator with vertical columns\" on your left side?", "option": [], "answer": "1", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the rich brown wooden cuboidal box\" and \"the white cast iron radiator with vertical columns\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_48018375/Validation_48018375_48018375_390120.836.png", "Validation_48018375/Validation_48018375_48018375_390122.835.png", "Validation_48018375/Validation_48018375_48018375_390124.834.png", "Validation_48018375/Validation_48018375_48018375_390126.833.png", "Validation_48018375/Validation_48018375_48018375_390128.832.png" ], "turn_id": 8, "scan_id": "Validation_48018375", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the large rectangular gradient carpet with horizontal stripes\" from your position?", "option": [], "answer": "45.78", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the large rectangular gradient carpet with horizontal stripes\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_48018375/Validation_48018375_48018375_390130.831.png", "Validation_48018375/Validation_48018375_48018375_390132.831.png", "Validation_48018375/Validation_48018375_48018375_390134.830.png", "Validation_48018375/Validation_48018375_48018375_390136.829.png", "Validation_48018375/Validation_48018375_48018375_390138.828.png" ], "turn_id": 9, "scan_id": "Validation_48018375", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the rich brown wooden cuboidal box\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the rich brown wooden cuboidal box\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_48018375/Validation_48018375_48018375_390140.827.png", "Validation_48018375/Validation_48018375_48018375_390142.827.png", "Validation_48018375/Validation_48018375_48018375_390144.842.png", "Validation_48018375/Validation_48018375_48018375_390146.841.png", "Validation_48018375/Validation_48018375_48018375_390148.841.png" ], "turn_id": 10, "scan_id": "Validation_48018375", "origin_question": "Assuming the direction you are facing at the end of the turn 4 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Left", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 4 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47430036/Validation_47430036_47430036_19499.302.png", "Validation_47430036/Validation_47430036_47430036_19501.285.png", "Validation_47430036/Validation_47430036_47430036_19503.284.png", "Validation_47430036/Validation_47430036_47430036_19505.283.png" ], "turn_id": 1, "scan_id": "Validation_47430036", "origin_question": "Remember, have you seen any book(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any book(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "Validation_47430036/Validation_47430036_47430036_19507.282.png", "Validation_47430036/Validation_47430036_47430036_19509.281.png", "Validation_47430036/Validation_47430036_47430036_19511.281.png", "Validation_47430036/Validation_47430036_47430036_19513.280.png" ], "turn_id": 2, "scan_id": "Validation_47430036", "origin_question": "Remember, how many remote control(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many remote control(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_47430036/Validation_47430036_47430036_19515.279.png", "Validation_47430036/Validation_47430036_47430036_19517.278.png", "Validation_47430036/Validation_47430036_47430036_19519.277.png", "Validation_47430036/Validation_47430036_47430036_19521.276.png" ], "turn_id": 3, "scan_id": "Validation_47430036", "origin_question": "Is the vibrant, colorful hardcover book with glossy finish to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the vibrant, colorful hardcover book with glossy finish to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_47430036/Validation_47430036_47430036_19523.276.png", "Validation_47430036/Validation_47430036_47430036_19525.275.png", "Validation_47430036/Validation_47430036_47430036_19527.274.png", "Validation_47430036/Validation_47430036_47430036_19529.273.png" ], "turn_id": 4, "scan_id": "Validation_47430036", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the vivid blue globe with the sturdy plastic base\" from your position?", "option": [], "answer": "119.65", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the vivid blue globe with the sturdy plastic base\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_47430036/Validation_47430036_47430036_19531.272.png", "Validation_47430036/Validation_47430036_47430036_19533.272.png", "Validation_47430036/Validation_47430036_47430036_19535.287.png", "Validation_47430036/Validation_47430036_47430036_19537.287.png" ], "turn_id": 5, "scan_id": "Validation_47430036", "origin_question": "Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Right", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_47430036/Validation_47430036_47430036_19539.286.png", "Validation_47430036/Validation_47430036_47430036_19541.285.png", "Validation_47430036/Validation_47430036_47430036_19543.284.png", "Validation_47430036/Validation_47430036_47430036_19545.283.png" ], "turn_id": 6, "scan_id": "Validation_47430036", "origin_question": "By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the vivid blue globe with the sturdy plastic base\", Y: \"the medium to large white radiator\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to Y and farther away from X", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the vivid blue globe with the sturdy plastic base\", Y: \"the medium to large white radiator\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_47430036/Validation_47430036_47430036_19547.283.png", "Validation_47430036/Validation_47430036_47430036_19549.282.png", "Validation_47430036/Validation_47430036_47430036_19551.281.png", "Validation_47430036/Validation_47430036_47430036_19553.280.png" ], "turn_id": 7, "scan_id": "Validation_47430036", "origin_question": "Please recall \"the vivid blue globe with the sturdy plastic base\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.2", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the vivid blue globe with the sturdy plastic base\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_47430036/Validation_47430036_47430036_19555.279.png", "Validation_47430036/Validation_47430036_47430036_19557.279.png", "Validation_47430036/Validation_47430036_47430036_19559.278.png", "Validation_47430036/Validation_47430036_47430036_19561.277.png" ], "turn_id": 8, "scan_id": "Validation_47430036", "origin_question": "When did you first discover \"the vivid blue globe with the sturdy plastic base\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the vivid blue globe with the sturdy plastic base\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_47430036/Validation_47430036_47430036_19563.276.png", "Validation_47430036/Validation_47430036_47430036_19565.275.png", "Validation_47430036/Validation_47430036_47430036_19567.274.png", "Validation_47430036/Validation_47430036_47430036_19569.274.png" ], "turn_id": 9, "scan_id": "Validation_47430036", "origin_question": "In which turn did you see both \"the vivid blue globe with the sturdy plastic base\" and \"the vibrant, colorful hardcover book with glossy finish\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 9 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the vivid blue globe with the sturdy plastic base\" and \"the vibrant, colorful hardcover book with glossy finish\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42898816/Validation_42898816_42898816_167353.919.png", "Validation_42898816/Validation_42898816_42898816_167355.901.png", "Validation_42898816/Validation_42898816_42898816_167357.900.png", "Validation_42898816/Validation_42898816_42898816_167359.899.png" ], "turn_id": 1, "scan_id": "Validation_42898816", "origin_question": "Remember, have you seen any air conditioner(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any air conditioner(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_42898816/Validation_42898816_42898816_167361.899.png", "Validation_42898816/Validation_42898816_42898816_167363.898.png", "Validation_42898816/Validation_42898816_42898816_167365.897.png", "Validation_42898816/Validation_42898816_42898816_167367.896.png" ], "turn_id": 2, "scan_id": "Validation_42898816", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the gray plastic dustpan with the wide scoop\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the gray plastic dustpan with the wide scoop\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_42898816/Validation_42898816_42898816_167369.895.png", "Validation_42898816/Validation_42898816_42898816_167371.894.png", "Validation_42898816/Validation_42898816_42898816_167373.894.png", "Validation_42898816/Validation_42898816_42898816_167375.893.png" ], "turn_id": 3, "scan_id": "Validation_42898816", "origin_question": "Is colorful clothes of various textures on the drying rack to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Is colorful clothes of various textures on the drying rack to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_42898816/Validation_42898816_42898816_167377.892.png", "Validation_42898816/Validation_42898816_42898816_167379.891.png", "Validation_42898816/Validation_42898816_42898816_167381.890.png", "Validation_42898816/Validation_42898816_42898816_167383.890.png" ], "turn_id": 4, "scan_id": "Validation_42898816", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "84.82", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_42898816/Validation_42898816_42898816_167385.889.png", "Validation_42898816/Validation_42898816_42898816_167387.888.png", "Validation_42898816/Validation_42898816_42898816_167389.887.png", "Validation_42898816/Validation_42898816_42898816_167391.903.png" ], "turn_id": 5, "scan_id": "Validation_42898816", "origin_question": "When did you first discover \"colorful clothes of various textures on the drying rack\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"colorful clothes of various textures on the drying rack\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_42898816/Validation_42898816_42898816_167393.902.png", "Validation_42898816/Validation_42898816_42898816_167395.901.png", "Validation_42898816/Validation_42898816_42898816_167397.900.png", "Validation_42898816/Validation_42898816_42898816_167399.900.png" ], "turn_id": 6, "scan_id": "Validation_42898816", "origin_question": "At the end of which turn were both of \"colorful clothes of various textures on the drying rack\" and \"the sink\" on your right side?", "option": [], "answer": "6", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"colorful clothes of various textures on the drying rack\" and \"the sink\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_41254382/Validation_41254382_41254382_94880.686.png", "Validation_41254382/Validation_41254382_41254382_94882.669.png", "Validation_41254382/Validation_41254382_41254382_94884.668.png", "Validation_41254382/Validation_41254382_41254382_94886.667.png" ], "turn_id": 1, "scan_id": "Validation_41254382", "origin_question": "Remember, have you seen any toilet(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any toilet(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_41254382/Validation_41254382_41254382_94888.666.png", "Validation_41254382/Validation_41254382_41254382_94890.665.png", "Validation_41254382/Validation_41254382_41254382_94892.665.png", "Validation_41254382/Validation_41254382_41254382_94894.664.png" ], "turn_id": 2, "scan_id": "Validation_41254382", "origin_question": "In which turn were you closest to \"the modern white porcelain toilet with the round bowl\"? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the modern white porcelain toilet with the round bowl\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_41254382/Validation_41254382_41254382_94896.663.png", "Validation_41254382/Validation_41254382_41254382_94898.662.png", "Validation_41254382/Validation_41254382_41254382_94900.661.png", "Validation_41254382/Validation_41254382_41254382_94902.661.png" ], "turn_id": 3, "scan_id": "Validation_41254382", "origin_question": "Which direction is the modern white porcelain toilet with the round bowl in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the modern white porcelain toilet with the round bowl in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_41254382/Validation_41254382_41254382_94904.660.png", "Validation_41254382/Validation_41254382_41254382_94906.659.png", "Validation_41254382/Validation_41254382_41254382_94908.675.png", "Validation_41254382/Validation_41254382_41254382_94910.674.png" ], "turn_id": 4, "scan_id": "Validation_41254382", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the modern white porcelain toilet with the round bowl\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the modern white porcelain toilet with the round bowl\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "Validation_41254382/Validation_41254382_41254382_94912.673.png", "Validation_41254382/Validation_41254382_41254382_94914.673.png", "Validation_41254382/Validation_41254382_41254382_94916.672.png", "Validation_41254382/Validation_41254382_41254382_94918.671.png" ], "turn_id": 5, "scan_id": "Validation_41254382", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the sleek, white wood shelf with vertical slats\"; \"the modern white porcelain toilet with the round bowl\"; \"an off-white, panel-detailed door\" ", "option": [ "the sleek, white wood shelf with vertical slats", "the modern white porcelain toilet with the round bowl", "an off-white, panel-detailed door" ], "answer": "the sleek, white wood shelf with vertical slats", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the sleek, white wood shelf with vertical slats\"; \"the modern white porcelain toilet with the round bowl\"; \"an off-white, panel-detailed door\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the sleek, white wood shelf with vertical slats', 'the modern white porcelain toilet with the round bowl', 'an off-white, panel-detailed door'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_41254382/Validation_41254382_41254382_94920.670.png", "Validation_41254382/Validation_41254382_41254382_94922.669.png", "Validation_41254382/Validation_41254382_41254382_94924.669.png", "Validation_41254382/Validation_41254382_41254382_94926.668.png" ], "turn_id": 6, "scan_id": "Validation_41254382", "origin_question": "Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_41254382/Validation_41254382_41254382_94928.667.png", "Validation_41254382/Validation_41254382_41254382_94930.666.png", "Validation_41254382/Validation_41254382_41254382_94932.665.png", "Validation_41254382/Validation_41254382_41254382_94934.665.png" ], "turn_id": 7, "scan_id": "Validation_41254382", "origin_question": "Is the sleek, rectangular white bathtub to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the sleek, rectangular white bathtub to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_41254382/Validation_41254382_41254382_94936.664.png", "Validation_41254382/Validation_41254382_41254382_94938.663.png", "Validation_41254382/Validation_41254382_41254382_94940.662.png", "Validation_41254382/Validation_41254382_41254382_94942.662.png" ], "turn_id": 8, "scan_id": "Validation_41254382", "origin_question": "Which direction is an off-white, panel-detailed door in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is an off-white, panel-detailed door in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_48018566/Validation_48018566_48018566_3857.455.png", "Validation_48018566/Validation_48018566_48018566_3858.938.png", "Validation_48018566/Validation_48018566_48018566_3860.438.png", "Validation_48018566/Validation_48018566_48018566_3861.937.png" ], "turn_id": 1, "scan_id": "Validation_48018566", "origin_question": "Remember, have you seen any statue(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any statue(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_48018566/Validation_48018566_48018566_3863.436.png", "Validation_48018566/Validation_48018566_48018566_3864.936.png", "Validation_48018566/Validation_48018566_48018566_3866.435.png", "Validation_48018566/Validation_48018566_48018566_3867.934.png" ], "turn_id": 2, "scan_id": "Validation_48018566", "origin_question": "In which turn did you see both \"the light beige rectangular two-seater couch\" and \"the sleek, minimalist cabinet with the light wood finish\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the light beige rectangular two-seater couch\" and \"the sleek, minimalist cabinet with the light wood finish\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_48018566/Validation_48018566_48018566_3869.434.png", "Validation_48018566/Validation_48018566_48018566_3870.933.png", "Validation_48018566/Validation_48018566_48018566_3872.433.png", "Validation_48018566/Validation_48018566_48018566_3873.932.png" ], "turn_id": 3, "scan_id": "Validation_48018566", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the sleek black flat-screen monitor with the slender frame\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the sleek black flat-screen monitor with the slender frame\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_48018566/Validation_48018566_48018566_3875.432.png", "Validation_48018566/Validation_48018566_48018566_3876.931.png", "Validation_48018566/Validation_48018566_48018566_3878.430.png", "Validation_48018566/Validation_48018566_48018566_3879.930.png" ], "turn_id": 4, "scan_id": "Validation_48018566", "origin_question": "In which turn did you see both \"the sleek, minimalist cabinet with the light wood finish\" and \"the sleek metal column radiator in the neutral color\" simultaneously? ", "option": [], "answer": "4", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the sleek, minimalist cabinet with the light wood finish\" and \"the sleek metal column radiator in the neutral color\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_48018566/Validation_48018566_48018566_3881.446.png", "Validation_48018566/Validation_48018566_48018566_3882.945.png", "Validation_48018566/Validation_48018566_48018566_3884.445.png", "Validation_48018566/Validation_48018566_48018566_3885.944.png" ], "turn_id": 5, "scan_id": "Validation_48018566", "origin_question": "What will be the first-time appearance order of the following categories: couch, flowerpot, table?", "option": [ "couch, table, flowerpot", "flowerpot, couch, table", "table, flowerpot, couch", "flowerpot, table, couch" ], "answer": "couch, table, flowerpot", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: couch, flowerpot, table? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['couch, table, flowerpot', 'flowerpot, couch, table', 'table, flowerpot, couch', 'flowerpot, table, couch'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_48018566/Validation_48018566_48018566_3887.443.png", "Validation_48018566/Validation_48018566_48018566_3888.943.png", "Validation_48018566/Validation_48018566_48018566_3890.442.png", "Validation_48018566/Validation_48018566_48018566_3891.941.png" ], "turn_id": 6, "scan_id": "Validation_48018566", "origin_question": "Please recall \"the sleek white ceramic cylindrical flowerpot\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.62", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the sleek white ceramic cylindrical flowerpot\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42897561/Validation_42897561_42897561_278913.707.png", "Validation_42897561/Validation_42897561_42897561_278915.206.png", "Validation_42897561/Validation_42897561_42897561_278916.705.png", "Validation_42897561/Validation_42897561_42897561_278918.205.png" ], "turn_id": 1, "scan_id": "Validation_42897561", "origin_question": "Remember, have you seen any chair(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any chair(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_42897561/Validation_42897561_42897561_278919.704.png", "Validation_42897561/Validation_42897561_42897561_278921.204.png", "Validation_42897561/Validation_42897561_42897561_278922.703.png", "Validation_42897561/Validation_42897561_42897561_278924.202.png" ], "turn_id": 2, "scan_id": "Validation_42897561", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the window\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the window\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_42897561/Validation_42897561_42897561_278925.702.png", "Validation_42897561/Validation_42897561_42897561_278927.201.png", "Validation_42897561/Validation_42897561_42897561_278928.717.png", "Validation_42897561/Validation_42897561_42897561_278930.216.png" ], "turn_id": 3, "scan_id": "Validation_42897561", "origin_question": "In which turn did you see both \"the tall white cabinet with metallic handles\" and \"the window\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the tall white cabinet with metallic handles\" and \"the window\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_42897561/Validation_42897561_42897561_278931.716.png", "Validation_42897561/Validation_42897561_42897561_278933.215.png", "Validation_42897561/Validation_42897561_42897561_278934.715.png", "Validation_42897561/Validation_42897561_42897561_278936.214.png" ], "turn_id": 4, "scan_id": "Validation_42897561", "origin_question": "Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "159.32", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42897561/Validation_42897561_42897561_278937.713.png", "Validation_42897561/Validation_42897561_42897561_278939.213.png", "Validation_42897561/Validation_42897561_42897561_278940.712.png", "Validation_42897561/Validation_42897561_42897561_278942.211.png" ], "turn_id": 5, "scan_id": "Validation_42897561", "origin_question": "Remember, have you seen any shelf(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any shelf(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_42897561/Validation_42897561_42897561_278943.711.png", "Validation_42897561/Validation_42897561_42897561_278945.210.png", "Validation_42897561/Validation_42897561_42897561_278946.710.png", "Validation_42897561/Validation_42897561_42897561_278948.209.png" ], "turn_id": 6, "scan_id": "Validation_42897561", "origin_question": "Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "85.97", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_42897561/Validation_42897561_42897561_278949.708.png", "Validation_42897561/Validation_42897561_42897561_278951.208.png", "Validation_42897561/Validation_42897561_42897561_278952.707.png", "Validation_42897561/Validation_42897561_42897561_278954.207.png" ], "turn_id": 7, "scan_id": "Validation_42897561", "origin_question": "Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "85.31", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_42897561/Validation_42897561_42897561_278955.706.png", "Validation_42897561/Validation_42897561_42897561_278957.205.png", "Validation_42897561/Validation_42897561_42897561_278958.705.png", "Validation_42897561/Validation_42897561_42897561_278960.204.png" ], "turn_id": 8, "scan_id": "Validation_42897561", "origin_question": "Please recall \"an elegant rectangular mirror with an ornate gold frame\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "3.02", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"an elegant rectangular mirror with an ornate gold frame\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_45663175/Validation_45663175_45663175_65872.641.png", "Validation_45663175/Validation_45663175_45663175_65874.124.png", "Validation_45663175/Validation_45663175_45663175_65875.624.png", "Validation_45663175/Validation_45663175_45663175_65877.123.png" ], "turn_id": 1, "scan_id": "Validation_45663175", "origin_question": "Remember, have you seen any clothes(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any clothes(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_45663175/Validation_45663175_45663175_65878.622.png", "Validation_45663175/Validation_45663175_45663175_65880.122.png", "Validation_45663175/Validation_45663175_45663175_65881.621.png", "Validation_45663175/Validation_45663175_45663175_65883.121.png" ], "turn_id": 2, "scan_id": "Validation_45663175", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular wooden chair with the high backrest\", Y: \"the rectangular white steel radiator with the smooth finish\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular wooden chair with the high backrest\", Y: \"the rectangular white steel radiator with the smooth finish\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_45663175/Validation_45663175_45663175_65884.620.png", "Validation_45663175/Validation_45663175_45663175_65886.119.png", "Validation_45663175/Validation_45663175_45663175_65887.619.png", "Validation_45663175/Validation_45663175_45663175_65889.118.png" ], "turn_id": 3, "scan_id": "Validation_45663175", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the charming room with wooden beams and yellow walls\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the charming room with wooden beams and yellow walls\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_45663175/Validation_45663175_45663175_65890.617.png", "Validation_45663175/Validation_45663175_45663175_65892.117.png", "Validation_45663175/Validation_45663175_45663175_65893.616.png", "Validation_45663175/Validation_45663175_45663175_65895.116.png" ], "turn_id": 4, "scan_id": "Validation_45663175", "origin_question": "What will be the first-time appearance order of the following categories: monitor, radiator, stove?", "option": [ "radiator, stove, monitor", "radiator, monitor, stove", "stove, monitor, radiator", "stove, radiator, monitor" ], "answer": "radiator, monitor, stove", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: monitor, radiator, stove? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['radiator, stove, monitor', 'radiator, monitor, stove', 'stove, monitor, radiator', 'stove, radiator, monitor'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_45663175/Validation_45663175_45663175_65896.615.png", "Validation_45663175/Validation_45663175_45663175_65898.115.png", "Validation_45663175/Validation_45663175_45663175_65899.614.png", "Validation_45663175/Validation_45663175_45663175_65901.113.png" ], "turn_id": 5, "scan_id": "Validation_45663175", "origin_question": "Please recall \"the wooden interior door with vertical paneling\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.78", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the wooden interior door with vertical paneling\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_45663175/Validation_45663175_45663175_65902.613.png", "Validation_45663175/Validation_45663175_45663175_65904.112.png", "Validation_45663175/Validation_45663175_45663175_65905.612.png", "Validation_45663175/Validation_45663175_45663175_65907.128.png" ], "turn_id": 6, "scan_id": "Validation_45663175", "origin_question": "Please recall \"the wooden interior door with vertical paneling\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.92", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the wooden interior door with vertical paneling\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_45663175/Validation_45663175_45663175_65908.627.png", "Validation_45663175/Validation_45663175_45663175_65910.126.png", "Validation_45663175/Validation_45663175_45663175_65911.626.png", "Validation_45663175/Validation_45663175_45663175_65913.125.png" ], "turn_id": 7, "scan_id": "Validation_45663175", "origin_question": "Using your orientation at the end of turn 6 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "47.07", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 6 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_45663175/Validation_45663175_45663175_65914.625.png", "Validation_45663175/Validation_45663175_45663175_65916.124.png", "Validation_45663175/Validation_45663175_45663175_65917.624.png", "Validation_45663175/Validation_45663175_45663175_65919.123.png" ], "turn_id": 8, "scan_id": "Validation_45663175", "origin_question": "", "option": [], "answer": "", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_41159571/Validation_41159571_41159571_34921.236.png", "Validation_41159571/Validation_41159571_41159571_34923.219.png", "Validation_41159571/Validation_41159571_41159571_34925.218.png", "Validation_41159571/Validation_41159571_41159571_34927.217.png", "Validation_41159571/Validation_41159571_41159571_34929.216.png" ], "turn_id": 1, "scan_id": "Validation_41159571", "origin_question": "Remember, have you seen any cabinet(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any cabinet(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_41159571/Validation_41159571_41159571_34931.215.png", "Validation_41159571/Validation_41159571_41159571_34933.215.png", "Validation_41159571/Validation_41159571_41159571_34935.214.png", "Validation_41159571/Validation_41159571_41159571_34937.213.png", "Validation_41159571/Validation_41159571_41159571_34939.212.png" ], "turn_id": 2, "scan_id": "Validation_41159571", "origin_question": "Is the rectangular white and cream pillow with polkthe dots to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the rectangular white and cream pillow with polkthe dots to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_41159571/Validation_41159571_41159571_34941.211.png", "Validation_41159571/Validation_41159571_41159571_34943.211.png", "Validation_41159571/Validation_41159571_41159571_34945.210.png", "Validation_41159571/Validation_41159571_41159571_34947.209.png", "Validation_41159571/Validation_41159571_41159571_34949.208.png" ], "turn_id": 3, "scan_id": "Validation_41159571", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the natural brown rattan chair with woven fiber\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the natural brown rattan chair with woven fiber\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_41159571/Validation_41159571_41159571_34951.224.png", "Validation_41159571/Validation_41159571_41159571_34953.223.png", "Validation_41159571/Validation_41159571_41159571_34955.223.png", "Validation_41159571/Validation_41159571_41159571_34957.222.png", "Validation_41159571/Validation_41159571_41159571_34959.221.png" ], "turn_id": 4, "scan_id": "Validation_41159571", "origin_question": "What will be the first-time appearance order of the following categories: cabinet, fireplace, stairs?", "option": [ "cabinet, stairs, fireplace", "cabinet, fireplace, stairs", "fireplace, stairs, cabinet", "stairs, cabinet, fireplace" ], "answer": "cabinet, stairs, fireplace", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: cabinet, fireplace, stairs? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['cabinet, stairs, fireplace', 'cabinet, fireplace, stairs', 'fireplace, stairs, cabinet', 'stairs, cabinet, fireplace'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_41159571/Validation_41159571_41159571_34961.220.png", "Validation_41159571/Validation_41159571_41159571_34963.219.png", "Validation_41159571/Validation_41159571_41159571_34965.219.png", "Validation_41159571/Validation_41159571_41159571_34967.218.png", "Validation_41159571/Validation_41159571_41159571_34969.217.png" ], "turn_id": 5, "scan_id": "Validation_41159571", "origin_question": "Which two objects are on the same side of you now? \"the rectangular window in the wooden frame\", \"the classic lamp with the cream lampshade\", \"the rustic rectangular brick fireplace with the sturdy mantle\". ", "option": [ "the rectangular window in the wooden frame, the classic lamp with the cream lampshade", "the rectangular window in the wooden frame, the rustic rectangular brick fireplace with the sturdy mantle", "the classic lamp with the cream lampshade, the rustic rectangular brick fireplace with the sturdy mantle" ], "answer": "the rectangular window in the wooden frame, the classic lamp with the cream lampshade", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the rectangular window in the wooden frame\", \"the classic lamp with the cream lampshade\", \"the rustic rectangular brick fireplace with the sturdy mantle\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular window in the wooden frame, the classic lamp with the cream lampshade', 'the rectangular window in the wooden frame, the rustic rectangular brick fireplace with the sturdy mantle', 'the classic lamp with the cream lampshade, the rustic rectangular brick fireplace with the sturdy mantle'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_41159571/Validation_41159571_41159571_34971.216.png", "Validation_41159571/Validation_41159571_41159571_34973.215.png", "Validation_41159571/Validation_41159571_41159571_34975.215.png", "Validation_41159571/Validation_41159571_41159571_34977.214.png", "Validation_41159571/Validation_41159571_41159571_34979.213.png" ], "turn_id": 6, "scan_id": "Validation_41159571", "origin_question": "By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the natural brown rattan chair with woven fiber\", Y: \"the rustic rectangular brick fireplace with the sturdy mantle\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the natural brown rattan chair with woven fiber\", Y: \"the rustic rectangular brick fireplace with the sturdy mantle\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_41159571/Validation_41159571_41159571_34981.212.png", "Validation_41159571/Validation_41159571_41159571_34983.211.png", "Validation_41159571/Validation_41159571_41159571_34985.211.png", "Validation_41159571/Validation_41159571_41159571_34987.210.png", "Validation_41159571/Validation_41159571_41159571_34989.209.png" ], "turn_id": 7, "scan_id": "Validation_41159571", "origin_question": "Please recall \"the medium-sized rectangular wooden cabinet\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "5.21", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the medium-sized rectangular wooden cabinet\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_41159571/Validation_41159571_41159571_34991.208.png", "Validation_41159571/Validation_41159571_41159571_34993.224.png", "Validation_41159571/Validation_41159571_41159571_34995.223.png", "Validation_41159571/Validation_41159571_41159571_34997.223.png", "Validation_41159571/Validation_41159571_41159571_34999.222.png" ], "turn_id": 8, "scan_id": "Validation_41159571", "origin_question": "Which direction is the rustic rectangular brick fireplace with the sturdy mantle in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the rustic rectangular brick fireplace with the sturdy mantle in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_41159571/Validation_41159571_41159571_35001.221.png", "Validation_41159571/Validation_41159571_41159571_35003.220.png", "Validation_41159571/Validation_41159571_41159571_35005.219.png", "Validation_41159571/Validation_41159571_41159571_35007.219.png", "Validation_41159571/Validation_41159571_41159571_35009.218.png" ], "turn_id": 9, "scan_id": "Validation_41159571", "origin_question": "In which turn did you see both \"the sleek glass top table with ornate dark legs\" and \"the rustic rectangular brick fireplace with the sturdy mantle\" simultaneously? ", "option": [], "answer": "7", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the sleek glass top table with ornate dark legs\" and \"the rustic rectangular brick fireplace with the sturdy mantle\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_41159571/Validation_41159571_41159571_35011.217.png", "Validation_41159571/Validation_41159571_41159571_35013.216.png", "Validation_41159571/Validation_41159571_41159571_35015.215.png", "Validation_41159571/Validation_41159571_41159571_35017.215.png", "Validation_41159571/Validation_41159571_41159571_35019.214.png" ], "turn_id": 10, "scan_id": "Validation_41159571", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the rustic rectangular brick fireplace with the sturdy mantle\" from your position?", "option": [], "answer": "81.24", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the rustic rectangular brick fireplace with the sturdy mantle\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47334120/Validation_47334120_47334120_94308.061.png", "Validation_47334120/Validation_47334120_47334120_94310.044.png", "Validation_47334120/Validation_47334120_47334120_94312.043.png", "Validation_47334120/Validation_47334120_47334120_94314.042.png", "Validation_47334120/Validation_47334120_47334120_94316.042.png" ], "turn_id": 1, "scan_id": "Validation_47334120", "origin_question": "Remember, have you seen any stapler(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any stapler(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47334120/Validation_47334120_47334120_94318.041.png", "Validation_47334120/Validation_47334120_47334120_94320.057.png", "Validation_47334120/Validation_47334120_47334120_94322.056.png", "Validation_47334120/Validation_47334120_47334120_94324.055.png", "Validation_47334120/Validation_47334120_47334120_94326.054.png" ], "turn_id": 2, "scan_id": "Validation_47334120", "origin_question": "Remember, have you seen any candlestick(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any candlestick(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47334120/Validation_47334120_47334120_94328.053.png", "Validation_47334120/Validation_47334120_47334120_94330.053.png", "Validation_47334120/Validation_47334120_47334120_94332.052.png", "Validation_47334120/Validation_47334120_47334120_94334.051.png", "Validation_47334120/Validation_47334120_47334120_94336.050.png" ], "turn_id": 3, "scan_id": "Validation_47334120", "origin_question": "In which turn were you closest to \"the blue ceramic cylindrical flowerpot with the tapered base\"? ", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the blue ceramic cylindrical flowerpot with the tapered base\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "Validation_47334120/Validation_47334120_47334120_94338.049.png", "Validation_47334120/Validation_47334120_47334120_94340.049.png", "Validation_47334120/Validation_47334120_47334120_94342.048.png", "Validation_47334120/Validation_47334120_47334120_94344.047.png", "Validation_47334120/Validation_47334120_47334120_94346.046.png" ], "turn_id": 4, "scan_id": "Validation_47334120", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the sleek cylindrical dark stool with metal legs\"; \"the tall rectangular matte cabinet in the neutral color\"; \"sleek silver rectangular picture frames with narrow borders\" ", "option": [ "the sleek cylindrical dark stool with metal legs", "the tall rectangular matte cabinet in the neutral color", "sleek silver rectangular picture frames with narrow borders" ], "answer": "the sleek cylindrical dark stool with metal legs", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the sleek cylindrical dark stool with metal legs\"; \"the tall rectangular matte cabinet in the neutral color\"; \"sleek silver rectangular picture frames with narrow borders\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the sleek cylindrical dark stool with metal legs', 'the tall rectangular matte cabinet in the neutral color', 'sleek silver rectangular picture frames with narrow borders'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47334120/Validation_47334120_47334120_94348.045.png", "Validation_47334120/Validation_47334120_47334120_94350.045.png", "Validation_47334120/Validation_47334120_47334120_94352.044.png", "Validation_47334120/Validation_47334120_47334120_94354.043.png", "Validation_47334120/Validation_47334120_47334120_94356.042.png" ], "turn_id": 5, "scan_id": "Validation_47334120", "origin_question": "", "option": [], "answer": "", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47334120/Validation_47334120_47334120_94358.041.png", "Validation_47334120/Validation_47334120_47334120_94360.041.png", "Validation_47334120/Validation_47334120_47334120_94362.056.png", "Validation_47334120/Validation_47334120_47334120_94364.056.png", "Validation_47334120/Validation_47334120_47334120_94366.055.png" ], "turn_id": 6, "scan_id": "Validation_47334120", "origin_question": "", "option": [], "answer": "", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47334120/Validation_47334120_47334120_94368.054.png", "Validation_47334120/Validation_47334120_47334120_94370.053.png", "Validation_47334120/Validation_47334120_47334120_94372.052.png", "Validation_47334120/Validation_47334120_47334120_94374.052.png", "Validation_47334120/Validation_47334120_47334120_94376.051.png" ], "turn_id": 7, "scan_id": "Validation_47334120", "origin_question": "In which turn were you closest to \"the teal bed with the sturdy wooden frame\"? ", "option": [], "answer": "6", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the teal bed with the sturdy wooden frame\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_47334120/Validation_47334120_47334120_94378.050.png", "Validation_47334120/Validation_47334120_47334120_94380.049.png", "Validation_47334120/Validation_47334120_47334120_94382.048.png", "Validation_47334120/Validation_47334120_47334120_94384.048.png", "Validation_47334120/Validation_47334120_47334120_94386.047.png" ], "turn_id": 8, "scan_id": "Validation_47334120", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the sleek cylindrical dark stool with metal legs\" from your position?", "option": [], "answer": "153.57", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the sleek cylindrical dark stool with metal legs\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_47334120/Validation_47334120_47334120_94388.046.png", "Validation_47334120/Validation_47334120_47334120_94390.045.png", "Validation_47334120/Validation_47334120_47334120_94392.044.png", "Validation_47334120/Validation_47334120_47334120_94394.043.png", "Validation_47334120/Validation_47334120_47334120_94396.043.png" ], "turn_id": 9, "scan_id": "Validation_47334120", "origin_question": "Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "153.01", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_47334120/Validation_47334120_47334120_94398.042.png", "Validation_47334120/Validation_47334120_47334120_94400.041.png", "Validation_47334120/Validation_47334120_47334120_94402.057.png", "Validation_47334120/Validation_47334120_47334120_94404.056.png", "Validation_47334120/Validation_47334120_47334120_94406.055.png" ], "turn_id": 10, "scan_id": "Validation_47334120", "origin_question": "Please recall \"sleek silver rectangular picture frames with narrow borders\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "3.33", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"sleek silver rectangular picture frames with narrow borders\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47115526/Validation_47115526_47115526_14536.007.png", "Validation_47115526/Validation_47115526_47115526_14537.989.png", "Validation_47115526/Validation_47115526_47115526_14539.989.png", "Validation_47115526/Validation_47115526_47115526_14541.988.png", "Validation_47115526/Validation_47115526_47115526_14543.987.png" ], "turn_id": 1, "scan_id": "Validation_47115526", "origin_question": "Remember, have you seen any glove(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any glove(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_47115526/Validation_47115526_47115526_14545.986.png", "Validation_47115526/Validation_47115526_47115526_14547.986.png", "Validation_47115526/Validation_47115526_47115526_14549.985.png", "Validation_47115526/Validation_47115526_47115526_14551.984.png", "Validation_47115526/Validation_47115526_47115526_14553.983.png" ], "turn_id": 2, "scan_id": "Validation_47115526", "origin_question": "In which turn did you see both \"the fireplace\" and \"the wooden door with glass panels and the metallic knob\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the fireplace\" and \"the wooden door with glass panels and the metallic knob\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47115526/Validation_47115526_47115526_14555.982.png", "Validation_47115526/Validation_47115526_47115526_14557.981.png", "Validation_47115526/Validation_47115526_47115526_14559.997.png", "Validation_47115526/Validation_47115526_47115526_14561.997.png", "Validation_47115526/Validation_47115526_47115526_14563.996.png" ], "turn_id": 3, "scan_id": "Validation_47115526", "origin_question": "In which turn were you farthest from \"the standard-sized beige and brown plaid couch\"? ", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the standard-sized beige and brown plaid couch\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_47115526/Validation_47115526_47115526_14565.995.png", "Validation_47115526/Validation_47115526_47115526_14567.994.png", "Validation_47115526/Validation_47115526_47115526_14569.993.png", "Validation_47115526/Validation_47115526_47115526_14571.993.png", "Validation_47115526/Validation_47115526_47115526_14573.992.png" ], "turn_id": 4, "scan_id": "Validation_47115526", "origin_question": "Is the wooden door with glass panels and the metallic knob to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the wooden door with glass panels and the metallic knob to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_47115526/Validation_47115526_47115526_14575.991.png", "Validation_47115526/Validation_47115526_47115526_14577.990.png", "Validation_47115526/Validation_47115526_47115526_14579.989.png", "Validation_47115526/Validation_47115526_47115526_14581.989.png", "Validation_47115526/Validation_47115526_47115526_14583.988.png" ], "turn_id": 5, "scan_id": "Validation_47115526", "origin_question": "By the end of the 3 turn, were you getting closer to or farther away from these two objects: X: \"the fireplace\", Y: \"the wooden door with glass panels and the metallic knob\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to X and farther away from Y", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 3 turn, were you getting closer to or farther away from these two objects: X: \"the fireplace\", Y: \"the wooden door with glass panels and the metallic knob\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47115526/Validation_47115526_47115526_14585.987.png", "Validation_47115526/Validation_47115526_47115526_14587.986.png", "Validation_47115526/Validation_47115526_47115526_14589.986.png", "Validation_47115526/Validation_47115526_47115526_14591.985.png", "Validation_47115526/Validation_47115526_47115526_14593.984.png" ], "turn_id": 6, "scan_id": "Validation_47115526", "origin_question": "Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the fireplace\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the fireplace\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47115526/Validation_47115526_47115526_14595.983.png", "Validation_47115526/Validation_47115526_47115526_14597.982.png", "Validation_47115526/Validation_47115526_47115526_14599.982.png", "Validation_47115526/Validation_47115526_47115526_14601.997.png", "Validation_47115526/Validation_47115526_47115526_14603.997.png" ], "turn_id": 7, "scan_id": "Validation_47115526", "origin_question": "Compared to your position at the end of the 5 turn, are you now closer or farther away from \"the fireplace\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 5 turn, are you now closer or farther away from \"the fireplace\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_47115526/Validation_47115526_47115526_14605.996.png", "Validation_47115526/Validation_47115526_47115526_14607.995.png", "Validation_47115526/Validation_47115526_47115526_14609.994.png", "Validation_47115526/Validation_47115526_47115526_14611.994.png", "Validation_47115526/Validation_47115526_47115526_14613.993.png" ], "turn_id": 8, "scan_id": "Validation_47115526", "origin_question": "At the end of which turn were both of \"the wooden door with glass panels and the metallic knob\" and \"the sturdy rectangular brown wooden table with the smooth surface\" on your right side?", "option": [], "answer": "4", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the wooden door with glass panels and the metallic knob\" and \"the sturdy rectangular brown wooden table with the smooth surface\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_47115526/Validation_47115526_47115526_14615.992.png", "Validation_47115526/Validation_47115526_47115526_14617.991.png", "Validation_47115526/Validation_47115526_47115526_14619.990.png", "Validation_47115526/Validation_47115526_47115526_14621.989.png", "Validation_47115526/Validation_47115526_47115526_14623.989.png" ], "turn_id": 9, "scan_id": "Validation_47115526", "origin_question": "Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Left", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_47115526/Validation_47115526_47115526_14625.988.png", "Validation_47115526/Validation_47115526_47115526_14627.987.png", "Validation_47115526/Validation_47115526_47115526_14629.986.png", "Validation_47115526/Validation_47115526_47115526_14631.986.png", "Validation_47115526/Validation_47115526_47115526_14633.985.png" ], "turn_id": 10, "scan_id": "Validation_47115526", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the wooden door with glass panels and the metallic knob\" from your position?", "option": [], "answer": "177.47", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the wooden door with glass panels and the metallic knob\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42446049/Validation_42446049_42446049_347774.096.png", "Validation_42446049/Validation_42446049_42446049_347776.079.png", "Validation_42446049/Validation_42446049_42446049_347778.078.png", "Validation_42446049/Validation_42446049_42446049_347780.077.png", "Validation_42446049/Validation_42446049_42446049_347782.076.png" ], "turn_id": 1, "scan_id": "Validation_42446049", "origin_question": "Remember, have you seen any glove(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any glove(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_42446049/Validation_42446049_42446049_347784.075.png", "Validation_42446049/Validation_42446049_42446049_347786.075.png", "Validation_42446049/Validation_42446049_42446049_347788.074.png", "Validation_42446049/Validation_42446049_42446049_347790.073.png", "Validation_42446049/Validation_42446049_42446049_347792.072.png" ], "turn_id": 2, "scan_id": "Validation_42446049", "origin_question": "Is the circular white ceramic plate to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the circular white ceramic plate to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_42446049/Validation_42446049_42446049_347794.071.png", "Validation_42446049/Validation_42446049_42446049_347796.070.png", "Validation_42446049/Validation_42446049_42446049_347798.070.png", "Validation_42446049/Validation_42446049_42446049_347800.086.png", "Validation_42446049/Validation_42446049_42446049_347802.085.png" ], "turn_id": 3, "scan_id": "Validation_42446049", "origin_question": "In which turn did you see both \"the oven\" and \"the medium-sized stainless steel pan with rounded shape\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the oven\" and \"the medium-sized stainless steel pan with rounded shape\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_42446049/Validation_42446049_42446049_347804.084.png", "Validation_42446049/Validation_42446049_42446049_347806.083.png", "Validation_42446049/Validation_42446049_42446049_347808.082.png", "Validation_42446049/Validation_42446049_42446049_347810.081.png", "Validation_42446049/Validation_42446049_42446049_347812.081.png" ], "turn_id": 4, "scan_id": "Validation_42446049", "origin_question": "Which direction is the circular white ceramic plate in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the circular white ceramic plate in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_42446049/Validation_42446049_42446049_347814.080.png", "Validation_42446049/Validation_42446049_42446049_347816.079.png", "Validation_42446049/Validation_42446049_42446049_347818.078.png", "Validation_42446049/Validation_42446049_42446049_347820.077.png", "Validation_42446049/Validation_42446049_42446049_347822.077.png" ], "turn_id": 5, "scan_id": "Validation_42446049", "origin_question": "", "option": [], "answer": "", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_42446049/Validation_42446049_42446049_347824.076.png", "Validation_42446049/Validation_42446049_42446049_347826.075.png", "Validation_42446049/Validation_42446049_42446049_347828.074.png", "Validation_42446049/Validation_42446049_42446049_347830.073.png", "Validation_42446049/Validation_42446049_42446049_347832.073.png" ], "turn_id": 6, "scan_id": "Validation_42446049", "origin_question": "When did you first discover \"the sleek rectangular window with the white pvc frame\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the sleek rectangular window with the white pvc frame\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_42446049/Validation_42446049_42446049_347834.072.png", "Validation_42446049/Validation_42446049_42446049_347836.071.png", "Validation_42446049/Validation_42446049_42446049_347838.070.png", "Validation_42446049/Validation_42446049_42446049_347840.069.png", "Validation_42446049/Validation_42446049_42446049_347842.068.png" ], "turn_id": 7, "scan_id": "Validation_42446049", "origin_question": "Using your orientation at the end of turn 5 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 5 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_42446049/Validation_42446049_42446049_347844.084.png", "Validation_42446049/Validation_42446049_42446049_347846.083.png", "Validation_42446049/Validation_42446049_42446049_347848.083.png", "Validation_42446049/Validation_42446049_42446049_347850.082.png", "Validation_42446049/Validation_42446049_42446049_347852.081.png" ], "turn_id": 8, "scan_id": "Validation_42446049", "origin_question": "Assuming the direction you are facing at the end of the turn 5 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Right", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 5 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "Validation_42897695/Validation_42897695_42897695_480232.603.png", "Validation_42897695/Validation_42897695_42897695_480234.602.png", "Validation_42897695/Validation_42897695_42897695_480236.601.png", "Validation_42897695/Validation_42897695_42897695_480238.601.png", "Validation_42897695/Validation_42897695_42897695_480240.600.png" ], "turn_id": 1, "scan_id": "Validation_42897695", "origin_question": "Remember, how many bed(s) have you seen so far? ", "option": [], "answer": "1", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many bed(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_42897695/Validation_42897695_42897695_480242.599.png", "Validation_42897695/Validation_42897695_42897695_480244.598.png", "Validation_42897695/Validation_42897695_42897695_480246.597.png", "Validation_42897695/Validation_42897695_42897695_480248.596.png", "Validation_42897695/Validation_42897695_42897695_480250.596.png" ], "turn_id": 2, "scan_id": "Validation_42897695", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the modern rectangular light wood shelf with compartments\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the modern rectangular light wood shelf with compartments\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_42897695/Validation_42897695_42897695_480252.595.png", "Validation_42897695/Validation_42897695_42897695_480254.594.png", "Validation_42897695/Validation_42897695_42897695_480256.593.png", "Validation_42897695/Validation_42897695_42897695_480258.592.png", "Validation_42897695/Validation_42897695_42897695_480260.591.png" ], "turn_id": 3, "scan_id": "Validation_42897695", "origin_question": "Please recall \"the modern rectangular light wood shelf with compartments\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "3.14", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the modern rectangular light wood shelf with compartments\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_42897695/Validation_42897695_42897695_480262.591.png", "Validation_42897695/Validation_42897695_42897695_480264.590.png", "Validation_42897695/Validation_42897695_42897695_480266.606.png", "Validation_42897695/Validation_42897695_42897695_480268.605.png", "Validation_42897695/Validation_42897695_42897695_480270.604.png" ], "turn_id": 4, "scan_id": "Validation_42897695", "origin_question": "Please recall \"the medium-sized lamp with the spherical glass base\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.12", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the medium-sized lamp with the spherical glass base\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_42897695/Validation_42897695_42897695_480272.603.png", "Validation_42897695/Validation_42897695_42897695_480274.602.png", "Validation_42897695/Validation_42897695_42897695_480276.602.png", "Validation_42897695/Validation_42897695_42897695_480278.601.png", "Validation_42897695/Validation_42897695_42897695_480280.600.png" ], "turn_id": 5, "scan_id": "Validation_42897695", "origin_question": "", "option": [], "answer": "", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "Validation_42897695/Validation_42897695_42897695_480282.599.png", "Validation_42897695/Validation_42897695_42897695_480284.598.png", "Validation_42897695/Validation_42897695_42897695_480286.597.png", "Validation_42897695/Validation_42897695_42897695_480288.597.png", "Validation_42897695/Validation_42897695_42897695_480290.596.png" ], "turn_id": 6, "scan_id": "Validation_42897695", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the rectangular monochromatic picture of trees\"; \"the modern rectangular light wood shelf with compartments\"; \"the rectangular window with the dark segmented frame\" ", "option": [ "the rectangular monochromatic picture of trees", "the modern rectangular light wood shelf with compartments", "the rectangular window with the dark segmented frame" ], "answer": "the rectangular monochromatic picture of trees", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the rectangular monochromatic picture of trees\"; \"the modern rectangular light wood shelf with compartments\"; \"the rectangular window with the dark segmented frame\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular monochromatic picture of trees', 'the modern rectangular light wood shelf with compartments', 'the rectangular window with the dark segmented frame'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_42897695/Validation_42897695_42897695_480292.595.png", "Validation_42897695/Validation_42897695_42897695_480294.594.png", "Validation_42897695/Validation_42897695_42897695_480296.593.png", "Validation_42897695/Validation_42897695_42897695_480298.593.png", "Validation_42897695/Validation_42897695_42897695_480300.592.png" ], "turn_id": 7, "scan_id": "Validation_42897695", "origin_question": "When did you first discover \"the rectangular monochromatic picture of trees\"? (index of the turn)", "option": [], "answer": "6", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the rectangular monochromatic picture of trees\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_42897695/Validation_42897695_42897695_480302.591.png", "Validation_42897695/Validation_42897695_42897695_480304.590.png", "Validation_42897695/Validation_42897695_42897695_480306.606.png", "Validation_42897695/Validation_42897695_42897695_480308.605.png", "Validation_42897695/Validation_42897695_42897695_480310.604.png" ], "turn_id": 8, "scan_id": "Validation_42897695", "origin_question": "When did you first discover \"the plush, oval-shaped soft green cushion\"? (index of the turn)", "option": [], "answer": "4", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the plush, oval-shaped soft green cushion\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_42897695/Validation_42897695_42897695_480312.603.png", "Validation_42897695/Validation_42897695_42897695_480314.603.png", "Validation_42897695/Validation_42897695_42897695_480316.602.png", "Validation_42897695/Validation_42897695_42897695_480318.601.png", "Validation_42897695/Validation_42897695_42897695_480320.600.png" ], "turn_id": 9, "scan_id": "Validation_42897695", "origin_question": "By the end of the 6 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular window with the dark segmented frame\", Y: \"the sleek rectangular wooden door with the white finish\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to X and farther away from Y", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 6 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular window with the dark segmented frame\", Y: \"the sleek rectangular wooden door with the white finish\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_42897695/Validation_42897695_42897695_480322.599.png", "Validation_42897695/Validation_42897695_42897695_480324.599.png", "Validation_42897695/Validation_42897695_42897695_480326.598.png", "Validation_42897695/Validation_42897695_42897695_480328.597.png", "Validation_42897695/Validation_42897695_42897695_480330.596.png" ], "turn_id": 10, "scan_id": "Validation_42897695", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47331316/Validation_47331316_47331316_177699.492.png", "Validation_47331316/Validation_47331316_47331316_177701.474.png", "Validation_47331316/Validation_47331316_47331316_177703.473.png", "Validation_47331316/Validation_47331316_47331316_177705.473.png" ], "turn_id": 1, "scan_id": "Validation_47331316", "origin_question": "Remember, have you seen any pillow(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any pillow(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_47331316/Validation_47331316_47331316_177707.472.png", "Validation_47331316/Validation_47331316_47331316_177709.471.png", "Validation_47331316/Validation_47331316_47331316_177711.470.png", "Validation_47331316/Validation_47331316_47331316_177713.469.png" ], "turn_id": 2, "scan_id": "Validation_47331316", "origin_question": "At the end of which turn were both of \"the modern rounded-triangle table with dark wood finish\" and \"the yellow plastic chair with rounded edges\" on your left side?", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the modern rounded-triangle table with dark wood finish\" and \"the yellow plastic chair with rounded edges\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47331316/Validation_47331316_47331316_177715.469.png", "Validation_47331316/Validation_47331316_47331316_177717.468.png", "Validation_47331316/Validation_47331316_47331316_177719.467.png", "Validation_47331316/Validation_47331316_47331316_177721.466.png" ], "turn_id": 3, "scan_id": "Validation_47331316", "origin_question": "In which turn were you closest to \"the yellow plastic chair with rounded edges\"? ", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the yellow plastic chair with rounded edges\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_47331316/Validation_47331316_47331316_177723.465.png", "Validation_47331316/Validation_47331316_47331316_177725.465.png", "Validation_47331316/Validation_47331316_47331316_177727.480.png", "Validation_47331316/Validation_47331316_47331316_177729.479.png" ], "turn_id": 4, "scan_id": "Validation_47331316", "origin_question": "In which turn did you see both \"the yellow plastic chair with rounded edges\" and \"the radiator\" simultaneously? ", "option": [], "answer": "4", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the yellow plastic chair with rounded edges\" and \"the radiator\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "Validation_47331316/Validation_47331316_47331316_177731.479.png", "Validation_47331316/Validation_47331316_47331316_177733.478.png", "Validation_47331316/Validation_47331316_47331316_177735.477.png", "Validation_47331316/Validation_47331316_47331316_177737.476.png" ], "turn_id": 5, "scan_id": "Validation_47331316", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the classic cylindrical flowerpot made of smooth ceramic\"; \"the standard-sized dark-colored book with minimalist design\"; \"the radiator\" ", "option": [ "the classic cylindrical flowerpot made of smooth ceramic", "the standard-sized dark-colored book with minimalist design", "the radiator" ], "answer": "the classic cylindrical flowerpot made of smooth ceramic", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the classic cylindrical flowerpot made of smooth ceramic\"; \"the standard-sized dark-colored book with minimalist design\"; \"the radiator\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the classic cylindrical flowerpot made of smooth ceramic', 'the standard-sized dark-colored book with minimalist design', 'the radiator'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_47331316/Validation_47331316_47331316_177739.476.png", "Validation_47331316/Validation_47331316_47331316_177741.475.png", "Validation_47331316/Validation_47331316_47331316_177743.474.png", "Validation_47331316/Validation_47331316_47331316_177745.473.png" ], "turn_id": 6, "scan_id": "Validation_47331316", "origin_question": "When did you first discover \"the modern rounded-triangle table with dark wood finish\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the modern rounded-triangle table with dark wood finish\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_45261144/Validation_45261144_45261144_4651.027.png", "Validation_45261144/Validation_45261144_45261144_4653.010.png", "Validation_45261144/Validation_45261144_45261144_4655.009.png", "Validation_45261144/Validation_45261144_45261144_4657.008.png" ], "turn_id": 1, "scan_id": "Validation_45261144", "origin_question": "Remember, have you seen any bed(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any bed(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_45261144/Validation_45261144_45261144_4659.007.png", "Validation_45261144/Validation_45261144_45261144_4661.006.png", "Validation_45261144/Validation_45261144_45261144_4663.005.png", "Validation_45261144/Validation_45261144_45261144_4665.005.png" ], "turn_id": 2, "scan_id": "Validation_45261144", "origin_question": "In which turn were you farthest from \"the pastel pink bed with soft fabric material\"? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the pastel pink bed with soft fabric material\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_45261144/Validation_45261144_45261144_4667.004.png", "Validation_45261144/Validation_45261144_45261144_4669.003.png", "Validation_45261144/Validation_45261144_45261144_4671.002.png", "Validation_45261144/Validation_45261144_45261144_4673.001.png" ], "turn_id": 3, "scan_id": "Validation_45261144", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "20.82", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_45261144/Validation_45261144_45261144_4675.000.png", "Validation_45261144/Validation_45261144_45261144_4677.000.png", "Validation_45261144/Validation_45261144_45261144_4678.999.png", "Validation_45261144/Validation_45261144_45261144_4680.998.png" ], "turn_id": 4, "scan_id": "Validation_45261144", "origin_question": "", "option": [], "answer": "", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_45261144/Validation_45261144_45261144_4682.997.png", "Validation_45261144/Validation_45261144_45261144_4684.996.png", "Validation_45261144/Validation_45261144_45261144_4687.012.png", "Validation_45261144/Validation_45261144_45261144_4689.011.png" ], "turn_id": 5, "scan_id": "Validation_45261144", "origin_question": "Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Left", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_45261144/Validation_45261144_45261144_4691.010.png", "Validation_45261144/Validation_45261144_45261144_4693.010.png", "Validation_45261144/Validation_45261144_45261144_4695.009.png", "Validation_45261144/Validation_45261144_45261144_4697.008.png" ], "turn_id": 6, "scan_id": "Validation_45261144", "origin_question": "", "option": [], "answer": "", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_45261144/Validation_45261144_45261144_4699.007.png", "Validation_45261144/Validation_45261144_45261144_4701.006.png", "Validation_45261144/Validation_45261144_45261144_4703.005.png", "Validation_45261144/Validation_45261144_45261144_4705.005.png" ], "turn_id": 7, "scan_id": "Validation_45261144", "origin_question": "Using your orientation at the end of turn 6 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 6 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_45261144/Validation_45261144_45261144_4707.004.png", "Validation_45261144/Validation_45261144_45261144_4709.003.png", "Validation_45261144/Validation_45261144_45261144_4711.002.png", "Validation_45261144/Validation_45261144_45261144_4713.001.png" ], "turn_id": 8, "scan_id": "Validation_45261144", "origin_question": "Assuming the direction you are facing at the end of the turn 5 is forward, did you move a certain distance forward or backward from that position?", "option": [ "Forward", "Backward" ], "answer": "Forward", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 5 is forward, did you move a certain distance forward or backward from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Forward', 'Backward'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_45261144/Validation_45261144_45261144_4715.000.png", "Validation_45261144/Validation_45261144_45261144_4717.000.png", "Validation_45261144/Validation_45261144_45261144_4718.999.png", "Validation_45261144/Validation_45261144_45261144_4720.998.png" ], "turn_id": 9, "scan_id": "Validation_45261144", "origin_question": "Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the pastel pink bed with soft fabric material\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 9 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the pastel pink bed with soft fabric material\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0029_region34/f8a73e17d842414a8fea443765cdc1be_i1_5.jpg", "1mp3d_0029_region34/f8a73e17d842414a8fea443765cdc1be_i2_5.jpg", "1mp3d_0029_region34/f8a73e17d842414a8fea443765cdc1be_i1_2.jpg", "1mp3d_0029_region34/f8a73e17d842414a8fea443765cdc1be_i2_2.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0029_region34", "origin_question": "Remember, have you seen any cosmetics(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any cosmetics(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "1mp3d_0029_region34/8180c5785c2b43b9933c6493c1e16520_i1_1.jpg", "1mp3d_0029_region34/8180c5785c2b43b9933c6493c1e16520_i2_1.jpg", "1mp3d_0029_region34/8180c5785c2b43b9933c6493c1e16520_i1_2.jpg", "1mp3d_0029_region34/8180c5785c2b43b9933c6493c1e16520_i2_2.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0029_region34", "origin_question": "In which turn were you farthest from \"the rectangular white wooden door with a simple design\"? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the rectangular white wooden door with a simple design\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "1mp3d_0029_region34/f8a73e17d842414a8fea443765cdc1be_i1_4.jpg", "1mp3d_0029_region34/f8a73e17d842414a8fea443765cdc1be_i2_4.jpg", "1mp3d_0029_region34/f8a73e17d842414a8fea443765cdc1be_i1_3.jpg", "1mp3d_0029_region34/f8a73e17d842414a8fea443765cdc1be_i2_3.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0029_region34", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "136.32", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "1mp3d_0029_region34/8180c5785c2b43b9933c6493c1e16520_i1_5.jpg", "1mp3d_0029_region34/8180c5785c2b43b9933c6493c1e16520_i2_5.jpg", "1mp3d_0029_region34/b37b1397ce5441199438878cfb1d9a64_i1_2.jpg", "1mp3d_0029_region34/b37b1397ce5441199438878cfb1d9a64_i2_2.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0029_region34", "origin_question": "When did you first discover \"the large white rectangular wooden window frame\"? (index of the turn)", "option": [], "answer": "3", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the large white rectangular wooden window frame\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "1mp3d_0029_region34/b37b1397ce5441199438878cfb1d9a64_i1_1.jpg", "1mp3d_0029_region34/b37b1397ce5441199438878cfb1d9a64_i2_1.jpg", "1mp3d_0029_region34/c298917e0a0c4567a0895d6243540479_i1_1.jpg", "1mp3d_0029_region34/c298917e0a0c4567a0895d6243540479_i2_1.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0029_region34", "origin_question": "Please recall \"the rectangular white wooden door with a simple design\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "3.85", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the rectangular white wooden door with a simple design\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0067_region27/be291fdeaee9413698d9da350207bc54_i1_0.jpg", "1mp3d_0067_region27/be291fdeaee9413698d9da350207bc54_i2_0.jpg", "1mp3d_0067_region27/94ef53dd5e024bd7a7c79c7f43731731_i2_1.jpg", "1mp3d_0067_region27/94ef53dd5e024bd7a7c79c7f43731731_i1_3.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0067_region27", "origin_question": "Remember, have you seen any bin(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any bin(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "1mp3d_0067_region27/94ef53dd5e024bd7a7c79c7f43731731_i2_3.jpg", "1mp3d_0067_region27/9f2ce45f125346de825f3be52f40aab0_i2_1.jpg", "1mp3d_0067_region27/c173f74f5a284b98b1ac17917c4c356e_i1_5.jpg", "1mp3d_0067_region27/c173f74f5a284b98b1ac17917c4c356e_i2_5.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0067_region27", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the large rectangular polished mirror with a reflective surface\", Y: \"the rectangular wooden tissue box with natural color\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "farther away from both X and Y", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the large rectangular polished mirror with a reflective surface\", Y: \"the rectangular wooden tissue box with natural color\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "1mp3d_0067_region27/c173f74f5a284b98b1ac17917c4c356e_i1_1.jpg", "1mp3d_0067_region27/c173f74f5a284b98b1ac17917c4c356e_i2_1.jpg", "1mp3d_0067_region27/0eae51fd57fc43a0b17c04a87dc559fb_i1_2.jpg", "1mp3d_0067_region27/0eae51fd57fc43a0b17c04a87dc559fb_i2_2.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0067_region27", "origin_question": "Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Right", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "1mp3d_0067_region27/06bc29df119749bd9c5be352b59d4451_i1_1.jpg", "1mp3d_0067_region27/06bc29df119749bd9c5be352b59d4451_i2_1.jpg", "1mp3d_0067_region27/06bc29df119749bd9c5be352b59d4451_i1_0.jpg", "1mp3d_0067_region27/06bc29df119749bd9c5be352b59d4451_i2_0.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0067_region27", "origin_question": "Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance forward or backward from that position?", "option": [ "Forward", "Backward" ], "answer": "Backward", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance forward or backward from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Forward', 'Backward'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "1mp3d_0067_region27/06bc29df119749bd9c5be352b59d4451_i1_5.jpg", "1mp3d_0067_region27/06bc29df119749bd9c5be352b59d4451_i2_5.jpg", "1mp3d_0067_region27/0eae51fd57fc43a0b17c04a87dc559fb_i1_5.jpg", "1mp3d_0067_region27/0eae51fd57fc43a0b17c04a87dc559fb_i2_5.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0067_region27", "origin_question": "Please recall \"the rectangular multicolored granite countertop with a polished surface\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.85", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the rectangular multicolored granite countertop with a polished surface\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "1mp3d_0067_region27/0eae51fd57fc43a0b17c04a87dc559fb_i1_0.jpg", "1mp3d_0067_region27/0eae51fd57fc43a0b17c04a87dc559fb_i2_0.jpg", "1mp3d_0067_region27/0eae51fd57fc43a0b17c04a87dc559fb_i1_1.jpg", "1mp3d_0067_region27/0eae51fd57fc43a0b17c04a87dc559fb_i2_1.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0067_region27", "origin_question": "Please recall \"the rectangular wooden tissue box with natural color\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.76", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the rectangular wooden tissue box with natural color\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "1mp3d_0067_region27/9d97383beaee4ec78b8b20f44ba72c3b_i1_5.jpg", "1mp3d_0067_region27/9d97383beaee4ec78b8b20f44ba72c3b_i2_5.jpg", "1mp3d_0067_region27/9d97383beaee4ec78b8b20f44ba72c3b_i1_0.jpg", "1mp3d_0067_region27/9d97383beaee4ec78b8b20f44ba72c3b_i2_0.jpg" ], "turn_id": 7, "scan_id": "1mp3d_0067_region27", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the toilet paper\" from your position?", "option": [], "answer": "65.09", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the toilet paper\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "1mp3d_0067_region27/9d97383beaee4ec78b8b20f44ba72c3b_i1_1.jpg", "1mp3d_0067_region27/9d97383beaee4ec78b8b20f44ba72c3b_i2_1.jpg", "1mp3d_0067_region27/626cac28fb084c17a09dfb81cfe01f22_i1_0.jpg", "1mp3d_0067_region27/626cac28fb084c17a09dfb81cfe01f22_i2_0.jpg" ], "turn_id": 8, "scan_id": "1mp3d_0067_region27", "origin_question": "Which direction is the rectangular glass window with a white frame in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the rectangular glass window with a white frame in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0545_01/00000.jpg", "scene0545_01/00030.jpg", "scene0545_01/00060.jpg", "scene0545_01/00090.jpg" ], "turn_id": 1, "scan_id": "scene0545_01", "origin_question": "Remember, have you seen any bread(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any bread(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0545_01/00120.jpg", "scene0545_01/00150.jpg", "scene0545_01/00180.jpg", "scene0545_01/00210.jpg" ], "turn_id": 2, "scan_id": "scene0545_01", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0545_01/00240.jpg", "scene0545_01/00270.jpg", "scene0545_01/00300.jpg", "scene0545_01/00330.jpg" ], "turn_id": 3, "scan_id": "scene0545_01", "origin_question": "Which direction is the light brown rectangular wooden shelf with two tiers in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the light brown rectangular wooden shelf with two tiers in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0545_01/00360.jpg", "scene0545_01/00390.jpg", "scene0545_01/00420.jpg", "scene0545_01/00450.jpg" ], "turn_id": 4, "scan_id": "scene0545_01", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the light brown rectangular wooden desk\" from your position?", "option": [], "answer": "148.2", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the light brown rectangular wooden desk\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0545_01/00480.jpg", "scene0545_01/00510.jpg", "scene0545_01/00540.jpg", "scene0545_01/00570.jpg" ], "turn_id": 5, "scan_id": "scene0545_01", "origin_question": "By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the sleek black chair with a curved backrest\", Y: \"the light brown rectangular wooden shelf with two tiers\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "farther away from both X and Y", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the sleek black chair with a curved backrest\", Y: \"the light brown rectangular wooden shelf with two tiers\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0545_01/00600.jpg", "scene0545_01/00630.jpg", "scene0545_01/00660.jpg", "scene0545_01/00690.jpg" ], "turn_id": 6, "scan_id": "scene0545_01", "origin_question": "At the end of which turn were both of \"the sleek black chair with a curved backrest\" and \"the light brown rectangular wooden shelf with two tiers\" on your right side?", "option": [], "answer": "5", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the sleek black chair with a curved backrest\" and \"the light brown rectangular wooden shelf with two tiers\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0030_00/00000.jpg", "scene0030_00/00040.jpg", "scene0030_00/00080.jpg", "scene0030_00/00120.jpg", "scene0030_00/00160.jpg" ], "turn_id": 1, "scan_id": "scene0030_00", "origin_question": "Remember, how many mop(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many mop(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0030_00/00200.jpg", "scene0030_00/00240.jpg", "scene0030_00/00280.jpg", "scene0030_00/00320.jpg", "scene0030_00/00360.jpg" ], "turn_id": 2, "scan_id": "scene0030_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0030_00/00400.jpg", "scene0030_00/00440.jpg", "scene0030_00/00480.jpg", "scene0030_00/00520.jpg", "scene0030_00/00560.jpg" ], "turn_id": 3, "scan_id": "scene0030_00", "origin_question": "How far is your current position from where you were at the end of the turn 1? (in meters)", "option": [], "answer": "1.62", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 1? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0030_00/00600.jpg", "scene0030_00/00640.jpg", "scene0030_00/00680.jpg", "scene0030_00/00720.jpg", "scene0030_00/00760.jpg" ], "turn_id": 4, "scan_id": "scene0030_00", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the brown paper bag with a round logo\", Y: \"the windowsill\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the brown paper bag with a round logo\", Y: \"the windowsill\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0030_00/00800.jpg", "scene0030_00/00840.jpg", "scene0030_00/00880.jpg", "scene0030_00/00920.jpg", "scene0030_00/00960.jpg" ], "turn_id": 5, "scan_id": "scene0030_00", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "2.82", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0030_00/01000.jpg", "scene0030_00/01040.jpg", "scene0030_00/01080.jpg", "scene0030_00/01120.jpg", "scene0030_00/01160.jpg" ], "turn_id": 6, "scan_id": "scene0030_00", "origin_question": "Which direction is the windowsill in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the windowsill in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0030_00/01200.jpg", "scene0030_00/01240.jpg", "scene0030_00/01280.jpg", "scene0030_00/01320.jpg", "scene0030_00/01360.jpg" ], "turn_id": 7, "scan_id": "scene0030_00", "origin_question": "Please recall \"the brown paper bag with a round logo\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "3.33", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the brown paper bag with a round logo\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "scene0030_00/01400.jpg", "scene0030_00/01440.jpg", "scene0030_00/01480.jpg", "scene0030_00/01520.jpg", "scene0030_00/01560.jpg" ], "turn_id": 8, "scan_id": "scene0030_00", "origin_question": "Among these three objects(\"the windowsill\",\"the medium-sized rectangular light beige door\",\"the small black plastic stapler\" ), which one is the closest to you?", "option": [ "the windowsill", "the medium-sized rectangular light beige door", "the small black plastic stapler" ], "answer": "the small black plastic stapler", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the windowsill\",\"the medium-sized rectangular light beige door\",\"the small black plastic stapler\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the windowsill', 'the medium-sized rectangular light beige door', 'the small black plastic stapler'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0030_00/01600.jpg", "scene0030_00/01640.jpg", "scene0030_00/01680.jpg", "scene0030_00/01720.jpg", "scene0030_00/01760.jpg" ], "turn_id": 9, "scan_id": "scene0030_00", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0030_00/01800.jpg", "scene0030_00/01840.jpg", "scene0030_00/01880.jpg", "scene0030_00/01920.jpg", "scene0030_00/01960.jpg" ], "turn_id": 10, "scan_id": "scene0030_00", "origin_question": "Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the small black plastic stapler\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the small black plastic stapler\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0203_02/00000.jpg", "scene0203_02/00040.jpg", "scene0203_02/00080.jpg", "scene0203_02/00120.jpg", "scene0203_02/00160.jpg" ], "turn_id": 1, "scan_id": "scene0203_02", "origin_question": "Remember, have you seen any stairs(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any stairs(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0203_02/00200.jpg", "scene0203_02/00240.jpg", "scene0203_02/00280.jpg", "scene0203_02/00320.jpg", "scene0203_02/00360.jpg" ], "turn_id": 2, "scan_id": "scene0203_02", "origin_question": "At the end of which turn were both of \"the low rectangular table with a dark smooth surface\" and \"the rectangular transparent plastic storage box\" on your left side?", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the low rectangular table with a dark smooth surface\" and \"the rectangular transparent plastic storage box\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0203_02/00400.jpg", "scene0203_02/00440.jpg", "scene0203_02/00480.jpg", "scene0203_02/00520.jpg", "scene0203_02/00560.jpg" ], "turn_id": 3, "scan_id": "scene0203_02", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "130.3", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "scene0203_02/00600.jpg", "scene0203_02/00640.jpg", "scene0203_02/00680.jpg", "scene0203_02/00720.jpg", "scene0203_02/00760.jpg" ], "turn_id": 4, "scan_id": "scene0203_02", "origin_question": "Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Left", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0203_02/00800.jpg", "scene0203_02/00840.jpg", "scene0203_02/00880.jpg", "scene0203_02/00920.jpg", "scene0203_02/00960.jpg" ], "turn_id": 5, "scan_id": "scene0203_02", "origin_question": "In which turn were you closest to \"the low rectangular table with a dark smooth surface\"? ", "option": [], "answer": "3", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the low rectangular table with a dark smooth surface\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0203_02/01000.jpg", "scene0203_02/01040.jpg", "scene0203_02/01080.jpg", "scene0203_02/01120.jpg", "scene0203_02/01160.jpg" ], "turn_id": 6, "scan_id": "scene0203_02", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the door\" from your position?", "option": [], "answer": "151.38", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the door\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0203_02/01200.jpg", "scene0203_02/01240.jpg", "scene0203_02/01280.jpg", "scene0203_02/01320.jpg", "scene0203_02/01380.jpg" ], "turn_id": 7, "scan_id": "scene0203_02", "origin_question": "Using your orientation at the end of turn 6 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "130.19", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 6 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0203_02/01420.jpg", "scene0203_02/01460.jpg", "scene0203_02/01500.jpg", "scene0203_02/01540.jpg", "scene0203_02/01580.jpg" ], "turn_id": 8, "scan_id": "scene0203_02", "origin_question": "How far is your current position from where you were at the end of the turn 7? (in meters)", "option": [], "answer": "1.36", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 7? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0203_02/01620.jpg", "scene0203_02/01660.jpg", "scene0203_02/01700.jpg", "scene0203_02/01740.jpg", "scene0203_02/01780.jpg" ], "turn_id": 9, "scan_id": "scene0203_02", "origin_question": "When did you first discover \"the low rectangular table with a dark smooth surface\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the low rectangular table with a dark smooth surface\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0203_02/01820.jpg", "scene0203_02/01860.jpg", "scene0203_02/01900.jpg", "scene0203_02/01940.jpg", "scene0203_02/01980.jpg" ], "turn_id": 10, "scan_id": "scene0203_02", "origin_question": "Compared to your position at the end of the 6 turn, are you now closer or farther away from \"the door\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 6 turn, are you now closer or farther away from \"the door\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0040_region23/5791a80c086b44a7a4e868d22fc374cc_i1_4.jpg", "1mp3d_0040_region23/5791a80c086b44a7a4e868d22fc374cc_i2_4.jpg", "1mp3d_0040_region23/3a0dc03705854da6abdd90b433887691_i1_0.jpg", "1mp3d_0040_region23/3a0dc03705854da6abdd90b433887691_i2_0.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0040_region23", "origin_question": "Remember, have you seen any fence(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any fence(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0040_region23/d5363662598b42b5b72a401065028040_i1_0.jpg", "1mp3d_0040_region23/d5363662598b42b5b72a401065028040_i2_0.jpg", "1mp3d_0040_region23/468e68b506364540972f4c718ca8c752_i1_2.jpg", "1mp3d_0040_region23/468e68b506364540972f4c718ca8c752_i2_2.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0040_region23", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "1mp3d_0040_region23/468e68b506364540972f4c718ca8c752_i1_4.jpg", "1mp3d_0040_region23/468e68b506364540972f4c718ca8c752_i2_4.jpg", "1mp3d_0040_region23/468e68b506364540972f4c718ca8c752_i1_3.jpg", "1mp3d_0040_region23/468e68b506364540972f4c718ca8c752_i2_3.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0040_region23", "origin_question": "In which turn were you farthest from \"the sleek, curved metallic faucet made of stainless steel\"? ", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the sleek, curved metallic faucet made of stainless steel\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "1mp3d_0040_region23/d5363662598b42b5b72a401065028040_i1_2.jpg", "1mp3d_0040_region23/d5363662598b42b5b72a401065028040_i2_2.jpg", "1mp3d_0040_region23/3a0dc03705854da6abdd90b433887691_i1_3.jpg", "1mp3d_0040_region23/3a0dc03705854da6abdd90b433887691_i2_3.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0040_region23", "origin_question": "Among these three objects(\"the rectangular gas stove with stainless steel decorations\",\"the rectangular picture with a yellow background\",\"the large circular stainless steel sink\" ), which one is the farthest from you?", "option": [ "the rectangular gas stove with stainless steel decorations", "the rectangular picture with a yellow background", "the large circular stainless steel sink" ], "answer": "the rectangular picture with a yellow background", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects(\"the rectangular gas stove with stainless steel decorations\",\"the rectangular picture with a yellow background\",\"the large circular stainless steel sink\" ), which one is the farthest from you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular gas stove with stainless steel decorations', 'the rectangular picture with a yellow background', 'the large circular stainless steel sink'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0040_region23/5791a80c086b44a7a4e868d22fc374cc_i1_1.jpg", "1mp3d_0040_region23/5791a80c086b44a7a4e868d22fc374cc_i2_1.jpg", "1mp3d_0040_region23/5791a80c086b44a7a4e868d22fc374cc_i1_2.jpg", "1mp3d_0040_region23/5791a80c086b44a7a4e868d22fc374cc_i2_2.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0040_region23", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the rectangular stainless steel range hood with a modern appearance\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the rectangular stainless steel range hood with a modern appearance\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "1mp3d_0040_region23/74ce15df381c460a8be6fac9709c6a58_i1_2.jpg", "1mp3d_0040_region23/74ce15df381c460a8be6fac9709c6a58_i2_2.jpg", "1mp3d_0040_region23/de9e1c3afd574dd79fb4c0a7282f7935_i1_2.jpg", "1mp3d_0040_region23/de9e1c3afd574dd79fb4c0a7282f7935_i2_2.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0040_region23", "origin_question": "When did you first discover \"the large circular stainless steel sink\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the large circular stainless steel sink\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "1mp3d_0040_region23/74ce15df381c460a8be6fac9709c6a58_i1_3.jpg", "1mp3d_0040_region23/74ce15df381c460a8be6fac9709c6a58_i2_3.jpg", "1mp3d_0040_region23/74ce15df381c460a8be6fac9709c6a58_i1_4.jpg", "1mp3d_0040_region23/74ce15df381c460a8be6fac9709c6a58_i2_4.jpg" ], "turn_id": 7, "scan_id": "1mp3d_0040_region23", "origin_question": "How far is your current position from where you were at the end of the turn 6? (in meters)", "option": [], "answer": "1.35", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 6? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "1mp3d_0040_region23/de9e1c3afd574dd79fb4c0a7282f7935_i1_4.jpg", "1mp3d_0040_region23/de9e1c3afd574dd79fb4c0a7282f7935_i2_4.jpg", "1mp3d_0040_region23/d509ae119b9f44daa01672e8e84a8e18_i1_3.jpg", "1mp3d_0040_region23/d509ae119b9f44daa01672e8e84a8e18_i2_3.jpg" ], "turn_id": 8, "scan_id": "1mp3d_0040_region23", "origin_question": "What will be the first-time appearance order of the following categories: chair, kitchen island, refrigerator?", "option": [ "chair, refrigerator, kitchen island", "kitchen island, chair, refrigerator", "refrigerator, kitchen island, chair", "kitchen island, refrigerator, chair" ], "answer": "kitchen island, chair, refrigerator", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: chair, kitchen island, refrigerator? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['chair, refrigerator, kitchen island', 'kitchen island, chair, refrigerator', 'refrigerator, kitchen island, chair', 'kitchen island, refrigerator, chair'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0406_01/00000.jpg", "scene0406_01/00040.jpg", "scene0406_01/00080.jpg", "scene0406_01/00120.jpg" ], "turn_id": 1, "scan_id": "scene0406_01", "origin_question": "Remember, have you seen any pool table(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any pool table(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0406_01/00160.jpg", "scene0406_01/00200.jpg", "scene0406_01/00240.jpg", "scene0406_01/00280.jpg" ], "turn_id": 2, "scan_id": "scene0406_01", "origin_question": "Please recall \"the modern metallic oval bin with a reflective surface\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.57", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the modern metallic oval bin with a reflective surface\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0406_01/00320.jpg", "scene0406_01/00360.jpg", "scene0406_01/00400.jpg", "scene0406_01/00440.jpg" ], "turn_id": 3, "scan_id": "scene0406_01", "origin_question": "In which turn did you see both \"the standard cylindrical roll of white toilet paper\" and \"the medium-sized white oval-shaped toilet\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the standard cylindrical roll of white toilet paper\" and \"the medium-sized white oval-shaped toilet\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0406_01/00480.jpg", "scene0406_01/00520.jpg", "scene0406_01/00560.jpg", "scene0406_01/00600.jpg" ], "turn_id": 4, "scan_id": "scene0406_01", "origin_question": "Is the modern metallic oval bin with a reflective surface to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the modern metallic oval bin with a reflective surface to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0406_01/00640.jpg", "scene0406_01/00680.jpg", "scene0406_01/00720.jpg", "scene0406_01/00760.jpg" ], "turn_id": 5, "scan_id": "scene0406_01", "origin_question": "At the end of which turn were both of \"the white rectangular porcelain bathtub\" and \"the medium-sized white oval-shaped toilet\" on your left side?", "option": [], "answer": "5", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the white rectangular porcelain bathtub\" and \"the medium-sized white oval-shaped toilet\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0406_01/00800.jpg", "scene0406_01/00840.jpg", "scene0406_01/00880.jpg", "scene0406_01/00920.jpg" ], "turn_id": 6, "scan_id": "scene0406_01", "origin_question": "Which two objects are on the same side of you now? \"the vertical beige curtain with a wave pattern\", \"the white rectangular porcelain bathtub\", \"the standard cylindrical roll of white toilet paper\". ", "option": [ "the vertical beige curtain with a wave pattern, the white rectangular porcelain bathtub", "the vertical beige curtain with a wave pattern, the standard cylindrical roll of white toilet paper", "the white rectangular porcelain bathtub, the standard cylindrical roll of white toilet paper" ], "answer": "the vertical beige curtain with a wave pattern, the white rectangular porcelain bathtub", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the vertical beige curtain with a wave pattern\", \"the white rectangular porcelain bathtub\", \"the standard cylindrical roll of white toilet paper\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the vertical beige curtain with a wave pattern, the white rectangular porcelain bathtub', 'the vertical beige curtain with a wave pattern, the standard cylindrical roll of white toilet paper', 'the white rectangular porcelain bathtub, the standard cylindrical roll of white toilet paper'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0406_01/00960.jpg", "scene0406_01/01000.jpg", "scene0406_01/01040.jpg", "scene0406_01/01080.jpg" ], "turn_id": 7, "scan_id": "scene0406_01", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "103.07", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0421_01/00000.jpg", "scene0421_01/00040.jpg", "scene0421_01/00080.jpg", "scene0421_01/00120.jpg", "scene0421_01/00160.jpg" ], "turn_id": 1, "scan_id": "scene0421_01", "origin_question": "Remember, have you seen any pillar(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any pillar(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0421_01/00200.jpg", "scene0421_01/00240.jpg", "scene0421_01/00280.jpg", "scene0421_01/00320.jpg", "scene0421_01/00360.jpg" ], "turn_id": 2, "scan_id": "scene0421_01", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0421_01/00400.jpg", "scene0421_01/00440.jpg", "scene0421_01/00480.jpg", "scene0421_01/00520.jpg", "scene0421_01/00560.jpg" ], "turn_id": 3, "scan_id": "scene0421_01", "origin_question": "Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "91.22", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0421_01/00600.jpg", "scene0421_01/00640.jpg", "scene0421_01/00680.jpg", "scene0421_01/00720.jpg", "scene0421_01/00760.jpg" ], "turn_id": 4, "scan_id": "scene0421_01", "origin_question": "In which turn did you see both \"the white rectangular wooden door, currently closed\" and \"the yellow broom with a black brush head\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the white rectangular wooden door, currently closed\" and \"the yellow broom with a black brush head\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0421_01/00800.jpg", "scene0421_01/00840.jpg", "scene0421_01/00880.jpg", "scene0421_01/00920.jpg", "scene0421_01/00960.jpg" ], "turn_id": 5, "scan_id": "scene0421_01", "origin_question": "What will be the first-time appearance order of the following categories: bottle, broom, washing machine?", "option": [ "bottle, washing machine, broom", "bottle, broom, washing machine", "broom, washing machine, bottle", "broom, bottle, washing machine" ], "answer": "broom, bottle, washing machine", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bottle, broom, washing machine? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['bottle, washing machine, broom', 'bottle, broom, washing machine', 'broom, washing machine, bottle', 'broom, bottle, washing machine'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0421_01/01000.jpg", "scene0421_01/01040.jpg", "scene0421_01/01080.jpg", "scene0421_01/01120.jpg", "scene0421_01/01160.jpg" ], "turn_id": 6, "scan_id": "scene0421_01", "origin_question": "Please recall \"the white rectangular wooden door, currently closed\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.44", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the white rectangular wooden door, currently closed\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0421_01/01200.jpg", "scene0421_01/01240.jpg", "scene0421_01/01280.jpg", "scene0421_01/01320.jpg", "scene0421_01/01360.jpg" ], "turn_id": 7, "scan_id": "scene0421_01", "origin_question": "Using your orientation at the end of turn 5 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 5 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0421_01/01400.jpg", "scene0421_01/01440.jpg", "scene0421_01/01480.jpg", "scene0421_01/01520.jpg", "scene0421_01/01560.jpg" ], "turn_id": 8, "scan_id": "scene0421_01", "origin_question": "In which turn did you see both \"the white rectangular wooden door, currently closed\" and \"the yellow broom with a black brush head\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the white rectangular wooden door, currently closed\" and \"the yellow broom with a black brush head\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0421_01/01600.jpg", "scene0421_01/01640.jpg", "scene0421_01/01680.jpg", "scene0421_01/01720.jpg", "scene0421_01/01760.jpg" ], "turn_id": 9, "scan_id": "scene0421_01", "origin_question": "Please recall \"the white rectangular wooden door, currently closed\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.97", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the white rectangular wooden door, currently closed\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0421_01/01800.jpg", "scene0421_01/01840.jpg", "scene0421_01/01880.jpg", "scene0421_01/01920.jpg", "scene0421_01/01960.jpg" ], "turn_id": 10, "scan_id": "scene0421_01", "origin_question": "When did you first discover \"the yellow broom with a black brush head\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the yellow broom with a black brush head\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0207_00/00000.jpg", "scene0207_00/00040.jpg", "scene0207_00/00080.jpg", "scene0207_00/00120.jpg", "scene0207_00/00160.jpg" ], "turn_id": 1, "scan_id": "scene0207_00", "origin_question": "Remember, how many pool(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many pool(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0207_00/00200.jpg", "scene0207_00/00240.jpg", "scene0207_00/00280.jpg", "scene0207_00/00320.jpg", "scene0207_00/00360.jpg" ], "turn_id": 2, "scan_id": "scene0207_00", "origin_question": "Is the stand to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the stand to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0207_00/00400.jpg", "scene0207_00/00440.jpg", "scene0207_00/00480.jpg", "scene0207_00/00520.jpg", "scene0207_00/00560.jpg" ], "turn_id": 3, "scan_id": "scene0207_00", "origin_question": "Is the radiator to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the radiator to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0207_00/00600.jpg", "scene0207_00/00640.jpg", "scene0207_00/00680.jpg", "scene0207_00/00720.jpg", "scene0207_00/00760.jpg" ], "turn_id": 4, "scan_id": "scene0207_00", "origin_question": "Please recall \"the medium-sized white cylindrical plastic basket\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.07", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the medium-sized white cylindrical plastic basket\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0207_00/00800.jpg", "scene0207_00/00840.jpg", "scene0207_00/00880.jpg", "scene0207_00/00920.jpg", "scene0207_00/00960.jpg" ], "turn_id": 5, "scan_id": "scene0207_00", "origin_question": "Which two objects are on the same side of you now? \"the radiator\", \"the sleek, black upright vacuum cleaner made of plastic\", \"the microwave\". ", "option": [ "the radiator, the sleek, black upright vacuum cleaner made of plastic", "the radiator, the microwave", "the sleek, black upright vacuum cleaner made of plastic, the microwave" ], "answer": "the sleek, black upright vacuum cleaner made of plastic, the microwave", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the radiator\", \"the sleek, black upright vacuum cleaner made of plastic\", \"the microwave\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the radiator, the sleek, black upright vacuum cleaner made of plastic', 'the radiator, the microwave', 'the sleek, black upright vacuum cleaner made of plastic, the microwave'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "scene0207_00/01000.jpg", "scene0207_00/01040.jpg", "scene0207_00/01080.jpg", "scene0207_00/01120.jpg", "scene0207_00/01160.jpg" ], "turn_id": 6, "scan_id": "scene0207_00", "origin_question": "Assuming the direction you are facing at the end of the turn 4 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Left", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 4 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0207_00/01200.jpg", "scene0207_00/01240.jpg", "scene0207_00/01280.jpg", "scene0207_00/01320.jpg", "scene0207_00/01360.jpg" ], "turn_id": 7, "scan_id": "scene0207_00", "origin_question": "What will be the first-time appearance order of the following categories: couch, shoe, toilet?", "option": [ "toilet, couch, shoe", "toilet, shoe, couch", "shoe, toilet, couch", "shoe, couch, toilet" ], "answer": "shoe, couch, toilet", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: couch, shoe, toilet? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['toilet, couch, shoe', 'toilet, shoe, couch', 'shoe, toilet, couch', 'shoe, couch, toilet'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0207_00/01400.jpg", "scene0207_00/01440.jpg", "scene0207_00/01480.jpg", "scene0207_00/01520.jpg", "scene0207_00/01560.jpg" ], "turn_id": 8, "scan_id": "scene0207_00", "origin_question": "Which two objects are on the same side of you now? \"the rectangular light brown wooden desk in good condition\", \"the dresser\", \"the sleek, black upright vacuum cleaner made of plastic\". ", "option": [ "the rectangular light brown wooden desk in good condition, the dresser", "the rectangular light brown wooden desk in good condition, the sleek, black upright vacuum cleaner made of plastic", "the dresser, the sleek, black upright vacuum cleaner made of plastic" ], "answer": "the rectangular light brown wooden desk in good condition, the dresser", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the rectangular light brown wooden desk in good condition\", \"the dresser\", \"the sleek, black upright vacuum cleaner made of plastic\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular light brown wooden desk in good condition, the dresser', 'the rectangular light brown wooden desk in good condition, the sleek, black upright vacuum cleaner made of plastic', 'the dresser, the sleek, black upright vacuum cleaner made of plastic'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0207_00/01600.jpg", "scene0207_00/01640.jpg", "scene0207_00/01680.jpg", "scene0207_00/01720.jpg", "scene0207_00/01760.jpg" ], "turn_id": 9, "scan_id": "scene0207_00", "origin_question": "Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "21.95", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0628_00/00000.jpg", "scene0628_00/00040.jpg", "scene0628_00/00080.jpg", "scene0628_00/00120.jpg" ], "turn_id": 1, "scan_id": "scene0628_00", "origin_question": "Remember, have you seen any paper(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any paper(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0628_00/00160.jpg", "scene0628_00/00200.jpg", "scene0628_00/00240.jpg", "scene0628_00/00280.jpg" ], "turn_id": 2, "scan_id": "scene0628_00", "origin_question": "When did you first discover \"the window\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the window\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0628_00/00320.jpg", "scene0628_00/00360.jpg", "scene0628_00/00400.jpg", "scene0628_00/00440.jpg" ], "turn_id": 3, "scan_id": "scene0628_00", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the white circular metal clock with a silver border\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the white circular metal clock with a silver border\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0628_00/00480.jpg", "scene0628_00/00520.jpg", "scene0628_00/00560.jpg", "scene0628_00/00600.jpg" ], "turn_id": 4, "scan_id": "scene0628_00", "origin_question": "Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "98.43", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0628_00/00640.jpg", "scene0628_00/00680.jpg", "scene0628_00/00720.jpg", "scene0628_00/00760.jpg" ], "turn_id": 5, "scan_id": "scene0628_00", "origin_question": "In which turn did you see both \"the medium-sized black backpack made of durable material\" and \"the large rectangular black blackboard made of stone\" simultaneously? ", "option": [], "answer": "3", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the medium-sized black backpack made of durable material\" and \"the large rectangular black blackboard made of stone\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0628_00/00800.jpg", "scene0628_00/00840.jpg", "scene0628_00/00880.jpg", "scene0628_00/00920.jpg" ], "turn_id": 6, "scan_id": "scene0628_00", "origin_question": "Please recall \"the white circular metal clock with a silver border\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.27", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the white circular metal clock with a silver border\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0628_00/00960.jpg", "scene0628_00/01000.jpg", "scene0628_00/01040.jpg", "scene0628_00/01080.jpg" ], "turn_id": 7, "scan_id": "scene0628_00", "origin_question": "Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the large rectangular black blackboard made of stone\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the large rectangular black blackboard made of stone\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "1mp3d_0074_region5/27b904430e4e48acb818e2ab80f183e3_i1_4.jpg", "1mp3d_0074_region5/27b904430e4e48acb818e2ab80f183e3_i2_4.jpg", "1mp3d_0074_region5/2ade9ff61be94782b425dd9f04d7847d_i1_3.jpg", "1mp3d_0074_region5/2ade9ff61be94782b425dd9f04d7847d_i2_3.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0074_region5", "origin_question": "Remember, how many light(s) have you seen so far? ", "option": [], "answer": "2", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many light(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0074_region5/0672a86275c14b148e8618ecde714e06_i1_4.jpg", "1mp3d_0074_region5/0672a86275c14b148e8618ecde714e06_i2_4.jpg", "1mp3d_0074_region5/0672a86275c14b148e8618ecde714e06_i1_3.jpg", "1mp3d_0074_region5/0672a86275c14b148e8618ecde714e06_i2_3.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0074_region5", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "1mp3d_0074_region5/2ade9ff61be94782b425dd9f04d7847d_i1_4.jpg", "1mp3d_0074_region5/2ade9ff61be94782b425dd9f04d7847d_i2_4.jpg", "1mp3d_0074_region5/27b904430e4e48acb818e2ab80f183e3_i1_3.jpg", "1mp3d_0074_region5/27b904430e4e48acb818e2ab80f183e3_i2_3.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0074_region5", "origin_question": "In which turn were you farthest from \"the rectangular white bed with blue borders\"? ", "option": [], "answer": "2", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the rectangular white bed with blue borders\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "1mp3d_0074_region5/27b904430e4e48acb818e2ab80f183e3_i1_2.jpg", "1mp3d_0074_region5/27b904430e4e48acb818e2ab80f183e3_i2_2.jpg", "1mp3d_0074_region5/c4ce11a276904bf38fe049e67b6c5f7a_i1_3.jpg", "1mp3d_0074_region5/c4ce11a276904bf38fe049e67b6c5f7a_i2_3.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0074_region5", "origin_question": "Which two objects are on the same side of you now? \"the white pillow with a blue border decoration\", \"the rectangular white bed with blue borders\", \"the long, textured gray curtain made of thick fabric\". ", "option": [ "the white pillow with a blue border decoration, the rectangular white bed with blue borders", "the white pillow with a blue border decoration, the long, textured gray curtain made of thick fabric", "the rectangular white bed with blue borders, the long, textured gray curtain made of thick fabric" ], "answer": "the white pillow with a blue border decoration, the rectangular white bed with blue borders", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the white pillow with a blue border decoration\", \"the rectangular white bed with blue borders\", \"the long, textured gray curtain made of thick fabric\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the white pillow with a blue border decoration, the rectangular white bed with blue borders', 'the white pillow with a blue border decoration, the long, textured gray curtain made of thick fabric', 'the rectangular white bed with blue borders, the long, textured gray curtain made of thick fabric'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0047_region1/af7ed758b8c04875a5ae383252e35332_i1_3.jpg", "1mp3d_0047_region1/af7ed758b8c04875a5ae383252e35332_i2_3.jpg", "1mp3d_0047_region1/bbfd4eb3b8d646a1bcb815d63dbf0fe2_i1_5.jpg", "1mp3d_0047_region1/bbfd4eb3b8d646a1bcb815d63dbf0fe2_i2_5.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0047_region1", "origin_question": "Remember, have you seen any stairs(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any stairs(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 2 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "1mp3d_0047_region1/bbfd4eb3b8d646a1bcb815d63dbf0fe2_i1_1.jpg", "1mp3d_0047_region1/bbfd4eb3b8d646a1bcb815d63dbf0fe2_i2_1.jpg", "1mp3d_0047_region1/bbfd4eb3b8d646a1bcb815d63dbf0fe2_i1_0.jpg", "1mp3d_0047_region1/bbfd4eb3b8d646a1bcb815d63dbf0fe2_i2_0.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0047_region1", "origin_question": "In which turn were you closest to \"Modern rectangular stainless steel stairs with a metallic sheen\"? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"Modern rectangular stainless steel stairs with a metallic sheen\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 2 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0616_01/00000.jpg", "scene0616_01/00040.jpg", "scene0616_01/00080.jpg", "scene0616_01/00120.jpg", "scene0616_01/00160.jpg" ], "turn_id": 1, "scan_id": "scene0616_01", "origin_question": "Remember, have you seen any chair(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any chair(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0616_01/00200.jpg", "scene0616_01/00240.jpg", "scene0616_01/00280.jpg", "scene0616_01/00320.jpg", "scene0616_01/00360.jpg" ], "turn_id": 2, "scan_id": "scene0616_01", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the deep brown rectangular wooden windowsill\"; \"the white rectangular metal radiator for household use\"; \"the rectangular, soft beige fabric couch with a curved back\" ", "option": [ "the deep brown rectangular wooden windowsill", "the white rectangular metal radiator for household use", "the rectangular, soft beige fabric couch with a curved back" ], "answer": "the deep brown rectangular wooden windowsill", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the deep brown rectangular wooden windowsill\"; \"the white rectangular metal radiator for household use\"; \"the rectangular, soft beige fabric couch with a curved back\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the deep brown rectangular wooden windowsill', 'the white rectangular metal radiator for household use', 'the rectangular, soft beige fabric couch with a curved back'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0616_01/00400.jpg", "scene0616_01/00440.jpg", "scene0616_01/00480.jpg", "scene0616_01/00520.jpg", "scene0616_01/00560.jpg" ], "turn_id": 3, "scan_id": "scene0616_01", "origin_question": "In which turn were you closest to \"the medium-sized stylish black vacuum cleaner with a red stripe\"? ", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the medium-sized stylish black vacuum cleaner with a red stripe\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "scene0616_01/00600.jpg", "scene0616_01/00640.jpg", "scene0616_01/00680.jpg", "scene0616_01/00720.jpg", "scene0616_01/00760.jpg" ], "turn_id": 4, "scan_id": "scene0616_01", "origin_question": "Among these three objects(\"the rectangular, soft beige fabric couch with a curved back\",\"the large rectangular brick fireplace with a herringbone pattern\",\"the deep brown rectangular wooden windowsill\" ), which one is the closest to you?", "option": [ "the rectangular, soft beige fabric couch with a curved back", "the large rectangular brick fireplace with a herringbone pattern", "the deep brown rectangular wooden windowsill" ], "answer": "the rectangular, soft beige fabric couch with a curved back", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the rectangular, soft beige fabric couch with a curved back\",\"the large rectangular brick fireplace with a herringbone pattern\",\"the deep brown rectangular wooden windowsill\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular, soft beige fabric couch with a curved back', 'the large rectangular brick fireplace with a herringbone pattern', 'the deep brown rectangular wooden windowsill'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0616_01/00800.jpg", "scene0616_01/00840.jpg", "scene0616_01/00880.jpg", "scene0616_01/00920.jpg", "scene0616_01/00960.jpg" ], "turn_id": 5, "scan_id": "scene0616_01", "origin_question": "At the end of which turn were both of \"the cylindrical white lampshade with a silver metal base\" and \"the deep brown rectangular wooden windowsill\" on your right side?", "option": [], "answer": "3", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the cylindrical white lampshade with a silver metal base\" and \"the deep brown rectangular wooden windowsill\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0616_01/01000.jpg", "scene0616_01/01040.jpg", "scene0616_01/01080.jpg", "scene0616_01/01120.jpg", "scene0616_01/01160.jpg" ], "turn_id": 6, "scan_id": "scene0616_01", "origin_question": "Is the large rectangular white-framed glass window to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the large rectangular white-framed glass window to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0616_01/01200.jpg", "scene0616_01/01240.jpg", "scene0616_01/01280.jpg", "scene0616_01/01320.jpg", "scene0616_01/01360.jpg" ], "turn_id": 7, "scan_id": "scene0616_01", "origin_question": "Compared to your position at the end of the 5 turn, are you now closer or farther away from \"the deep brown rectangular wooden windowsill\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 5 turn, are you now closer or farther away from \"the deep brown rectangular wooden windowsill\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0616_01/01400.jpg", "scene0616_01/01440.jpg", "scene0616_01/01480.jpg", "scene0616_01/01520.jpg", "scene0616_01/01560.jpg" ], "turn_id": 8, "scan_id": "scene0616_01", "origin_question": "Which direction is the large rectangular brick fireplace with a herringbone pattern in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the large rectangular brick fireplace with a herringbone pattern in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0616_01/01600.jpg", "scene0616_01/01640.jpg", "scene0616_01/01680.jpg", "scene0616_01/01720.jpg", "scene0616_01/01760.jpg" ], "turn_id": 9, "scan_id": "scene0616_01", "origin_question": "In which turn did you see both \"the cylindrical white lampshade with a silver metal base\" and \"the large rectangular white-framed glass window\" simultaneously? ", "option": [], "answer": "4", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the cylindrical white lampshade with a silver metal base\" and \"the large rectangular white-framed glass window\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0616_01/01800.jpg", "scene0616_01/01840.jpg", "scene0616_01/01880.jpg", "scene0616_01/01920.jpg", "scene0616_01/01960.jpg" ], "turn_id": 10, "scan_id": "scene0616_01", "origin_question": "Please recall \"the medium-sized stylish black vacuum cleaner with a red stripe\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.86", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the medium-sized stylish black vacuum cleaner with a red stripe\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0664_00/00000.jpg", "scene0664_00/00040.jpg", "scene0664_00/00080.jpg", "scene0664_00/00120.jpg" ], "turn_id": 1, "scan_id": "scene0664_00", "origin_question": "Remember, have you seen any toilet paper(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any toilet paper(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0664_00/00160.jpg", "scene0664_00/00200.jpg", "scene0664_00/00240.jpg", "scene0664_00/00280.jpg" ], "turn_id": 2, "scan_id": "scene0664_00", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the circular silver wall-mounted shower\"; \"the medium-sized white oval toilet\"; \"the bathtub\" ", "option": [ "the circular silver wall-mounted shower", "the medium-sized white oval toilet", "the bathtub" ], "answer": "the circular silver wall-mounted shower", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the circular silver wall-mounted shower\"; \"the medium-sized white oval toilet\"; \"the bathtub\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the circular silver wall-mounted shower', 'the medium-sized white oval toilet', 'the bathtub'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0664_00/00320.jpg", "scene0664_00/00360.jpg", "scene0664_00/00400.jpg", "scene0664_00/00440.jpg" ], "turn_id": 3, "scan_id": "scene0664_00", "origin_question": "In which turn did you see both \"the bathtub\" and \"the rectangular framed picture with a black frame\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the bathtub\" and \"the rectangular framed picture with a black frame\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0664_00/00480.jpg", "scene0664_00/00520.jpg", "scene0664_00/00560.jpg", "scene0664_00/00600.jpg" ], "turn_id": 4, "scan_id": "scene0664_00", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the bathtub\"; \"the smooth rectangular beige door\"; \"the medium-sized white oval toilet\" ", "option": [ "the bathtub", "the smooth rectangular beige door", "the medium-sized white oval toilet" ], "answer": "the smooth rectangular beige door", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the bathtub\"; \"the smooth rectangular beige door\"; \"the medium-sized white oval toilet\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the bathtub', 'the smooth rectangular beige door', 'the medium-sized white oval toilet'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0664_00/00640.jpg", "scene0664_00/00680.jpg", "scene0664_00/00720.jpg", "scene0664_00/00760.jpg" ], "turn_id": 5, "scan_id": "scene0664_00", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the modern cylindrical stainless steel bin in light silver\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the modern cylindrical stainless steel bin in light silver\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0664_00/00800.jpg", "scene0664_00/00840.jpg", "scene0664_00/00880.jpg", "scene0664_00/00920.jpg" ], "turn_id": 6, "scan_id": "scene0664_00", "origin_question": "Which direction is the modern cylindrical stainless steel bin in light silver in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the modern cylindrical stainless steel bin in light silver in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0664_00/00960.jpg", "scene0664_00/01000.jpg", "scene0664_00/01040.jpg", "scene0664_00/01080.jpg" ], "turn_id": 7, "scan_id": "scene0664_00", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the rectangular framed picture with a black frame\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the rectangular framed picture with a black frame\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0664_00/01120.jpg", "scene0664_00/01160.jpg", "scene0664_00/01200.jpg", "scene0664_00/01240.jpg" ], "turn_id": 8, "scan_id": "scene0664_00", "origin_question": "Please recall \"the rectangular framed picture with a black frame\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.02", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the rectangular framed picture with a black frame\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0664_00/01280.jpg", "scene0664_00/01320.jpg", "scene0664_00/01360.jpg", "scene0664_00/01400.jpg" ], "turn_id": 9, "scan_id": "scene0664_00", "origin_question": "Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 9 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0509_01/00000.jpg", "scene0509_01/00040.jpg", "scene0509_01/00080.jpg", "scene0509_01/00120.jpg" ], "turn_id": 1, "scan_id": "scene0509_01", "origin_question": "Remember, have you seen any curtain(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any curtain(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0509_01/00160.jpg", "scene0509_01/00200.jpg", "scene0509_01/00240.jpg", "scene0509_01/00280.jpg" ], "turn_id": 2, "scan_id": "scene0509_01", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the hair dryer\"; \"the large rectangular mirror with a silver frame\"; \"the white, square vanity with a silver faucet\" ", "option": [ "the hair dryer", "the large rectangular mirror with a silver frame", "the white, square vanity with a silver faucet" ], "answer": "the large rectangular mirror with a silver frame", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the hair dryer\"; \"the large rectangular mirror with a silver frame\"; \"the white, square vanity with a silver faucet\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the hair dryer', 'the large rectangular mirror with a silver frame', 'the white, square vanity with a silver faucet'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0509_01/00320.jpg", "scene0509_01/00360.jpg", "scene0509_01/00400.jpg", "scene0509_01/00440.jpg" ], "turn_id": 3, "scan_id": "scene0509_01", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the small, silver metal bin\" from your position?", "option": [], "answer": "162.8", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the small, silver metal bin\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0509_01/00480.jpg", "scene0509_01/00520.jpg", "scene0509_01/00560.jpg", "scene0509_01/00600.jpg" ], "turn_id": 4, "scan_id": "scene0509_01", "origin_question": "Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "50.71", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0509_01/00640.jpg", "scene0509_01/00680.jpg", "scene0509_01/00720.jpg", "scene0509_01/00760.jpg" ], "turn_id": 5, "scan_id": "scene0509_01", "origin_question": "Using your orientation at the end of turn 4 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "79.88", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 4 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0509_01/00800.jpg", "scene0509_01/00840.jpg", "scene0509_01/00880.jpg", "scene0509_01/00920.jpg" ], "turn_id": 6, "scan_id": "scene0509_01", "origin_question": "Which two objects are on the same side of you now? \"the medium-sized polished bar\", \"the hair dryer\", \"the rectangular white plastic switch with a beige edge\". ", "option": [ "the medium-sized polished bar, the hair dryer", "the medium-sized polished bar, the rectangular white plastic switch with a beige edge", "the hair dryer, the rectangular white plastic switch with a beige edge" ], "answer": "the hair dryer, the rectangular white plastic switch with a beige edge", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the medium-sized polished bar\", \"the hair dryer\", \"the rectangular white plastic switch with a beige edge\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the medium-sized polished bar, the hair dryer', 'the medium-sized polished bar, the rectangular white plastic switch with a beige edge', 'the hair dryer, the rectangular white plastic switch with a beige edge'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0047_region24/4890b24fbb59414892ca854895896126_i1_3.jpg", "1mp3d_0047_region24/4890b24fbb59414892ca854895896126_i2_3.jpg", "1mp3d_0047_region24/37cb64e0cddc4f20b309abc4ff5344c2_i1_0.jpg", "1mp3d_0047_region24/37cb64e0cddc4f20b309abc4ff5344c2_i2_0.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0047_region24", "origin_question": "Remember, have you seen any curtain(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any curtain(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 2 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "1mp3d_0047_region24/37cb64e0cddc4f20b309abc4ff5344c2_i1_5.jpg", "1mp3d_0047_region24/37cb64e0cddc4f20b309abc4ff5344c2_i2_5.jpg", "1mp3d_0047_region24/4890b24fbb59414892ca854895896126_i1_5.jpg", "1mp3d_0047_region24/4890b24fbb59414892ca854895896126_i2_5.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0047_region24", "origin_question": "When did you first discover \"the vertical curtain with a silver frame\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the vertical curtain with a silver frame\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 2 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0334_01/00000.jpg", "scene0334_01/00040.jpg", "scene0334_01/00080.jpg", "scene0334_01/00120.jpg" ], "turn_id": 1, "scan_id": "scene0334_01", "origin_question": "Remember, have you seen any table(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any table(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0334_01/00160.jpg", "scene0334_01/00200.jpg", "scene0334_01/00240.jpg", "scene0334_01/00280.jpg" ], "turn_id": 2, "scan_id": "scene0334_01", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0334_01/00320.jpg", "scene0334_01/00360.jpg", "scene0334_01/00400.jpg", "scene0334_01/00440.jpg" ], "turn_id": 3, "scan_id": "scene0334_01", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0334_01/00480.jpg", "scene0334_01/00520.jpg", "scene0334_01/00560.jpg", "scene0334_01/00600.jpg" ], "turn_id": 4, "scan_id": "scene0334_01", "origin_question": "Remember, how many bench(s) have you seen so far? ", "option": [], "answer": "2", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many bench(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0334_01/00640.jpg", "scene0334_01/00680.jpg", "scene0334_01/00720.jpg", "scene0334_01/00760.jpg" ], "turn_id": 5, "scan_id": "scene0334_01", "origin_question": "In which turn were you closest to \"the tall white cylindrical pillar with a smooth surface\"? ", "option": [], "answer": "5", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the tall white cylindrical pillar with a smooth surface\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0334_01/00800.jpg", "scene0334_01/00840.jpg", "scene0334_01/00880.jpg", "scene0334_01/00920.jpg" ], "turn_id": 6, "scan_id": "scene0334_01", "origin_question": "Using your orientation at the end of turn 5 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "103.62", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 5 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0558_02/00000.jpg", "scene0558_02/00040.jpg", "scene0558_02/00080.jpg", "scene0558_02/00120.jpg", "scene0558_02/00160.jpg" ], "turn_id": 1, "scan_id": "scene0558_02", "origin_question": "Remember, have you seen any book(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any book(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0558_02/00200.jpg", "scene0558_02/00240.jpg", "scene0558_02/00280.jpg", "scene0558_02/00320.jpg", "scene0558_02/00360.jpg" ], "turn_id": 2, "scan_id": "scene0558_02", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0558_02/00400.jpg", "scene0558_02/00440.jpg", "scene0558_02/00480.jpg", "scene0558_02/00520.jpg", "scene0558_02/00560.jpg" ], "turn_id": 3, "scan_id": "scene0558_02", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the round, smooth light wooden table with elegant legs\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the round, smooth light wooden table with elegant legs\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0558_02/00600.jpg", "scene0558_02/00640.jpg", "scene0558_02/00680.jpg", "scene0558_02/00720.jpg", "scene0558_02/00760.jpg" ], "turn_id": 4, "scan_id": "scene0558_02", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the rectangular canvas picture with a bright red frame\" from your position?", "option": [], "answer": "105.82", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the rectangular canvas picture with a bright red frame\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0558_02/00800.jpg", "scene0558_02/00840.jpg", "scene0558_02/00880.jpg", "scene0558_02/00920.jpg", "scene0558_02/00960.jpg" ], "turn_id": 5, "scan_id": "scene0558_02", "origin_question": "What will be the first-time appearance order of the following categories: plant, shelf, window?", "option": [ "window, shelf, plant", "shelf, plant, window", "plant, shelf, window", "shelf, window, plant" ], "answer": "shelf, window, plant", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: plant, shelf, window? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['window, shelf, plant', 'shelf, plant, window', 'plant, shelf, window', 'shelf, window, plant'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0558_02/01000.jpg", "scene0558_02/01040.jpg", "scene0558_02/01080.jpg", "scene0558_02/01120.jpg", "scene0558_02/01160.jpg" ], "turn_id": 6, "scan_id": "scene0558_02", "origin_question": "Is the round, smooth light wooden table with elegant legs to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the round, smooth light wooden table with elegant legs to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0558_02/01320.jpg", "scene0558_02/01360.jpg", "scene0558_02/01400.jpg", "scene0558_02/01440.jpg", "scene0558_02/01480.jpg" ], "turn_id": 7, "scan_id": "scene0558_02", "origin_question": "Using your orientation at the end of turn 5 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "87.63", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 5 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0558_02/01520.jpg", "scene0558_02/01560.jpg", "scene0558_02/01600.jpg", "scene0558_02/01640.jpg", "scene0558_02/01680.jpg" ], "turn_id": 8, "scan_id": "scene0558_02", "origin_question": "By the end of the 6 turn, were you getting closer to or farther away from these two objects: X: \"the round, smooth light wooden table with elegant legs\", Y: \"the rectangular canvas picture with a bright red frame\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 6 turn, were you getting closer to or farther away from these two objects: X: \"the round, smooth light wooden table with elegant legs\", Y: \"the rectangular canvas picture with a bright red frame\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0553_01/00000.jpg", "scene0553_01/00040.jpg", "scene0553_01/00080.jpg", "scene0553_01/00120.jpg" ], "turn_id": 1, "scan_id": "scene0553_01", "origin_question": "Remember, have you seen any toilet(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any toilet(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0553_01/00160.jpg", "scene0553_01/00200.jpg", "scene0553_01/00240.jpg", "scene0553_01/00280.jpg" ], "turn_id": 2, "scan_id": "scene0553_01", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the small silver metal garbage bin with an open door\"; \"the slim stainless steel bar with a shiny surface\"; \"the toilet paper\" ", "option": [ "the small silver metal garbage bin with an open door", "the slim stainless steel bar with a shiny surface", "the toilet paper" ], "answer": "the small silver metal garbage bin with an open door", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the small silver metal garbage bin with an open door\"; \"the slim stainless steel bar with a shiny surface\"; \"the toilet paper\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the small silver metal garbage bin with an open door', 'the slim stainless steel bar with a shiny surface', 'the toilet paper'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0553_01/00320.jpg", "scene0553_01/00360.jpg", "scene0553_01/00400.jpg", "scene0553_01/00440.jpg" ], "turn_id": 3, "scan_id": "scene0553_01", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the white ceramic standalone toilet with a matching seat\" from your position?", "option": [], "answer": "117.92", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the white ceramic standalone toilet with a matching seat\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0553_01/00480.jpg", "scene0553_01/00520.jpg", "scene0553_01/00560.jpg", "scene0553_01/00600.jpg" ], "turn_id": 4, "scan_id": "scene0553_01", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0553_01/00640.jpg", "scene0553_01/00680.jpg", "scene0553_01/00720.jpg", "scene0553_01/00760.jpg" ], "turn_id": 5, "scan_id": "scene0553_01", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the toilet paper\" from your position?", "option": [], "answer": "138.8", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the toilet paper\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0553_01/00800.jpg", "scene0553_01/00840.jpg", "scene0553_01/00880.jpg", "scene0553_01/00920.jpg" ], "turn_id": 6, "scan_id": "scene0553_01", "origin_question": "Which two objects are on the same side of you now? \"the white ceramic standalone toilet with a matching seat\", \"the dispenser\", \"the medium-sized black and gray rectangular backpack\". ", "option": [ "the white ceramic standalone toilet with a matching seat, the dispenser", "the white ceramic standalone toilet with a matching seat, the medium-sized black and gray rectangular backpack", "the dispenser, the medium-sized black and gray rectangular backpack" ], "answer": "the white ceramic standalone toilet with a matching seat, the dispenser", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the white ceramic standalone toilet with a matching seat\", \"the dispenser\", \"the medium-sized black and gray rectangular backpack\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the white ceramic standalone toilet with a matching seat, the dispenser', 'the white ceramic standalone toilet with a matching seat, the medium-sized black and gray rectangular backpack', 'the dispenser, the medium-sized black and gray rectangular backpack'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0553_01/00960.jpg", "scene0553_01/01000.jpg", "scene0553_01/01040.jpg", "scene0553_01/01080.jpg" ], "turn_id": 7, "scan_id": "scene0553_01", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the slim stainless steel bar with a shiny surface\" from your position?", "option": [], "answer": "77.87", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the slim stainless steel bar with a shiny surface\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0553_01/01120.jpg", "scene0553_01/01160.jpg", "scene0553_01/01200.jpg", "scene0553_01/01240.jpg" ], "turn_id": 8, "scan_id": "scene0553_01", "origin_question": "Using your orientation at the end of turn 6 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "86.35", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 6 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0118_00/00000.jpg", "scene0118_00/00030.jpg", "scene0118_00/00060.jpg", "scene0118_00/00090.jpg" ], "turn_id": 1, "scan_id": "scene0118_00", "origin_question": "Remember, have you seen any magazine(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any magazine(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0118_00/00120.jpg", "scene0118_00/00150.jpg", "scene0118_00/00180.jpg", "scene0118_00/00210.jpg" ], "turn_id": 2, "scan_id": "scene0118_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0118_00/00240.jpg", "scene0118_00/00270.jpg", "scene0118_00/00300.jpg", "scene0118_00/00330.jpg" ], "turn_id": 3, "scan_id": "scene0118_00", "origin_question": "What will be the first-time appearance order of the following categories: book, jacket, sink?", "option": [ "jacket, book, sink", "book, sink, jacket", "sink, jacket, book", "book, jacket, sink" ], "answer": "jacket, book, sink", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: book, jacket, sink? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['jacket, book, sink', 'book, sink, jacket', 'sink, jacket, book', 'book, jacket, sink'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0118_00/00360.jpg", "scene0118_00/00390.jpg", "scene0118_00/00420.jpg", "scene0118_00/00450.jpg" ], "turn_id": 4, "scan_id": "scene0118_00", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the compact rectangular book with a vibrant red cover\" from your position?", "option": [], "answer": "131.95", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the compact rectangular book with a vibrant red cover\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0118_00/00480.jpg", "scene0118_00/00510.jpg", "scene0118_00/00540.jpg", "scene0118_00/00570.jpg" ], "turn_id": 5, "scan_id": "scene0118_00", "origin_question": "What will be the first-time appearance order of the following categories: book, sink, table?", "option": [ "sink, book, table", "book, sink, table", "book, table, sink", "table, book, sink" ], "answer": "table, book, sink", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: book, sink, table? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['sink, book, table', 'book, sink, table', 'book, table, sink', 'table, book, sink'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "1mp3d_0044_region40/9dfff33c1f0b46589e3cbd1def57bbde_i1_2.jpg", "1mp3d_0044_region40/9dfff33c1f0b46589e3cbd1def57bbde_i2_2.jpg", "1mp3d_0044_region40/72f5a95c93054a44a76b806392b36c2d_i1_1.jpg", "1mp3d_0044_region40/72f5a95c93054a44a76b806392b36c2d_i2_1.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0044_region40", "origin_question": "Remember, how many dresser(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many dresser(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "1mp3d_0044_region40/26fa42364e8f4343893b313829a8721b_i1_0.jpg", "1mp3d_0044_region40/26fa42364e8f4343893b313829a8721b_i2_0.jpg", "1mp3d_0044_region40/72f5a95c93054a44a76b806392b36c2d_i1_2.jpg", "1mp3d_0044_region40/72f5a95c93054a44a76b806392b36c2d_i2_2.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0044_region40", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the smooth, round decoration made of dark material\"; \"the door\"; \"the large rectangular white stone fireplace\" ", "option": [ "the smooth, round decoration made of dark material", "the door", "the large rectangular white stone fireplace" ], "answer": "the door", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the smooth, round decoration made of dark material\"; \"the door\"; \"the large rectangular white stone fireplace\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the smooth, round decoration made of dark material', 'the door', 'the large rectangular white stone fireplace'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0044_region40/26fa42364e8f4343893b313829a8721b_i1_2.jpg", "1mp3d_0044_region40/26fa42364e8f4343893b313829a8721b_i2_2.jpg", "1mp3d_0044_region40/26fa42364e8f4343893b313829a8721b_i1_1.jpg", "1mp3d_0044_region40/26fa42364e8f4343893b313829a8721b_i2_1.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0044_region40", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the large rectangular white stone fireplace\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the large rectangular white stone fireplace\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "1mp3d_0044_region40/ce117fb11ae8486dbc4c9d47a0cd829d_i1_3.jpg", "1mp3d_0044_region40/ce117fb11ae8486dbc4c9d47a0cd829d_i2_3.jpg", "1mp3d_0044_region40/ce117fb11ae8486dbc4c9d47a0cd829d_i1_2.jpg", "1mp3d_0044_region40/ce117fb11ae8486dbc4c9d47a0cd829d_i2_2.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0044_region40", "origin_question": "Which direction is the door in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the door in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "1mp3d_0044_region40/cef845328ceb4a548007c3b5cfb2642f_i1_4.jpg", "1mp3d_0044_region40/cef845328ceb4a548007c3b5cfb2642f_i2_4.jpg", "1mp3d_0044_region40/cef845328ceb4a548007c3b5cfb2642f_i2_5.jpg", "1mp3d_0044_region40/cef845328ceb4a548007c3b5cfb2642f_i1_0.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0044_region40", "origin_question": "In which turn did you see both \"the smooth, round decoration made of dark material\" and \"the door\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the smooth, round decoration made of dark material\" and \"the door\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "1mp3d_0044_region40/cef845328ceb4a548007c3b5cfb2642f_i2_0.jpg", "1mp3d_0044_region40/6f616e80892645bd9030aff59f9c36ba_i1_0.jpg", "1mp3d_0044_region40/6f616e80892645bd9030aff59f9c36ba_i2_0.jpg", "1mp3d_0044_region40/14646d17b7ea4433b23a5562158dcb52_i1_1.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0044_region40", "origin_question": "Please recall \"the mat\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "7.12", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the mat\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "1mp3d_0044_region40/14646d17b7ea4433b23a5562158dcb52_i2_1.jpg", "1mp3d_0044_region40/95f92aee4646442280f6bd962e64641b_i1_3.jpg", "1mp3d_0044_region40/95f92aee4646442280f6bd962e64641b_i2_3.jpg", "1mp3d_0044_region40/95f92aee4646442280f6bd962e64641b_i1_4.jpg" ], "turn_id": 7, "scan_id": "1mp3d_0044_region40", "origin_question": "Using your orientation at the end of turn 6 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "76.37", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 6 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "1mp3d_0044_region40/95f92aee4646442280f6bd962e64641b_i2_4.jpg", "1mp3d_0044_region40/14646d17b7ea4433b23a5562158dcb52_i1_3.jpg", "1mp3d_0044_region40/14646d17b7ea4433b23a5562158dcb52_i2_3.jpg", "1mp3d_0044_region40/14646d17b7ea4433b23a5562158dcb52_i1_2.jpg" ], "turn_id": 8, "scan_id": "1mp3d_0044_region40", "origin_question": "Which direction is the door in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the door in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0423_01/00000.jpg", "scene0423_01/00030.jpg", "scene0423_01/00060.jpg", "scene0423_01/00090.jpg" ], "turn_id": 1, "scan_id": "scene0423_01", "origin_question": "Remember, have you seen any sack(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any sack(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0423_01/00120.jpg", "scene0423_01/00150.jpg", "scene0423_01/00180.jpg", "scene0423_01/00210.jpg" ], "turn_id": 2, "scan_id": "scene0423_01", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0423_01/00240.jpg", "scene0423_01/00270.jpg", "scene0423_01/00300.jpg", "scene0423_01/00330.jpg" ], "turn_id": 3, "scan_id": "scene0423_01", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "scene0423_01/00360.jpg", "scene0423_01/00390.jpg", "scene0423_01/00420.jpg", "scene0423_01/00450.jpg" ], "turn_id": 4, "scan_id": "scene0423_01", "origin_question": "Assuming the direction you are facing at the end of the turn 3 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Right", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 3 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0423_01/00480.jpg", "scene0423_01/00510.jpg", "scene0423_01/00540.jpg", "scene0423_01/00570.jpg" ], "turn_id": 5, "scan_id": "scene0423_01", "origin_question": "", "option": [], "answer": "", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0040_region39/a1fac5ca305e46c6b778fa32e244868b_i1_2.jpg", "1mp3d_0040_region39/a1fac5ca305e46c6b778fa32e244868b_i2_2.jpg", "1mp3d_0040_region39/a1fac5ca305e46c6b778fa32e244868b_i1_1.jpg", "1mp3d_0040_region39/a1fac5ca305e46c6b778fa32e244868b_i2_1.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0040_region39", "origin_question": "Remember, have you seen any excercise equipment(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any excercise equipment(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0040_region39/a1fac5ca305e46c6b778fa32e244868b_i1_3.jpg", "1mp3d_0040_region39/a1fac5ca305e46c6b778fa32e244868b_i2_3.jpg", "1mp3d_0040_region39/52771bb222044d689a1676d9d2085b85_i1_2.jpg", "1mp3d_0040_region39/52771bb222044d689a1676d9d2085b85_i2_2.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0040_region39", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0040_region39/c9de4124b5664e17b2de92e43314390f_i1_2.jpg", "1mp3d_0040_region39/c9de4124b5664e17b2de92e43314390f_i2_2.jpg", "1mp3d_0040_region39/b2841aff8c1d4b4fae73a440cb628aa4_i1_2.jpg", "1mp3d_0040_region39/b2841aff8c1d4b4fae73a440cb628aa4_i2_2.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0040_region39", "origin_question": "", "option": [], "answer": "", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "1mp3d_0040_region39/b2841aff8c1d4b4fae73a440cb628aa4_i1_1.jpg", "1mp3d_0040_region39/b2841aff8c1d4b4fae73a440cb628aa4_i2_1.jpg", "1mp3d_0040_region39/c9de4124b5664e17b2de92e43314390f_i1_0.jpg", "1mp3d_0040_region39/c9de4124b5664e17b2de92e43314390f_i2_0.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0040_region39", "origin_question": "Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance forward or backward from that position?", "option": [ "Forward", "Backward" ], "answer": "Backward", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance forward or backward from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Forward', 'Backward'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0040_region39/52771bb222044d689a1676d9d2085b85_i1_0.jpg", "1mp3d_0040_region39/52771bb222044d689a1676d9d2085b85_i2_0.jpg", "1mp3d_0040_region39/52771bb222044d689a1676d9d2085b85_i1_5.jpg", "1mp3d_0040_region39/52771bb222044d689a1676d9d2085b85_i2_5.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0040_region39", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the rectangular white plastic thermostat with a digital display\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the rectangular white plastic thermostat with a digital display\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "1mp3d_0040_region39/c9de4124b5664e17b2de92e43314390f_i1_5.jpg", "1mp3d_0040_region39/c9de4124b5664e17b2de92e43314390f_i2_5.jpg", "1mp3d_0040_region39/b2841aff8c1d4b4fae73a440cb628aa4_i1_0.jpg", "1mp3d_0040_region39/b2841aff8c1d4b4fae73a440cb628aa4_i2_0.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0040_region39", "origin_question": "Please recall \"the rectangular white plastic thermostat with a digital display\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.44", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the rectangular white plastic thermostat with a digital display\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0680_01/00000.jpg", "scene0680_01/00030.jpg", "scene0680_01/00060.jpg", "scene0680_01/00090.jpg" ], "turn_id": 1, "scan_id": "scene0680_01", "origin_question": "Remember, have you seen any chair(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any chair(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0680_01/00120.jpg", "scene0680_01/00150.jpg", "scene0680_01/00180.jpg", "scene0680_01/00210.jpg" ], "turn_id": 2, "scan_id": "scene0680_01", "origin_question": "Remember, have you seen any blackboard(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any blackboard(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0680_01/00240.jpg", "scene0680_01/00270.jpg", "scene0680_01/00300.jpg", "scene0680_01/00330.jpg" ], "turn_id": 3, "scan_id": "scene0680_01", "origin_question": "In which turn were you closest to \"the rectangular light wood table of moderate size\"? ", "option": [], "answer": "2", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the rectangular light wood table of moderate size\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0680_01/00360.jpg", "scene0680_01/00390.jpg", "scene0680_01/00420.jpg", "scene0680_01/00450.jpg" ], "turn_id": 4, "scan_id": "scene0680_01", "origin_question": "Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "79.2", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0680_01/00480.jpg", "scene0680_01/00510.jpg", "scene0680_01/00540.jpg", "scene0680_01/00570.jpg" ], "turn_id": 5, "scan_id": "scene0680_01", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the modern flat-screen TV with a dark matte surface\" from your position?", "option": [], "answer": "79.66", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the modern flat-screen TV with a dark matte surface\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0680_01/00600.jpg", "scene0680_01/01070.jpg", "scene0680_01/01100.jpg", "scene0680_01/01130.jpg" ], "turn_id": 6, "scan_id": "scene0680_01", "origin_question": "At the end of which turn were both of \"the rectangular glass window with a clear appearance\" and \"the modern flat-screen TV with a dark matte surface\" on your left side?", "option": [], "answer": "3", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the rectangular glass window with a clear appearance\" and \"the modern flat-screen TV with a dark matte surface\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0680_01/01160.jpg", "scene0680_01/01190.jpg", "scene0680_01/01220.jpg", "scene0680_01/01250.jpg" ], "turn_id": 7, "scan_id": "scene0680_01", "origin_question": "When did you first discover \"the rectangular glass window with a clear appearance\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the rectangular glass window with a clear appearance\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "1mp3d_0072_region1/ecf39d39cf7745eebe103ce9c4b3cbe0_i1_2.jpg", "1mp3d_0072_region1/ecf39d39cf7745eebe103ce9c4b3cbe0_i2_2.jpg", "1mp3d_0072_region1/ab5c93de69a745bbb1ec09e3c4f2c0ae_i1_2.jpg", "1mp3d_0072_region1/ab5c93de69a745bbb1ec09e3c4f2c0ae_i2_2.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0072_region1", "origin_question": "Remember, how many light(s) have you seen so far? ", "option": [], "answer": "3", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many light(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "1mp3d_0072_region1/f1a43d69cfc04812b3099e08fc27d56c_i1_2.jpg", "1mp3d_0072_region1/f1a43d69cfc04812b3099e08fc27d56c_i2_2.jpg", "1mp3d_0072_region1/b6aff1ec4dd447c0ab21f2a6bf52e592_i1_2.jpg", "1mp3d_0072_region1/b6aff1ec4dd447c0ab21f2a6bf52e592_i2_2.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0072_region1", "origin_question": "In which turn were you farthest from \"the rectangular single bed with a white cotton bedsheet\"? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the rectangular single bed with a white cotton bedsheet\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "1mp3d_0072_region1/f1a43d69cfc04812b3099e08fc27d56c_i1_1.jpg", "1mp3d_0072_region1/f1a43d69cfc04812b3099e08fc27d56c_i2_1.jpg", "1mp3d_0072_region1/f1a43d69cfc04812b3099e08fc27d56c_i1_0.jpg", "1mp3d_0072_region1/f1a43d69cfc04812b3099e08fc27d56c_i2_0.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0072_region1", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "116.37", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "1mp3d_0072_region1/ab5c93de69a745bbb1ec09e3c4f2c0ae_i1_5.jpg", "1mp3d_0072_region1/ab5c93de69a745bbb1ec09e3c4f2c0ae_i2_5.jpg", "1mp3d_0072_region1/ab5c93de69a745bbb1ec09e3c4f2c0ae_i1_4.jpg", "1mp3d_0072_region1/ab5c93de69a745bbb1ec09e3c4f2c0ae_i2_4.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0072_region1", "origin_question": "At the end of which turn were both of \"the large rectangular striped mat in orange and grey\" and \"the cylindrical woven bin made of light-colored material\" on your right side?", "option": [], "answer": "4", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the large rectangular striped mat in orange and grey\" and \"the cylindrical woven bin made of light-colored material\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "1mp3d_0072_region1/ab5c93de69a745bbb1ec09e3c4f2c0ae_i1_3.jpg", "1mp3d_0072_region1/ab5c93de69a745bbb1ec09e3c4f2c0ae_i2_3.jpg", "1mp3d_0072_region1/3faf8726ba1a4adea46d7f63d8948bd3_i1_2.jpg", "1mp3d_0072_region1/3faf8726ba1a4adea46d7f63d8948bd3_i2_2.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0072_region1", "origin_question": "Which direction is the large rectangular window in a dark wooden frame in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the large rectangular window in a dark wooden frame in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "1mp3d_0072_region1/3faf8726ba1a4adea46d7f63d8948bd3_i1_3.jpg", "1mp3d_0072_region1/3faf8726ba1a4adea46d7f63d8948bd3_i2_3.jpg", "1mp3d_0072_region1/3faf8726ba1a4adea46d7f63d8948bd3_i1_4.jpg", "1mp3d_0072_region1/3faf8726ba1a4adea46d7f63d8948bd3_i2_4.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0072_region1", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the medium-sized rectangular white ottoman\" from your position?", "option": [], "answer": "179.16", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the medium-sized rectangular white ottoman\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "1mp3d_0072_region1/3faf8726ba1a4adea46d7f63d8948bd3_i1_5.jpg", "1mp3d_0072_region1/3faf8726ba1a4adea46d7f63d8948bd3_i2_5.jpg", "1mp3d_0072_region1/3faf8726ba1a4adea46d7f63d8948bd3_i1_0.jpg", "1mp3d_0072_region1/3faf8726ba1a4adea46d7f63d8948bd3_i2_0.jpg" ], "turn_id": 7, "scan_id": "1mp3d_0072_region1", "origin_question": "At the end of which turn were both of \"the tall, slender black lamp with a rectangular lampshade\" and \"the large rectangular striped mat in orange and grey\" on your right side?", "option": [], "answer": "3", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the tall, slender black lamp with a rectangular lampshade\" and \"the large rectangular striped mat in orange and grey\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "1mp3d_0072_region1/ebe1e9c4b5bb42cc99287a05f4ee5526_i1_0.jpg", "1mp3d_0072_region1/ebe1e9c4b5bb42cc99287a05f4ee5526_i2_0.jpg", "1mp3d_0072_region1/e57c3575f3674cc782b1155b1de1d110_i1_4.jpg", "1mp3d_0072_region1/e57c3575f3674cc782b1155b1de1d110_i2_4.jpg" ], "turn_id": 8, "scan_id": "1mp3d_0072_region1", "origin_question": "Which direction is the tall, slender black lamp with a rectangular lampshade in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the tall, slender black lamp with a rectangular lampshade in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "1mp3d_0072_region1/e57c3575f3674cc782b1155b1de1d110_i1_5.jpg", "1mp3d_0072_region1/e57c3575f3674cc782b1155b1de1d110_i2_5.jpg", "1mp3d_0072_region1/ebe1e9c4b5bb42cc99287a05f4ee5526_i1_2.jpg", "1mp3d_0072_region1/ebe1e9c4b5bb42cc99287a05f4ee5526_i2_2.jpg" ], "turn_id": 9, "scan_id": "1mp3d_0072_region1", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the flat-panel TV with a glossy black coating\" from your position?", "option": [], "answer": "138.08", "user_message": "For the 9 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the flat-panel TV with a glossy black coating\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0067_01/00000.jpg", "scene0067_01/00030.jpg", "scene0067_01/00060.jpg", "scene0067_01/00160.jpg" ], "turn_id": 1, "scan_id": "scene0067_01", "origin_question": "Remember, have you seen any grab bar(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any grab bar(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0067_01/00190.jpg", "scene0067_01/00220.jpg", "scene0067_01/00250.jpg", "scene0067_01/00280.jpg" ], "turn_id": 2, "scan_id": "scene0067_01", "origin_question": "In which turn did you see both \"the grab bar\" and \"the dispenser\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the grab bar\" and \"the dispenser\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0067_01/00320.jpg", "scene0067_01/00550.jpg", "scene0067_01/00580.jpg", "scene0067_01/00610.jpg" ], "turn_id": 3, "scan_id": "scene0067_01", "origin_question": "Please recall \"the door\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.0", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the door\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0067_01/00640.jpg", "scene0067_01/00670.jpg", "scene0067_01/00700.jpg", "scene0067_01/00730.jpg" ], "turn_id": 4, "scan_id": "scene0067_01", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the door\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the door\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0365_02/00000.jpg", "scene0365_02/00040.jpg", "scene0365_02/00080.jpg", "scene0365_02/00120.jpg" ], "turn_id": 1, "scan_id": "scene0365_02", "origin_question": "Remember, have you seen any bin(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any bin(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0365_02/00160.jpg", "scene0365_02/00200.jpg", "scene0365_02/00240.jpg", "scene0365_02/00280.jpg" ], "turn_id": 2, "scan_id": "scene0365_02", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the dark cylindrical plastic soap dispenser\" from your position?", "option": [], "answer": "104.44", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the dark cylindrical plastic soap dispenser\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0365_02/00320.jpg", "scene0365_02/00360.jpg", "scene0365_02/00400.jpg", "scene0365_02/00440.jpg" ], "turn_id": 3, "scan_id": "scene0365_02", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the medium-sized white rectangular wall-mounted sink\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the medium-sized white rectangular wall-mounted sink\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0365_02/00480.jpg", "scene0365_02/00520.jpg", "scene0365_02/00560.jpg", "scene0365_02/00600.jpg" ], "turn_id": 4, "scan_id": "scene0365_02", "origin_question": "At the end of which turn were both of \"the sleek vertical black dispenser\" and \"the dark cylindrical plastic soap dispenser\" on your right side?", "option": [], "answer": "4", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the sleek vertical black dispenser\" and \"the dark cylindrical plastic soap dispenser\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0365_02/00640.jpg", "scene0365_02/00680.jpg", "scene0365_02/00720.jpg", "scene0365_02/00760.jpg" ], "turn_id": 5, "scan_id": "scene0365_02", "origin_question": "At the end of which turn were both of \"the medium-sized white rectangular wall-mounted sink\" and \"the sleek vertical black dispenser\" on your left side?", "option": [], "answer": "2", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the medium-sized white rectangular wall-mounted sink\" and \"the sleek vertical black dispenser\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0365_02/00800.jpg", "scene0365_02/00840.jpg", "scene0365_02/00880.jpg", "scene0365_02/00920.jpg" ], "turn_id": 6, "scan_id": "scene0365_02", "origin_question": "Which two objects are on the same side of you now? \"the medium-sized white rectangular wall-mounted sink\", \"the dark cylindrical plastic soap dispenser\", \"the plant\". ", "option": [ "the medium-sized white rectangular wall-mounted sink, the dark cylindrical plastic soap dispenser", "the medium-sized white rectangular wall-mounted sink, the plant", "the dark cylindrical plastic soap dispenser, the plant" ], "answer": "the medium-sized white rectangular wall-mounted sink, the dark cylindrical plastic soap dispenser", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the medium-sized white rectangular wall-mounted sink\", \"the dark cylindrical plastic soap dispenser\", \"the plant\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the medium-sized white rectangular wall-mounted sink, the dark cylindrical plastic soap dispenser', 'the medium-sized white rectangular wall-mounted sink, the plant', 'the dark cylindrical plastic soap dispenser, the plant'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0365_02/00960.jpg", "scene0365_02/01000.jpg", "scene0365_02/01040.jpg", "scene0365_02/01080.jpg" ], "turn_id": 7, "scan_id": "scene0365_02", "origin_question": "Please recall \"the medium-sized white rectangular wall-mounted sink\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.54", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the medium-sized white rectangular wall-mounted sink\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0365_02/01120.jpg", "scene0365_02/01160.jpg", "scene0365_02/01200.jpg", "scene0365_02/01240.jpg" ], "turn_id": 8, "scan_id": "scene0365_02", "origin_question": "Please recall \"the medium-sized white rectangular wall-mounted sink\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.05", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the medium-sized white rectangular wall-mounted sink\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0004_region23/f68d84f143034de196b5b7f5b91fb28f_i1_4.jpg", "1mp3d_0004_region23/f68d84f143034de196b5b7f5b91fb28f_i2_4.jpg", "1mp3d_0004_region23/f68d84f143034de196b5b7f5b91fb28f_i2_3.jpg", "1mp3d_0004_region23/3cace39abf5f4a92a2abbc21d51b25bf_i1_2.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0004_region23", "origin_question": "Remember, have you seen any clothes(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any clothes(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 2 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "1mp3d_0004_region23/3cace39abf5f4a92a2abbc21d51b25bf_i2_2.jpg", "1mp3d_0004_region23/3cace39abf5f4a92a2abbc21d51b25bf_i1_3.jpg", "1mp3d_0004_region23/3cace39abf5f4a92a2abbc21d51b25bf_i2_3.jpg", "1mp3d_0004_region23/3cace39abf5f4a92a2abbc21d51b25bf_i1_4.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0004_region23", "origin_question": "When did you first discover \"the door\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the door\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 2 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0278_00/00000.jpg", "scene0278_00/00030.jpg", "scene0278_00/00060.jpg", "scene0278_00/00090.jpg" ], "turn_id": 1, "scan_id": "scene0278_00", "origin_question": "Remember, have you seen any counter(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any counter(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0278_00/00120.jpg", "scene0278_00/00150.jpg", "scene0278_00/00180.jpg", "scene0278_00/00210.jpg" ], "turn_id": 2, "scan_id": "scene0278_00", "origin_question": "When did you first discover \"the counter\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the counter\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0278_00/00240.jpg", "scene0278_00/00270.jpg", "scene0278_00/00300.jpg", "scene0278_00/00330.jpg" ], "turn_id": 3, "scan_id": "scene0278_00", "origin_question": "Please recall \"the large, modern gray copier made of metal and plastic\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.62", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the large, modern gray copier made of metal and plastic\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0278_00/00360.jpg", "scene0278_00/00390.jpg", "scene0278_00/00420.jpg", "scene0278_00/00450.jpg" ], "turn_id": 4, "scan_id": "scene0278_00", "origin_question": "Please recall \"the rectangular broom with a long black handle\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.02", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the rectangular broom with a long black handle\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0278_00/00480.jpg", "scene0278_00/00510.jpg", "scene0278_00/00540.jpg", "scene0278_00/00570.jpg" ], "turn_id": 5, "scan_id": "scene0278_00", "origin_question": "Remember, how many chair(s) have you seen so far? ", "option": [], "answer": "2", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many chair(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0278_00/00600.jpg", "scene0278_00/00630.jpg", "scene0278_00/00660.jpg", "scene0278_00/00690.jpg" ], "turn_id": 6, "scan_id": "scene0278_00", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the medium-sized white cabinet with four doors\" from your position?", "option": [], "answer": "100.07", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the medium-sized white cabinet with four doors\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0451_00/00000.jpg", "scene0451_00/00040.jpg", "scene0451_00/00080.jpg", "scene0451_00/00120.jpg", "scene0451_00/00160.jpg" ], "turn_id": 1, "scan_id": "scene0451_00", "origin_question": "Remember, have you seen any bench(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any bench(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0451_00/00200.jpg", "scene0451_00/00240.jpg", "scene0451_00/00280.jpg", "scene0451_00/00320.jpg", "scene0451_00/00360.jpg" ], "turn_id": 2, "scan_id": "scene0451_00", "origin_question": "In which turn were you closest to \"the clothes\"? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the clothes\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0451_00/00400.jpg", "scene0451_00/00440.jpg", "scene0451_00/00480.jpg", "scene0451_00/00520.jpg", "scene0451_00/00560.jpg" ], "turn_id": 3, "scan_id": "scene0451_00", "origin_question": "Which two objects are on the same side of you now? \"the white plastic bag made of translucent material\", \"the white rectangular metal radiator\", \"the cup\". ", "option": [ "the white plastic bag made of translucent material, the white rectangular metal radiator", "the white plastic bag made of translucent material, the cup", "the white rectangular metal radiator, the cup" ], "answer": "the white rectangular metal radiator, the cup", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the white plastic bag made of translucent material\", \"the white rectangular metal radiator\", \"the cup\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the white plastic bag made of translucent material, the white rectangular metal radiator', 'the white plastic bag made of translucent material, the cup', 'the white rectangular metal radiator, the cup'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "scene0451_00/00600.jpg", "scene0451_00/00640.jpg", "scene0451_00/00680.jpg", "scene0451_00/00720.jpg", "scene0451_00/00760.jpg" ], "turn_id": 4, "scan_id": "scene0451_00", "origin_question": "Among these three objects(\"the clothes\",\"the small rectangular wooden bench with a deep brown color\",\"the cup\" ), which one is the farthest from you?", "option": [ "the clothes", "the small rectangular wooden bench with a deep brown color", "the cup" ], "answer": "the cup", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the clothes\",\"the small rectangular wooden bench with a deep brown color\",\"the cup\" ), which one is the farthest from you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the clothes', 'the small rectangular wooden bench with a deep brown color', 'the cup'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0451_00/00800.jpg", "scene0451_00/00840.jpg", "scene0451_00/00880.jpg", "scene0451_00/00920.jpg", "scene0451_00/00960.jpg" ], "turn_id": 5, "scan_id": "scene0451_00", "origin_question": "What will be the first-time appearance order of the following categories: chair, clothes, table?", "option": [ "table, chair, clothes", "clothes, table, chair", "chair, clothes, table", "table, clothes, chair" ], "answer": "chair, clothes, table", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: chair, clothes, table? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['table, chair, clothes', 'clothes, table, chair', 'chair, clothes, table', 'table, clothes, chair'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0451_00/01000.jpg", "scene0451_00/01040.jpg", "scene0451_00/01080.jpg", "scene0451_00/01120.jpg", "scene0451_00/01180.jpg" ], "turn_id": 6, "scan_id": "scene0451_00", "origin_question": "How far is your current position from where you were at the end of the turn 4? (in meters)", "option": [], "answer": "2.53", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 4? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0451_00/01220.jpg", "scene0451_00/01260.jpg", "scene0451_00/01300.jpg", "scene0451_00/01340.jpg", "scene0451_00/01380.jpg" ], "turn_id": 7, "scan_id": "scene0451_00", "origin_question": "Which direction is the tall, cylindrical brown wooden stand in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the tall, cylindrical brown wooden stand in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0451_00/01420.jpg", "scene0451_00/01460.jpg", "scene0451_00/01500.jpg", "scene0451_00/01540.jpg", "scene0451_00/01580.jpg" ], "turn_id": 8, "scan_id": "scene0451_00", "origin_question": "Which two objects are on the same side of you now? \"the clothes\", \"the large rectangular white window with a lace curtain\", \"the tall, cylindrical brown wooden stand\". ", "option": [ "the clothes, the large rectangular white window with a lace curtain", "the clothes, the tall, cylindrical brown wooden stand", "the large rectangular white window with a lace curtain, the tall, cylindrical brown wooden stand" ], "answer": "the clothes, the large rectangular white window with a lace curtain", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the clothes\", \"the large rectangular white window with a lace curtain\", \"the tall, cylindrical brown wooden stand\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the clothes, the large rectangular white window with a lace curtain', 'the clothes, the tall, cylindrical brown wooden stand', 'the large rectangular white window with a lace curtain, the tall, cylindrical brown wooden stand'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0451_00/01620.jpg", "scene0451_00/01660.jpg", "scene0451_00/01700.jpg", "scene0451_00/01740.jpg", "scene0451_00/01780.jpg" ], "turn_id": 9, "scan_id": "scene0451_00", "origin_question": "Is the small rectangular wooden bench with a deep brown color to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the small rectangular wooden bench with a deep brown color to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0451_00/01820.jpg", "scene0451_00/01860.jpg", "scene0451_00/01900.jpg", "scene0451_00/01940.jpg", "scene0451_00/01980.jpg" ], "turn_id": 10, "scan_id": "scene0451_00", "origin_question": "Which direction is the tall, cylindrical brown wooden stand in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the tall, cylindrical brown wooden stand in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0249_00/00000.jpg", "scene0249_00/00040.jpg", "scene0249_00/00080.jpg", "scene0249_00/00120.jpg", "scene0249_00/00160.jpg" ], "turn_id": 1, "scan_id": "scene0249_00", "origin_question": "Remember, how many air conditioner(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many air conditioner(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0249_00/00200.jpg", "scene0249_00/00240.jpg", "scene0249_00/00280.jpg", "scene0249_00/00320.jpg", "scene0249_00/00360.jpg" ], "turn_id": 2, "scan_id": "scene0249_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0249_00/00400.jpg", "scene0249_00/00440.jpg", "scene0249_00/00480.jpg", "scene0249_00/00520.jpg", "scene0249_00/00560.jpg" ], "turn_id": 3, "scan_id": "scene0249_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0249_00/00600.jpg", "scene0249_00/00640.jpg", "scene0249_00/00680.jpg", "scene0249_00/00720.jpg", "scene0249_00/00760.jpg" ], "turn_id": 4, "scan_id": "scene0249_00", "origin_question": "Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "76.27", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0249_00/00800.jpg", "scene0249_00/00840.jpg", "scene0249_00/00880.jpg", "scene0249_00/00920.jpg", "scene0249_00/00960.jpg" ], "turn_id": 5, "scan_id": "scene0249_00", "origin_question": "What will be the first-time appearance order of the following categories: remote control, table, window?", "option": [ "remote control, window, table", "window, table, remote control", "table, window, remote control", "window, remote control, table" ], "answer": "table, window, remote control", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: remote control, table, window? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['remote control, window, table', 'window, table, remote control', 'table, window, remote control', 'window, remote control, table'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0249_00/01000.jpg", "scene0249_00/01040.jpg", "scene0249_00/01080.jpg", "scene0249_00/01120.jpg", "scene0249_00/01160.jpg" ], "turn_id": 6, "scan_id": "scene0249_00", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "scene0249_00/01200.jpg", "scene0249_00/01240.jpg", "scene0249_00/01280.jpg", "scene0249_00/01320.jpg", "scene0249_00/01360.jpg" ], "turn_id": 7, "scan_id": "scene0249_00", "origin_question": "Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Left", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0249_00/01400.jpg", "scene0249_00/01440.jpg", "scene0249_00/01480.jpg", "scene0249_00/01520.jpg", "scene0249_00/01560.jpg" ], "turn_id": 8, "scan_id": "scene0249_00", "origin_question": "What will be the first-time appearance order of the following categories: bin, remote control, window?", "option": [ "bin, window, remote control", "window, bin, remote control", "remote control, window, bin", "bin, remote control, window" ], "answer": "bin, window, remote control", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bin, remote control, window? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['bin, window, remote control', 'window, bin, remote control', 'remote control, window, bin', 'bin, remote control, window'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0249_00/01600.jpg", "scene0249_00/01640.jpg", "scene0249_00/01680.jpg", "scene0249_00/01720.jpg", "scene0249_00/01760.jpg" ], "turn_id": 9, "scan_id": "scene0249_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "scene0249_00/01810.jpg", "scene0249_00/01850.jpg", "scene0249_00/01890.jpg", "scene0249_00/01930.jpg", "scene0249_00/01970.jpg" ], "turn_id": 10, "scan_id": "scene0249_00", "origin_question": "Assuming the direction you are facing at the end of the turn 3 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Right", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 3 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0579_00/00000.jpg", "scene0579_00/00030.jpg", "scene0579_00/00060.jpg", "scene0579_00/00090.jpg" ], "turn_id": 1, "scan_id": "scene0579_00", "origin_question": "Remember, have you seen any sink(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any sink(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0579_00/00120.jpg", "scene0579_00/00150.jpg", "scene0579_00/00180.jpg", "scene0579_00/00210.jpg" ], "turn_id": 2, "scan_id": "scene0579_00", "origin_question": "Remember, how many mirror(s) have you seen so far? ", "option": [], "answer": "3", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many mirror(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0579_00/00240.jpg", "scene0579_00/00270.jpg", "scene0579_00/00300.jpg", "scene0579_00/00330.jpg" ], "turn_id": 3, "scan_id": "scene0579_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0579_00/00360.jpg", "scene0579_00/00390.jpg", "scene0579_00/00420.jpg", "scene0579_00/00450.jpg" ], "turn_id": 4, "scan_id": "scene0579_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0579_00/00480.jpg", "scene0579_00/00510.jpg", "scene0579_00/00540.jpg", "scene0579_00/00570.jpg" ], "turn_id": 5, "scan_id": "scene0579_00", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "153.64", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0579_00/00600.jpg", "scene0579_00/00630.jpg", "scene0579_00/00660.jpg", "scene0579_00/00690.jpg" ], "turn_id": 6, "scan_id": "scene0579_00", "origin_question": "Using your orientation at the end of turn 5 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 5 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0030_region7/0fc522280a134941bf49510757261f48_i1_3.jpg", "1mp3d_0030_region7/0fc522280a134941bf49510757261f48_i2_3.jpg", "1mp3d_0030_region7/63333423642f49caac4871521e93cc45_i2_4.jpg", "1mp3d_0030_region7/63333423642f49caac4871521e93cc45_i2_1.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0030_region7", "origin_question": "Remember, have you seen any cabinet(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any cabinet(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0030_region7/63333423642f49caac4871521e93cc45_i1_2.jpg", "1mp3d_0030_region7/63333423642f49caac4871521e93cc45_i2_2.jpg", "1mp3d_0030_region7/0fc522280a134941bf49510757261f48_i1_2.jpg", "1mp3d_0030_region7/0fc522280a134941bf49510757261f48_i2_2.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0030_region7", "origin_question": "Remember, have you seen any toaster(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any toaster(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "1mp3d_0030_region7/5965958b0a55454d88d45f21d97bc3fc_i1_2.jpg", "1mp3d_0030_region7/5965958b0a55454d88d45f21d97bc3fc_i2_2.jpg", "1mp3d_0030_region7/1527831f765e40c1b5210e52248427f7_i1_1.jpg", "1mp3d_0030_region7/1527831f765e40c1b5210e52248427f7_i2_1.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0030_region7", "origin_question": "When did you first discover \"the tall decorative plant with slender green leaves\"? (index of the turn)", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the tall decorative plant with slender green leaves\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "1mp3d_0030_region7/51328414aea04771890766f436b40021_i1_1.jpg", "1mp3d_0030_region7/51328414aea04771890766f436b40021_i2_1.jpg", "1mp3d_0030_region7/d9b5196ff58c494d8bebdeb3b6ccd061_i1_2.jpg", "1mp3d_0030_region7/d9b5196ff58c494d8bebdeb3b6ccd061_i2_2.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0030_region7", "origin_question": "In which turn were you closest to \"the tall decorative plant with slender green leaves\"? ", "option": [], "answer": "1", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the tall decorative plant with slender green leaves\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "1mp3d_0030_region7/1527831f765e40c1b5210e52248427f7_i1_0.jpg", "1mp3d_0030_region7/1527831f765e40c1b5210e52248427f7_i2_0.jpg", "1mp3d_0030_region7/1527831f765e40c1b5210e52248427f7_i1_2.jpg", "1mp3d_0030_region7/1527831f765e40c1b5210e52248427f7_i2_2.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0030_region7", "origin_question": "Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance forward or backward from that position?", "option": [ "Forward", "Backward" ], "answer": "Backward", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance forward or backward from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Forward', 'Backward'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0067_region18/fabd7c77b41e48eda0855efd12ecb1ec_i1_1.jpg", "1mp3d_0067_region18/fabd7c77b41e48eda0855efd12ecb1ec_i2_1.jpg", "1mp3d_0067_region18/ea60a28a151f4aa286b51f9f9819aefd_i1_1.jpg", "1mp3d_0067_region18/ea60a28a151f4aa286b51f9f9819aefd_i2_1.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0067_region18", "origin_question": "Remember, have you seen any towel(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any towel(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "1mp3d_0067_region18/4bd7d1a0c683471babf8c7abab4121a3_i1_2.jpg", "1mp3d_0067_region18/4bd7d1a0c683471babf8c7abab4121a3_i2_2.jpg", "1mp3d_0067_region18/4bd7d1a0c683471babf8c7abab4121a3_i2_4.jpg", "1mp3d_0067_region18/ea60a28a151f4aa286b51f9f9819aefd_i1_5.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0067_region18", "origin_question": "In which turn were you farthest from \"the medium-sized, natural brown woven basket\"? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the medium-sized, natural brown woven basket\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "1mp3d_0067_region18/ea60a28a151f4aa286b51f9f9819aefd_i2_5.jpg", "1mp3d_0067_region18/fabd7c77b41e48eda0855efd12ecb1ec_i1_0.jpg", "1mp3d_0067_region18/fabd7c77b41e48eda0855efd12ecb1ec_i2_0.jpg", "1mp3d_0067_region18/ea60a28a151f4aa286b51f9f9819aefd_i1_4.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0067_region18", "origin_question": "Please recall \"the window\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.16", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the window\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "1mp3d_0067_region18/ea60a28a151f4aa286b51f9f9819aefd_i2_4.jpg", "1mp3d_0067_region18/ea60a28a151f4aa286b51f9f9819aefd_i1_3.jpg", "1mp3d_0067_region18/ea60a28a151f4aa286b51f9f9819aefd_i2_3.jpg", "1mp3d_0067_region18/ea60a28a151f4aa286b51f9f9819aefd_i2_2.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0067_region18", "origin_question": "At the end of which turn were both of \"the white rectangular bathtub made of glossy material\" and \"the window frame\" on your left side?", "option": [], "answer": "3", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the white rectangular bathtub made of glossy material\" and \"the window frame\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "1mp3d_0067_region18/fabd7c77b41e48eda0855efd12ecb1ec_i2_3.jpg", "1mp3d_0067_region18/fabd7c77b41e48eda0855efd12ecb1ec_i1_4.jpg", "1mp3d_0067_region18/fabd7c77b41e48eda0855efd12ecb1ec_i2_4.jpg", "1mp3d_0067_region18/fabd7c77b41e48eda0855efd12ecb1ec_i1_2.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0067_region18", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0171_00/00000.jpg", "scene0171_00/00030.jpg", "scene0171_00/00060.jpg", "scene0171_00/00090.jpg" ], "turn_id": 1, "scan_id": "scene0171_00", "origin_question": "Remember, have you seen any chair(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any chair(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0171_00/00120.jpg", "scene0171_00/00150.jpg", "scene0171_00/00180.jpg", "scene0171_00/00210.jpg" ], "turn_id": 2, "scan_id": "scene0171_00", "origin_question": "In which turn were you farthest from \"the white rectangular conference table made of laminated panels\"? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the white rectangular conference table made of laminated panels\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0171_00/00240.jpg", "scene0171_00/00270.jpg", "scene0171_00/00300.jpg", "scene0171_00/00330.jpg" ], "turn_id": 3, "scan_id": "scene0171_00", "origin_question": "In which turn did you see both \"the white rectangular conference table made of laminated panels\" and \"the blackboard\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the white rectangular conference table made of laminated panels\" and \"the blackboard\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0171_00/00360.jpg", "scene0171_00/00390.jpg", "scene0171_00/00420.jpg", "scene0171_00/00450.jpg" ], "turn_id": 4, "scan_id": "scene0171_00", "origin_question": "In which turn were you closest to \"the cylindrical silver metal bin\"? ", "option": [], "answer": "1", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the cylindrical silver metal bin\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0171_00/00480.jpg", "scene0171_00/00510.jpg", "scene0171_00/00540.jpg", "scene0171_00/00570.jpg" ], "turn_id": 5, "scan_id": "scene0171_00", "origin_question": "By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the white rectangular conference table made of laminated panels\", Y: \"the cylindrical silver metal bin\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the white rectangular conference table made of laminated panels\", Y: \"the cylindrical silver metal bin\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0171_00/00600.jpg", "scene0171_00/00630.jpg", "scene0171_00/00660.jpg", "scene0171_00/00690.jpg" ], "turn_id": 6, "scan_id": "scene0171_00", "origin_question": "In which turn did you see both \"the blackboard\" and \"the door\" simultaneously? ", "option": [], "answer": "6", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the blackboard\" and \"the door\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0171_00/00720.jpg", "scene0171_00/00750.jpg", "scene0171_00/00780.jpg", "scene0171_00/00810.jpg" ], "turn_id": 7, "scan_id": "scene0171_00", "origin_question": "In which turn did you see both \"the blackboard\" and \"the door\" simultaneously? ", "option": [], "answer": "6", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the blackboard\" and \"the door\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0150_02/00000.jpg", "scene0150_02/00030.jpg", "scene0150_02/00060.jpg", "scene0150_02/00090.jpg" ], "turn_id": 1, "scan_id": "scene0150_02", "origin_question": "Remember, have you seen any table(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any table(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 1 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0078_region4/591d3c367a09478b8a863f17635749e4_i1_0.jpg", "1mp3d_0078_region4/591d3c367a09478b8a863f17635749e4_i2_0.jpg", "1mp3d_0078_region4/3a24d35a28604e7392f6f7d36e73922c_i1_3.jpg", "1mp3d_0078_region4/3a24d35a28604e7392f6f7d36e73922c_i2_3.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0078_region4", "origin_question": "Remember, have you seen any dustpan(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any dustpan(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "1mp3d_0078_region4/591d3c367a09478b8a863f17635749e4_i1_1.jpg", "1mp3d_0078_region4/591d3c367a09478b8a863f17635749e4_i2_1.jpg", "1mp3d_0078_region4/591d3c367a09478b8a863f17635749e4_i1_2.jpg", "1mp3d_0078_region4/591d3c367a09478b8a863f17635749e4_i2_2.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0078_region4", "origin_question": "At the end of which turn were both of \"the modern chair with a curved backrest and wooden legs\" and \"the small round light brown wooden table\" on your left side?", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the modern chair with a curved backrest and wooden legs\" and \"the small round light brown wooden table\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "1mp3d_0078_region4/3a24d35a28604e7392f6f7d36e73922c_i1_4.jpg", "1mp3d_0078_region4/3a24d35a28604e7392f6f7d36e73922c_i2_4.jpg", "1mp3d_0078_region4/3a24d35a28604e7392f6f7d36e73922c_i1_5.jpg", "1mp3d_0078_region4/3a24d35a28604e7392f6f7d36e73922c_i2_5.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0078_region4", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "1mp3d_0078_region4/3a24d35a28604e7392f6f7d36e73922c_i1_0.jpg", "1mp3d_0078_region4/3a24d35a28604e7392f6f7d36e73922c_i2_0.jpg", "1mp3d_0078_region4/009a121be20f46a6a41981c3974ab276_i1_0.jpg", "1mp3d_0078_region4/009a121be20f46a6a41981c3974ab276_i2_0.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0078_region4", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "138.02", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "1mp3d_0078_region4/009a121be20f46a6a41981c3974ab276_i1_5.jpg", "1mp3d_0078_region4/009a121be20f46a6a41981c3974ab276_i2_5.jpg", "1mp3d_0078_region4/009a121be20f46a6a41981c3974ab276_i1_4.jpg", "1mp3d_0078_region4/009a121be20f46a6a41981c3974ab276_i2_4.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0078_region4", "origin_question": "How far is your current position from where you were at the end of the turn 1? (in meters)", "option": [], "answer": "1.34", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 1? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "1mp3d_0078_region4/4b82c138b8b743479d535513aabfabf4_i1_4.jpg", "1mp3d_0078_region4/4b82c138b8b743479d535513aabfabf4_i2_4.jpg", "1mp3d_0078_region4/e538501778d5481f80d3557b1571f33f_i1_4.jpg", "1mp3d_0078_region4/e538501778d5481f80d3557b1571f33f_i2_4.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0078_region4", "origin_question": "How far is your current position from where you were at the end of the turn 4? (in meters)", "option": [], "answer": "2.29", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 4? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "1mp3d_0078_region4/e538501778d5481f80d3557b1571f33f_i1_3.jpg", "1mp3d_0078_region4/e538501778d5481f80d3557b1571f33f_i2_3.jpg", "1mp3d_0078_region4/4b82c138b8b743479d535513aabfabf4_i1_0.jpg", "1mp3d_0078_region4/4b82c138b8b743479d535513aabfabf4_i2_0.jpg" ], "turn_id": 7, "scan_id": "1mp3d_0078_region4", "origin_question": "At the end of which turn were both of \"the medium-sized, rectangular light brown wooden bed\" and \"the window\" on your right side?", "option": [], "answer": "5", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the medium-sized, rectangular light brown wooden bed\" and \"the window\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "1mp3d_0078_region4/c9030e71fc29447599a5c21d90261ca5_i1_4.jpg", "1mp3d_0078_region4/c9030e71fc29447599a5c21d90261ca5_i2_4.jpg", "1mp3d_0078_region4/c9030e71fc29447599a5c21d90261ca5_i1_3.jpg", "1mp3d_0078_region4/c9030e71fc29447599a5c21d90261ca5_i2_3.jpg" ], "turn_id": 8, "scan_id": "1mp3d_0078_region4", "origin_question": "What will be the first-time appearance order of the following categories: bed, picture, window?", "option": [ "picture, window, bed", "bed, window, picture", "picture, bed, window", "bed, picture, window" ], "answer": "bed, window, picture", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bed, picture, window? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['picture, window, bed', 'bed, window, picture', 'picture, bed, window', 'bed, picture, window'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0412_01/00000.jpg", "scene0412_01/00040.jpg", "scene0412_01/00080.jpg", "scene0412_01/00120.jpg" ], "turn_id": 1, "scan_id": "scene0412_01", "origin_question": "Remember, have you seen any bin(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any bin(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0412_01/00160.jpg", "scene0412_01/00200.jpg", "scene0412_01/00240.jpg", "scene0412_01/00280.jpg" ], "turn_id": 2, "scan_id": "scene0412_01", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0412_01/00320.jpg", "scene0412_01/00360.jpg", "scene0412_01/00400.jpg", "scene0412_01/00440.jpg" ], "turn_id": 3, "scan_id": "scene0412_01", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the rectangular stainless steel sink with multiple faucets\"; \"the medium-sized, silver metallic ladder\"; \"the rectangular shiny stainless steel counter\" ", "option": [ "the rectangular stainless steel sink with multiple faucets", "the medium-sized, silver metallic ladder", "the rectangular shiny stainless steel counter" ], "answer": "the medium-sized, silver metallic ladder", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the rectangular stainless steel sink with multiple faucets\"; \"the medium-sized, silver metallic ladder\"; \"the rectangular shiny stainless steel counter\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular stainless steel sink with multiple faucets', 'the medium-sized, silver metallic ladder', 'the rectangular shiny stainless steel counter'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0412_01/00480.jpg", "scene0412_01/00520.jpg", "scene0412_01/00560.jpg", "scene0412_01/00600.jpg" ], "turn_id": 4, "scan_id": "scene0412_01", "origin_question": "", "option": [], "answer": "", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0412_01/00640.jpg", "scene0412_01/00680.jpg", "scene0412_01/00720.jpg", "scene0412_01/00760.jpg" ], "turn_id": 5, "scan_id": "scene0412_01", "origin_question": "Which direction is the rectangular shiny stainless steel counter in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the rectangular shiny stainless steel counter in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0412_01/00800.jpg", "scene0412_01/00840.jpg", "scene0412_01/00880.jpg", "scene0412_01/00920.jpg" ], "turn_id": 6, "scan_id": "scene0412_01", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "98.23", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0412_01/00960.jpg", "scene0412_01/01000.jpg", "scene0412_01/01040.jpg", "scene0412_01/01080.jpg" ], "turn_id": 7, "scan_id": "scene0412_01", "origin_question": "How far is your current position from where you were at the end of the turn 5? (in meters)", "option": [], "answer": "2.91", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 5? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0412_01/01120.jpg", "scene0412_01/01160.jpg", "scene0412_01/01200.jpg", "scene0412_01/01240.jpg" ], "turn_id": 8, "scan_id": "scene0412_01", "origin_question": "Using your orientation at the end of turn 7 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 7 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0651_02/00000.jpg", "scene0651_02/00030.jpg", "scene0651_02/00060.jpg", "scene0651_02/00090.jpg" ], "turn_id": 1, "scan_id": "scene0651_02", "origin_question": "Remember, have you seen any stove(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any stove(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0651_02/00120.jpg", "scene0651_02/00150.jpg", "scene0651_02/00180.jpg", "scene0651_02/00210.jpg" ], "turn_id": 2, "scan_id": "scene0651_02", "origin_question": "In which turn did you see both \"the white double-door refrigerator in good condition\" and \"the rectangular stainless steel single basin sink\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the white double-door refrigerator in good condition\" and \"the rectangular stainless steel single basin sink\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0651_02/00240.jpg", "scene0651_02/00270.jpg", "scene0651_02/00300.jpg", "scene0651_02/00330.jpg" ], "turn_id": 3, "scan_id": "scene0651_02", "origin_question": "Which direction is the stylish white rectangular range hood in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the stylish white rectangular range hood in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "scene0651_02/00360.jpg", "scene0651_02/00390.jpg", "scene0651_02/00420.jpg", "scene0651_02/00450.jpg" ], "turn_id": 4, "scan_id": "scene0651_02", "origin_question": "Among these three objects(\"the rectangular white electric stove with six burners\",\"the white double-door refrigerator in good condition\",\"the rectangular stainless steel single basin sink\" ), which one is the farthest from you?", "option": [ "the rectangular white electric stove with six burners", "the white double-door refrigerator in good condition", "the rectangular stainless steel single basin sink" ], "answer": "the rectangular stainless steel single basin sink", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects(\"the rectangular white electric stove with six burners\",\"the white double-door refrigerator in good condition\",\"the rectangular stainless steel single basin sink\" ), which one is the farthest from you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular white electric stove with six burners', 'the white double-door refrigerator in good condition', 'the rectangular stainless steel single basin sink'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0651_02/00480.jpg", "scene0651_02/00510.jpg", "scene0651_02/00540.jpg", "scene0651_02/00570.jpg" ], "turn_id": 5, "scan_id": "scene0651_02", "origin_question": "By the end of the 4 turn, were you getting closer to or farther away from these two objects: X: \"the stylish white rectangular range hood\", Y: \"the rectangular stainless steel single basin sink\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "farther away from both X and Y", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 4 turn, were you getting closer to or farther away from these two objects: X: \"the stylish white rectangular range hood\", Y: \"the rectangular stainless steel single basin sink\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0651_02/00600.jpg", "scene0651_02/00630.jpg", "scene0651_02/00660.jpg", "scene0651_02/00690.jpg" ], "turn_id": 6, "scan_id": "scene0651_02", "origin_question": "In which turn did you see both \"the rectangular white electric stove with six burners\" and \"the rectangular stainless steel single basin sink\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the rectangular white electric stove with six burners\" and \"the rectangular stainless steel single basin sink\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0651_02/00720.jpg", "scene0651_02/00750.jpg", "scene0651_02/00780.jpg", "scene0651_02/00810.jpg" ], "turn_id": 7, "scan_id": "scene0651_02", "origin_question": "Which two objects are on the same side of you now? \"the white double-door refrigerator in good condition\", \"the couch\", \"the bin\". ", "option": [ "the white double-door refrigerator in good condition, the couch", "the white double-door refrigerator in good condition, the bin", "the couch, the bin" ], "answer": "the white double-door refrigerator in good condition, the couch", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the white double-door refrigerator in good condition\", \"the couch\", \"the bin\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the white double-door refrigerator in good condition, the couch', 'the white double-door refrigerator in good condition, the bin', 'the couch, the bin'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0047_region12/36a2419151544f299c99cc0f247b2bb2_i2_3.jpg", "1mp3d_0047_region12/36a2419151544f299c99cc0f247b2bb2_i1_2.jpg", "1mp3d_0047_region12/36a2419151544f299c99cc0f247b2bb2_i2_2.jpg", "1mp3d_0047_region12/0e7a7b8f99a84a719128d3d0982e9b59_i2_3.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0047_region12", "origin_question": "", "option": [], "answer": "", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 2 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "1mp3d_0047_region12/0e7a7b8f99a84a719128d3d0982e9b59_i1_2.jpg", "1mp3d_0047_region12/0e7a7b8f99a84a719128d3d0982e9b59_i2_2.jpg", "1mp3d_0047_region12/36a2419151544f299c99cc0f247b2bb2_i1_1.jpg", "1mp3d_0047_region12/36a2419151544f299c99cc0f247b2bb2_i2_1.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0047_region12", "origin_question": "When did you first discover \"the modern, silver stainless steel rail\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the modern, silver stainless steel rail\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 2 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0510_00/00000.jpg", "scene0510_00/00040.jpg", "scene0510_00/00080.jpg", "scene0510_00/00120.jpg", "scene0510_00/00160.jpg" ], "turn_id": 1, "scan_id": "scene0510_00", "origin_question": "Remember, how many stopcock(s) have you seen so far? ", "option": [], "answer": "2", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many stopcock(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0510_00/00200.jpg", "scene0510_00/00240.jpg", "scene0510_00/00280.jpg", "scene0510_00/00320.jpg", "scene0510_00/00360.jpg" ], "turn_id": 2, "scan_id": "scene0510_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0510_00/00400.jpg", "scene0510_00/00440.jpg", "scene0510_00/00480.jpg", "scene0510_00/00520.jpg", "scene0510_00/00560.jpg" ], "turn_id": 3, "scan_id": "scene0510_00", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the toilet paper\", Y: \"the shelf\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the toilet paper\", Y: \"the shelf\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0510_00/00600.jpg", "scene0510_00/00640.jpg", "scene0510_00/00680.jpg", "scene0510_00/00720.jpg", "scene0510_00/00760.jpg" ], "turn_id": 4, "scan_id": "scene0510_00", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the black soft fabric jacket with a zipper\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the black soft fabric jacket with a zipper\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0510_00/00800.jpg", "scene0510_00/00840.jpg", "scene0510_00/00880.jpg", "scene0510_00/00920.jpg", "scene0510_00/00960.jpg" ], "turn_id": 5, "scan_id": "scene0510_00", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0510_00/01000.jpg", "scene0510_00/01040.jpg", "scene0510_00/01080.jpg", "scene0510_00/01120.jpg", "scene0510_00/01160.jpg" ], "turn_id": 6, "scan_id": "scene0510_00", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "1.14", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0510_00/01200.jpg", "scene0510_00/01240.jpg", "scene0510_00/01280.jpg", "scene0510_00/01320.jpg", "scene0510_00/01360.jpg" ], "turn_id": 7, "scan_id": "scene0510_00", "origin_question": "Which two objects are on the same side of you now? \"the counter\", \"the medium gray plastic bucket with a red handle\", \"the rectangular gray metal dispenser with a white label\". ", "option": [ "the counter, the medium gray plastic bucket with a red handle", "the counter, the rectangular gray metal dispenser with a white label", "the medium gray plastic bucket with a red handle, the rectangular gray metal dispenser with a white label" ], "answer": "the medium gray plastic bucket with a red handle, the rectangular gray metal dispenser with a white label", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the counter\", \"the medium gray plastic bucket with a red handle\", \"the rectangular gray metal dispenser with a white label\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the counter, the medium gray plastic bucket with a red handle', 'the counter, the rectangular gray metal dispenser with a white label', 'the medium gray plastic bucket with a red handle, the rectangular gray metal dispenser with a white label'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0510_00/01400.jpg", "scene0510_00/01440.jpg", "scene0510_00/01480.jpg", "scene0510_00/01520.jpg", "scene0510_00/01560.jpg" ], "turn_id": 8, "scan_id": "scene0510_00", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0015_00/00000.jpg", "scene0015_00/00040.jpg", "scene0015_00/00080.jpg", "scene0015_00/00120.jpg", "scene0015_00/00160.jpg" ], "turn_id": 1, "scan_id": "scene0015_00", "origin_question": "Remember, how many ladder(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many ladder(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0015_00/00200.jpg", "scene0015_00/00240.jpg", "scene0015_00/00280.jpg", "scene0015_00/00320.jpg", "scene0015_00/00360.jpg" ], "turn_id": 2, "scan_id": "scene0015_00", "origin_question": "In which turn did you see both \"the rectangular light brown wooden windowsill\" and \"the transparent plastic bottle with a silver cap\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the rectangular light brown wooden windowsill\" and \"the transparent plastic bottle with a silver cap\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0015_00/00400.jpg", "scene0015_00/00440.jpg", "scene0015_00/00480.jpg", "scene0015_00/00520.jpg", "scene0015_00/00560.jpg" ], "turn_id": 3, "scan_id": "scene0015_00", "origin_question": "Is the transparent plastic bottle with a silver cap to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the transparent plastic bottle with a silver cap to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0015_00/00600.jpg", "scene0015_00/00640.jpg", "scene0015_00/00680.jpg", "scene0015_00/00720.jpg", "scene0015_00/00760.jpg" ], "turn_id": 4, "scan_id": "scene0015_00", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0015_00/00800.jpg", "scene0015_00/00840.jpg", "scene0015_00/00880.jpg", "scene0015_00/00920.jpg", "scene0015_00/00960.jpg" ], "turn_id": 5, "scan_id": "scene0015_00", "origin_question": "Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the backpack\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the backpack\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0015_00/01000.jpg", "scene0015_00/01040.jpg", "scene0015_00/01080.jpg", "scene0015_00/01120.jpg", "scene0015_00/01160.jpg" ], "turn_id": 6, "scan_id": "scene0015_00", "origin_question": "At the end of which turn were both of \"the rectangular light brown wooden windowsill\" and \"the transparent plastic bottle with a silver cap\" on your right side?", "option": [], "answer": "3", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the rectangular light brown wooden windowsill\" and \"the transparent plastic bottle with a silver cap\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0015_00/01200.jpg", "scene0015_00/01240.jpg", "scene0015_00/01280.jpg", "scene0015_00/01320.jpg", "scene0015_00/01360.jpg" ], "turn_id": 7, "scan_id": "scene0015_00", "origin_question": "Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "123.37", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "scene0015_00/01400.jpg", "scene0015_00/01440.jpg", "scene0015_00/01480.jpg", "scene0015_00/01520.jpg", "scene0015_00/01560.jpg" ], "turn_id": 8, "scan_id": "scene0015_00", "origin_question": "Assuming the direction you are facing at the end of the turn 4 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Right", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 4 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0015_00/01600.jpg", "scene0015_00/01650.jpg", "scene0015_00/01690.jpg", "scene0015_00/01730.jpg", "scene0015_00/01770.jpg" ], "turn_id": 9, "scan_id": "scene0015_00", "origin_question": "By the end of the 7 turn, were you getting closer to or farther away from these two objects: X: \"the transparent plastic bottle with a silver cap\", Y: \"the backpack\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 7 turn, were you getting closer to or farther away from these two objects: X: \"the transparent plastic bottle with a silver cap\", Y: \"the backpack\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0040_region57/e27a1e250624457f8cbff1781dc8695b_i1_5.jpg", "1mp3d_0040_region57/e27a1e250624457f8cbff1781dc8695b_i2_5.jpg", "1mp3d_0040_region57/53b98ee4dfc1481ab5750f44ebdee37c_i1_4.jpg", "1mp3d_0040_region57/53b98ee4dfc1481ab5750f44ebdee37c_i2_4.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0040_region57", "origin_question": "Remember, have you seen any plant(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any plant(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "1mp3d_0040_region57/53b98ee4dfc1481ab5750f44ebdee37c_i1_5.jpg", "1mp3d_0040_region57/53b98ee4dfc1481ab5750f44ebdee37c_i2_5.jpg", "1mp3d_0040_region57/e27a1e250624457f8cbff1781dc8695b_i1_0.jpg", "1mp3d_0040_region57/e27a1e250624457f8cbff1781dc8695b_i2_0.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0040_region57", "origin_question": "Please recall \"the green potted plant with broad heart-shaped leaves\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.07", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the green potted plant with broad heart-shaped leaves\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "1mp3d_0040_region57/e27a1e250624457f8cbff1781dc8695b_i1_1.jpg", "1mp3d_0040_region57/e27a1e250624457f8cbff1781dc8695b_i2_1.jpg", "1mp3d_0040_region57/3a6ace8feba54c03b6803ffcee7e4a57_i1_1.jpg", "1mp3d_0040_region57/3a6ace8feba54c03b6803ffcee7e4a57_i2_1.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0040_region57", "origin_question": "In which turn did you see both \"the green potted plant with broad heart-shaped leaves\" and \"the rectangular, polished brown wooden vanity\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the green potted plant with broad heart-shaped leaves\" and \"the rectangular, polished brown wooden vanity\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "1mp3d_0040_region57/3a6ace8feba54c03b6803ffcee7e4a57_i1_2.jpg", "1mp3d_0040_region57/3a6ace8feba54c03b6803ffcee7e4a57_i2_2.jpg", "1mp3d_0040_region57/3a6ace8feba54c03b6803ffcee7e4a57_i1_3.jpg", "1mp3d_0040_region57/3a6ace8feba54c03b6803ffcee7e4a57_i2_3.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0040_region57", "origin_question": "Which direction is the rectangular, polished brown wooden vanity in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the rectangular, polished brown wooden vanity in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "1mp3d_0040_region57/cd657625baf8410f88fe3a3bd8e5a771_i1_4.jpg", "1mp3d_0040_region57/cd657625baf8410f88fe3a3bd8e5a771_i2_4.jpg", "1mp3d_0040_region57/059728445f604624a12b40fc093044cf_i1_3.jpg", "1mp3d_0040_region57/059728445f604624a12b40fc093044cf_i2_3.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0040_region57", "origin_question": "Which direction is the small, brown and gold patterned wood chair in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the small, brown and gold patterned wood chair in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "1mp3d_0040_region57/cd657625baf8410f88fe3a3bd8e5a771_i1_3.jpg", "1mp3d_0040_region57/cd657625baf8410f88fe3a3bd8e5a771_i2_3.jpg", "1mp3d_0040_region57/cd657625baf8410f88fe3a3bd8e5a771_i1_2.jpg", "1mp3d_0040_region57/cd657625baf8410f88fe3a3bd8e5a771_i2_2.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0040_region57", "origin_question": "What will be the first-time appearance order of the following categories: chair, light, plant?", "option": [ "plant, chair, light", "light, chair, plant", "light, plant, chair", "plant, light, chair" ], "answer": "plant, chair, light", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: chair, light, plant? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['plant, chair, light', 'light, chair, plant', 'light, plant, chair', 'plant, light, chair'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "1mp3d_0040_region57/059728445f604624a12b40fc093044cf_i1_1.jpg", "1mp3d_0040_region57/059728445f604624a12b40fc093044cf_i2_1.jpg", "1mp3d_0040_region57/059728445f604624a12b40fc093044cf_i1_2.jpg", "1mp3d_0040_region57/059728445f604624a12b40fc093044cf_i2_2.jpg" ], "turn_id": 7, "scan_id": "1mp3d_0040_region57", "origin_question": "By the end of the 3 turn, were you getting closer to or farther away from these two objects: X: \"the green potted plant with broad heart-shaped leaves\", Y: \"the white soft absorbent cotton towel\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "farther away from both X and Y", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 3 turn, were you getting closer to or farther away from these two objects: X: \"the green potted plant with broad heart-shaped leaves\", Y: \"the white soft absorbent cotton towel\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "1mp3d_0040_region57/4d0c96681f204c87987c717aa563cdd6_i1_5.jpg", "1mp3d_0040_region57/4d0c96681f204c87987c717aa563cdd6_i2_5.jpg", "1mp3d_0040_region57/4d0c96681f204c87987c717aa563cdd6_i1_4.jpg", "1mp3d_0040_region57/4d0c96681f204c87987c717aa563cdd6_i2_4.jpg" ], "turn_id": 8, "scan_id": "1mp3d_0040_region57", "origin_question": "What will be the first-time appearance order of the following categories: chair, light, towel?", "option": [ "chair, light, towel", "chair, towel, light", "light, chair, towel", "towel, light, chair" ], "answer": "chair, light, towel", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: chair, light, towel? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['chair, light, towel', 'chair, towel, light', 'light, chair, towel', 'towel, light, chair'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "1mp3d_0040_region57/c0dcda75061042248d0367f1a1f70cda_i1_4.jpg", "1mp3d_0040_region57/c0dcda75061042248d0367f1a1f70cda_i2_4.jpg", "1mp3d_0040_region57/c0dcda75061042248d0367f1a1f70cda_i1_5.jpg", "1mp3d_0040_region57/c0dcda75061042248d0367f1a1f70cda_i2_5.jpg" ], "turn_id": 9, "scan_id": "1mp3d_0040_region57", "origin_question": "Using your orientation at the end of turn 8 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "64.51", "user_message": "For the 9 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 8 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0448_01/00000.jpg", "scene0448_01/00040.jpg", "scene0448_01/00080.jpg", "scene0448_01/00120.jpg", "scene0448_01/00160.jpg" ], "turn_id": 1, "scan_id": "scene0448_01", "origin_question": "Remember, have you seen any flowerpot(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any flowerpot(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0448_01/00200.jpg", "scene0448_01/00240.jpg", "scene0448_01/00280.jpg", "scene0448_01/00320.jpg", "scene0448_01/00360.jpg" ], "turn_id": 2, "scan_id": "scene0448_01", "origin_question": "Which direction is the window in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the window in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0448_01/00400.jpg", "scene0448_01/00440.jpg", "scene0448_01/00480.jpg", "scene0448_01/00520.jpg", "scene0448_01/00560.jpg" ], "turn_id": 3, "scan_id": "scene0448_01", "origin_question": "What will be the first-time appearance order of the following categories: stopcock, toilet, toilet paper?", "option": [ "toilet, toilet paper, stopcock", "stopcock, toilet paper, toilet", "toilet paper, stopcock, toilet", "stopcock, toilet, toilet paper" ], "answer": "toilet, toilet paper, stopcock", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: stopcock, toilet, toilet paper? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['toilet, toilet paper, stopcock', 'stopcock, toilet paper, toilet', 'toilet paper, stopcock, toilet', 'stopcock, toilet, toilet paper'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0448_01/00600.jpg", "scene0448_01/00640.jpg", "scene0448_01/00680.jpg", "scene0448_01/00720.jpg", "scene0448_01/00760.jpg" ], "turn_id": 4, "scan_id": "scene0448_01", "origin_question": "Remember, have you seen any excercise equipment(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any excercise equipment(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0448_01/00800.jpg", "scene0448_01/00840.jpg", "scene0448_01/00880.jpg", "scene0448_01/00920.jpg", "scene0448_01/00980.jpg" ], "turn_id": 5, "scan_id": "scene0448_01", "origin_question": "Please recall \"the sturdy silver metal rod\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.96", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the sturdy silver metal rod\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0448_01/01020.jpg", "scene0448_01/01060.jpg", "scene0448_01/01100.jpg", "scene0448_01/01140.jpg", "scene0448_01/01180.jpg" ], "turn_id": 6, "scan_id": "scene0448_01", "origin_question": "Which two objects are on the same side of you now? \"the table\", \"the sturdy silver metal rod\", \"the white oval-shaped porcelain toilet\". ", "option": [ "the table, the sturdy silver metal rod", "the table, the white oval-shaped porcelain toilet", "the sturdy silver metal rod, the white oval-shaped porcelain toilet" ], "answer": "the table, the sturdy silver metal rod", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the table\", \"the sturdy silver metal rod\", \"the white oval-shaped porcelain toilet\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the table, the sturdy silver metal rod', 'the table, the white oval-shaped porcelain toilet', 'the sturdy silver metal rod, the white oval-shaped porcelain toilet'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0448_01/01220.jpg", "scene0448_01/01260.jpg", "scene0448_01/01300.jpg", "scene0448_01/01340.jpg", "scene0448_01/01380.jpg" ], "turn_id": 7, "scan_id": "scene0448_01", "origin_question": "In which turn did you see both \"the white oval-shaped porcelain toilet\" and \"the sturdy silver metal rod\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the white oval-shaped porcelain toilet\" and \"the sturdy silver metal rod\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0448_01/01420.jpg", "scene0448_01/01460.jpg", "scene0448_01/01500.jpg", "scene0448_01/01540.jpg", "scene0448_01/01580.jpg" ], "turn_id": 8, "scan_id": "scene0448_01", "origin_question": "Please recall \"the round off-white speckled sink\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.22", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the round off-white speckled sink\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "scene0448_01/01620.jpg", "scene0448_01/01660.jpg", "scene0448_01/01700.jpg", "scene0448_01/01740.jpg", "scene0448_01/01780.jpg" ], "turn_id": 9, "scan_id": "scene0448_01", "origin_question": "Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance forward or backward from that position?", "option": [ "Forward", "Backward" ], "answer": "Forward", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance forward or backward from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Forward', 'Backward'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0448_01/01820.jpg", "scene0448_01/01860.jpg", "scene0448_01/01900.jpg", "scene0448_01/01940.jpg", "scene0448_01/02000.jpg" ], "turn_id": 10, "scan_id": "scene0448_01", "origin_question": "Using your orientation at the end of turn 7 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "158.77", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 7 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0641_00/00000.jpg", "scene0641_00/00040.jpg", "scene0641_00/00080.jpg", "scene0641_00/00120.jpg" ], "turn_id": 1, "scan_id": "scene0641_00", "origin_question": "Remember, have you seen any table(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any table(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0641_00/00160.jpg", "scene0641_00/00200.jpg", "scene0641_00/00240.jpg", "scene0641_00/00280.jpg" ], "turn_id": 2, "scan_id": "scene0641_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0641_00/00320.jpg", "scene0641_00/00360.jpg", "scene0641_00/00400.jpg", "scene0641_00/00440.jpg" ], "turn_id": 3, "scan_id": "scene0641_00", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0641_00/00480.jpg", "scene0641_00/00520.jpg", "scene0641_00/00560.jpg", "scene0641_00/00600.jpg" ], "turn_id": 4, "scan_id": "scene0641_00", "origin_question": "Remember, how many table(s) have you seen so far? ", "option": [], "answer": "3", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many table(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0641_00/00640.jpg", "scene0641_00/00680.jpg", "scene0641_00/00720.jpg", "scene0641_00/00760.jpg" ], "turn_id": 5, "scan_id": "scene0641_00", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0641_00/00800.jpg", "scene0641_00/00840.jpg", "scene0641_00/00880.jpg", "scene0641_00/00920.jpg" ], "turn_id": 6, "scan_id": "scene0641_00", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "5.7", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0641_00/00960.jpg", "scene0641_00/01000.jpg", "scene0641_00/01040.jpg", "scene0641_00/01080.jpg" ], "turn_id": 7, "scan_id": "scene0641_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0641_00/01120.jpg", "scene0641_00/01160.jpg", "scene0641_00/01200.jpg", "scene0641_00/01240.jpg" ], "turn_id": 8, "scan_id": "scene0641_00", "origin_question": "How far is your current position from where you were at the end of the turn 6? (in meters)", "option": [], "answer": "3.37", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 6? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0356_02/00000.jpg", "scene0356_02/00030.jpg", "scene0356_02/00060.jpg", "scene0356_02/00090.jpg" ], "turn_id": 1, "scan_id": "scene0356_02", "origin_question": "Remember, have you seen any hat(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any hat(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0356_02/00120.jpg", "scene0356_02/00150.jpg", "scene0356_02/00180.jpg", "scene0356_02/00210.jpg" ], "turn_id": 2, "scan_id": "scene0356_02", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0356_02/00240.jpg", "scene0356_02/00270.jpg", "scene0356_02/00300.jpg", "scene0356_02/00330.jpg" ], "turn_id": 3, "scan_id": "scene0356_02", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0356_02/00360.jpg", "scene0356_02/00390.jpg", "scene0356_02/00420.jpg", "scene0356_02/00450.jpg" ], "turn_id": 4, "scan_id": "scene0356_02", "origin_question": "Which two objects are on the same side of you now? \"the large, light-colored wooden shelf with two shelves\", \"the radiator\", \"the white corded telephone with a simple design\". ", "option": [ "the large, light-colored wooden shelf with two shelves, the radiator", "the large, light-colored wooden shelf with two shelves, the white corded telephone with a simple design", "the radiator, the white corded telephone with a simple design" ], "answer": "the radiator, the white corded telephone with a simple design", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the large, light-colored wooden shelf with two shelves\", \"the radiator\", \"the white corded telephone with a simple design\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the large, light-colored wooden shelf with two shelves, the radiator', 'the large, light-colored wooden shelf with two shelves, the white corded telephone with a simple design', 'the radiator, the white corded telephone with a simple design'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0356_02/00480.jpg", "scene0356_02/00510.jpg", "scene0356_02/00540.jpg", "scene0356_02/00570.jpg" ], "turn_id": 5, "scan_id": "scene0356_02", "origin_question": "When did you first discover \"the bed\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the bed\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0356_02/00600.jpg", "scene0356_02/00630.jpg", "scene0356_02/00660.jpg", "scene0356_02/00690.jpg" ], "turn_id": 6, "scan_id": "scene0356_02", "origin_question": "Is the bed to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the bed to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0356_02/00720.jpg", "scene0356_02/00750.jpg", "scene0356_02/00780.jpg", "scene0356_02/00810.jpg" ], "turn_id": 7, "scan_id": "scene0356_02", "origin_question": "What will be the first-time appearance order of the following categories: door, telephone, window?", "option": [ "door, telephone, window", "window, door, telephone", "window, telephone, door", "telephone, window, door" ], "answer": "window, door, telephone", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: door, telephone, window? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['door, telephone, window', 'window, door, telephone', 'window, telephone, door', 'telephone, window, door'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0643_00/00000.jpg", "scene0643_00/00040.jpg", "scene0643_00/00080.jpg", "scene0643_00/00120.jpg", "scene0643_00/00160.jpg" ], "turn_id": 1, "scan_id": "scene0643_00", "origin_question": "Remember, have you seen any monitor(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any monitor(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0643_00/00200.jpg", "scene0643_00/00240.jpg", "scene0643_00/00280.jpg", "scene0643_00/00320.jpg", "scene0643_00/00360.jpg" ], "turn_id": 2, "scan_id": "scene0643_00", "origin_question": "Remember, have you seen any desk(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any desk(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0643_00/00400.jpg", "scene0643_00/00440.jpg", "scene0643_00/00480.jpg", "scene0643_00/00520.jpg", "scene0643_00/00560.jpg" ], "turn_id": 3, "scan_id": "scene0643_00", "origin_question": "Which direction is the red vertical calendar with white text and numbers in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the red vertical calendar with white text and numbers in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0643_00/00600.jpg", "scene0643_00/00640.jpg", "scene0643_00/00680.jpg", "scene0643_00/00720.jpg", "scene0643_00/00760.jpg" ], "turn_id": 4, "scan_id": "scene0643_00", "origin_question": "Please recall \"the red vertical calendar with white text and numbers\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.36", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the red vertical calendar with white text and numbers\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0643_00/00800.jpg", "scene0643_00/00840.jpg", "scene0643_00/00880.jpg", "scene0643_00/00920.jpg", "scene0643_00/00960.jpg" ], "turn_id": 5, "scan_id": "scene0643_00", "origin_question": "Which two objects are on the same side of you now? \"the red vertical calendar with white text and numbers\", \"the medium-sized black book with a red cover\", \"the standard-sized, brown wooden door\". ", "option": [ "the red vertical calendar with white text and numbers, the medium-sized black book with a red cover", "the red vertical calendar with white text and numbers, the standard-sized, brown wooden door", "the medium-sized black book with a red cover, the standard-sized, brown wooden door" ], "answer": "the red vertical calendar with white text and numbers, the medium-sized black book with a red cover", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the red vertical calendar with white text and numbers\", \"the medium-sized black book with a red cover\", \"the standard-sized, brown wooden door\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the red vertical calendar with white text and numbers, the medium-sized black book with a red cover', 'the red vertical calendar with white text and numbers, the standard-sized, brown wooden door', 'the medium-sized black book with a red cover, the standard-sized, brown wooden door'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0643_00/01000.jpg", "scene0643_00/01040.jpg", "scene0643_00/01080.jpg", "scene0643_00/01120.jpg", "scene0643_00/01160.jpg" ], "turn_id": 6, "scan_id": "scene0643_00", "origin_question": "Which direction is the pair of white headphones with a black headband in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the pair of white headphones with a black headband in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0643_00/01200.jpg", "scene0643_00/01240.jpg", "scene0643_00/01280.jpg", "scene0643_00/01320.jpg", "scene0643_00/01360.jpg" ], "turn_id": 7, "scan_id": "scene0643_00", "origin_question": "Please recall \"the medium-sized black book with a red cover\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.25", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the medium-sized black book with a red cover\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0643_00/01400.jpg", "scene0643_00/01440.jpg", "scene0643_00/01480.jpg", "scene0643_00/01520.jpg", "scene0643_00/01560.jpg" ], "turn_id": 8, "scan_id": "scene0643_00", "origin_question": "Using your orientation at the end of turn 7 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 7 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0643_00/01600.jpg", "scene0643_00/01640.jpg", "scene0643_00/01680.jpg", "scene0643_00/01720.jpg", "scene0643_00/01760.jpg" ], "turn_id": 9, "scan_id": "scene0643_00", "origin_question": "Please recall \"the white oval-shaped wireless mouse made of plastic\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.47", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the white oval-shaped wireless mouse made of plastic\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0146_02/00000.jpg", "scene0146_02/00040.jpg", "scene0146_02/00080.jpg", "scene0146_02/00120.jpg" ], "turn_id": 1, "scan_id": "scene0146_02", "origin_question": "Remember, have you seen any pool(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any pool(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0146_02/00160.jpg", "scene0146_02/00200.jpg", "scene0146_02/00240.jpg", "scene0146_02/00280.jpg" ], "turn_id": 2, "scan_id": "scene0146_02", "origin_question": "In which turn did you see both \"the bin\" and \"the white, modern ceramic toilet with a round bowl\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the bin\" and \"the white, modern ceramic toilet with a round bowl\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0146_02/00320.jpg", "scene0146_02/00360.jpg", "scene0146_02/00400.jpg", "scene0146_02/00440.jpg" ], "turn_id": 3, "scan_id": "scene0146_02", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0146_02/00480.jpg", "scene0146_02/00520.jpg", "scene0146_02/00560.jpg", "scene0146_02/00600.jpg" ], "turn_id": 4, "scan_id": "scene0146_02", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the black backpack with a gray stripe and sturdy material\"; \"the large rectangular mirror with a silver frame\"; \"the white rectangular porcelain sink with smooth corners\" ", "option": [ "the black backpack with a gray stripe and sturdy material", "the large rectangular mirror with a silver frame", "the white rectangular porcelain sink with smooth corners" ], "answer": "the black backpack with a gray stripe and sturdy material", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the black backpack with a gray stripe and sturdy material\"; \"the large rectangular mirror with a silver frame\"; \"the white rectangular porcelain sink with smooth corners\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the black backpack with a gray stripe and sturdy material', 'the large rectangular mirror with a silver frame', 'the white rectangular porcelain sink with smooth corners'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0146_02/00640.jpg", "scene0146_02/00680.jpg", "scene0146_02/00720.jpg", "scene0146_02/00760.jpg" ], "turn_id": 5, "scan_id": "scene0146_02", "origin_question": "In which turn did you see both \"the white, modern ceramic toilet with a round bowl\" and \"the white rectangular porcelain sink with smooth corners\" simultaneously? ", "option": [], "answer": "3", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the white, modern ceramic toilet with a round bowl\" and \"the white rectangular porcelain sink with smooth corners\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0146_02/00800.jpg", "scene0146_02/00840.jpg", "scene0146_02/00880.jpg", "scene0146_02/00960.jpg" ], "turn_id": 6, "scan_id": "scene0146_02", "origin_question": "When did you first discover \"the rectangular black soap dispenser with a smooth exterior\"? (index of the turn)", "option": [], "answer": "3", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the rectangular black soap dispenser with a smooth exterior\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0146_02/01000.jpg", "scene0146_02/01040.jpg", "scene0146_02/01080.jpg", "scene0146_02/01120.jpg" ], "turn_id": 7, "scan_id": "scene0146_02", "origin_question": "How far is your current position from where you were at the end of the turn 4? (in meters)", "option": [], "answer": "1.27", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 4? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0146_02/01160.jpg", "scene0146_02/01200.jpg", "scene0146_02/01240.jpg", "scene0146_02/01280.jpg" ], "turn_id": 8, "scan_id": "scene0146_02", "origin_question": "Which direction is the black backpack with a gray stripe and sturdy material in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the black backpack with a gray stripe and sturdy material in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0670_00/00000.jpg", "scene0670_00/00040.jpg", "scene0670_00/00080.jpg", "scene0670_00/00120.jpg", "scene0670_00/00160.jpg" ], "turn_id": 1, "scan_id": "scene0670_00", "origin_question": "Remember, have you seen any bowl(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any bowl(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0670_00/00200.jpg", "scene0670_00/00240.jpg", "scene0670_00/00280.jpg", "scene0670_00/00320.jpg", "scene0670_00/00360.jpg" ], "turn_id": 2, "scan_id": "scene0670_00", "origin_question": "At the end of which turn were both of \"the medium-sized, round silver stainless steel pot\" and \"the rectangular wooden board with a light brown hue\" on your left side?", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the medium-sized, round silver stainless steel pot\" and \"the rectangular wooden board with a light brown hue\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0670_00/00400.jpg", "scene0670_00/00440.jpg", "scene0670_00/00480.jpg", "scene0670_00/00520.jpg", "scene0670_00/00560.jpg" ], "turn_id": 3, "scan_id": "scene0670_00", "origin_question": "Is the large rectangular stainless steel sink in silver-gray to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the large rectangular stainless steel sink in silver-gray to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0670_00/00600.jpg", "scene0670_00/00640.jpg", "scene0670_00/00680.jpg", "scene0670_00/00720.jpg", "scene0670_00/00760.jpg" ], "turn_id": 4, "scan_id": "scene0670_00", "origin_question": "Which two objects are on the same side of you now? \"the large shiny metal bowl\", \"the counter\", \"the small rectangular brown cardboard box\". ", "option": [ "the large shiny metal bowl, the counter", "the large shiny metal bowl, the small rectangular brown cardboard box", "the counter, the small rectangular brown cardboard box" ], "answer": "the large shiny metal bowl, the counter", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the large shiny metal bowl\", \"the counter\", \"the small rectangular brown cardboard box\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the large shiny metal bowl, the counter', 'the large shiny metal bowl, the small rectangular brown cardboard box', 'the counter, the small rectangular brown cardboard box'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "scene0670_00/00800.jpg", "scene0670_00/00840.jpg", "scene0670_00/00880.jpg", "scene0670_00/00920.jpg", "scene0670_00/00960.jpg" ], "turn_id": 5, "scan_id": "scene0670_00", "origin_question": "Among these three objects(\"the large rectangular stainless steel sink in silver-gray\",\"the rectangular stainless steel range hood with a stylish design\",\"the modern black and silver stove with gas burners\" ), which one is the closest to you?", "option": [ "the large rectangular stainless steel sink in silver-gray", "the rectangular stainless steel range hood with a stylish design", "the modern black and silver stove with gas burners" ], "answer": "the large rectangular stainless steel sink in silver-gray", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the large rectangular stainless steel sink in silver-gray\",\"the rectangular stainless steel range hood with a stylish design\",\"the modern black and silver stove with gas burners\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the large rectangular stainless steel sink in silver-gray', 'the rectangular stainless steel range hood with a stylish design', 'the modern black and silver stove with gas burners'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0670_00/01000.jpg", "scene0670_00/01040.jpg", "scene0670_00/01080.jpg", "scene0670_00/01120.jpg", "scene0670_00/01160.jpg" ], "turn_id": 6, "scan_id": "scene0670_00", "origin_question": "Using your orientation at the end of turn 5 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "62.34", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 5 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0670_00/01200.jpg", "scene0670_00/01240.jpg", "scene0670_00/01280.jpg", "scene0670_00/01320.jpg", "scene0670_00/01360.jpg" ], "turn_id": 7, "scan_id": "scene0670_00", "origin_question": "What will be the first-time appearance order of the following categories: calendar, dishwasher, picture?", "option": [ "dishwasher, calendar, picture", "dishwasher, picture, calendar", "picture, dishwasher, calendar", "calendar, dishwasher, picture" ], "answer": "dishwasher, picture, calendar", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: calendar, dishwasher, picture? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['dishwasher, calendar, picture', 'dishwasher, picture, calendar', 'picture, dishwasher, calendar', 'calendar, dishwasher, picture'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0670_00/01400.jpg", "scene0670_00/01440.jpg", "scene0670_00/01480.jpg", "scene0670_00/01520.jpg", "scene0670_00/01560.jpg" ], "turn_id": 8, "scan_id": "scene0670_00", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the counter\"; \"the round wooden clock with a white face\"; \"the sleek rectangular stainless steel dishwasher\" ", "option": [ "the counter", "the round wooden clock with a white face", "the sleek rectangular stainless steel dishwasher" ], "answer": "the round wooden clock with a white face", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the counter\"; \"the round wooden clock with a white face\"; \"the sleek rectangular stainless steel dishwasher\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the counter', 'the round wooden clock with a white face', 'the sleek rectangular stainless steel dishwasher'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0670_00/01600.jpg", "scene0670_00/01640.jpg", "scene0670_00/01680.jpg", "scene0670_00/01720.jpg", "scene0670_00/01760.jpg" ], "turn_id": 9, "scan_id": "scene0670_00", "origin_question": "Please recall \"the round wooden clock with a white face\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.78", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the round wooden clock with a white face\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0670_00/01800.jpg", "scene0670_00/01840.jpg", "scene0670_00/01880.jpg", "scene0670_00/01920.jpg", "scene0670_00/01960.jpg" ], "turn_id": 10, "scan_id": "scene0670_00", "origin_question": "By the end of the 5 turn, were you getting closer to or farther away from these two objects: X: \"the large rectangular window with a white frame\", Y: \"the vibrant rectangular paper calendar with a monkey design\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to Y and farther away from X", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 5 turn, were you getting closer to or farther away from these two objects: X: \"the large rectangular window with a white frame\", Y: \"the vibrant rectangular paper calendar with a monkey design\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0057_region12/a426fc0b709e42e1a8ee8ede0de335f7_i1_2.jpg", "1mp3d_0057_region12/a426fc0b709e42e1a8ee8ede0de335f7_i2_2.jpg", "1mp3d_0057_region12/a426fc0b709e42e1a8ee8ede0de335f7_i1_3.jpg", "1mp3d_0057_region12/a426fc0b709e42e1a8ee8ede0de335f7_i2_3.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0057_region12", "origin_question": "Remember, have you seen any coat hanger(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any coat hanger(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "1mp3d_0057_region12/4955496ce75d4347bab8ed9dfa24780b_i1_4.jpg", "1mp3d_0057_region12/4955496ce75d4347bab8ed9dfa24780b_i2_4.jpg", "1mp3d_0057_region12/4955496ce75d4347bab8ed9dfa24780b_i1_0.jpg", "1mp3d_0057_region12/4955496ce75d4347bab8ed9dfa24780b_i2_0.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0057_region12", "origin_question": "When did you first discover \"the white rectangular ceramic bathtub in good condition\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the white rectangular ceramic bathtub in good condition\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "1mp3d_0057_region12/4955496ce75d4347bab8ed9dfa24780b_i2_1.jpg", "1mp3d_0057_region12/01cb3e6fb48a4a178834426cf4b340c4_i1_1.jpg", "1mp3d_0057_region12/01cb3e6fb48a4a178834426cf4b340c4_i2_1.jpg", "1mp3d_0057_region12/01cb3e6fb48a4a178834426cf4b340c4_i1_2.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0057_region12", "origin_question": "When did you first discover \"the white rectangular ceramic bathtub in good condition\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the white rectangular ceramic bathtub in good condition\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "1mp3d_0057_region12/01cb3e6fb48a4a178834426cf4b340c4_i2_2.jpg", "1mp3d_0057_region12/01cb3e6fb48a4a178834426cf4b340c4_i2_5.jpg", "1mp3d_0057_region12/01cb3e6fb48a4a178834426cf4b340c4_i1_3.jpg", "1mp3d_0057_region12/01cb3e6fb48a4a178834426cf4b340c4_i2_3.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0057_region12", "origin_question": "At the end of which turn were both of \"the white rectangular ceramic bathtub in good condition\" and \"the soft, white towel neatly folded\" on your left side?", "option": [], "answer": "2", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the white rectangular ceramic bathtub in good condition\" and \"the soft, white towel neatly folded\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "1mp3d_0057_region12/01cb3e6fb48a4a178834426cf4b340c4_i1_0.jpg", "1mp3d_0057_region12/01cb3e6fb48a4a178834426cf4b340c4_i2_0.jpg", "1mp3d_0057_region12/01cb3e6fb48a4a178834426cf4b340c4_i1_4.jpg", "1mp3d_0057_region12/01cb3e6fb48a4a178834426cf4b340c4_i2_4.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0057_region12", "origin_question": "Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Left", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0088_03/00000.jpg", "scene0088_03/00040.jpg", "scene0088_03/00080.jpg", "scene0088_03/00120.jpg", "scene0088_03/00160.jpg" ], "turn_id": 1, "scan_id": "scene0088_03", "origin_question": "Remember, how many chair(s) have you seen so far? ", "option": [], "answer": "19", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many chair(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0088_03/00200.jpg", "scene0088_03/00240.jpg", "scene0088_03/00280.jpg", "scene0088_03/00320.jpg", "scene0088_03/00360.jpg" ], "turn_id": 2, "scan_id": "scene0088_03", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0088_03/00400.jpg", "scene0088_03/00440.jpg", "scene0088_03/00480.jpg", "scene0088_03/00520.jpg", "scene0088_03/00560.jpg" ], "turn_id": 3, "scan_id": "scene0088_03", "origin_question": "Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "53.98", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0088_03/00600.jpg", "scene0088_03/00640.jpg", "scene0088_03/00680.jpg", "scene0088_03/00720.jpg", "scene0088_03/00760.jpg" ], "turn_id": 4, "scan_id": "scene0088_03", "origin_question": "When did you first discover \"the rectangular white plastic board with a metal frame\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the rectangular white plastic board with a metal frame\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0088_03/00800.jpg", "scene0088_03/00840.jpg", "scene0088_03/00880.jpg", "scene0088_03/00920.jpg", "scene0088_03/00960.jpg" ], "turn_id": 5, "scan_id": "scene0088_03", "origin_question": "Which direction is the light brown rectangular wooden table in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the light brown rectangular wooden table in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0088_03/01000.jpg", "scene0088_03/01040.jpg", "scene0088_03/01080.jpg", "scene0088_03/01120.jpg", "scene0088_03/01160.jpg" ], "turn_id": 6, "scan_id": "scene0088_03", "origin_question": "How far is your current position from where you were at the end of the turn 3? (in meters)", "option": [], "answer": "1.32", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 3? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0088_03/01200.jpg", "scene0088_03/01240.jpg", "scene0088_03/01280.jpg", "scene0088_03/01320.jpg", "scene0088_03/01360.jpg" ], "turn_id": 7, "scan_id": "scene0088_03", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "1.91", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0088_03/01400.jpg", "scene0088_03/01440.jpg", "scene0088_03/01480.jpg", "scene0088_03/01520.jpg", "scene0088_03/01560.jpg" ], "turn_id": 8, "scan_id": "scene0088_03", "origin_question": "Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0040_region61/f83d0032ce684933a4d3e8784a9fd0bf_i1_2.jpg", "1mp3d_0040_region61/f83d0032ce684933a4d3e8784a9fd0bf_i2_2.jpg", "1mp3d_0040_region61/f83d0032ce684933a4d3e8784a9fd0bf_i1_1.jpg", "1mp3d_0040_region61/f83d0032ce684933a4d3e8784a9fd0bf_i2_1.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0040_region61", "origin_question": "Remember, have you seen any door(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any door(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 2 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "1mp3d_0040_region61/f83d0032ce684933a4d3e8784a9fd0bf_i1_3.jpg", "1mp3d_0040_region61/f83d0032ce684933a4d3e8784a9fd0bf_i2_3.jpg", "1mp3d_0040_region61/28efb6e3427f427d989566cd1aa10b8d_i1_3.jpg", "1mp3d_0040_region61/28efb6e3427f427d989566cd1aa10b8d_i2_3.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0040_region61", "origin_question": "In which turn did you see both \"the light\" and \"the sink\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the light\" and \"the sink\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 2 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0044_region6/81259373b5f748a18863513b926d2c0a_i1_4.jpg", "1mp3d_0044_region6/81259373b5f748a18863513b926d2c0a_i2_4.jpg", "1mp3d_0044_region6/b97390fe8c0a4d6d8adb523def8c77f5_i1_2.jpg", "1mp3d_0044_region6/b97390fe8c0a4d6d8adb523def8c77f5_i2_2.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0044_region6", "origin_question": "Remember, have you seen any toaster(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any toaster(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0044_region6/81259373b5f748a18863513b926d2c0a_i1_2.jpg", "1mp3d_0044_region6/81259373b5f748a18863513b926d2c0a_i2_2.jpg", "1mp3d_0044_region6/81259373b5f748a18863513b926d2c0a_i1_3.jpg", "1mp3d_0044_region6/81259373b5f748a18863513b926d2c0a_i2_3.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0044_region6", "origin_question": "Remember, have you seen any door(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any door(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "1mp3d_0044_region6/81259373b5f748a18863513b926d2c0a_i1_1.jpg", "1mp3d_0044_region6/81259373b5f748a18863513b926d2c0a_i2_1.jpg", "1mp3d_0044_region6/7372e123d5d94035a6d77c5f77bf0ff6_i1_2.jpg", "1mp3d_0044_region6/7372e123d5d94035a6d77c5f77bf0ff6_i2_2.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0044_region6", "origin_question": "Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Left", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "1mp3d_0044_region6/1dbb0638f04f486fa9226a986b243873_i1_0.jpg", "1mp3d_0044_region6/1dbb0638f04f486fa9226a986b243873_i2_0.jpg", "1mp3d_0044_region6/7372e123d5d94035a6d77c5f77bf0ff6_i1_1.jpg", "1mp3d_0044_region6/7372e123d5d94035a6d77c5f77bf0ff6_i2_1.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0044_region6", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "1.47", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0631_01/00000.jpg", "scene0631_01/00040.jpg", "scene0631_01/00080.jpg", "scene0631_01/00120.jpg" ], "turn_id": 1, "scan_id": "scene0631_01", "origin_question": "Remember, have you seen any lamp(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any lamp(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0631_01/00160.jpg", "scene0631_01/00200.jpg", "scene0631_01/00240.jpg", "scene0631_01/00280.jpg" ], "turn_id": 2, "scan_id": "scene0631_01", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the tall black floor lamp with a white lampshade\", Y: \"the clear glass bottle with a silver lid\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the tall black floor lamp with a white lampshade\", Y: \"the clear glass bottle with a silver lid\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0631_01/00320.jpg", "scene0631_01/00360.jpg", "scene0631_01/00400.jpg", "scene0631_01/00440.jpg" ], "turn_id": 3, "scan_id": "scene0631_01", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the rectangular soft beige wooden dresser with multiple drawers\"; \"the white wooden four-poster bed\"; \"the light beige rectangular wooden stand\" ", "option": [ "the rectangular soft beige wooden dresser with multiple drawers", "the white wooden four-poster bed", "the light beige rectangular wooden stand" ], "answer": "the rectangular soft beige wooden dresser with multiple drawers", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the rectangular soft beige wooden dresser with multiple drawers\"; \"the white wooden four-poster bed\"; \"the light beige rectangular wooden stand\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular soft beige wooden dresser with multiple drawers', 'the white wooden four-poster bed', 'the light beige rectangular wooden stand'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0631_01/00480.jpg", "scene0631_01/00520.jpg", "scene0631_01/00560.jpg", "scene0631_01/00600.jpg" ], "turn_id": 4, "scan_id": "scene0631_01", "origin_question": "Which two objects are on the same side of you now? \"the white wooden four-poster bed\", \"the clear glass bottle with a silver lid\", \"the shelf\". ", "option": [ "the white wooden four-poster bed, the clear glass bottle with a silver lid", "the white wooden four-poster bed, the shelf", "the clear glass bottle with a silver lid, the shelf" ], "answer": "the white wooden four-poster bed, the clear glass bottle with a silver lid", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the white wooden four-poster bed\", \"the clear glass bottle with a silver lid\", \"the shelf\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the white wooden four-poster bed, the clear glass bottle with a silver lid', 'the white wooden four-poster bed, the shelf', 'the clear glass bottle with a silver lid, the shelf'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0631_01/00640.jpg", "scene0631_01/00680.jpg", "scene0631_01/00720.jpg", "scene0631_01/00760.jpg" ], "turn_id": 5, "scan_id": "scene0631_01", "origin_question": "What will be the first-time appearance order of the following categories: bed, clothes, door?", "option": [ "bed, door, clothes", "door, bed, clothes", "clothes, bed, door", "clothes, door, bed" ], "answer": "bed, door, clothes", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bed, clothes, door? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['bed, door, clothes', 'door, bed, clothes', 'clothes, bed, door', 'clothes, door, bed'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0631_01/00800.jpg", "scene0631_01/00850.jpg", "scene0631_01/00890.jpg", "scene0631_01/00930.jpg" ], "turn_id": 6, "scan_id": "scene0631_01", "origin_question": "When did you first discover \"the light beige rectangular wooden stand\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the light beige rectangular wooden stand\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0631_01/00970.jpg", "scene0631_01/01010.jpg", "scene0631_01/01050.jpg", "scene0631_01/01090.jpg" ], "turn_id": 7, "scan_id": "scene0631_01", "origin_question": "When did you first discover \"the black flat-screen TV made of plastic\"? (index of the turn)", "option": [], "answer": "3", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the black flat-screen TV made of plastic\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0044_region27/2e96ca977e674555a39c6b369e6bd3df_i1_5.jpg", "1mp3d_0044_region27/2e96ca977e674555a39c6b369e6bd3df_i2_5.jpg", "1mp3d_0044_region27/f938e19896a6413da590ef3e9e055b95_i1_5.jpg", "1mp3d_0044_region27/f938e19896a6413da590ef3e9e055b95_i2_5.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0044_region27", "origin_question": "Remember, have you seen any curtain(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any curtain(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0044_region27/2e96ca977e674555a39c6b369e6bd3df_i1_3.jpg", "1mp3d_0044_region27/2e96ca977e674555a39c6b369e6bd3df_i2_3.jpg", "1mp3d_0044_region27/2e96ca977e674555a39c6b369e6bd3df_i1_0.jpg", "1mp3d_0044_region27/2e96ca977e674555a39c6b369e6bd3df_i2_0.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0044_region27", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the rectangular box with brown and beige patterns\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the rectangular box with brown and beige patterns\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "1mp3d_0044_region27/2e96ca977e674555a39c6b369e6bd3df_i1_1.jpg", "1mp3d_0044_region27/2e96ca977e674555a39c6b369e6bd3df_i2_1.jpg", "1mp3d_0044_region27/2e96ca977e674555a39c6b369e6bd3df_i1_2.jpg", "1mp3d_0044_region27/2e96ca977e674555a39c6b369e6bd3df_i2_2.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0044_region27", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the glossy black flat-panel TV with rounded corners\" from your position?", "option": [], "answer": "40.63", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the glossy black flat-panel TV with rounded corners\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "1mp3d_0044_region27/f938e19896a6413da590ef3e9e055b95_i1_0.jpg", "1mp3d_0044_region27/f938e19896a6413da590ef3e9e055b95_i2_0.jpg", "1mp3d_0044_region27/f90f7e3981ac4510906239ff3422c9e8_i1_5.jpg", "1mp3d_0044_region27/f90f7e3981ac4510906239ff3422c9e8_i2_5.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0044_region27", "origin_question": "By the end of the 3 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular box with brown and beige patterns\", Y: \"the small, lush green plant with vibrant leaves\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to Y and farther away from X", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 3 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular box with brown and beige patterns\", Y: \"the small, lush green plant with vibrant leaves\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0044_region27/b00097db86d14f95b0d7179849fa183d_i1_5.jpg", "1mp3d_0044_region27/b00097db86d14f95b0d7179849fa183d_i2_5.jpg", "1mp3d_0044_region27/b00097db86d14f95b0d7179849fa183d_i1_4.jpg", "1mp3d_0044_region27/b00097db86d14f95b0d7179849fa183d_i2_4.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0044_region27", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the white plastic rod of standard size\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the white plastic rod of standard size\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "1mp3d_0044_region27/b00097db86d14f95b0d7179849fa183d_i1_3.jpg", "1mp3d_0044_region27/b00097db86d14f95b0d7179849fa183d_i2_3.jpg", "1mp3d_0044_region27/f90f7e3981ac4510906239ff3422c9e8_i1_1.jpg", "1mp3d_0044_region27/f90f7e3981ac4510906239ff3422c9e8_i2_1.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0044_region27", "origin_question": "Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "149.22", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0645_02/00000.jpg", "scene0645_02/00040.jpg", "scene0645_02/00080.jpg", "scene0645_02/00120.jpg", "scene0645_02/00160.jpg" ], "turn_id": 1, "scan_id": "scene0645_02", "origin_question": "Remember, have you seen any lamp(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any lamp(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0645_02/00200.jpg", "scene0645_02/00240.jpg", "scene0645_02/00280.jpg", "scene0645_02/00320.jpg", "scene0645_02/00360.jpg" ], "turn_id": 2, "scan_id": "scene0645_02", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0645_02/00400.jpg", "scene0645_02/00440.jpg", "scene0645_02/00480.jpg", "scene0645_02/00520.jpg", "scene0645_02/00560.jpg" ], "turn_id": 3, "scan_id": "scene0645_02", "origin_question": "What will be the first-time appearance order of the following categories: door, stand, tv?", "option": [ "stand, door, tv", "door, tv, stand", "door, stand, tv", "tv, stand, door" ], "answer": "stand, door, tv", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: door, stand, tv? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['stand, door, tv', 'door, tv, stand', 'door, stand, tv', 'tv, stand, door'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0645_02/00600.jpg", "scene0645_02/00640.jpg", "scene0645_02/00680.jpg", "scene0645_02/00720.jpg", "scene0645_02/00760.jpg" ], "turn_id": 4, "scan_id": "scene0645_02", "origin_question": "In which turn did you see both \"the classic brass lamp with a wide white lampshade\" and \"the traditional dark brown wooden cabinet with raised doors\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the classic brass lamp with a wide white lampshade\" and \"the traditional dark brown wooden cabinet with raised doors\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0645_02/00800.jpg", "scene0645_02/00840.jpg", "scene0645_02/00880.jpg", "scene0645_02/00920.jpg", "scene0645_02/00960.jpg" ], "turn_id": 5, "scan_id": "scene0645_02", "origin_question": "Which direction is the triangular white plastic bag with red markings in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the triangular white plastic bag with red markings in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0645_02/01000.jpg", "scene0645_02/01040.jpg", "scene0645_02/01080.jpg", "scene0645_02/01120.jpg", "scene0645_02/01160.jpg" ], "turn_id": 6, "scan_id": "scene0645_02", "origin_question": "Which two objects are on the same side of you now? \"the triangular white plastic bag with red markings\", \"the white corded telephone with a round base\", \"the wooden chair with a soft cushioned seat\". ", "option": [ "the triangular white plastic bag with red markings, the white corded telephone with a round base", "the triangular white plastic bag with red markings, the wooden chair with a soft cushioned seat", "the white corded telephone with a round base, the wooden chair with a soft cushioned seat" ], "answer": "the white corded telephone with a round base, the wooden chair with a soft cushioned seat", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the triangular white plastic bag with red markings\", \"the white corded telephone with a round base\", \"the wooden chair with a soft cushioned seat\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the triangular white plastic bag with red markings, the white corded telephone with a round base', 'the triangular white plastic bag with red markings, the wooden chair with a soft cushioned seat', 'the white corded telephone with a round base, the wooden chair with a soft cushioned seat'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0645_02/01200.jpg", "scene0645_02/01240.jpg", "scene0645_02/01280.jpg", "scene0645_02/01320.jpg", "scene0645_02/01360.jpg" ], "turn_id": 7, "scan_id": "scene0645_02", "origin_question": "Which direction is the large flat-screen TV with a black screen in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the large flat-screen TV with a black screen in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0645_02/01400.jpg", "scene0645_02/01440.jpg", "scene0645_02/01480.jpg", "scene0645_02/01520.jpg", "scene0645_02/01560.jpg" ], "turn_id": 8, "scan_id": "scene0645_02", "origin_question": "Is the rectangular dark wooden stand with closed doors to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the rectangular dark wooden stand with closed doors to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "scene0645_02/01600.jpg", "scene0645_02/01640.jpg", "scene0645_02/01680.jpg", "scene0645_02/01720.jpg", "scene0645_02/01760.jpg" ], "turn_id": 9, "scan_id": "scene0645_02", "origin_question": "Assuming the direction you are facing at the end of the turn 5 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Left", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 5 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0645_02/01800.jpg", "scene0645_02/01840.jpg", "scene0645_02/01880.jpg", "scene0645_02/01920.jpg", "scene0645_02/01960.jpg" ], "turn_id": 10, "scan_id": "scene0645_02", "origin_question": "At the end of which turn were both of \"the clothes\" and \"the white corded telephone with a round base\" on your right side?", "option": [], "answer": "9", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the clothes\" and \"the white corded telephone with a round base\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0655_00/00000.jpg", "scene0655_00/00030.jpg", "scene0655_00/00060.jpg", "scene0655_00/00090.jpg" ], "turn_id": 1, "scan_id": "scene0655_00", "origin_question": "Remember, have you seen any table(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any table(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0655_00/00160.jpg", "scene0655_00/00190.jpg", "scene0655_00/00220.jpg", "scene0655_00/00250.jpg" ], "turn_id": 2, "scan_id": "scene0655_00", "origin_question": "At the end of which turn were both of \"the rectangular white table with a black metal frame\" and \"the large rectangular blackboard\" on your left side?", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the rectangular white table with a black metal frame\" and \"the large rectangular blackboard\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0655_00/00280.jpg", "scene0655_00/00310.jpg", "scene0655_00/00340.jpg", "scene0655_00/00370.jpg" ], "turn_id": 3, "scan_id": "scene0655_00", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "59.07", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0655_00/00400.jpg", "scene0655_00/00430.jpg", "scene0655_00/00460.jpg", "scene0655_00/00490.jpg" ], "turn_id": 4, "scan_id": "scene0655_00", "origin_question": "Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "67.1", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0655_00/00520.jpg", "scene0655_00/00550.jpg", "scene0655_00/00580.jpg", "scene0655_00/00610.jpg" ], "turn_id": 5, "scan_id": "scene0655_00", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the large rectangular blackboard\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the large rectangular blackboard\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0236_00/00000.jpg", "scene0236_00/00040.jpg", "scene0236_00/00080.jpg", "scene0236_00/00120.jpg", "scene0236_00/00160.jpg" ], "turn_id": 1, "scan_id": "scene0236_00", "origin_question": "Remember, have you seen any window(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any window(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0236_00/00200.jpg", "scene0236_00/00240.jpg", "scene0236_00/00280.jpg", "scene0236_00/00320.jpg", "scene0236_00/00360.jpg" ], "turn_id": 2, "scan_id": "scene0236_00", "origin_question": "In which turn did you see both \"the bin\" and \"the rectangular blue cardboard tissue box\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the bin\" and \"the rectangular blue cardboard tissue box\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0236_00/00400.jpg", "scene0236_00/00440.jpg", "scene0236_00/00480.jpg", "scene0236_00/00520.jpg", "scene0236_00/00560.jpg" ], "turn_id": 3, "scan_id": "scene0236_00", "origin_question": "Remember, have you seen any camera(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any camera(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0236_00/00600.jpg", "scene0236_00/00640.jpg", "scene0236_00/00680.jpg", "scene0236_00/00720.jpg", "scene0236_00/00760.jpg" ], "turn_id": 4, "scan_id": "scene0236_00", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the mirror\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the mirror\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0236_00/00800.jpg", "scene0236_00/00840.jpg", "scene0236_00/00880.jpg", "scene0236_00/00920.jpg", "scene0236_00/00960.jpg" ], "turn_id": 5, "scan_id": "scene0236_00", "origin_question": "Which two objects are on the same side of you now? \"the mirror\", \"the long, slender, bright red plastic hair dryer\", \"the pink frame with a white border holding photos\". ", "option": [ "the mirror, the long, slender, bright red plastic hair dryer", "the mirror, the pink frame with a white border holding photos", "the long, slender, bright red plastic hair dryer, the pink frame with a white border holding photos" ], "answer": "the mirror, the long, slender, bright red plastic hair dryer", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the mirror\", \"the long, slender, bright red plastic hair dryer\", \"the pink frame with a white border holding photos\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the mirror, the long, slender, bright red plastic hair dryer', 'the mirror, the pink frame with a white border holding photos', 'the long, slender, bright red plastic hair dryer, the pink frame with a white border holding photos'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0236_00/01000.jpg", "scene0236_00/01040.jpg", "scene0236_00/01080.jpg", "scene0236_00/01120.jpg", "scene0236_00/01160.jpg" ], "turn_id": 6, "scan_id": "scene0236_00", "origin_question": "By the end of the 4 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular blue cardboard tissue box\", Y: \"the mirror\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to Y and farther away from X", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 4 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular blue cardboard tissue box\", Y: \"the mirror\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0236_00/01200.jpg", "scene0236_00/01240.jpg", "scene0236_00/01280.jpg", "scene0236_00/01320.jpg", "scene0236_00/01360.jpg" ], "turn_id": 7, "scan_id": "scene0236_00", "origin_question": "Which direction is the round, light-colored cotton hat in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the round, light-colored cotton hat in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0236_00/01400.jpg", "scene0236_00/01440.jpg", "scene0236_00/01500.jpg", "scene0236_00/01590.jpg", "scene0236_00/01630.jpg" ], "turn_id": 8, "scan_id": "scene0236_00", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the rectangular light-colored wooden dresser with multiple drawers\"; \"the soft gray rectangular tote bag made of synthetic material\"; \"the long, slender, bright red plastic hair dryer\" ", "option": [ "the rectangular light-colored wooden dresser with multiple drawers", "the soft gray rectangular tote bag made of synthetic material", "the long, slender, bright red plastic hair dryer" ], "answer": "the soft gray rectangular tote bag made of synthetic material", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the rectangular light-colored wooden dresser with multiple drawers\"; \"the soft gray rectangular tote bag made of synthetic material\"; \"the long, slender, bright red plastic hair dryer\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular light-colored wooden dresser with multiple drawers', 'the soft gray rectangular tote bag made of synthetic material', 'the long, slender, bright red plastic hair dryer'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0438_00/00000.jpg", "scene0438_00/00040.jpg", "scene0438_00/00080.jpg", "scene0438_00/00120.jpg" ], "turn_id": 1, "scan_id": "scene0438_00", "origin_question": "Remember, have you seen any bed(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any bed(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0438_00/00160.jpg", "scene0438_00/00200.jpg", "scene0438_00/00240.jpg", "scene0438_00/00280.jpg" ], "turn_id": 2, "scan_id": "scene0438_00", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the standard rectangular double bed with a wooden frame\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the standard rectangular double bed with a wooden frame\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0438_00/00320.jpg", "scene0438_00/00360.jpg", "scene0438_00/00400.jpg", "scene0438_00/00440.jpg" ], "turn_id": 3, "scan_id": "scene0438_00", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the gray medium-sized cotton shirt\" from your position?", "option": [], "answer": "79.66", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the gray medium-sized cotton shirt\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0438_00/00480.jpg", "scene0438_00/00520.jpg", "scene0438_00/00560.jpg", "scene0438_00/00600.jpg" ], "turn_id": 4, "scan_id": "scene0438_00", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the gray medium-sized cotton shirt\"; \"the rectangular light-colored wooden closest door\"; \"the curtain\" ", "option": [ "the gray medium-sized cotton shirt", "the rectangular light-colored wooden closet door", "the curtain" ], "answer": "the rectangular light-colored wooden closet door", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the gray medium-sized cotton shirt\"; \"the rectangular light-colored wooden closest door\"; \"the curtain\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the gray medium-sized cotton shirt', 'the rectangular light-colored wooden closet door', 'the curtain'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "scene0438_00/00690.jpg", "scene0438_00/00730.jpg", "scene0438_00/00770.jpg", "scene0438_00/00810.jpg" ], "turn_id": 5, "scan_id": "scene0438_00", "origin_question": "Among these three objects(\"the tall rectangular wooden dresser in rich brown\",\"the striped rectangular bag in blue and white\",\"the rectangular wooden ottoman with a deep brown surface\" ), which one is the closest to you?", "option": [ "the tall rectangular wooden dresser in rich brown", "the striped rectangular bag in blue and white", "the rectangular wooden ottoman with a deep brown surface" ], "answer": "the rectangular wooden ottoman with a deep brown surface", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects(\"the tall rectangular wooden dresser in rich brown\",\"the striped rectangular bag in blue and white\",\"the rectangular wooden ottoman with a deep brown surface\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the tall rectangular wooden dresser in rich brown', 'the striped rectangular bag in blue and white', 'the rectangular wooden ottoman with a deep brown surface'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0438_00/00850.jpg", "scene0438_00/00890.jpg", "scene0438_00/00930.jpg", "scene0438_00/00970.jpg" ], "turn_id": 6, "scan_id": "scene0438_00", "origin_question": "Please recall \"the deep brown wooden stand with three drawers\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.49", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the deep brown wooden stand with three drawers\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0438_00/01010.jpg", "scene0438_00/01050.jpg", "scene0438_00/01090.jpg", "scene0438_00/01130.jpg" ], "turn_id": 7, "scan_id": "scene0438_00", "origin_question": "Please recall \"the gray medium-sized cotton shirt\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.2", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the gray medium-sized cotton shirt\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0438_00/01170.jpg", "scene0438_00/01210.jpg", "scene0438_00/01250.jpg", "scene0438_00/01290.jpg" ], "turn_id": 8, "scan_id": "scene0438_00", "origin_question": "At the end of which turn were both of \"the standard rectangular double bed with a wooden frame\" and \"the window\" on your left side?", "option": [], "answer": "4", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the standard rectangular double bed with a wooden frame\" and \"the window\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0435_01/00000.jpg", "scene0435_01/00040.jpg", "scene0435_01/00080.jpg", "scene0435_01/00120.jpg", "scene0435_01/00160.jpg" ], "turn_id": 1, "scan_id": "scene0435_01", "origin_question": "Remember, how many refrigerator(s) have you seen so far? ", "option": [], "answer": "1", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many refrigerator(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0435_01/00200.jpg", "scene0435_01/00240.jpg", "scene0435_01/00280.jpg", "scene0435_01/00320.jpg", "scene0435_01/00360.jpg" ], "turn_id": 2, "scan_id": "scene0435_01", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the modern chair with a black mesh seat and backrest\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the modern chair with a black mesh seat and backrest\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0435_01/00400.jpg", "scene0435_01/00440.jpg", "scene0435_01/00480.jpg", "scene0435_01/00520.jpg", "scene0435_01/00560.jpg" ], "turn_id": 3, "scan_id": "scene0435_01", "origin_question": "By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the stylish black refrigerator with a modern design\", Y: \"the bottle\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the stylish black refrigerator with a modern design\", Y: \"the bottle\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0435_01/00600.jpg", "scene0435_01/00640.jpg", "scene0435_01/00680.jpg", "scene0435_01/00720.jpg", "scene0435_01/00760.jpg" ], "turn_id": 4, "scan_id": "scene0435_01", "origin_question": "Please recall \"the modern chair with a black mesh seat and backrest\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.95", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the modern chair with a black mesh seat and backrest\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0435_01/00800.jpg", "scene0435_01/00840.jpg", "scene0435_01/00880.jpg", "scene0435_01/00920.jpg", "scene0435_01/00960.jpg" ], "turn_id": 5, "scan_id": "scene0435_01", "origin_question": "In which turn were you closest to \"the stylish black refrigerator with a modern design\"? ", "option": [], "answer": "4", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the stylish black refrigerator with a modern design\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0435_01/01000.jpg", "scene0435_01/01040.jpg", "scene0435_01/01080.jpg", "scene0435_01/01120.jpg", "scene0435_01/01160.jpg" ], "turn_id": 6, "scan_id": "scene0435_01", "origin_question": "Compared to your position at the end of the 5 turn, are you now closer or farther away from \"the modern chair with a black mesh seat and backrest\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 5 turn, are you now closer or farther away from \"the modern chair with a black mesh seat and backrest\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0435_01/01200.jpg", "scene0435_01/01240.jpg", "scene0435_01/01280.jpg", "scene0435_01/01320.jpg", "scene0435_01/01360.jpg" ], "turn_id": 7, "scan_id": "scene0435_01", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the light brown fabric couch with a rectangular shape\" from your position?", "option": [], "answer": "163.05", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the light brown fabric couch with a rectangular shape\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0435_01/01400.jpg", "scene0435_01/01440.jpg", "scene0435_01/01480.jpg", "scene0435_01/01520.jpg", "scene0435_01/01560.jpg" ], "turn_id": 8, "scan_id": "scene0435_01", "origin_question": "Which two objects are on the same side of you now? \"the modern chair with a black mesh seat and backrest\", \"the light brown fabric couch with a rectangular shape\", \"the flat-screen TV with a black glossy frame\". ", "option": [ "the modern chair with a black mesh seat and backrest, the light brown fabric couch with a rectangular shape", "the modern chair with a black mesh seat and backrest, the flat-screen TV with a black glossy frame", "the light brown fabric couch with a rectangular shape, the flat-screen TV with a black glossy frame" ], "answer": "the modern chair with a black mesh seat and backrest, the light brown fabric couch with a rectangular shape", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the modern chair with a black mesh seat and backrest\", \"the light brown fabric couch with a rectangular shape\", \"the flat-screen TV with a black glossy frame\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the modern chair with a black mesh seat and backrest, the light brown fabric couch with a rectangular shape', 'the modern chair with a black mesh seat and backrest, the flat-screen TV with a black glossy frame', 'the light brown fabric couch with a rectangular shape, the flat-screen TV with a black glossy frame'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0435_01/01600.jpg", "scene0435_01/01640.jpg", "scene0435_01/01680.jpg", "scene0435_01/01720.jpg", "scene0435_01/01760.jpg" ], "turn_id": 9, "scan_id": "scene0435_01", "origin_question": "Please recall \"the light brown fabric couch with a rectangular shape\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "3.45", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the light brown fabric couch with a rectangular shape\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0435_01/01800.jpg", "scene0435_01/01840.jpg", "scene0435_01/01880.jpg", "scene0435_01/01920.jpg", "scene0435_01/01960.jpg" ], "turn_id": 10, "scan_id": "scene0435_01", "origin_question": "By the end of the 7 turn, were you getting closer to or farther away from these two objects: X: \"the small rectangular dark wooden desk\", Y: \"the medium-sized black nylon backpack with zippers\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "farther away from both X and Y", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 7 turn, were you getting closer to or farther away from these two objects: X: \"the small rectangular dark wooden desk\", Y: \"the medium-sized black nylon backpack with zippers\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0156_00/00000.jpg", "scene0156_00/00030.jpg", "scene0156_00/00060.jpg", "scene0156_00/00090.jpg" ], "turn_id": 1, "scan_id": "scene0156_00", "origin_question": "Remember, have you seen any chair(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any chair(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0156_00/00120.jpg", "scene0156_00/00150.jpg", "scene0156_00/00180.jpg", "scene0156_00/00210.jpg" ], "turn_id": 2, "scan_id": "scene0156_00", "origin_question": "Please recall \"the light brown rectangular plastic bin\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.06", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the light brown rectangular plastic bin\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0156_00/00240.jpg", "scene0156_00/00270.jpg", "scene0156_00/00300.jpg", "scene0156_00/00330.jpg" ], "turn_id": 3, "scan_id": "scene0156_00", "origin_question": "In which turn did you see both \"the large rectangular wooden table with geometric design\" and \"the light brown rectangular plastic bin\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the large rectangular wooden table with geometric design\" and \"the light brown rectangular plastic bin\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0156_00/00360.jpg", "scene0156_00/00390.jpg", "scene0156_00/00420.jpg", "scene0156_00/00450.jpg" ], "turn_id": 4, "scan_id": "scene0156_00", "origin_question": "Please recall \"the light brown rectangular plastic bin\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.94", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the light brown rectangular plastic bin\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0156_00/00480.jpg", "scene0156_00/00510.jpg", "scene0156_00/00540.jpg", "scene0156_00/00570.jpg" ], "turn_id": 5, "scan_id": "scene0156_00", "origin_question": "Please recall \"the light brown rectangular plastic bin\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "4.43", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the light brown rectangular plastic bin\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0377_00/00000.jpg", "scene0377_00/00030.jpg", "scene0377_00/00060.jpg", "scene0377_00/00090.jpg" ], "turn_id": 1, "scan_id": "scene0377_00", "origin_question": "Remember, have you seen any broom(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any broom(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0377_00/00120.jpg", "scene0377_00/00150.jpg", "scene0377_00/00180.jpg", "scene0377_00/00210.jpg" ], "turn_id": 2, "scan_id": "scene0377_00", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the rectangular white storage box with a flat top\"; \"the wooden chair with orange cushions and vertical slats\"; \"the medium-sized white cylindrical bucket\" ", "option": [ "the rectangular white storage box with a flat top", "the wooden chair with orange cushions and vertical slats", "the medium-sized white cylindrical bucket" ], "answer": "the medium-sized white cylindrical bucket", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the rectangular white storage box with a flat top\"; \"the wooden chair with orange cushions and vertical slats\"; \"the medium-sized white cylindrical bucket\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular white storage box with a flat top', 'the wooden chair with orange cushions and vertical slats', 'the medium-sized white cylindrical bucket'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0377_00/00240.jpg", "scene0377_00/00270.jpg", "scene0377_00/00300.jpg", "scene0377_00/00330.jpg" ], "turn_id": 3, "scan_id": "scene0377_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0377_00/00360.jpg", "scene0377_00/00390.jpg", "scene0377_00/00420.jpg", "scene0377_00/00450.jpg" ], "turn_id": 4, "scan_id": "scene0377_00", "origin_question": "When did you first discover \"the rectangular white storage box with a flat top\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the rectangular white storage box with a flat top\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0377_00/00480.jpg", "scene0377_00/00510.jpg", "scene0377_00/00540.jpg", "scene0377_00/00570.jpg" ], "turn_id": 5, "scan_id": "scene0377_00", "origin_question": "Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0377_00/00600.jpg", "scene0377_00/00630.jpg", "scene0377_00/00660.jpg", "scene0377_00/00690.jpg" ], "turn_id": 6, "scan_id": "scene0377_00", "origin_question": "What will be the first-time appearance order of the following categories: bucket, chair, shelf?", "option": [ "shelf, chair, bucket", "shelf, bucket, chair", "chair, bucket, shelf", "chair, shelf, bucket" ], "answer": "chair, shelf, bucket", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bucket, chair, shelf? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['shelf, chair, bucket', 'shelf, bucket, chair', 'chair, bucket, shelf', 'chair, shelf, bucket'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0377_00/00720.jpg", "scene0377_00/00750.jpg", "scene0377_00/00780.jpg", "scene0377_00/00810.jpg" ], "turn_id": 7, "scan_id": "scene0377_00", "origin_question": "Is the window to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the window to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0044_region11/6f18db4cdf70471dac7f545b9fb08163_i1_4.jpg", "1mp3d_0044_region11/8630d0d97d4c4767b37488dfbe97f51b_i1_1.jpg", "1mp3d_0044_region11/8630d0d97d4c4767b37488dfbe97f51b_i2_1.jpg", "1mp3d_0044_region11/51d93d4c184f4197bd84f9b082b54e2b_i1_0.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0044_region11", "origin_question": "Remember, have you seen any letter(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any letter(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "1mp3d_0044_region11/32240afd8005415a998d5aedd2827dc1_i1_4.jpg", "1mp3d_0044_region11/310b6efab43c4dc99c2b6eabaff9b675_i1_1.jpg", "1mp3d_0044_region11/32240afd8005415a998d5aedd2827dc1_i1_3.jpg", "1mp3d_0044_region11/17ffcca2452e4cfd9d38c7485919d9d2_i1_0.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0044_region11", "origin_question": "Remember, how many tool(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many tool(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "1mp3d_0044_region11/17a99656493c47abbb5ceb214af3e8ea_i1_2.jpg", "1mp3d_0044_region11/17a99656493c47abbb5ceb214af3e8ea_i2_2.jpg", "1mp3d_0044_region11/17ffcca2452e4cfd9d38c7485919d9d2_i1_1.jpg", "1mp3d_0044_region11/5ff2309e75414df998435045bcb8c2ed_i1_1.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0044_region11", "origin_question": "How far is your current position from where you were at the end of the turn 1? (in meters)", "option": [], "answer": "4.12", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 1? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "1mp3d_0044_region11/5ff2309e75414df998435045bcb8c2ed_i2_1.jpg", "1mp3d_0044_region11/5ff2309e75414df998435045bcb8c2ed_i1_0.jpg", "1mp3d_0044_region11/5ff2309e75414df998435045bcb8c2ed_i2_0.jpg", "1mp3d_0044_region11/8282f8aa72164546855a2ab00dbad3a3_i1_3.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0044_region11", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0044_region11/8282f8aa72164546855a2ab00dbad3a3_i2_3.jpg", "1mp3d_0044_region11/8282f8aa72164546855a2ab00dbad3a3_i1_2.jpg", "1mp3d_0044_region11/8282f8aa72164546855a2ab00dbad3a3_i2_2.jpg", "1mp3d_0044_region11/c2ff31961cd0464d9fd12ba5208ffa52_i1_2.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0044_region11", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the rectangular wooden door with rich brown tones\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the rectangular wooden door with rich brown tones\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "1mp3d_0044_region11/c2ff31961cd0464d9fd12ba5208ffa52_i2_2.jpg", "1mp3d_0044_region11/144f0c921b1b4211a2b593a095a76b5a_i1_4.jpg", "1mp3d_0044_region11/144f0c921b1b4211a2b593a095a76b5a_i2_4.jpg", "1mp3d_0044_region11/144f0c921b1b4211a2b593a095a76b5a_i1_3.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0044_region11", "origin_question": "How far is your current position from where you were at the end of the turn 1? (in meters)", "option": [], "answer": "9.02", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 1? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0044_region11/144f0c921b1b4211a2b593a095a76b5a_i2_3.jpg", "1mp3d_0044_region11/144f0c921b1b4211a2b593a095a76b5a_i1_2.jpg", "1mp3d_0044_region11/144f0c921b1b4211a2b593a095a76b5a_i2_2.jpg", "1mp3d_0044_region11/c2ff31961cd0464d9fd12ba5208ffa52_i1_0.jpg" ], "turn_id": 7, "scan_id": "1mp3d_0044_region11", "origin_question": "Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the rectangular wooden door with rich brown tones\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the rectangular wooden door with rich brown tones\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0081_00/00000.jpg", "scene0081_00/00030.jpg", "scene0081_00/00060.jpg", "scene0081_00/00090.jpg" ], "turn_id": 1, "scan_id": "scene0081_00", "origin_question": "Remember, how many couch(s) have you seen so far? ", "option": [], "answer": "3", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many couch(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 2 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0081_00/00120.jpg", "scene0081_00/00150.jpg", "scene0081_00/00180.jpg", "scene0081_00/00210.jpg" ], "turn_id": 2, "scan_id": "scene0081_00", "origin_question": "Remember, how many ottoman(s) have you seen so far? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many ottoman(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 2 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0291_00/00000.jpg", "scene0291_00/00040.jpg", "scene0291_00/00080.jpg", "scene0291_00/00120.jpg", "scene0291_00/00160.jpg" ], "turn_id": 1, "scan_id": "scene0291_00", "origin_question": "Remember, have you seen any body loofah(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any body loofah(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0291_00/00200.jpg", "scene0291_00/00240.jpg", "scene0291_00/00280.jpg", "scene0291_00/00320.jpg", "scene0291_00/00360.jpg" ], "turn_id": 2, "scan_id": "scene0291_00", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular black rubber eraser with a smooth texture\", Y: \"the large rectangular polished wooden table\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular black rubber eraser with a smooth texture\", Y: \"the large rectangular polished wooden table\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0291_00/00400.jpg", "scene0291_00/00440.jpg", "scene0291_00/00480.jpg", "scene0291_00/00520.jpg", "scene0291_00/00560.jpg" ], "turn_id": 3, "scan_id": "scene0291_00", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "2.02", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0291_00/00600.jpg", "scene0291_00/00640.jpg", "scene0291_00/00680.jpg", "scene0291_00/00720.jpg", "scene0291_00/00760.jpg" ], "turn_id": 4, "scan_id": "scene0291_00", "origin_question": "In which turn did you see both \"the rectangular black rubber eraser with a smooth texture\" and \"the rectangular light brown wooden door\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the rectangular black rubber eraser with a smooth texture\" and \"the rectangular light brown wooden door\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0291_00/00800.jpg", "scene0291_00/00840.jpg", "scene0291_00/00880.jpg", "scene0291_00/00920.jpg", "scene0291_00/00960.jpg" ], "turn_id": 5, "scan_id": "scene0291_00", "origin_question": "Please recall \"the rectangular dark-colored storage box\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.79", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the rectangular dark-colored storage box\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0291_00/01000.jpg", "scene0291_00/01040.jpg", "scene0291_00/01080.jpg", "scene0291_00/01120.jpg", "scene0291_00/01160.jpg" ], "turn_id": 6, "scan_id": "scene0291_00", "origin_question": "Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "155.74", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0291_00/01200.jpg", "scene0291_00/01240.jpg", "scene0291_00/01280.jpg", "scene0291_00/01320.jpg", "scene0291_00/01360.jpg" ], "turn_id": 7, "scan_id": "scene0291_00", "origin_question": "In which turn were you closest to \"the rectangular dark-colored storage box\"? ", "option": [], "answer": "3", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the rectangular dark-colored storage box\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0291_00/01400.jpg", "scene0291_00/01440.jpg", "scene0291_00/01480.jpg", "scene0291_00/01520.jpg", "scene0291_00/01560.jpg" ], "turn_id": 8, "scan_id": "scene0291_00", "origin_question": "Compared to your position at the end of the 7 turn, are you now closer or farther away from \"the rectangular glass window framed in white\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 7 turn, are you now closer or farther away from \"the rectangular glass window framed in white\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0291_00/01600.jpg", "scene0291_00/01640.jpg", "scene0291_00/01680.jpg", "scene0291_00/01720.jpg", "scene0291_00/01760.jpg" ], "turn_id": 9, "scan_id": "scene0291_00", "origin_question": "Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0460_00/00000.jpg", "scene0460_00/00030.jpg", "scene0460_00/00060.jpg", "scene0460_00/00090.jpg" ], "turn_id": 1, "scan_id": "scene0460_00", "origin_question": "Remember, have you seen any toilet(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any toilet(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0460_00/00120.jpg", "scene0460_00/00150.jpg", "scene0460_00/00180.jpg", "scene0460_00/00210.jpg" ], "turn_id": 2, "scan_id": "scene0460_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0460_00/00240.jpg", "scene0460_00/00270.jpg", "scene0460_00/00300.jpg", "scene0460_00/00330.jpg" ], "turn_id": 3, "scan_id": "scene0460_00", "origin_question": "Please recall \"the toilet\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.68", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the toilet\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0460_00/00360.jpg", "scene0460_00/00390.jpg", "scene0460_00/00420.jpg", "scene0460_00/00450.jpg" ], "turn_id": 4, "scan_id": "scene0460_00", "origin_question": "Remember, have you seen any sink(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any sink(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0460_00/00480.jpg", "scene0460_00/00510.jpg", "scene0460_00/00540.jpg", "scene0460_00/00570.jpg" ], "turn_id": 5, "scan_id": "scene0460_00", "origin_question": "Which direction is the toilet in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the toilet in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0460_00/00600.jpg", "scene0460_00/00630.jpg", "scene0460_00/00670.jpg", "scene0460_00/00700.jpg" ], "turn_id": 6, "scan_id": "scene0460_00", "origin_question": "What will be the first-time appearance order of the following categories: curtain, mirror, toilet?", "option": [ "mirror, toilet, curtain", "toilet, mirror, curtain", "mirror, curtain, toilet", "toilet, curtain, mirror" ], "answer": "toilet, curtain, mirror", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: curtain, mirror, toilet? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['mirror, toilet, curtain', 'toilet, mirror, curtain', 'mirror, curtain, toilet', 'toilet, curtain, mirror'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0460_00/00730.jpg", "scene0460_00/00760.jpg", "scene0460_00/00790.jpg", "scene0460_00/00820.jpg" ], "turn_id": 7, "scan_id": "scene0460_00", "origin_question": "At the end of which turn were both of \"the toilet\" and \"the round, white ceramic sink with a silver faucet\" on your left side?", "option": [], "answer": "6", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the toilet\" and \"the round, white ceramic sink with a silver faucet\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0512_00/00000.jpg", "scene0512_00/00030.jpg", "scene0512_00/00060.jpg", "scene0512_00/00090.jpg" ], "turn_id": 1, "scan_id": "scene0512_00", "origin_question": "Remember, how many table(s) have you seen so far? ", "option": [], "answer": "2", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many table(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0512_00/00120.jpg", "scene0512_00/00150.jpg", "scene0512_00/00180.jpg", "scene0512_00/00210.jpg" ], "turn_id": 2, "scan_id": "scene0512_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0512_00/00240.jpg", "scene0512_00/00270.jpg", "scene0512_00/00300.jpg", "scene0512_00/00330.jpg" ], "turn_id": 3, "scan_id": "scene0512_00", "origin_question": "How far is your current position from where you were at the end of the turn 1? (in meters)", "option": [], "answer": "3.85", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 1? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0512_00/00360.jpg", "scene0512_00/00390.jpg", "scene0512_00/00420.jpg", "scene0512_00/00450.jpg" ], "turn_id": 4, "scan_id": "scene0512_00", "origin_question": "Remember, how many sack(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many sack(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0633_01/00000.jpg", "scene0633_01/00040.jpg", "scene0633_01/00080.jpg", "scene0633_01/00120.jpg" ], "turn_id": 1, "scan_id": "scene0633_01", "origin_question": "Remember, have you seen any bin(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any bin(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0633_01/00160.jpg", "scene0633_01/00200.jpg", "scene0633_01/00240.jpg", "scene0633_01/00280.jpg" ], "turn_id": 2, "scan_id": "scene0633_01", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the small rectangular black bin made of plastic\"; \"the round analog clock with a black border\"; \"the small rectangular wooden cabinet in muted brown\" ", "option": [ "the small rectangular black bin made of plastic", "the round analog clock with a black border", "the small rectangular wooden cabinet in muted brown" ], "answer": "the round analog clock with a black border", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the small rectangular black bin made of plastic\"; \"the round analog clock with a black border\"; \"the small rectangular wooden cabinet in muted brown\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the small rectangular black bin made of plastic', 'the round analog clock with a black border', 'the small rectangular wooden cabinet in muted brown'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0633_01/00320.jpg", "scene0633_01/00360.jpg", "scene0633_01/00400.jpg", "scene0633_01/00440.jpg" ], "turn_id": 3, "scan_id": "scene0633_01", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the medium-sized white rectangular wooden table with a drawer\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the medium-sized white rectangular wooden table with a drawer\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "scene0633_01/00480.jpg", "scene0633_01/00520.jpg", "scene0633_01/00560.jpg", "scene0633_01/00600.jpg" ], "turn_id": 4, "scan_id": "scene0633_01", "origin_question": "Among these three objects(\"the white wooden wardrobe with a double-door design\",\"the round analog clock with a black border\",\"the medium-sized white rectangular wooden table with a drawer\" ), which one is the closest to you?", "option": [ "the white wooden wardrobe with a double-door design", "the round analog clock with a black border", "the medium-sized white rectangular wooden table with a drawer" ], "answer": "the white wooden wardrobe with a double-door design", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects(\"the white wooden wardrobe with a double-door design\",\"the round analog clock with a black border\",\"the medium-sized white rectangular wooden table with a drawer\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the white wooden wardrobe with a double-door design', 'the round analog clock with a black border', 'the medium-sized white rectangular wooden table with a drawer'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0633_01/00640.jpg", "scene0633_01/00680.jpg", "scene0633_01/00720.jpg", "scene0633_01/00760.jpg" ], "turn_id": 5, "scan_id": "scene0633_01", "origin_question": "", "option": [], "answer": "", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0633_01/00800.jpg", "scene0633_01/00840.jpg", "scene0633_01/00880.jpg", "scene0633_01/00920.jpg" ], "turn_id": 6, "scan_id": "scene0633_01", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the round analog clock with a black border\" from your position?", "option": [], "answer": "53.97", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the round analog clock with a black border\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0633_01/00960.jpg", "scene0633_01/01000.jpg", "scene0633_01/01040.jpg", "scene0633_01/01080.jpg" ], "turn_id": 7, "scan_id": "scene0633_01", "origin_question": "Which direction is the white rectangular wooden shelf with three open shelves in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the white rectangular wooden shelf with three open shelves in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0633_01/01120.jpg", "scene0633_01/01160.jpg", "scene0633_01/01200.jpg", "scene0633_01/01240.jpg" ], "turn_id": 8, "scan_id": "scene0633_01", "origin_question": "What will be the first-time appearance order of the following categories: bin, paper, wardrobe?", "option": [ "wardrobe, bin, paper", "bin, wardrobe, paper", "bin, paper, wardrobe", "paper, bin, wardrobe" ], "answer": "bin, wardrobe, paper", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bin, paper, wardrobe? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['wardrobe, bin, paper', 'bin, wardrobe, paper', 'bin, paper, wardrobe', 'paper, bin, wardrobe'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0277_01/00000.jpg", "scene0277_01/00030.jpg", "scene0277_01/00060.jpg", "scene0277_01/00090.jpg" ], "turn_id": 1, "scan_id": "scene0277_01", "origin_question": "Remember, have you seen any dresser(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any dresser(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0277_01/00120.jpg", "scene0277_01/00150.jpg", "scene0277_01/00180.jpg", "scene0277_01/00210.jpg" ], "turn_id": 2, "scan_id": "scene0277_01", "origin_question": "In which turn did you see both \"the rectangular single bed with a deep blue mattress\" and \"the simple white rectangular wooden shelf\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the rectangular single bed with a deep blue mattress\" and \"the simple white rectangular wooden shelf\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0277_01/00240.jpg", "scene0277_01/00270.jpg", "scene0277_01/00300.jpg", "scene0277_01/00330.jpg" ], "turn_id": 3, "scan_id": "scene0277_01", "origin_question": "When did you first discover \"the simple white rectangular wooden shelf\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the simple white rectangular wooden shelf\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0277_01/00360.jpg", "scene0277_01/00390.jpg", "scene0277_01/00420.jpg", "scene0277_01/00450.jpg" ], "turn_id": 4, "scan_id": "scene0277_01", "origin_question": "", "option": [], "answer": "", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0277_01/00480.jpg", "scene0277_01/00510.jpg", "scene0277_01/00540.jpg", "scene0277_01/00570.jpg" ], "turn_id": 5, "scan_id": "scene0277_01", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the white rectangular metal radiator with horizontal partitions\" from your position?", "option": [], "answer": "145.46", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the white rectangular metal radiator with horizontal partitions\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0277_01/00600.jpg", "scene0277_01/00630.jpg", "scene0277_01/00660.jpg", "scene0277_01/00690.jpg" ], "turn_id": 6, "scan_id": "scene0277_01", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the black chair with a curved backrest\" from your position?", "option": [], "answer": "79.46", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the black chair with a curved backrest\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0277_01/00720.jpg", "scene0277_01/00750.jpg", "scene0277_01/00780.jpg", "scene0277_01/00810.jpg" ], "turn_id": 7, "scan_id": "scene0277_01", "origin_question": "In which turn were you closest to \"the simple white rectangular wooden shelf\"? ", "option": [], "answer": "4", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the simple white rectangular wooden shelf\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0129_00/00000.jpg", "scene0129_00/00040.jpg", "scene0129_00/00080.jpg", "scene0129_00/00120.jpg", "scene0129_00/00160.jpg" ], "turn_id": 1, "scan_id": "scene0129_00", "origin_question": "Remember, have you seen any chair(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any chair(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0129_00/00200.jpg", "scene0129_00/00240.jpg", "scene0129_00/00280.jpg", "scene0129_00/00320.jpg", "scene0129_00/00360.jpg" ], "turn_id": 2, "scan_id": "scene0129_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0129_00/00400.jpg", "scene0129_00/00440.jpg", "scene0129_00/00480.jpg", "scene0129_00/00520.jpg", "scene0129_00/00560.jpg" ], "turn_id": 3, "scan_id": "scene0129_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0129_00/00600.jpg", "scene0129_00/00640.jpg", "scene0129_00/00680.jpg", "scene0129_00/00720.jpg", "scene0129_00/00760.jpg" ], "turn_id": 4, "scan_id": "scene0129_00", "origin_question": "When did you first discover \"the small, round wooden table with a dark brown finish\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the small, round wooden table with a dark brown finish\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0129_00/00800.jpg", "scene0129_00/00840.jpg", "scene0129_00/00880.jpg", "scene0129_00/00920.jpg", "scene0129_00/00960.jpg" ], "turn_id": 5, "scan_id": "scene0129_00", "origin_question": "Is the small, round wooden table with a dark brown finish to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the small, round wooden table with a dark brown finish to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0129_00/01000.jpg", "scene0129_00/01040.jpg", "scene0129_00/01080.jpg", "scene0129_00/01120.jpg", "scene0129_00/01160.jpg" ], "turn_id": 6, "scan_id": "scene0129_00", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the rectangular white plastic switch with a digital display\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the rectangular white plastic switch with a digital display\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0129_00/01200.jpg", "scene0129_00/01240.jpg", "scene0129_00/01280.jpg", "scene0129_00/01320.jpg", "scene0129_00/01360.jpg" ], "turn_id": 7, "scan_id": "scene0129_00", "origin_question": "What will be the first-time appearance order of the following categories: door, switch, window?", "option": [ "switch, window, door", "window, door, switch", "switch, door, window", "window, switch, door" ], "answer": "window, switch, door", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: door, switch, window? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['switch, window, door', 'window, door, switch', 'switch, door, window', 'window, switch, door'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0129_00/01400.jpg", "scene0129_00/01440.jpg", "scene0129_00/01480.jpg", "scene0129_00/01520.jpg", "scene0129_00/01560.jpg" ], "turn_id": 8, "scan_id": "scene0129_00", "origin_question": "Using your orientation at the end of turn 6 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "82.75", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 6 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0129_00/01600.jpg", "scene0129_00/01640.jpg", "scene0129_00/01680.jpg", "scene0129_00/01720.jpg", "scene0129_00/01760.jpg" ], "turn_id": 9, "scan_id": "scene0129_00", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the rectangular white plastic switch with a digital display\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the rectangular white plastic switch with a digital display\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0129_00/01800.jpg", "scene0129_00/01840.jpg", "scene0129_00/01880.jpg", "scene0129_00/01920.jpg", "scene0129_00/01960.jpg" ], "turn_id": 10, "scan_id": "scene0129_00", "origin_question": "Using your orientation at the end of turn 7 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "148.14", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 7 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "1mp3d_0044_region34/e57c3ad712794b6f84b64925df90afae_i1_0.jpg", "1mp3d_0044_region34/e57c3ad712794b6f84b64925df90afae_i2_0.jpg", "1mp3d_0044_region34/e57c3ad712794b6f84b64925df90afae_i1_5.jpg", "1mp3d_0044_region34/e57c3ad712794b6f84b64925df90afae_i2_5.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0044_region34", "origin_question": "Remember, how many projector(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many projector(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "1mp3d_0044_region34/e57c3ad712794b6f84b64925df90afae_i1_4.jpg", "1mp3d_0044_region34/e57c3ad712794b6f84b64925df90afae_i2_4.jpg", "1mp3d_0044_region34/b545ce32aeff42ab9282887c7e50a4f7_i1_4.jpg", "1mp3d_0044_region34/b545ce32aeff42ab9282887c7e50a4f7_i2_4.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0044_region34", "origin_question": "When did you first discover \"the small pink chair with a patterned backrest\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the small pink chair with a patterned backrest\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "1mp3d_0044_region34/3bf4f616b03141f69004b618911f1261_i2_1.jpg", "1mp3d_0044_region34/3bf4f616b03141f69004b618911f1261_i1_2.jpg", "1mp3d_0044_region34/3bf4f616b03141f69004b618911f1261_i2_2.jpg", "1mp3d_0044_region34/ddbfa25a4d2f4270be337edfab59ddcb_i1_1.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0044_region34", "origin_question": "How far is your current position from where you were at the end of the turn 1? (in meters)", "option": [], "answer": "2.87", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 1? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "1mp3d_0044_region34/ddbfa25a4d2f4270be337edfab59ddcb_i2_1.jpg", "1mp3d_0044_region34/ddbfa25a4d2f4270be337edfab59ddcb_i1_0.jpg", "1mp3d_0044_region34/ddbfa25a4d2f4270be337edfab59ddcb_i2_0.jpg", "1mp3d_0044_region34/ddbfa25a4d2f4270be337edfab59ddcb_i1_5.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0044_region34", "origin_question": "In which turn did you see both \"the large round mirror with a golden frame\" and \"the small pink chair with a patterned backrest\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the large round mirror with a golden frame\" and \"the small pink chair with a patterned backrest\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "1mp3d_0044_region34/ddbfa25a4d2f4270be337edfab59ddcb_i2_5.jpg", "1mp3d_0044_region34/3bf4f616b03141f69004b618911f1261_i1_5.jpg", "1mp3d_0044_region34/3bf4f616b03141f69004b618911f1261_i1_0.jpg", "1mp3d_0044_region34/3bf4f616b03141f69004b618911f1261_i2_0.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0044_region34", "origin_question": "Which direction is the black wall-mounted flat-screen TV of medium size in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the black wall-mounted flat-screen TV of medium size in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0462_00/00000.jpg", "scene0462_00/00040.jpg", "scene0462_00/00080.jpg", "scene0462_00/00120.jpg" ], "turn_id": 1, "scan_id": "scene0462_00", "origin_question": "Remember, have you seen any player(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any player(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0462_00/00160.jpg", "scene0462_00/00200.jpg", "scene0462_00/00240.jpg", "scene0462_00/00280.jpg" ], "turn_id": 2, "scan_id": "scene0462_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0462_00/00320.jpg", "scene0462_00/00360.jpg", "scene0462_00/00400.jpg", "scene0462_00/00440.jpg" ], "turn_id": 3, "scan_id": "scene0462_00", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the large white copier with multiple trays and papers\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the large white copier with multiple trays and papers\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0462_00/00480.jpg", "scene0462_00/00520.jpg", "scene0462_00/00560.jpg", "scene0462_00/00600.jpg" ], "turn_id": 4, "scan_id": "scene0462_00", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the standard-sized white plastic and metal printer\" from your position?", "option": [], "answer": "178.87", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the standard-sized white plastic and metal printer\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0462_00/00640.jpg", "scene0462_00/00680.jpg", "scene0462_00/00720.jpg", "scene0462_00/00760.jpg" ], "turn_id": 5, "scan_id": "scene0462_00", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the standard upright refrigerator in light grey color\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the standard upright refrigerator in light grey color\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0462_00/00800.jpg", "scene0462_00/00840.jpg", "scene0462_00/00880.jpg", "scene0462_00/00920.jpg" ], "turn_id": 6, "scan_id": "scene0462_00", "origin_question": "In which turn did you see both \"the large white copier with multiple trays and papers\" and \"the standard-sized white plastic and metal printer\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the large white copier with multiple trays and papers\" and \"the standard-sized white plastic and metal printer\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0462_00/00960.jpg", "scene0462_00/01000.jpg", "scene0462_00/01040.jpg", "scene0462_00/01080.jpg" ], "turn_id": 7, "scan_id": "scene0462_00", "origin_question": "In which turn did you see both \"the large white copier with multiple trays and papers\" and \"the rectangular beige cork board\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the large white copier with multiple trays and papers\" and \"the rectangular beige cork board\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0387_01/00000.jpg", "scene0387_01/00040.jpg", "scene0387_01/00080.jpg", "scene0387_01/00120.jpg" ], "turn_id": 1, "scan_id": "scene0387_01", "origin_question": "Remember, have you seen any bin(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any bin(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0387_01/00160.jpg", "scene0387_01/00200.jpg", "scene0387_01/00240.jpg", "scene0387_01/00280.jpg" ], "turn_id": 2, "scan_id": "scene0387_01", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0387_01/00320.jpg", "scene0387_01/00360.jpg", "scene0387_01/00400.jpg", "scene0387_01/00440.jpg" ], "turn_id": 3, "scan_id": "scene0387_01", "origin_question": "Remember, have you seen any brush(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any brush(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0387_01/00480.jpg", "scene0387_01/00520.jpg", "scene0387_01/00560.jpg", "scene0387_01/00600.jpg" ], "turn_id": 4, "scan_id": "scene0387_01", "origin_question": "", "option": [], "answer": "", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0387_01/00640.jpg", "scene0387_01/00680.jpg", "scene0387_01/00720.jpg", "scene0387_01/00760.jpg" ], "turn_id": 5, "scan_id": "scene0387_01", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "3.64", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "scene0387_01/00800.jpg", "scene0387_01/00840.jpg", "scene0387_01/00880.jpg", "scene0387_01/00920.jpg" ], "turn_id": 6, "scan_id": "scene0387_01", "origin_question": "Assuming the direction you are facing at the end of the turn 4 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Right", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 4 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0057_region8/65b7fe76c12f44f1b67c5ad7b03ba63b_i1_5.jpg", "1mp3d_0057_region8/65b7fe76c12f44f1b67c5ad7b03ba63b_i2_5.jpg", "1mp3d_0057_region8/8e5ae8d7ee55420f80bf433f33e790fe_i1_0.jpg", "1mp3d_0057_region8/8e5ae8d7ee55420f80bf433f33e790fe_i2_0.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0057_region8", "origin_question": "Remember, have you seen any door(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any door(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "1mp3d_0057_region8/301baa2b53d54369899be99587972b37_i1_0.jpg", "1mp3d_0057_region8/301baa2b53d54369899be99587972b37_i2_0.jpg", "1mp3d_0057_region8/301baa2b53d54369899be99587972b37_i1_1.jpg", "1mp3d_0057_region8/301baa2b53d54369899be99587972b37_i2_1.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0057_region8", "origin_question": "At the end of which turn were both of \"the round light brown wooden fan\" and \"the black flat-screen television on a wooden shelf\" on your left side?", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the round light brown wooden fan\" and \"the black flat-screen television on a wooden shelf\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "1mp3d_0057_region8/8e5ae8d7ee55420f80bf433f33e790fe_i1_2.jpg", "1mp3d_0057_region8/8e5ae8d7ee55420f80bf433f33e790fe_i2_2.jpg", "1mp3d_0057_region8/8e5ae8d7ee55420f80bf433f33e790fe_i1_1.jpg", "1mp3d_0057_region8/8e5ae8d7ee55420f80bf433f33e790fe_i2_1.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0057_region8", "origin_question": "At the end of which turn were both of \"the round light brown wooden fan\" and \"the single bed with wooden headboard and footboard\" on your right side?", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the round light brown wooden fan\" and \"the single bed with wooden headboard and footboard\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "1mp3d_0057_region8/65b7fe76c12f44f1b67c5ad7b03ba63b_i1_4.jpg", "1mp3d_0057_region8/65b7fe76c12f44f1b67c5ad7b03ba63b_i2_4.jpg", "1mp3d_0057_region8/65b7fe76c12f44f1b67c5ad7b03ba63b_i1_3.jpg", "1mp3d_0057_region8/65b7fe76c12f44f1b67c5ad7b03ba63b_i2_3.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0057_region8", "origin_question": "Among these three objects(\"the round light brown wooden fan\",\"the light brown rectangular door\",\"the black flat-screen television on a wooden shelf\" ), which one is the farthest from you?", "option": [ "the round light brown wooden fan", "the light brown rectangular door", "the black flat-screen television on a wooden shelf" ], "answer": "the light brown rectangular door", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects(\"the round light brown wooden fan\",\"the light brown rectangular door\",\"the black flat-screen television on a wooden shelf\" ), which one is the farthest from you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the round light brown wooden fan', 'the light brown rectangular door', 'the black flat-screen television on a wooden shelf'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0057_region8/9fb720a0582945cc81a6e6d2b63b6b30_i1_3.jpg", "1mp3d_0057_region8/9fb720a0582945cc81a6e6d2b63b6b30_i2_3.jpg", "1mp3d_0057_region8/54f491ad92e347c892e35a0d48d497d2_i1_4.jpg", "1mp3d_0057_region8/54f491ad92e347c892e35a0d48d497d2_i2_4.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0057_region8", "origin_question": "Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the black flat-screen television on a wooden shelf\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the black flat-screen television on a wooden shelf\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "1mp3d_0057_region8/9fb720a0582945cc81a6e6d2b63b6b30_i1_4.jpg", "1mp3d_0057_region8/9fb720a0582945cc81a6e6d2b63b6b30_i2_4.jpg", "1mp3d_0057_region8/54f491ad92e347c892e35a0d48d497d2_i1_3.jpg", "1mp3d_0057_region8/54f491ad92e347c892e35a0d48d497d2_i2_3.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0057_region8", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "3.59", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0222_00/00000.jpg", "scene0222_00/00040.jpg", "scene0222_00/00080.jpg", "scene0222_00/00120.jpg", "scene0222_00/00160.jpg" ], "turn_id": 1, "scan_id": "scene0222_00", "origin_question": "Remember, have you seen any shower(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any shower(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0222_00/00200.jpg", "scene0222_00/00240.jpg", "scene0222_00/00280.jpg", "scene0222_00/00320.jpg", "scene0222_00/00360.jpg" ], "turn_id": 2, "scan_id": "scene0222_00", "origin_question": "Please recall \"the medium-sized cylindrical plastic cup with a blue lid\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.75", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the medium-sized cylindrical plastic cup with a blue lid\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0222_00/00400.jpg", "scene0222_00/00440.jpg", "scene0222_00/00480.jpg", "scene0222_00/00520.jpg", "scene0222_00/00560.jpg" ], "turn_id": 3, "scan_id": "scene0222_00", "origin_question": "Please recall \"the classic wooden dresser with a warm brown color\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.08", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the classic wooden dresser with a warm brown color\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0222_00/00600.jpg", "scene0222_00/00640.jpg", "scene0222_00/00680.jpg", "scene0222_00/00720.jpg", "scene0222_00/00760.jpg" ], "turn_id": 4, "scan_id": "scene0222_00", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the classic wooden dresser with a warm brown color\" from your position?", "option": [], "answer": "81.16", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the classic wooden dresser with a warm brown color\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "scene0222_00/00800.jpg", "scene0222_00/00840.jpg", "scene0222_00/00880.jpg", "scene0222_00/00920.jpg", "scene0222_00/00960.jpg" ], "turn_id": 5, "scan_id": "scene0222_00", "origin_question": "Among these three objects(\"the grey cylindrical fan made of plastic\",\"the horizontal, white metal radiator\",\"the backpack\" ), which one is the closest to you?", "option": [ "the grey cylindrical fan made of plastic", "the horizontal, white metal radiator", "the backpack" ], "answer": "the grey cylindrical fan made of plastic", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the grey cylindrical fan made of plastic\",\"the horizontal, white metal radiator\",\"the backpack\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the grey cylindrical fan made of plastic', 'the horizontal, white metal radiator', 'the backpack'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0222_00/01000.jpg", "scene0222_00/01040.jpg", "scene0222_00/01080.jpg", "scene0222_00/01120.jpg", "scene0222_00/01160.jpg" ], "turn_id": 6, "scan_id": "scene0222_00", "origin_question": "Which direction is the backpack in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the backpack in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0222_00/01200.jpg", "scene0222_00/01240.jpg", "scene0222_00/01280.jpg", "scene0222_00/01320.jpg", "scene0222_00/01360.jpg" ], "turn_id": 7, "scan_id": "scene0222_00", "origin_question": "In which turn did you see both \"the medium yellow plastic bag with wrinkled edges\" and \"the rectangular cushion with gray and white geometric patterns\" simultaneously? ", "option": [], "answer": "5", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the medium yellow plastic bag with wrinkled edges\" and \"the rectangular cushion with gray and white geometric patterns\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0222_00/01400.jpg", "scene0222_00/01440.jpg", "scene0222_00/01480.jpg", "scene0222_00/01520.jpg", "scene0222_00/01560.jpg" ], "turn_id": 8, "scan_id": "scene0222_00", "origin_question": "Compared to your position at the end of the 6 turn, are you now closer or farther away from \"the classic wooden dresser with a warm brown color\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 6 turn, are you now closer or farther away from \"the classic wooden dresser with a warm brown color\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0222_00/01600.jpg", "scene0222_00/01640.jpg", "scene0222_00/01680.jpg", "scene0222_00/01720.jpg", "scene0222_00/01770.jpg" ], "turn_id": 9, "scan_id": "scene0222_00", "origin_question": "Using your orientation at the end of turn 6 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 6 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0222_00/01810.jpg", "scene0222_00/01850.jpg", "scene0222_00/01890.jpg", "scene0222_00/01930.jpg", "scene0222_00/01970.jpg" ], "turn_id": 10, "scan_id": "scene0222_00", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the backpack\" from your position?", "option": [], "answer": "129.94", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the backpack\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0307_01/00000.jpg", "scene0307_01/00040.jpg", "scene0307_01/00080.jpg", "scene0307_01/00120.jpg", "scene0307_01/00160.jpg" ], "turn_id": 1, "scan_id": "scene0307_01", "origin_question": "Remember, have you seen any towel(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any towel(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0307_01/00200.jpg", "scene0307_01/00240.jpg", "scene0307_01/00280.jpg", "scene0307_01/00320.jpg", "scene0307_01/00360.jpg" ], "turn_id": 2, "scan_id": "scene0307_01", "origin_question": "In which turn were you closest to \"the rectangular woven basket in beige or light brown\"? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the rectangular woven basket in beige or light brown\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0307_01/00400.jpg", "scene0307_01/00440.jpg", "scene0307_01/00480.jpg", "scene0307_01/00520.jpg", "scene0307_01/00560.jpg" ], "turn_id": 3, "scan_id": "scene0307_01", "origin_question": "", "option": [], "answer": "", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0307_01/00600.jpg", "scene0307_01/00640.jpg", "scene0307_01/00680.jpg", "scene0307_01/00720.jpg", "scene0307_01/00760.jpg" ], "turn_id": 4, "scan_id": "scene0307_01", "origin_question": "Is the cabinet to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the cabinet to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0307_01/00800.jpg", "scene0307_01/00840.jpg", "scene0307_01/00880.jpg", "scene0307_01/00920.jpg", "scene0307_01/00960.jpg" ], "turn_id": 5, "scan_id": "scene0307_01", "origin_question": "Is the wooden chair with a green fabric draped over it to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the wooden chair with a green fabric draped over it to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0307_01/01000.jpg", "scene0307_01/01040.jpg", "scene0307_01/01080.jpg", "scene0307_01/01120.jpg", "scene0307_01/01160.jpg" ], "turn_id": 6, "scan_id": "scene0307_01", "origin_question": "Which direction is the white front-loading washing machine made of metal in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the white front-loading washing machine made of metal in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0307_01/01200.jpg", "scene0307_01/01240.jpg", "scene0307_01/01280.jpg", "scene0307_01/01320.jpg", "scene0307_01/01360.jpg" ], "turn_id": 7, "scan_id": "scene0307_01", "origin_question": "Is the wooden chair with a green fabric draped over it to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the wooden chair with a green fabric draped over it to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0307_01/01400.jpg", "scene0307_01/01440.jpg", "scene0307_01/01480.jpg", "scene0307_01/01520.jpg", "scene0307_01/01560.jpg" ], "turn_id": 8, "scan_id": "scene0307_01", "origin_question": "", "option": [], "answer": "", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0307_01/01600.jpg", "scene0307_01/01640.jpg", "scene0307_01/01680.jpg", "scene0307_01/01720.jpg", "scene0307_01/01760.jpg" ], "turn_id": 9, "scan_id": "scene0307_01", "origin_question": "Please recall \"the refrigerator\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "4.05", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the refrigerator\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "scene0307_01/01800.jpg", "scene0307_01/01840.jpg", "scene0307_01/01880.jpg", "scene0307_01/01920.jpg", "scene0307_01/01960.jpg" ], "turn_id": 10, "scan_id": "scene0307_01", "origin_question": "Among these three objects(\"the white front-loading washing machine made of metal\",\"the cabinet\",\"the dark cylindrical pipe\" ), which one is the closest to you?", "option": [ "the white front-loading washing machine made of metal", "the cabinet", "the dark cylindrical pipe" ], "answer": "the white front-loading washing machine made of metal", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the white front-loading washing machine made of metal\",\"the cabinet\",\"the dark cylindrical pipe\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the white front-loading washing machine made of metal', 'the cabinet', 'the dark cylindrical pipe'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "1mp3d_0084_region8/5125c4a0a99f45788e694d5a73ba99c7_i1_3.jpg", "1mp3d_0084_region8/5125c4a0a99f45788e694d5a73ba99c7_i2_3.jpg", "1mp3d_0084_region8/7743b72f7c3d40749363873307eef9c9_i1_5.jpg", "1mp3d_0084_region8/7743b72f7c3d40749363873307eef9c9_i2_5.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0084_region8", "origin_question": "Remember, how many plant(s) have you seen so far? ", "option": [], "answer": "2", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many plant(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "1mp3d_0084_region8/5125c4a0a99f45788e694d5a73ba99c7_i1_4.jpg", "1mp3d_0084_region8/5125c4a0a99f45788e694d5a73ba99c7_i2_4.jpg", "1mp3d_0084_region8/20f70ab4efec496c86529075648caef2_i1_4.jpg", "1mp3d_0084_region8/20f70ab4efec496c86529075648caef2_i2_4.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0084_region8", "origin_question": "When did you first discover \"the small arched white fireplace with a black interior\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the small arched white fireplace with a black interior\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "1mp3d_0084_region8/1b321779a4374c2b952c51820daa9e6c_i1_5.jpg", "1mp3d_0084_region8/1b321779a4374c2b952c51820daa9e6c_i2_5.jpg", "1mp3d_0084_region8/d70997cce0464616a65afcf28d32b41e_i1_5.jpg", "1mp3d_0084_region8/d70997cce0464616a65afcf28d32b41e_i2_5.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0084_region8", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "1mp3d_0084_region8/1b321779a4374c2b952c51820daa9e6c_i1_4.jpg", "1mp3d_0084_region8/1b321779a4374c2b952c51820daa9e6c_i2_4.jpg", "1mp3d_0084_region8/1b321779a4374c2b952c51820daa9e6c_i1_3.jpg", "1mp3d_0084_region8/1b321779a4374c2b952c51820daa9e6c_i2_3.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0084_region8", "origin_question": "How far is your current position from where you were at the end of the turn 3? (in meters)", "option": [], "answer": "1.44", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 3? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "1mp3d_0084_region8/20f70ab4efec496c86529075648caef2_i1_3.jpg", "1mp3d_0084_region8/20f70ab4efec496c86529075648caef2_i2_3.jpg", "1mp3d_0084_region8/acfc7a38863c412bbde815478879ed54_i1_1.jpg", "1mp3d_0084_region8/acfc7a38863c412bbde815478879ed54_i2_1.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0084_region8", "origin_question": "When did you first discover \"the round wooden table with a rich brown color\"? (index of the turn)", "option": [], "answer": "5", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the round wooden table with a rich brown color\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0084_region8/519b19d242994285bd99db910f4717f6_i1_2.jpg", "1mp3d_0084_region8/519b19d242994285bd99db910f4717f6_i2_2.jpg", "1mp3d_0084_region8/aa4cfd0126dd4c6a9c533ca9cb4a033d_i1_2.jpg", "1mp3d_0084_region8/aa4cfd0126dd4c6a9c533ca9cb4a033d_i2_2.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0084_region8", "origin_question": "", "option": [], "answer": "", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "1mp3d_0084_region8/c29e99f090194613b5b11af906c47dab_i1_4.jpg", "1mp3d_0084_region8/c29e99f090194613b5b11af906c47dab_i2_4.jpg", "1mp3d_0084_region8/c29e99f090194613b5b11af906c47dab_i1_5.jpg", "1mp3d_0084_region8/c29e99f090194613b5b11af906c47dab_i2_5.jpg" ], "turn_id": 7, "scan_id": "1mp3d_0084_region8", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the small arched white fireplace with a black interior\"; \"the rectangular, light brown wooden bench\"; \"the yellow bowl with handles containing fruit\" ", "option": [ "the small arched white fireplace with a black interior", "the rectangular, light brown wooden bench", "the yellow bowl with handles containing fruit" ], "answer": "the rectangular, light brown wooden bench", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the small arched white fireplace with a black interior\"; \"the rectangular, light brown wooden bench\"; \"the yellow bowl with handles containing fruit\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the small arched white fireplace with a black interior', 'the rectangular, light brown wooden bench', 'the yellow bowl with handles containing fruit'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "1mp3d_0084_region8/aa4cfd0126dd4c6a9c533ca9cb4a033d_i1_4.jpg", "1mp3d_0084_region8/aa4cfd0126dd4c6a9c533ca9cb4a033d_i2_4.jpg", "1mp3d_0084_region8/519b19d242994285bd99db910f4717f6_i1_4.jpg", "1mp3d_0084_region8/519b19d242994285bd99db910f4717f6_i2_4.jpg" ], "turn_id": 8, "scan_id": "1mp3d_0084_region8", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "125.82", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "1mp3d_0084_region8/519b19d242994285bd99db910f4717f6_i1_5.jpg", "1mp3d_0084_region8/519b19d242994285bd99db910f4717f6_i2_5.jpg", "1mp3d_0084_region8/519b19d242994285bd99db910f4717f6_i1_0.jpg", "1mp3d_0084_region8/519b19d242994285bd99db910f4717f6_i2_0.jpg" ], "turn_id": 9, "scan_id": "1mp3d_0084_region8", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the small arched white fireplace with a black interior\" from your position?", "option": [], "answer": "75.9", "user_message": "For the 9 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the small arched white fireplace with a black interior\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0029_region18/1e7c16e6cb054c9fbbeb15160bc11f73_i1_4.jpg", "1mp3d_0029_region18/1e7c16e6cb054c9fbbeb15160bc11f73_i2_4.jpg", "1mp3d_0029_region18/112edb40f34e470da3a5b04599e71211_i1_5.jpg", "1mp3d_0029_region18/112edb40f34e470da3a5b04599e71211_i2_5.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0029_region18", "origin_question": "Remember, have you seen any sink(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any sink(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "1mp3d_0029_region18/1e7c16e6cb054c9fbbeb15160bc11f73_i1_3.jpg", "1mp3d_0029_region18/1e7c16e6cb054c9fbbeb15160bc11f73_i2_3.jpg", "1mp3d_0029_region18/112edb40f34e470da3a5b04599e71211_i1_0.jpg", "1mp3d_0029_region18/112edb40f34e470da3a5b04599e71211_i2_0.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0029_region18", "origin_question": "In which turn did you see both \"the medium-sized hardcover book with a white cover\" and \"the small rectangular white range hood with a glossy finish\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the medium-sized hardcover book with a white cover\" and \"the small rectangular white range hood with a glossy finish\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "1mp3d_0029_region18/afad225acfbe48f39cf2894c61d9e714_i1_3.jpg", "1mp3d_0029_region18/afad225acfbe48f39cf2894c61d9e714_i2_3.jpg", "1mp3d_0029_region18/afad225acfbe48f39cf2894c61d9e714_i1_2.jpg", "1mp3d_0029_region18/afad225acfbe48f39cf2894c61d9e714_i2_2.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0029_region18", "origin_question": "Is the small rectangular white range hood with a glossy finish to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the small rectangular white range hood with a glossy finish to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "1mp3d_0029_region18/afad225acfbe48f39cf2894c61d9e714_i1_1.jpg", "1mp3d_0029_region18/afad225acfbe48f39cf2894c61d9e714_i2_1.jpg", "1mp3d_0029_region18/543bd11ef03b4d3e87b0d9572db01ee2_i1_0.jpg", "1mp3d_0029_region18/543bd11ef03b4d3e87b0d9572db01ee2_i2_0.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0029_region18", "origin_question": "Please recall \"the white kitchen towel with blue stripes\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.37", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the white kitchen towel with blue stripes\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "1mp3d_0029_region18/543bd11ef03b4d3e87b0d9572db01ee2_i1_1.jpg", "1mp3d_0029_region18/543bd11ef03b4d3e87b0d9572db01ee2_i2_1.jpg", "1mp3d_0029_region18/543bd11ef03b4d3e87b0d9572db01ee2_i1_2.jpg", "1mp3d_0029_region18/543bd11ef03b4d3e87b0d9572db01ee2_i2_2.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0029_region18", "origin_question": "Which two objects are on the same side of you now? \"the white rectangular door with a glass panel\", \"the medium-sized hardcover book with a white cover\", \"the small rectangular white range hood with a glossy finish\". ", "option": [ "the white rectangular door with a glass panel, the medium-sized hardcover book with a white cover", "the white rectangular door with a glass panel, the small rectangular white range hood with a glossy finish", "the medium-sized hardcover book with a white cover, the small rectangular white range hood with a glossy finish" ], "answer": "the white rectangular door with a glass panel, the medium-sized hardcover book with a white cover", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the white rectangular door with a glass panel\", \"the medium-sized hardcover book with a white cover\", \"the small rectangular white range hood with a glossy finish\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the white rectangular door with a glass panel, the medium-sized hardcover book with a white cover', 'the white rectangular door with a glass panel, the small rectangular white range hood with a glossy finish', 'the medium-sized hardcover book with a white cover, the small rectangular white range hood with a glossy finish'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "1mp3d_0029_region18/b4ac79acffd24a838cc44bb094d0a5cb_i1_2.jpg", "1mp3d_0029_region18/b4ac79acffd24a838cc44bb094d0a5cb_i2_2.jpg", "1mp3d_0029_region18/f0db0d83f9d646b997d099d7eae80dd1_i1_1.jpg", "1mp3d_0029_region18/f0db0d83f9d646b997d099d7eae80dd1_i2_1.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0029_region18", "origin_question": "Among these three objects(\"the white kitchen towel with blue stripes\",\"the medium-sized hardcover book with a white cover\",\"the small rectangular white range hood with a glossy finish\" ), which one is the closest to you?", "option": [ "the white kitchen towel with blue stripes", "the medium-sized hardcover book with a white cover", "the small rectangular white range hood with a glossy finish" ], "answer": "the medium-sized hardcover book with a white cover", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects(\"the white kitchen towel with blue stripes\",\"the medium-sized hardcover book with a white cover\",\"the small rectangular white range hood with a glossy finish\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the white kitchen towel with blue stripes', 'the medium-sized hardcover book with a white cover', 'the small rectangular white range hood with a glossy finish'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "1mp3d_0029_region18/b4ac79acffd24a838cc44bb094d0a5cb_i1_1.jpg", "1mp3d_0029_region18/b4ac79acffd24a838cc44bb094d0a5cb_i2_1.jpg", "1mp3d_0029_region18/f0db0d83f9d646b997d099d7eae80dd1_i1_0.jpg", "1mp3d_0029_region18/f0db0d83f9d646b997d099d7eae80dd1_i2_0.jpg" ], "turn_id": 7, "scan_id": "1mp3d_0029_region18", "origin_question": "Which direction is the medium-sized hardcover book with a white cover in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the medium-sized hardcover book with a white cover in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "1mp3d_0029_region18/f0db0d83f9d646b997d099d7eae80dd1_i1_5.jpg", "1mp3d_0029_region18/f0db0d83f9d646b997d099d7eae80dd1_i2_5.jpg", "1mp3d_0029_region18/f6a03fba4dfd415a8b65cce84ee21c19_i1_5.jpg", "1mp3d_0029_region18/f6a03fba4dfd415a8b65cce84ee21c19_i2_5.jpg" ], "turn_id": 8, "scan_id": "1mp3d_0029_region18", "origin_question": "At the end of which turn were both of \"the white rectangular door with a glass panel\" and \"the white kitchen towel with blue stripes\" on your left side?", "option": [], "answer": "2", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the white rectangular door with a glass panel\" and \"the white kitchen towel with blue stripes\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "1mp3d_0029_region18/8bb3222421334f02a07273f14a9b2d32_i1_4.jpg", "1mp3d_0029_region18/8bb3222421334f02a07273f14a9b2d32_i2_4.jpg", "1mp3d_0029_region18/8bb3222421334f02a07273f14a9b2d32_i1_5.jpg", "1mp3d_0029_region18/8bb3222421334f02a07273f14a9b2d32_i2_5.jpg" ], "turn_id": 9, "scan_id": "1mp3d_0029_region18", "origin_question": "Is the large stainless steel refrigerator with a vertical design to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 9 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the large stainless steel refrigerator with a vertical design to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0574_01/00000.jpg", "scene0574_01/00040.jpg", "scene0574_01/00080.jpg", "scene0574_01/00120.jpg", "scene0574_01/00160.jpg" ], "turn_id": 1, "scan_id": "scene0574_01", "origin_question": "Remember, have you seen any sink(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any sink(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0574_01/00200.jpg", "scene0574_01/00240.jpg", "scene0574_01/00280.jpg", "scene0574_01/00320.jpg", "scene0574_01/00360.jpg" ], "turn_id": 2, "scan_id": "scene0574_01", "origin_question": "At the end of which turn were both of \"the round white ceramic sink with a smooth surface\" and \"the vanity\" on your left side?", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the round white ceramic sink with a smooth surface\" and \"the vanity\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0574_01/00400.jpg", "scene0574_01/00440.jpg", "scene0574_01/00480.jpg", "scene0574_01/00520.jpg", "scene0574_01/00560.jpg" ], "turn_id": 3, "scan_id": "scene0574_01", "origin_question": "At the end of which turn were both of \"the round white ceramic sink with a smooth surface\" and \"the vanity\" on your right side?", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the round white ceramic sink with a smooth surface\" and \"the vanity\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0574_01/00600.jpg", "scene0574_01/00640.jpg", "scene0574_01/00680.jpg", "scene0574_01/00720.jpg", "scene0574_01/00760.jpg" ], "turn_id": 4, "scan_id": "scene0574_01", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0574_01/00800.jpg", "scene0574_01/00840.jpg", "scene0574_01/00880.jpg", "scene0574_01/00920.jpg", "scene0574_01/00960.jpg" ], "turn_id": 5, "scan_id": "scene0574_01", "origin_question": "In which turn were you farthest from \"the vanity\"? ", "option": [], "answer": "1", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the vanity\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0574_01/01000.jpg", "scene0574_01/01040.jpg", "scene0574_01/01080.jpg", "scene0574_01/01120.jpg", "scene0574_01/01160.jpg" ], "turn_id": 6, "scan_id": "scene0574_01", "origin_question": "What will be the first-time appearance order of the following categories: bench, bin, shoe?", "option": [ "bench, bin, shoe", "bin, shoe, bench", "shoe, bench, bin", "bin, bench, shoe" ], "answer": "bin, bench, shoe", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bench, bin, shoe? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['bench, bin, shoe', 'bin, shoe, bench', 'shoe, bench, bin', 'bin, bench, shoe'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0574_01/01200.jpg", "scene0574_01/01240.jpg", "scene0574_01/01280.jpg", "scene0574_01/01320.jpg", "scene0574_01/01360.jpg" ], "turn_id": 7, "scan_id": "scene0574_01", "origin_question": "Is the round white ceramic sink with a smooth surface to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the round white ceramic sink with a smooth surface to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0574_01/01400.jpg", "scene0574_01/01440.jpg", "scene0574_01/01480.jpg", "scene0574_01/01520.jpg", "scene0574_01/01560.jpg" ], "turn_id": 8, "scan_id": "scene0574_01", "origin_question": "How far is your current position from where you were at the end of the turn 7? (in meters)", "option": [], "answer": "1.28", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 7? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0625_01/00000.jpg", "scene0625_01/00030.jpg", "scene0625_01/00060.jpg", "scene0625_01/00090.jpg" ], "turn_id": 1, "scan_id": "scene0625_01", "origin_question": "Remember, have you seen any bathtub(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any bathtub(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0625_01/00120.jpg", "scene0625_01/00150.jpg", "scene0625_01/00180.jpg", "scene0625_01/00210.jpg" ], "turn_id": 2, "scan_id": "scene0625_01", "origin_question": "When did you first discover \"the bathtub\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the bathtub\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0625_01/00240.jpg", "scene0625_01/00270.jpg", "scene0625_01/00300.jpg", "scene0625_01/00330.jpg" ], "turn_id": 3, "scan_id": "scene0625_01", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0625_01/00360.jpg", "scene0625_01/00390.jpg", "scene0625_01/00420.jpg", "scene0625_01/00450.jpg" ], "turn_id": 4, "scan_id": "scene0625_01", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "91.16", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0625_01/00480.jpg", "scene0625_01/00510.jpg", "scene0625_01/00540.jpg", "scene0625_01/00570.jpg" ], "turn_id": 5, "scan_id": "scene0625_01", "origin_question": "Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0625_01/00600.jpg", "scene0625_01/00630.jpg", "scene0625_01/00660.jpg", "scene0625_01/00690.jpg" ], "turn_id": 6, "scan_id": "scene0625_01", "origin_question": "", "option": [], "answer": "", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0625_01/00720.jpg", "scene0625_01/00750.jpg", "scene0625_01/00780.jpg", "scene0625_01/00810.jpg" ], "turn_id": 7, "scan_id": "scene0625_01", "origin_question": "At the end of which turn were both of \"the curtain\" and \"the rectangular gray plastic bin\" on your left side?", "option": [], "answer": "4", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the curtain\" and \"the rectangular gray plastic bin\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0625_01/00840.jpg", "scene0625_01/00870.jpg", "scene0625_01/00900.jpg", "scene0625_01/00930.jpg" ], "turn_id": 8, "scan_id": "scene0625_01", "origin_question": "What will be the first-time appearance order of the following categories: stopcock, telephone, toilet?", "option": [ "stopcock, toilet, telephone", "toilet, telephone, stopcock", "stopcock, telephone, toilet", "toilet, stopcock, telephone" ], "answer": "stopcock, toilet, telephone", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: stopcock, telephone, toilet? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['stopcock, toilet, telephone', 'toilet, telephone, stopcock', 'stopcock, telephone, toilet', 'toilet, stopcock, telephone'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0329_02/00000.jpg", "scene0329_02/00040.jpg", "scene0329_02/00080.jpg", "scene0329_02/00120.jpg" ], "turn_id": 1, "scan_id": "scene0329_02", "origin_question": "Remember, have you seen any luggage(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any luggage(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0329_02/00160.jpg", "scene0329_02/00200.jpg", "scene0329_02/00240.jpg", "scene0329_02/00280.jpg" ], "turn_id": 2, "scan_id": "scene0329_02", "origin_question": "In which turn were you closest to \"the medium-sized black rectangular backpack with silver zippers\"? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the medium-sized black rectangular backpack with silver zippers\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0329_02/00320.jpg", "scene0329_02/00360.jpg", "scene0329_02/00400.jpg", "scene0329_02/00440.jpg" ], "turn_id": 3, "scan_id": "scene0329_02", "origin_question": "Please recall \"the medium-sized black rectangular backpack with silver zippers\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.91", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the medium-sized black rectangular backpack with silver zippers\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0329_02/00480.jpg", "scene0329_02/00520.jpg", "scene0329_02/00560.jpg", "scene0329_02/00600.jpg" ], "turn_id": 4, "scan_id": "scene0329_02", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the lamp\", Y: \"the picture\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to Y and farther away from X", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the lamp\", Y: \"the picture\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0329_02/00640.jpg", "scene0329_02/00680.jpg", "scene0329_02/00720.jpg", "scene0329_02/00760.jpg" ], "turn_id": 5, "scan_id": "scene0329_02", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "2.81", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0329_02/00800.jpg", "scene0329_02/00840.jpg", "scene0329_02/00880.jpg", "scene0329_02/00920.jpg" ], "turn_id": 6, "scan_id": "scene0329_02", "origin_question": "Please recall \"the standard red and silver metal fire extinguisher\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "3.99", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the standard red and silver metal fire extinguisher\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0329_02/00960.jpg", "scene0329_02/01000.jpg", "scene0329_02/01040.jpg", "scene0329_02/01080.jpg" ], "turn_id": 7, "scan_id": "scene0329_02", "origin_question": "How far is your current position from where you were at the end of the turn 5? (in meters)", "option": [], "answer": "2.44", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 5? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0329_02/01120.jpg", "scene0329_02/01160.jpg", "scene0329_02/01200.jpg", "scene0329_02/01240.jpg" ], "turn_id": 8, "scan_id": "scene0329_02", "origin_question": "Using your orientation at the end of turn 6 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 6 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0084_region1/9afe7c95718743e288c6f3dceb252faf_i1_5.jpg", "1mp3d_0084_region1/9afe7c95718743e288c6f3dceb252faf_i2_5.jpg", "1mp3d_0084_region1/2b9fc7a9ebdd4114b3e1f78a8abd1c61_i1_2.jpg", "1mp3d_0084_region1/2b9fc7a9ebdd4114b3e1f78a8abd1c61_i2_2.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0084_region1", "origin_question": "Remember, have you seen any shelf(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any shelf(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0084_region1/6b27862316444c83b4ab9f4e19ddec77_i1_1.jpg", "1mp3d_0084_region1/6b27862316444c83b4ab9f4e19ddec77_i2_1.jpg", "1mp3d_0084_region1/2b9fc7a9ebdd4114b3e1f78a8abd1c61_i1_3.jpg", "1mp3d_0084_region1/2b9fc7a9ebdd4114b3e1f78a8abd1c61_i2_3.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0084_region1", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "1mp3d_0084_region1/8329085eff624e2383ac389479ebb608_i1_5.jpg", "1mp3d_0084_region1/8329085eff624e2383ac389479ebb608_i2_5.jpg", "1mp3d_0084_region1/9afe7c95718743e288c6f3dceb252faf_i1_4.jpg", "1mp3d_0084_region1/9afe7c95718743e288c6f3dceb252faf_i2_4.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0084_region1", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "1mp3d_0084_region1/db7a229a929043b290d5d9510bfb3c2b_i1_4.jpg", "1mp3d_0084_region1/db7a229a929043b290d5d9510bfb3c2b_i2_4.jpg", "1mp3d_0084_region1/9afe7c95718743e288c6f3dceb252faf_i2_3.jpg", "1mp3d_0084_region1/8329085eff624e2383ac389479ebb608_i2_2.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0084_region1", "origin_question": "In which turn did you see both \"the rectangular, rough-textured stone shelf\" and \"the grey rectangular metal stool with a handle\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the rectangular, rough-textured stone shelf\" and \"the grey rectangular metal stool with a handle\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0057_region24/b8bb3aadfe61498bbaab510af3a999ac_i1_1.jpg", "1mp3d_0057_region24/c9c9dd651f25409c915f962a60efb7fb_i1_1.jpg", "1mp3d_0057_region24/c9c9dd651f25409c915f962a60efb7fb_i2_1.jpg", "1mp3d_0057_region24/416d35ec584141e093766986f1b74144_i1_0.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0057_region24", "origin_question": "Remember, have you seen any air conditioner(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any air conditioner(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "1mp3d_0057_region24/416d35ec584141e093766986f1b74144_i2_0.jpg", "1mp3d_0057_region24/416d35ec584141e093766986f1b74144_i1_1.jpg", "1mp3d_0057_region24/416d35ec584141e093766986f1b74144_i2_1.jpg", "1mp3d_0057_region24/c9c9dd651f25409c915f962a60efb7fb_i1_0.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0057_region24", "origin_question": "In which turn were you closest to \"the rectangular wooden pool table with a blue surface\"? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the rectangular wooden pool table with a blue surface\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "1mp3d_0057_region24/c9c9dd651f25409c915f962a60efb7fb_i2_0.jpg", "1mp3d_0057_region24/b8bb3aadfe61498bbaab510af3a999ac_i1_2.jpg", "1mp3d_0057_region24/b8bb3aadfe61498bbaab510af3a999ac_i2_2.jpg", "1mp3d_0057_region24/b8bb3aadfe61498bbaab510af3a999ac_i1_0.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0057_region24", "origin_question": "In which turn did you see both \"the rectangular wooden pool table with a blue surface\" and \"the large brown wooden fan with light fixture\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the rectangular wooden pool table with a blue surface\" and \"the large brown wooden fan with light fixture\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0057_region24/b8bb3aadfe61498bbaab510af3a999ac_i2_0.jpg", "1mp3d_0057_region24/b4939ad76d9c43d38c082e5dff3c6e22_i1_4.jpg", "1mp3d_0057_region24/abea76767e384adb9cbc473b4011806f_i1_5.jpg", "1mp3d_0057_region24/abea76767e384adb9cbc473b4011806f_i2_5.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0057_region24", "origin_question": "", "option": [], "answer": "", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0057_region24/aea99822f5aa4e3faab5431696c13cfa_i1_0.jpg", "1mp3d_0057_region24/aea99822f5aa4e3faab5431696c13cfa_i2_0.jpg", "1mp3d_0057_region24/aea99822f5aa4e3faab5431696c13cfa_i1_1.jpg", "1mp3d_0057_region24/abea76767e384adb9cbc473b4011806f_i1_0.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0057_region24", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the rectangular wooden pool table with a blue surface\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the rectangular wooden pool table with a blue surface\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "1mp3d_0057_region24/b4939ad76d9c43d38c082e5dff3c6e22_i1_3.jpg", "1mp3d_0057_region24/b4939ad76d9c43d38c082e5dff3c6e22_i2_3.jpg", "1mp3d_0057_region24/e0cf6a5f87384560a5a1087eeb714de0_i1_1.jpg", "1mp3d_0057_region24/e0cf6a5f87384560a5a1087eeb714de0_i2_1.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0057_region24", "origin_question": "In which turn were you closest to \"the large brown wooden fan with light fixture\"? ", "option": [], "answer": "4", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the large brown wooden fan with light fixture\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "1mp3d_0057_region24/e0cf6a5f87384560a5a1087eeb714de0_i1_0.jpg", "1mp3d_0057_region24/87533ca2d23b444ab0f7da5d7a4c17b4_i1_2.jpg", "1mp3d_0057_region24/87533ca2d23b444ab0f7da5d7a4c17b4_i1_3.jpg", "1mp3d_0057_region24/4e9aca10b4f142a48cde60ca007e614d_i1_2.jpg" ], "turn_id": 7, "scan_id": "1mp3d_0057_region24", "origin_question": "Assuming the direction you are facing at the end of the turn 5 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Left", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 5 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0535_00/00000.jpg", "scene0535_00/00030.jpg", "scene0535_00/00060.jpg", "scene0535_00/00090.jpg" ], "turn_id": 1, "scan_id": "scene0535_00", "origin_question": "Remember, how many shelf(s) have you seen so far? ", "option": [], "answer": "2", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many shelf(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0535_00/00120.jpg", "scene0535_00/00150.jpg", "scene0535_00/00180.jpg", "scene0535_00/00210.jpg" ], "turn_id": 2, "scan_id": "scene0535_00", "origin_question": "In which turn were you closest to \"the chunky knitted jacket with a black and white pattern\"? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the chunky knitted jacket with a black and white pattern\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0535_00/00240.jpg", "scene0535_00/00270.jpg", "scene0535_00/00300.jpg", "scene0535_00/00330.jpg" ], "turn_id": 3, "scan_id": "scene0535_00", "origin_question": "In which turn were you closest to \"the chunky knitted jacket with a black and white pattern\"? ", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the chunky knitted jacket with a black and white pattern\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0535_00/00360.jpg", "scene0535_00/00390.jpg", "scene0535_00/00420.jpg", "scene0535_00/00450.jpg" ], "turn_id": 4, "scan_id": "scene0535_00", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the chunky knitted jacket with a black and white pattern\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the chunky knitted jacket with a black and white pattern\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0535_00/00480.jpg", "scene0535_00/00510.jpg", "scene0535_00/00540.jpg", "scene0535_00/00570.jpg" ], "turn_id": 5, "scan_id": "scene0535_00", "origin_question": "What will be the first-time appearance order of the following categories: paper, picture, plant?", "option": [ "picture, plant, paper", "plant, picture, paper", "paper, picture, plant", "plant, paper, picture" ], "answer": "picture, plant, paper", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: paper, picture, plant? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['picture, plant, paper', 'plant, picture, paper', 'paper, picture, plant', 'plant, paper, picture'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0699_00/00000.jpg", "scene0699_00/00040.jpg", "scene0699_00/00080.jpg", "scene0699_00/00120.jpg", "scene0699_00/00160.jpg" ], "turn_id": 1, "scan_id": "scene0699_00", "origin_question": "Remember, have you seen any bag(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any bag(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0699_00/00200.jpg", "scene0699_00/00240.jpg", "scene0699_00/00280.jpg", "scene0699_00/00320.jpg", "scene0699_00/00360.jpg" ], "turn_id": 2, "scan_id": "scene0699_00", "origin_question": "Which direction is the white rectangular plastic remote control with buttons in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the white rectangular plastic remote control with buttons in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0699_00/00400.jpg", "scene0699_00/00440.jpg", "scene0699_00/00480.jpg", "scene0699_00/00520.jpg", "scene0699_00/00560.jpg" ], "turn_id": 3, "scan_id": "scene0699_00", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the petite light brown wooden box with a rectangular shape\"; \"the small wooden stand with a glass tabletop\"; \"the black, standard size backpack made of durable material\" ", "option": [ "the petite light brown wooden box with a rectangular shape", "the small wooden stand with a glass tabletop", "the black, standard size backpack made of durable material" ], "answer": "the black, standard size backpack made of durable material", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the petite light brown wooden box with a rectangular shape\"; \"the small wooden stand with a glass tabletop\"; \"the black, standard size backpack made of durable material\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the petite light brown wooden box with a rectangular shape', 'the small wooden stand with a glass tabletop', 'the black, standard size backpack made of durable material'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0699_00/00600.jpg", "scene0699_00/00640.jpg", "scene0699_00/00680.jpg", "scene0699_00/00720.jpg", "scene0699_00/00760.jpg" ], "turn_id": 4, "scan_id": "scene0699_00", "origin_question": "Which two objects are on the same side of you now? \"the square white and blue patterned tissue box\", \"the framed rectangular picture with a white frame\", \"the medium-sized black rectangular leather purse\". ", "option": [ "the square white and blue patterned tissue box, the framed rectangular picture with a white frame", "the square white and blue patterned tissue box, the medium-sized black rectangular leather purse", "the framed rectangular picture with a white frame, the medium-sized black rectangular leather purse" ], "answer": "the framed rectangular picture with a white frame, the medium-sized black rectangular leather purse", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the square white and blue patterned tissue box\", \"the framed rectangular picture with a white frame\", \"the medium-sized black rectangular leather purse\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the square white and blue patterned tissue box, the framed rectangular picture with a white frame', 'the square white and blue patterned tissue box, the medium-sized black rectangular leather purse', 'the framed rectangular picture with a white frame, the medium-sized black rectangular leather purse'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0699_00/00800.jpg", "scene0699_00/00840.jpg", "scene0699_00/00880.jpg", "scene0699_00/00920.jpg", "scene0699_00/00960.jpg" ], "turn_id": 5, "scan_id": "scene0699_00", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the rectangular bed with a white bedsheet\" from your position?", "option": [], "answer": "107.0", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the rectangular bed with a white bedsheet\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0699_00/01000.jpg", "scene0699_00/01040.jpg", "scene0699_00/01080.jpg", "scene0699_00/01120.jpg", "scene0699_00/01160.jpg" ], "turn_id": 6, "scan_id": "scene0699_00", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the square white and blue patterned tissue box\" from your position?", "option": [], "answer": "167.55", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the square white and blue patterned tissue box\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0699_00/01200.jpg", "scene0699_00/01240.jpg", "scene0699_00/01280.jpg", "scene0699_00/01320.jpg", "scene0699_00/01360.jpg" ], "turn_id": 7, "scan_id": "scene0699_00", "origin_question": "Which two objects are on the same side of you now? \"the white rectangular plastic remote control with buttons\", \"the small green and black wrap\", \"the round black and gold machine\". ", "option": [ "the white rectangular plastic remote control with buttons, the small green and black wrap", "the white rectangular plastic remote control with buttons, the round black and gold machine", "the small green and black wrap, the round black and gold machine" ], "answer": "the small green and black wrap, the round black and gold machine", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the white rectangular plastic remote control with buttons\", \"the small green and black wrap\", \"the round black and gold machine\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the white rectangular plastic remote control with buttons, the small green and black wrap', 'the white rectangular plastic remote control with buttons, the round black and gold machine', 'the small green and black wrap, the round black and gold machine'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0699_00/01400.jpg", "scene0699_00/01440.jpg", "scene0699_00/01480.jpg", "scene0699_00/01520.jpg", "scene0699_00/01560.jpg" ], "turn_id": 8, "scan_id": "scene0699_00", "origin_question": "In which turn were you closest to \"the framed rectangular picture with a white frame\"? ", "option": [], "answer": "4", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the framed rectangular picture with a white frame\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0699_00/01600.jpg", "scene0699_00/01640.jpg", "scene0699_00/01680.jpg", "scene0699_00/01720.jpg", "scene0699_00/01760.jpg" ], "turn_id": 9, "scan_id": "scene0699_00", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "57.85", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0699_00/01800.jpg", "scene0699_00/01840.jpg", "scene0699_00/01880.jpg", "scene0699_00/01920.jpg", "scene0699_00/01960.jpg" ], "turn_id": 10, "scan_id": "scene0699_00", "origin_question": "Is the small green and black wrap to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the small green and black wrap to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0474_00/00000.jpg", "scene0474_00/00040.jpg", "scene0474_00/00080.jpg", "scene0474_00/00120.jpg", "scene0474_00/00160.jpg" ], "turn_id": 1, "scan_id": "scene0474_00", "origin_question": "Remember, have you seen any fireplace(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any fireplace(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0474_00/00200.jpg", "scene0474_00/00240.jpg", "scene0474_00/00280.jpg", "scene0474_00/00320.jpg", "scene0474_00/00360.jpg" ], "turn_id": 2, "scan_id": "scene0474_00", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the black rectangular metal desktop monitor\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the black rectangular metal desktop monitor\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0474_00/00400.jpg", "scene0474_00/00440.jpg", "scene0474_00/00480.jpg", "scene0474_00/00520.jpg", "scene0474_00/00560.jpg" ], "turn_id": 3, "scan_id": "scene0474_00", "origin_question": "By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular blue plastic bin\", Y: \"the clothes\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to Y and farther away from X", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular blue plastic bin\", Y: \"the clothes\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0474_00/00600.jpg", "scene0474_00/00640.jpg", "scene0474_00/00680.jpg", "scene0474_00/00720.jpg", "scene0474_00/00760.jpg" ], "turn_id": 4, "scan_id": "scene0474_00", "origin_question": "What will be the first-time appearance order of the following categories: bin, clothes, monitor?", "option": [ "bin, monitor, clothes", "monitor, bin, clothes", "clothes, bin, monitor", "monitor, clothes, bin" ], "answer": "bin, monitor, clothes", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bin, clothes, monitor? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['bin, monitor, clothes', 'monitor, bin, clothes', 'clothes, bin, monitor', 'monitor, clothes, bin'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0474_00/00800.jpg", "scene0474_00/00840.jpg", "scene0474_00/00880.jpg", "scene0474_00/00920.jpg", "scene0474_00/00960.jpg" ], "turn_id": 5, "scan_id": "scene0474_00", "origin_question": "Which direction is the modern silver aluminum laptop with a standard size in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the modern silver aluminum laptop with a standard size in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0474_00/01000.jpg", "scene0474_00/01040.jpg", "scene0474_00/01080.jpg", "scene0474_00/01120.jpg", "scene0474_00/01160.jpg" ], "turn_id": 6, "scan_id": "scene0474_00", "origin_question": "Which two objects are on the same side of you now? \"the rectangular brown wooden shelf\", \"the couch\", \"the black and silver wired telephone made of plastic\". ", "option": [ "the rectangular brown wooden shelf, the couch", "the rectangular brown wooden shelf, the black and silver wired telephone made of plastic", "the couch, the black and silver wired telephone made of plastic" ], "answer": "the couch, the black and silver wired telephone made of plastic", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the rectangular brown wooden shelf\", \"the couch\", \"the black and silver wired telephone made of plastic\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular brown wooden shelf, the couch', 'the rectangular brown wooden shelf, the black and silver wired telephone made of plastic', 'the couch, the black and silver wired telephone made of plastic'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0474_00/01200.jpg", "scene0474_00/01240.jpg", "scene0474_00/01280.jpg", "scene0474_00/01320.jpg", "scene0474_00/01360.jpg" ], "turn_id": 7, "scan_id": "scene0474_00", "origin_question": "By the end of the 5 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular brown wooden shelf\", Y: \"the rectangular wood office desk in good condition\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "farther away from both X and Y", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 5 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular brown wooden shelf\", Y: \"the rectangular wood office desk in good condition\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0474_00/01400.jpg", "scene0474_00/01440.jpg", "scene0474_00/01480.jpg", "scene0474_00/01520.jpg", "scene0474_00/01560.jpg" ], "turn_id": 8, "scan_id": "scene0474_00", "origin_question": "Remember, have you seen any bin(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any bin(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0614_01/00000.jpg", "scene0614_01/00040.jpg", "scene0614_01/00080.jpg", "scene0614_01/00120.jpg" ], "turn_id": 1, "scan_id": "scene0614_01", "origin_question": "Remember, how many desk(s) have you seen so far? ", "option": [], "answer": "2", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many desk(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0614_01/00160.jpg", "scene0614_01/00200.jpg", "scene0614_01/00240.jpg", "scene0614_01/00280.jpg" ], "turn_id": 2, "scan_id": "scene0614_01", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular blackboard with a wooden frame\", Y: \"the red ceramic cup with a white interior\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to X and farther away from Y", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular blackboard with a wooden frame\", Y: \"the red ceramic cup with a white interior\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0614_01/00320.jpg", "scene0614_01/00360.jpg", "scene0614_01/00400.jpg", "scene0614_01/00440.jpg" ], "turn_id": 3, "scan_id": "scene0614_01", "origin_question": "Which two objects are on the same side of you now? \"the rectangular picture with a vibrant red background\", \"the clock\", \"the brown wooden door with a rectangular window\". ", "option": [ "the rectangular picture with a vibrant red background, the clock", "the rectangular picture with a vibrant red background, the brown wooden door with a rectangular window", "the clock, the brown wooden door with a rectangular window" ], "answer": "the rectangular picture with a vibrant red background, the clock", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the rectangular picture with a vibrant red background\", \"the clock\", \"the brown wooden door with a rectangular window\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular picture with a vibrant red background, the clock', 'the rectangular picture with a vibrant red background, the brown wooden door with a rectangular window', 'the clock, the brown wooden door with a rectangular window'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0614_01/00480.jpg", "scene0614_01/00520.jpg", "scene0614_01/00560.jpg", "scene0614_01/00600.jpg" ], "turn_id": 4, "scan_id": "scene0614_01", "origin_question": "Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "111.6", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0614_01/00640.jpg", "scene0614_01/00680.jpg", "scene0614_01/00720.jpg", "scene0614_01/00760.jpg" ], "turn_id": 5, "scan_id": "scene0614_01", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the brown wooden door with a rectangular window\" from your position?", "option": [], "answer": "134.29", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the brown wooden door with a rectangular window\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0614_01/00800.jpg", "scene0614_01/00840.jpg", "scene0614_01/00880.jpg", "scene0614_01/00920.jpg" ], "turn_id": 6, "scan_id": "scene0614_01", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0040_01/00000.jpg", "scene0040_01/00040.jpg", "scene0040_01/00080.jpg", "scene0040_01/00120.jpg", "scene0040_01/00160.jpg" ], "turn_id": 1, "scan_id": "scene0040_01", "origin_question": "Remember, how many globe(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many globe(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0040_01/00200.jpg", "scene0040_01/00240.jpg", "scene0040_01/00280.jpg", "scene0040_01/00320.jpg", "scene0040_01/00360.jpg" ], "turn_id": 2, "scan_id": "scene0040_01", "origin_question": "At the end of which turn were both of \"the door\" and \"the small, purple, round toy\" on your left side?", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the door\" and \"the small, purple, round toy\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0040_01/00400.jpg", "scene0040_01/00440.jpg", "scene0040_01/00480.jpg", "scene0040_01/00520.jpg", "scene0040_01/00560.jpg" ], "turn_id": 3, "scan_id": "scene0040_01", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the black chair with a grey seat and armrests\"; \"the door\"; \"the rectangular black and white plastic switch\" ", "option": [ "the black chair with a grey seat and armrests", "the door", "the rectangular black and white plastic switch" ], "answer": "the black chair with a grey seat and armrests", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the black chair with a grey seat and armrests\"; \"the door\"; \"the rectangular black and white plastic switch\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the black chair with a grey seat and armrests', 'the door', 'the rectangular black and white plastic switch'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0040_01/00600.jpg", "scene0040_01/00640.jpg", "scene0040_01/00680.jpg", "scene0040_01/00720.jpg", "scene0040_01/00760.jpg" ], "turn_id": 4, "scan_id": "scene0040_01", "origin_question": "Which direction is the door in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the door in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0040_01/00800.jpg", "scene0040_01/00840.jpg", "scene0040_01/00880.jpg", "scene0040_01/00920.jpg", "scene0040_01/00960.jpg" ], "turn_id": 5, "scan_id": "scene0040_01", "origin_question": "By the end of the 4 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular black and white plastic switch\", Y: \"the black rectangular table with a modern design\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to Y and farther away from X", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 4 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular black and white plastic switch\", Y: \"the black rectangular table with a modern design\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0040_01/01000.jpg", "scene0040_01/01040.jpg", "scene0040_01/01080.jpg", "scene0040_01/01120.jpg", "scene0040_01/01160.jpg" ], "turn_id": 6, "scan_id": "scene0040_01", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the window\"; \"the black rectangular table with a modern design\"; \"Black medium-sized headphones with plastic and leather ear cups\" ", "option": [ "the window", "the black rectangular table with a modern design", "Black medium-sized headphones with plastic and leather ear cups" ], "answer": "the window", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the window\"; \"the black rectangular table with a modern design\"; \"Black medium-sized headphones with plastic and leather ear cups\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the window', 'the black rectangular table with a modern design', 'Black medium-sized headphones with plastic and leather ear cups'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0040_01/01200.jpg", "scene0040_01/01240.jpg", "scene0040_01/01280.jpg", "scene0040_01/01320.jpg", "scene0040_01/01360.jpg" ], "turn_id": 7, "scan_id": "scene0040_01", "origin_question": "In which turn were you closest to \"the black rectangular table with a modern design\"? ", "option": [], "answer": "7", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the black rectangular table with a modern design\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0040_01/01400.jpg", "scene0040_01/01440.jpg", "scene0040_01/01480.jpg", "scene0040_01/01520.jpg", "scene0040_01/01560.jpg" ], "turn_id": 8, "scan_id": "scene0040_01", "origin_question": "Is the door to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the door to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0040_01/01600.jpg", "scene0040_01/01640.jpg", "scene0040_01/01680.jpg", "scene0040_01/01720.jpg", "scene0040_01/01760.jpg" ], "turn_id": 9, "scan_id": "scene0040_01", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0040_01/01800.jpg", "scene0040_01/01840.jpg", "scene0040_01/01880.jpg", "scene0040_01/01920.jpg", "scene0040_01/01960.jpg" ], "turn_id": 10, "scan_id": "scene0040_01", "origin_question": "Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the rectangular black and white plastic switch\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the rectangular black and white plastic switch\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0287_00/00000.jpg", "scene0287_00/00030.jpg", "scene0287_00/00060.jpg", "scene0287_00/00090.jpg" ], "turn_id": 1, "scan_id": "scene0287_00", "origin_question": "Remember, have you seen any spoon(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any spoon(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0287_00/00120.jpg", "scene0287_00/00150.jpg", "scene0287_00/00180.jpg", "scene0287_00/00210.jpg" ], "turn_id": 2, "scan_id": "scene0287_00", "origin_question": "In which turn did you see both \"the light rectangular wooden cabinet with two doors\" and \"the horizontal, white metal radiator\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the light rectangular wooden cabinet with two doors\" and \"the horizontal, white metal radiator\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0287_00/00240.jpg", "scene0287_00/00270.jpg", "scene0287_00/00300.jpg", "scene0287_00/00330.jpg" ], "turn_id": 3, "scan_id": "scene0287_00", "origin_question": "At the end of which turn were both of \"the desk\" and \"the horizontal, white metal radiator\" on your left side?", "option": [], "answer": "2", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the desk\" and \"the horizontal, white metal radiator\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0287_00/00360.jpg", "scene0287_00/00390.jpg", "scene0287_00/00420.jpg", "scene0287_00/00450.jpg" ], "turn_id": 4, "scan_id": "scene0287_00", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "1.25", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0287_00/00480.jpg", "scene0287_00/00510.jpg", "scene0287_00/00540.jpg", "scene0287_00/00570.jpg" ], "turn_id": 5, "scan_id": "scene0287_00", "origin_question": "Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the desk\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the desk\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0287_00/00630.jpg", "scene0287_00/00660.jpg", "scene0287_00/00690.jpg", "scene0287_00/00720.jpg" ], "turn_id": 6, "scan_id": "scene0287_00", "origin_question": "Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0700_00/00000.jpg", "scene0700_00/00040.jpg", "scene0700_00/00080.jpg", "scene0700_00/00120.jpg", "scene0700_00/00160.jpg" ], "turn_id": 1, "scan_id": "scene0700_00", "origin_question": "Remember, how many clothes(s) have you seen so far? ", "option": [], "answer": "3", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many clothes(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0700_00/00200.jpg", "scene0700_00/00240.jpg", "scene0700_00/00280.jpg", "scene0700_00/00320.jpg", "scene0700_00/00360.jpg" ], "turn_id": 2, "scan_id": "scene0700_00", "origin_question": "Is the rectangular blue cardboard box to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the rectangular blue cardboard box to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0700_00/00400.jpg", "scene0700_00/00440.jpg", "scene0700_00/00480.jpg", "scene0700_00/00520.jpg", "scene0700_00/00560.jpg" ], "turn_id": 3, "scan_id": "scene0700_00", "origin_question": "Which direction is the rectangular black keyboard made of plastic in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the rectangular black keyboard made of plastic in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0700_00/00600.jpg", "scene0700_00/00640.jpg", "scene0700_00/00680.jpg", "scene0700_00/00720.jpg", "scene0700_00/00760.jpg" ], "turn_id": 4, "scan_id": "scene0700_00", "origin_question": "In which turn did you see both \"the door\" and \"the rectangular sponge with a green top and yellow bottom\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the door\" and \"the rectangular sponge with a green top and yellow bottom\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0700_00/00800.jpg", "scene0700_00/00840.jpg", "scene0700_00/00880.jpg", "scene0700_00/00920.jpg", "scene0700_00/00960.jpg" ], "turn_id": 5, "scan_id": "scene0700_00", "origin_question": "Is the small black and yellow plastic pen to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the small black and yellow plastic pen to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0700_00/01000.jpg", "scene0700_00/01040.jpg", "scene0700_00/01080.jpg", "scene0700_00/01120.jpg", "scene0700_00/01160.jpg" ], "turn_id": 6, "scan_id": "scene0700_00", "origin_question": "Which direction is the radiator in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the radiator in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0700_00/01200.jpg", "scene0700_00/01240.jpg", "scene0700_00/01280.jpg", "scene0700_00/01320.jpg", "scene0700_00/01360.jpg" ], "turn_id": 7, "scan_id": "scene0700_00", "origin_question": "Please recall \"the rectangular sponge with a green top and yellow bottom\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.12", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the rectangular sponge with a green top and yellow bottom\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0700_00/01400.jpg", "scene0700_00/01440.jpg", "scene0700_00/01480.jpg", "scene0700_00/01520.jpg", "scene0700_00/01560.jpg" ], "turn_id": 8, "scan_id": "scene0700_00", "origin_question": "Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the rectangular blue cardboard box\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the rectangular blue cardboard box\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "scene0700_00/01600.jpg", "scene0700_00/01640.jpg", "scene0700_00/01680.jpg", "scene0700_00/01720.jpg", "scene0700_00/01760.jpg" ], "turn_id": 9, "scan_id": "scene0700_00", "origin_question": "Among these three objects(\"the medium rectangular green plastic container\",\"the radiator\",\"the small black and yellow plastic pen\" ), which one is the farthest from you?", "option": [ "the medium rectangular green plastic container", "the radiator", "the small black and yellow plastic pen" ], "answer": "the radiator", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the medium rectangular green plastic container\",\"the radiator\",\"the small black and yellow plastic pen\" ), which one is the farthest from you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the medium rectangular green plastic container', 'the radiator', 'the small black and yellow plastic pen'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0700_00/01800.jpg", "scene0700_00/01840.jpg", "scene0700_00/01880.jpg", "scene0700_00/01920.jpg", "scene0700_00/01960.jpg" ], "turn_id": 10, "scan_id": "scene0700_00", "origin_question": "Which two objects are on the same side of you now? \"the rectangular metal shelf with a black base\", \"the radiator\", \"the rectangular black keyboard made of plastic\". ", "option": [ "the rectangular metal shelf with a black base, the radiator", "the rectangular metal shelf with a black base, the rectangular black keyboard made of plastic", "the radiator, the rectangular black keyboard made of plastic" ], "answer": "the radiator, the rectangular black keyboard made of plastic", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the rectangular metal shelf with a black base\", \"the radiator\", \"the rectangular black keyboard made of plastic\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular metal shelf with a black base, the radiator', 'the rectangular metal shelf with a black base, the rectangular black keyboard made of plastic', 'the radiator, the rectangular black keyboard made of plastic'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0057_region1/40aeaed616e04b9ea572e0fa9af1ff93_i1_1.jpg", "1mp3d_0057_region1/40aeaed616e04b9ea572e0fa9af1ff93_i2_1.jpg", "1mp3d_0057_region1/6f1036ee53464cad80aec93d2214617f_i1_0.jpg", "1mp3d_0057_region1/6f1036ee53464cad80aec93d2214617f_i2_0.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0057_region1", "origin_question": "Remember, have you seen any toilet(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any toilet(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0057_region1/40aeaed616e04b9ea572e0fa9af1ff93_i1_0.jpg", "1mp3d_0057_region1/40aeaed616e04b9ea572e0fa9af1ff93_i2_0.jpg", "1mp3d_0057_region1/616b7f836d8e4f04b50cab0e6d15efb6_i1_0.jpg", "1mp3d_0057_region1/616b7f836d8e4f04b50cab0e6d15efb6_i2_0.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0057_region1", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the cylindrical black metal bin\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the cylindrical black metal bin\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0057_region1/ff14bec47e17474dbdbfeccfe40e5d39_i1_5.jpg", "1mp3d_0057_region1/ff14bec47e17474dbdbfeccfe40e5d39_i2_5.jpg", "1mp3d_0057_region1/ff14bec47e17474dbdbfeccfe40e5d39_i1_0.jpg", "1mp3d_0057_region1/ff14bec47e17474dbdbfeccfe40e5d39_i2_0.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0057_region1", "origin_question": "", "option": [], "answer": "", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "1mp3d_0057_region1/616b7f836d8e4f04b50cab0e6d15efb6_i1_1.jpg", "1mp3d_0057_region1/616b7f836d8e4f04b50cab0e6d15efb6_i2_1.jpg", "1mp3d_0057_region1/0b107165285b48e5a2899b70cd478d92_i1_1.jpg", "1mp3d_0057_region1/0b107165285b48e5a2899b70cd478d92_i2_1.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0057_region1", "origin_question": "In which turn did you see both \"the cylindrical black metal bin\" and \"the small rectangular white holder\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the cylindrical black metal bin\" and \"the small rectangular white holder\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "1mp3d_0057_region1/0b107165285b48e5a2899b70cd478d92_i1_0.jpg", "1mp3d_0057_region1/0b107165285b48e5a2899b70cd478d92_i2_0.jpg", "1mp3d_0057_region1/616b7f836d8e4f04b50cab0e6d15efb6_i1_5.jpg", "1mp3d_0057_region1/616b7f836d8e4f04b50cab0e6d15efb6_i2_5.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0057_region1", "origin_question": "Which direction is the small rectangular white holder in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the small rectangular white holder in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "1mp3d_0057_region1/40aeaed616e04b9ea572e0fa9af1ff93_i1_4.jpg", "1mp3d_0057_region1/40aeaed616e04b9ea572e0fa9af1ff93_i2_4.jpg", "1mp3d_0057_region1/40aeaed616e04b9ea572e0fa9af1ff93_i1_3.jpg", "1mp3d_0057_region1/40aeaed616e04b9ea572e0fa9af1ff93_i2_3.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0057_region1", "origin_question": "Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0378_01/00000.jpg", "scene0378_01/00040.jpg", "scene0378_01/00080.jpg", "scene0378_01/00120.jpg", "scene0378_01/00160.jpg" ], "turn_id": 1, "scan_id": "scene0378_01", "origin_question": "Remember, have you seen any knife(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any knife(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0378_01/00200.jpg", "scene0378_01/00240.jpg", "scene0378_01/00280.jpg", "scene0378_01/00320.jpg", "scene0378_01/00360.jpg" ], "turn_id": 2, "scan_id": "scene0378_01", "origin_question": "In which turn were you closest to \"the cabinet\"? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the cabinet\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0378_01/00400.jpg", "scene0378_01/00440.jpg", "scene0378_01/00480.jpg", "scene0378_01/00520.jpg", "scene0378_01/00560.jpg" ], "turn_id": 3, "scan_id": "scene0378_01", "origin_question": "What will be the first-time appearance order of the following categories: blackboard, chair, cup?", "option": [ "blackboard, chair, cup", "blackboard, cup, chair", "chair, blackboard, cup", "chair, cup, blackboard" ], "answer": "chair, cup, blackboard", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: blackboard, chair, cup? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['blackboard, chair, cup', 'blackboard, cup, chair', 'chair, blackboard, cup', 'chair, cup, blackboard'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0378_01/00600.jpg", "scene0378_01/00640.jpg", "scene0378_01/00680.jpg", "scene0378_01/00720.jpg", "scene0378_01/00760.jpg" ], "turn_id": 4, "scan_id": "scene0378_01", "origin_question": "Which direction is the medium-sized rectangular blue plastic bin in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the medium-sized rectangular blue plastic bin in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0378_01/00800.jpg", "scene0378_01/00840.jpg", "scene0378_01/00880.jpg", "scene0378_01/00920.jpg", "scene0378_01/00960.jpg" ], "turn_id": 5, "scan_id": "scene0378_01", "origin_question": "Please recall \"the window\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.44", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the window\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0378_01/01000.jpg", "scene0378_01/01040.jpg", "scene0378_01/01080.jpg", "scene0378_01/01120.jpg", "scene0378_01/01160.jpg" ], "turn_id": 6, "scan_id": "scene0378_01", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the dark rectangular blackboard with writing and drawings\"; \"the blue shoe with white accents\"; \"the light yellow rectangular wooden door\" ", "option": [ "the dark rectangular blackboard with writing and drawings", "the blue shoe with white accents", "the light yellow rectangular wooden door" ], "answer": "the light yellow rectangular wooden door", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the dark rectangular blackboard with writing and drawings\"; \"the blue shoe with white accents\"; \"the light yellow rectangular wooden door\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the dark rectangular blackboard with writing and drawings', 'the blue shoe with white accents', 'the light yellow rectangular wooden door'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0378_01/01200.jpg", "scene0378_01/01240.jpg", "scene0378_01/01280.jpg", "scene0378_01/01320.jpg", "scene0378_01/01360.jpg" ], "turn_id": 7, "scan_id": "scene0378_01", "origin_question": "Which two objects are on the same side of you now? \"the medium-sized rectangular blue plastic bin\", \"the compact, light grey multimedia speaker\", \"the light yellow rectangular wooden door\". ", "option": [ "the medium-sized rectangular blue plastic bin, the compact, light grey multimedia speaker", "the medium-sized rectangular blue plastic bin, the light yellow rectangular wooden door", "the compact, light grey multimedia speaker, the light yellow rectangular wooden door" ], "answer": "the medium-sized rectangular blue plastic bin, the compact, light grey multimedia speaker", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the medium-sized rectangular blue plastic bin\", \"the compact, light grey multimedia speaker\", \"the light yellow rectangular wooden door\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the medium-sized rectangular blue plastic bin, the compact, light grey multimedia speaker', 'the medium-sized rectangular blue plastic bin, the light yellow rectangular wooden door', 'the compact, light grey multimedia speaker, the light yellow rectangular wooden door'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0378_01/01400.jpg", "scene0378_01/01440.jpg", "scene0378_01/01480.jpg", "scene0378_01/01520.jpg", "scene0378_01/01560.jpg" ], "turn_id": 8, "scan_id": "scene0378_01", "origin_question": "Compared to your position at the end of the 5 turn, are you now closer or farther away from \"the window\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 5 turn, are you now closer or farther away from \"the window\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0378_01/01600.jpg", "scene0378_01/01640.jpg", "scene0378_01/01680.jpg", "scene0378_01/01720.jpg", "scene0378_01/01760.jpg" ], "turn_id": 9, "scan_id": "scene0378_01", "origin_question": "Which direction is the light yellow rectangular wooden door in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the light yellow rectangular wooden door in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0378_01/01800.jpg", "scene0378_01/01840.jpg", "scene0378_01/01880.jpg", "scene0378_01/01920.jpg", "scene0378_01/01960.jpg" ], "turn_id": 10, "scan_id": "scene0378_01", "origin_question": "Which two objects are on the same side of you now? \"the window\", \"the medium-sized rectangular blue plastic bin\", \"the light yellow rectangular wooden door\". ", "option": [ "the window, the medium-sized rectangular blue plastic bin", "the window, the light yellow rectangular wooden door", "the medium-sized rectangular blue plastic bin, the light yellow rectangular wooden door" ], "answer": "the window, the medium-sized rectangular blue plastic bin", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the window\", \"the medium-sized rectangular blue plastic bin\", \"the light yellow rectangular wooden door\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the window, the medium-sized rectangular blue plastic bin', 'the window, the light yellow rectangular wooden door', 'the medium-sized rectangular blue plastic bin, the light yellow rectangular wooden door'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0702_00/00000.jpg", "scene0702_00/00030.jpg", "scene0702_00/00060.jpg", "scene0702_00/00090.jpg" ], "turn_id": 1, "scan_id": "scene0702_00", "origin_question": "Remember, have you seen any bathtub(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any bathtub(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0702_00/00120.jpg", "scene0702_00/00150.jpg", "scene0702_00/00180.jpg", "scene0702_00/00220.jpg" ], "turn_id": 2, "scan_id": "scene0702_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0702_00/00250.jpg", "scene0702_00/00280.jpg", "scene0702_00/00310.jpg", "scene0702_00/00340.jpg" ], "turn_id": 3, "scan_id": "scene0702_00", "origin_question": "Please recall \"the silver metal bar made of iron\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.36", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the silver metal bar made of iron\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0702_00/00370.jpg", "scene0702_00/00400.jpg", "scene0702_00/00430.jpg", "scene0702_00/00460.jpg" ], "turn_id": 4, "scan_id": "scene0702_00", "origin_question": "Which direction is the small rectangular white ceramic soap dish in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the small rectangular white ceramic soap dish in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0702_00/00490.jpg", "scene0702_00/00520.jpg", "scene0702_00/00550.jpg", "scene0702_00/00580.jpg" ], "turn_id": 5, "scan_id": "scene0702_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0702_00/00610.jpg", "scene0702_00/00640.jpg", "scene0702_00/00690.jpg", "scene0702_00/00720.jpg" ], "turn_id": 6, "scan_id": "scene0702_00", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the curtain\" from your position?", "option": [], "answer": "83.86", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the curtain\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0040_region30/21f3a12071494948a2b79c77a32a266d_i1_5.jpg", "1mp3d_0040_region30/21f3a12071494948a2b79c77a32a266d_i2_5.jpg", "1mp3d_0040_region30/8789cbb1b52b44038196d39d43465947_i1_5.jpg", "1mp3d_0040_region30/8789cbb1b52b44038196d39d43465947_i2_5.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0040_region30", "origin_question": "Remember, have you seen any panel(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any panel(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "1mp3d_0040_region30/084c9684a25a4c4f8cdea7afa21d6851_i1_5.jpg", "1mp3d_0040_region30/084c9684a25a4c4f8cdea7afa21d6851_i2_5.jpg", "1mp3d_0040_region30/8789cbb1b52b44038196d39d43465947_i2_0.jpg", "1mp3d_0040_region30/230d9b641a9d4b129304032061a90427_i1_4.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0040_region30", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the large potted plant with various colored flowers\"; \"the medium-sized double-handled deep purple vase\"; \"the round wooden table with a glass tabletop\" ", "option": [ "the large potted plant with various colored flowers", "the medium-sized double-handled deep purple vase", "the round wooden table with a glass tabletop" ], "answer": "the medium-sized double-handled deep purple vase", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the large potted plant with various colored flowers\"; \"the medium-sized double-handled deep purple vase\"; \"the round wooden table with a glass tabletop\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the large potted plant with various colored flowers', 'the medium-sized double-handled deep purple vase', 'the round wooden table with a glass tabletop'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "1mp3d_0040_region30/230d9b641a9d4b129304032061a90427_i2_4.jpg", "1mp3d_0040_region30/230d9b641a9d4b129304032061a90427_i1_3.jpg", "1mp3d_0040_region30/230d9b641a9d4b129304032061a90427_i2_3.jpg", "1mp3d_0040_region30/ad048e19d6cc41d2b559988b26aea08a_i1_4.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0040_region30", "origin_question": "What will be the first-time appearance order of the following categories: chair, plant, vase?", "option": [ "chair, vase, plant", "vase, plant, chair", "plant, vase, chair", "vase, chair, plant" ], "answer": "plant, vase, chair", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: chair, plant, vase? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['chair, vase, plant', 'vase, plant, chair', 'plant, vase, chair', 'vase, chair, plant'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "1mp3d_0040_region30/ad048e19d6cc41d2b559988b26aea08a_i2_4.jpg", "1mp3d_0040_region30/86d522ead0d0483f94fb6fa567ddc997_i1_5.jpg", "1mp3d_0040_region30/86d522ead0d0483f94fb6fa567ddc997_i2_5.jpg", "1mp3d_0040_region30/86d522ead0d0483f94fb6fa567ddc997_i1_4.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0040_region30", "origin_question": "What will be the first-time appearance order of the following categories: chandelier, plant, vase?", "option": [ "chandelier, vase, plant", "plant, vase, chandelier", "plant, chandelier, vase", "chandelier, plant, vase" ], "answer": "plant, vase, chandelier", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: chandelier, plant, vase? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['chandelier, vase, plant', 'plant, vase, chandelier', 'plant, chandelier, vase', 'chandelier, plant, vase'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "1mp3d_0040_region30/86d522ead0d0483f94fb6fa567ddc997_i2_4.jpg", "1mp3d_0040_region30/86d522ead0d0483f94fb6fa567ddc997_i1_3.jpg", "1mp3d_0040_region30/86d522ead0d0483f94fb6fa567ddc997_i2_3.jpg", "1mp3d_0040_region30/86d522ead0d0483f94fb6fa567ddc997_i1_2.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0040_region30", "origin_question": "When did you first discover \"the rectangular brown wooden door with a metal grille\"? (index of the turn)", "option": [], "answer": "3", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the rectangular brown wooden door with a metal grille\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "1mp3d_0040_region30/86d522ead0d0483f94fb6fa567ddc997_i2_2.jpg", "1mp3d_0040_region30/86d522ead0d0483f94fb6fa567ddc997_i1_1.jpg", "1mp3d_0040_region30/86d522ead0d0483f94fb6fa567ddc997_i2_1.jpg", "1mp3d_0040_region30/c948e0adde424b22af74a18d3cf14ee6_i1_2.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0040_region30", "origin_question": "By the end of the 4 turn, were you getting closer to or farther away from these two objects: X: \"the round wooden table with a glass tabletop\", Y: \"the rack\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "farther away from both X and Y", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 4 turn, were you getting closer to or farther away from these two objects: X: \"the round wooden table with a glass tabletop\", Y: \"the rack\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "1mp3d_0040_region30/c948e0adde424b22af74a18d3cf14ee6_i2_2.jpg", "1mp3d_0040_region30/a8f4d86fe29043e3b6eee87bd8612432_i1_4.jpg", "1mp3d_0040_region30/a8f4d86fe29043e3b6eee87bd8612432_i2_4.jpg", "1mp3d_0040_region30/a8f4d86fe29043e3b6eee87bd8612432_i1_5.jpg" ], "turn_id": 7, "scan_id": "1mp3d_0040_region30", "origin_question": "Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance forward or backward from that position?", "option": [ "Forward", "Backward" ], "answer": "Forward", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance forward or backward from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Forward', 'Backward'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0040_region30/a8f4d86fe29043e3b6eee87bd8612432_i2_5.jpg", "1mp3d_0040_region30/e211c7a914cc4880a869fd9a4de997e1_i1_4.jpg", "1mp3d_0040_region30/e211c7a914cc4880a869fd9a4de997e1_i2_4.jpg", "1mp3d_0040_region30/14f0a9cef49a4c79b9db3a87121e2aeb_i1_5.jpg" ], "turn_id": 8, "scan_id": "1mp3d_0040_region30", "origin_question": "Compared to your position at the end of the 6 turn, are you now closer or farther away from \"the round wooden table with a glass tabletop\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 6 turn, are you now closer or farther away from \"the round wooden table with a glass tabletop\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0609_02/00000.jpg", "scene0609_02/00040.jpg", "scene0609_02/00080.jpg", "scene0609_02/00120.jpg", "scene0609_02/00160.jpg" ], "turn_id": 1, "scan_id": "scene0609_02", "origin_question": "Remember, how many chair(s) have you seen so far? ", "option": [], "answer": "7", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many chair(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0609_02/00200.jpg", "scene0609_02/00240.jpg", "scene0609_02/00280.jpg", "scene0609_02/00320.jpg", "scene0609_02/00360.jpg" ], "turn_id": 2, "scan_id": "scene0609_02", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0609_02/00400.jpg", "scene0609_02/00440.jpg", "scene0609_02/00480.jpg", "scene0609_02/00520.jpg", "scene0609_02/00560.jpg" ], "turn_id": 3, "scan_id": "scene0609_02", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "69.67", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "scene0609_02/00600.jpg", "scene0609_02/00640.jpg", "scene0609_02/00680.jpg", "scene0609_02/00720.jpg", "scene0609_02/00760.jpg" ], "turn_id": 4, "scan_id": "scene0609_02", "origin_question": "Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Left", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0609_02/00800.jpg", "scene0609_02/00840.jpg", "scene0609_02/00880.jpg", "scene0609_02/00920.jpg", "scene0609_02/00960.jpg" ], "turn_id": 5, "scan_id": "scene0609_02", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "2.78", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0609_02/01000.jpg", "scene0609_02/01040.jpg", "scene0609_02/01080.jpg", "scene0609_02/01120.jpg", "scene0609_02/01160.jpg" ], "turn_id": 6, "scan_id": "scene0609_02", "origin_question": "Using your orientation at the end of turn 4 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "90.73", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 4 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0609_02/01200.jpg", "scene0609_02/01240.jpg", "scene0609_02/01280.jpg", "scene0609_02/01320.jpg", "scene0609_02/01360.jpg" ], "turn_id": 7, "scan_id": "scene0609_02", "origin_question": "Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0609_02/01400.jpg", "scene0609_02/01440.jpg", "scene0609_02/01480.jpg", "scene0609_02/01520.jpg", "scene0609_02/01560.jpg" ], "turn_id": 8, "scan_id": "scene0609_02", "origin_question": "Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the black cylindrical stool made of plastic or metal\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the black cylindrical stool made of plastic or metal\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0017_00/00000.jpg", "scene0017_00/00040.jpg", "scene0017_00/00080.jpg", "scene0017_00/00120.jpg" ], "turn_id": 1, "scan_id": "scene0017_00", "origin_question": "Remember, how many chair(s) have you seen so far? ", "option": [], "answer": "3", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many chair(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0017_00/00160.jpg", "scene0017_00/00200.jpg", "scene0017_00/00240.jpg", "scene0017_00/00280.jpg" ], "turn_id": 2, "scan_id": "scene0017_00", "origin_question": "Which direction is the rectangular white window with a glass frame in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the rectangular white window with a glass frame in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0017_00/00320.jpg", "scene0017_00/00360.jpg", "scene0017_00/00400.jpg", "scene0017_00/00440.jpg" ], "turn_id": 3, "scan_id": "scene0017_00", "origin_question": "Is the rectangular dark-colored container made of plastic or metal to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the rectangular dark-colored container made of plastic or metal to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0017_00/00480.jpg", "scene0017_00/00520.jpg", "scene0017_00/00560.jpg", "scene0017_00/00600.jpg" ], "turn_id": 4, "scan_id": "scene0017_00", "origin_question": "Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0017_00/00640.jpg", "scene0017_00/00680.jpg", "scene0017_00/00720.jpg", "scene0017_00/00760.jpg" ], "turn_id": 5, "scan_id": "scene0017_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0017_00/00800.jpg", "scene0017_00/00840.jpg", "scene0017_00/00880.jpg", "scene0017_00/00920.jpg" ], "turn_id": 6, "scan_id": "scene0017_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0017_00/00960.jpg", "scene0017_00/01000.jpg", "scene0017_00/01040.jpg", "scene0017_00/01080.jpg" ], "turn_id": 7, "scan_id": "scene0017_00", "origin_question": "What will be the first-time appearance order of the following categories: bin, picture, plant?", "option": [ "picture, bin, plant", "plant, bin, picture", "plant, picture, bin", "bin, picture, plant" ], "answer": "plant, picture, bin", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bin, picture, plant? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['picture, bin, plant', 'plant, bin, picture', 'plant, picture, bin', 'bin, picture, plant'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0017_00/01120.jpg", "scene0017_00/01160.jpg", "scene0017_00/01200.jpg", "scene0017_00/01240.jpg" ], "turn_id": 8, "scan_id": "scene0017_00", "origin_question": "In which turn did you see both \"the medium-sized black circular plastic fan\" and \"the bright blue rectangular wooden door\" simultaneously? ", "option": [], "answer": "5", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the medium-sized black circular plastic fan\" and \"the bright blue rectangular wooden door\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0040_region44/20ebc6f329244c5aad279fd6aacff577_i1_2.jpg", "1mp3d_0040_region44/20ebc6f329244c5aad279fd6aacff577_i2_2.jpg", "1mp3d_0040_region44/0bda68aefc4347a1bf992849bd7a616a_i1_3.jpg", "1mp3d_0040_region44/0bda68aefc4347a1bf992849bd7a616a_i2_3.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0040_region44", "origin_question": "Remember, have you seen any folder(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any folder(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 2 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "1mp3d_0040_region44/0bda68aefc4347a1bf992849bd7a616a_i1_4.jpg", "1mp3d_0040_region44/0bda68aefc4347a1bf992849bd7a616a_i2_4.jpg", "1mp3d_0040_region44/0bda68aefc4347a1bf992849bd7a616a_i1_0.jpg", "1mp3d_0040_region44/0bda68aefc4347a1bf992849bd7a616a_i2_0.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0040_region44", "origin_question": "Is the rectangular mirror with a decorative metallic frame to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the rectangular mirror with a decorative metallic frame to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 2 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0044_region18/94a3a6a06fff4322a07a98f5afffcefd_i1_2.jpg", "1mp3d_0044_region18/94a3a6a06fff4322a07a98f5afffcefd_i2_2.jpg", "1mp3d_0044_region18/b112e4279c88446ea024fb39a53439c9_i1_2.jpg", "1mp3d_0044_region18/b112e4279c88446ea024fb39a53439c9_i2_2.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0044_region18", "origin_question": "Remember, have you seen any microwave(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any microwave(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "1mp3d_0044_region18/6c6a1962787d4f429c7e0134404c5ccc_i1_0.jpg", "1mp3d_0044_region18/6c6a1962787d4f429c7e0134404c5ccc_i2_0.jpg", "1mp3d_0044_region18/6c6a1962787d4f429c7e0134404c5ccc_i1_5.jpg", "1mp3d_0044_region18/6c6a1962787d4f429c7e0134404c5ccc_i2_5.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0044_region18", "origin_question": "In which turn were you closest to \"the rectangular light blue plastic switch\"? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the rectangular light blue plastic switch\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "1mp3d_0044_region18/6c6a1962787d4f429c7e0134404c5ccc_i1_4.jpg", "1mp3d_0044_region18/b112e4279c88446ea024fb39a53439c9_i1_1.jpg", "1mp3d_0044_region18/94a3a6a06fff4322a07a98f5afffcefd_i1_1.jpg", "1mp3d_0044_region18/94a3a6a06fff4322a07a98f5afffcefd_i2_1.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0044_region18", "origin_question": "When did you first discover \"the small, white rounded fireplace with a smooth finish\"? (index of the turn)", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the small, white rounded fireplace with a smooth finish\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "1mp3d_0044_region18/5d5f0e1444064e2096d63884b975d173_i1_1.jpg", "1mp3d_0044_region18/5d5f0e1444064e2096d63884b975d173_i2_1.jpg", "1mp3d_0044_region18/50529218991d46c8ba8f6acbc0732ceb_i1_2.jpg", "1mp3d_0044_region18/50529218991d46c8ba8f6acbc0732ceb_i2_2.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0044_region18", "origin_question": "Which direction is the brown wooden and metal fan in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the brown wooden and metal fan in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "1mp3d_0044_region18/326a10e701e5484ca3c73bf6b142ebfe_i1_5.jpg", "1mp3d_0044_region18/326a10e701e5484ca3c73bf6b142ebfe_i2_5.jpg", "1mp3d_0044_region18/326a10e701e5484ca3c73bf6b142ebfe_i1_4.jpg", "1mp3d_0044_region18/326a10e701e5484ca3c73bf6b142ebfe_i2_4.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0044_region18", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0361_00/00000.jpg", "scene0361_00/00040.jpg", "scene0361_00/00080.jpg", "scene0361_00/00120.jpg" ], "turn_id": 1, "scan_id": "scene0361_00", "origin_question": "Remember, how many chair(s) have you seen so far? ", "option": [], "answer": "2", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many chair(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0361_00/00160.jpg", "scene0361_00/00200.jpg", "scene0361_00/00240.jpg", "scene0361_00/00280.jpg" ], "turn_id": 2, "scan_id": "scene0361_00", "origin_question": "Remember, how many cart(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many cart(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0361_00/00320.jpg", "scene0361_00/00360.jpg", "scene0361_00/00400.jpg", "scene0361_00/00440.jpg" ], "turn_id": 3, "scan_id": "scene0361_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0361_00/00480.jpg", "scene0361_00/00520.jpg", "scene0361_00/00560.jpg", "scene0361_00/00600.jpg" ], "turn_id": 4, "scan_id": "scene0361_00", "origin_question": "Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "54.02", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0361_00/00640.jpg", "scene0361_00/00680.jpg", "scene0361_00/00720.jpg", "scene0361_00/00760.jpg" ], "turn_id": 5, "scan_id": "scene0361_00", "origin_question": "Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "62.15", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0361_00/00800.jpg", "scene0361_00/00840.jpg", "scene0361_00/00880.jpg", "scene0361_00/00920.jpg" ], "turn_id": 6, "scan_id": "scene0361_00", "origin_question": "Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0361_00/00960.jpg", "scene0361_00/01000.jpg", "scene0361_00/01040.jpg", "scene0361_00/01080.jpg" ], "turn_id": 7, "scan_id": "scene0361_00", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0067_region11/7eb32b1e57c644f1b5a5801dbc6401cc_i1_2.jpg", "1mp3d_0067_region11/7eb32b1e57c644f1b5a5801dbc6401cc_i2_2.jpg", "1mp3d_0067_region11/7eb32b1e57c644f1b5a5801dbc6401cc_i1_3.jpg", "1mp3d_0067_region11/7eb32b1e57c644f1b5a5801dbc6401cc_i2_3.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0067_region11", "origin_question": "Remember, have you seen any sink(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any sink(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "1mp3d_0067_region11/9b4cb016c2ae494f91549fef70da01da_i1_5.jpg", "1mp3d_0067_region11/9b4cb016c2ae494f91549fef70da01da_i2_5.jpg", "1mp3d_0067_region11/7eb32b1e57c644f1b5a5801dbc6401cc_i1_4.jpg", "1mp3d_0067_region11/7eb32b1e57c644f1b5a5801dbc6401cc_i2_4.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0067_region11", "origin_question": "In which turn did you see both \"the white ceramic standalone toilet with a closed lid\" and \"the small rectangular wooden tissue box with a natural finish\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the white ceramic standalone toilet with a closed lid\" and \"the small rectangular wooden tissue box with a natural finish\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "1mp3d_0067_region11/7eb32b1e57c644f1b5a5801dbc6401cc_i1_5.jpg", "1mp3d_0067_region11/7eb32b1e57c644f1b5a5801dbc6401cc_i2_5.jpg", "1mp3d_0067_region11/9b4cb016c2ae494f91549fef70da01da_i1_1.jpg", "1mp3d_0067_region11/9b4cb016c2ae494f91549fef70da01da_i2_1.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0067_region11", "origin_question": "Which direction is the large, round gray sink made of smooth material in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the large, round gray sink made of smooth material in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "1mp3d_0067_region11/9b4cb016c2ae494f91549fef70da01da_i1_0.jpg", "1mp3d_0067_region11/9b4cb016c2ae494f91549fef70da01da_i2_0.jpg", "1mp3d_0067_region11/982920829a0b433880410222539f240e_i1_0.jpg", "1mp3d_0067_region11/982920829a0b433880410222539f240e_i2_0.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0067_region11", "origin_question": "By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the large rectangular wooden cabinet with closed doors\", Y: \"the large, round gray sink made of smooth material\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the large rectangular wooden cabinet with closed doors\", Y: \"the large, round gray sink made of smooth material\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "1mp3d_0067_region11/4ff62efbc0934e888120522e4c84e712_i1_2.jpg", "1mp3d_0067_region11/4ff62efbc0934e888120522e4c84e712_i2_2.jpg", "1mp3d_0067_region11/4ff62efbc0934e888120522e4c84e712_i1_1.jpg", "1mp3d_0067_region11/4ff62efbc0934e888120522e4c84e712_i2_1.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0067_region11", "origin_question": "Which direction is the small rectangular wooden tissue box with a natural finish in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the small rectangular wooden tissue box with a natural finish in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0067_region11/982920829a0b433880410222539f240e_i1_5.jpg", "1mp3d_0067_region11/982920829a0b433880410222539f240e_i2_5.jpg", "1mp3d_0067_region11/4ff62efbc0934e888120522e4c84e712_i1_3.jpg", "1mp3d_0067_region11/4ff62efbc0934e888120522e4c84e712_i2_3.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0067_region11", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the small cylindrical light brown wooden bin\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the small cylindrical light brown wooden bin\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0095_00/00000.jpg", "scene0095_00/00040.jpg", "scene0095_00/00080.jpg", "scene0095_00/00120.jpg", "scene0095_00/00160.jpg" ], "turn_id": 1, "scan_id": "scene0095_00", "origin_question": "Remember, how many keyboard(s) have you seen so far? ", "option": [], "answer": "3", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many keyboard(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0095_00/00200.jpg", "scene0095_00/00240.jpg", "scene0095_00/00280.jpg", "scene0095_00/00320.jpg", "scene0095_00/00360.jpg" ], "turn_id": 2, "scan_id": "scene0095_00", "origin_question": "When did you first discover \"the small rectangular dark blackboard with a smooth surface\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the small rectangular dark blackboard with a smooth surface\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0095_00/00400.jpg", "scene0095_00/00440.jpg", "scene0095_00/00480.jpg", "scene0095_00/00550.jpg", "scene0095_00/00590.jpg" ], "turn_id": 3, "scan_id": "scene0095_00", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "2.21", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0095_00/00630.jpg", "scene0095_00/00670.jpg", "scene0095_00/00710.jpg", "scene0095_00/00750.jpg", "scene0095_00/00790.jpg" ], "turn_id": 4, "scan_id": "scene0095_00", "origin_question": "Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the medium-sized ergonomic black mouse\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the medium-sized ergonomic black mouse\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0095_00/00830.jpg", "scene0095_00/00870.jpg", "scene0095_00/00910.jpg", "scene0095_00/00950.jpg", "scene0095_00/00990.jpg" ], "turn_id": 5, "scan_id": "scene0095_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0095_00/01030.jpg", "scene0095_00/01070.jpg", "scene0095_00/01110.jpg", "scene0095_00/01150.jpg", "scene0095_00/01190.jpg" ], "turn_id": 6, "scan_id": "scene0095_00", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0095_00/01230.jpg", "scene0095_00/01270.jpg", "scene0095_00/01310.jpg", "scene0095_00/01350.jpg", "scene0095_00/01390.jpg" ], "turn_id": 7, "scan_id": "scene0095_00", "origin_question": "What will be the first-time appearance order of the following categories: blackboard, cabinet, radiator?", "option": [ "radiator, cabinet, blackboard", "radiator, blackboard, cabinet", "blackboard, cabinet, radiator", "cabinet, radiator, blackboard" ], "answer": "radiator, cabinet, blackboard", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: blackboard, cabinet, radiator? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['radiator, cabinet, blackboard', 'radiator, blackboard, cabinet', 'blackboard, cabinet, radiator', 'cabinet, radiator, blackboard'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0095_00/01430.jpg", "scene0095_00/01470.jpg", "scene0095_00/01510.jpg", "scene0095_00/01550.jpg", "scene0095_00/01590.jpg" ], "turn_id": 8, "scan_id": "scene0095_00", "origin_question": "Remember, have you seen any grab bar(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any grab bar(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0067_region2/e74b6fff96c74d018e07e1452bafa49d_i1_3.jpg", "1mp3d_0067_region2/e74b6fff96c74d018e07e1452bafa49d_i2_3.jpg", "1mp3d_0067_region2/cd249ef1708b416e8523ec3e332c0753_i2_1.jpg", "1mp3d_0067_region2/e74b6fff96c74d018e07e1452bafa49d_i1_2.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0067_region2", "origin_question": "Remember, have you seen any printer(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any printer(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0067_region2/e74b6fff96c74d018e07e1452bafa49d_i2_2.jpg", "1mp3d_0067_region2/cd249ef1708b416e8523ec3e332c0753_i1_2.jpg", "1mp3d_0067_region2/cd249ef1708b416e8523ec3e332c0753_i2_2.jpg", "1mp3d_0067_region2/cd249ef1708b416e8523ec3e332c0753_i1_3.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0067_region2", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "1mp3d_0067_region2/cd249ef1708b416e8523ec3e332c0753_i2_3.jpg", "1mp3d_0067_region2/cd249ef1708b416e8523ec3e332c0753_i1_4.jpg", "1mp3d_0067_region2/cd249ef1708b416e8523ec3e332c0753_i2_4.jpg", "1mp3d_0067_region2/cd249ef1708b416e8523ec3e332c0753_i1_5.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0067_region2", "origin_question": "Which two objects are on the same side of you now? \"the rectangular brown basket made of light material\", \"the single roll of white cylindrical toilet paper\", \"the rectangular window with a deep brown frame\". ", "option": [ "the rectangular brown basket made of light material, the single roll of white cylindrical toilet paper", "the rectangular brown basket made of light material, the rectangular window with a deep brown frame", "the single roll of white cylindrical toilet paper, the rectangular window with a deep brown frame" ], "answer": "the single roll of white cylindrical toilet paper, the rectangular window with a deep brown frame", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the rectangular brown basket made of light material\", \"the single roll of white cylindrical toilet paper\", \"the rectangular window with a deep brown frame\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular brown basket made of light material, the single roll of white cylindrical toilet paper', 'the rectangular brown basket made of light material, the rectangular window with a deep brown frame', 'the single roll of white cylindrical toilet paper, the rectangular window with a deep brown frame'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "1mp3d_0067_region2/cd249ef1708b416e8523ec3e332c0753_i2_5.jpg", "1mp3d_0067_region2/e74b6fff96c74d018e07e1452bafa49d_i1_0.jpg", "1mp3d_0067_region2/e74b6fff96c74d018e07e1452bafa49d_i2_0.jpg", "1mp3d_0067_region2/2291e23cde8e4eaebe6b9fe9cbdeb888_i1_5.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0067_region2", "origin_question": "Which direction is the standard white toilet with a closed lid in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the standard white toilet with a closed lid in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "1mp3d_0067_region2/2291e23cde8e4eaebe6b9fe9cbdeb888_i2_5.jpg", "1mp3d_0067_region2/2291e23cde8e4eaebe6b9fe9cbdeb888_i1_4.jpg", "1mp3d_0067_region2/2291e23cde8e4eaebe6b9fe9cbdeb888_i2_4.jpg", "1mp3d_0067_region2/2291e23cde8e4eaebe6b9fe9cbdeb888_i1_3.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0067_region2", "origin_question": "Is the rectangular beige wooden bin to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the rectangular beige wooden bin to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0589_01/00000.jpg", "scene0589_01/00040.jpg", "scene0589_01/00080.jpg", "scene0589_01/00120.jpg", "scene0589_01/00160.jpg" ], "turn_id": 1, "scan_id": "scene0589_01", "origin_question": "Remember, have you seen any case(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any case(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "scene0589_01/00200.jpg", "scene0589_01/00240.jpg", "scene0589_01/00280.jpg", "scene0589_01/00320.jpg", "scene0589_01/00360.jpg" ], "turn_id": 2, "scan_id": "scene0589_01", "origin_question": "Among these three objects(\"the cylindrical metal or plastic light emitting yellow glow\",\"the modern, conical yellow or gold metal lamp\",\"the rectangular wooden stand in warm brown\" ), which one is the closest to you?", "option": [ "the cylindrical metal or plastic light emitting yellow glow", "the modern, conical yellow or gold metal lamp", "the rectangular wooden stand in warm brown" ], "answer": "the cylindrical metal or plastic light emitting yellow glow", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the cylindrical metal or plastic light emitting yellow glow\",\"the modern, conical yellow or gold metal lamp\",\"the rectangular wooden stand in warm brown\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the cylindrical metal or plastic light emitting yellow glow', 'the modern, conical yellow or gold metal lamp', 'the rectangular wooden stand in warm brown'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0589_01/00400.jpg", "scene0589_01/00440.jpg", "scene0589_01/00480.jpg", "scene0589_01/00520.jpg", "scene0589_01/00560.jpg" ], "turn_id": 3, "scan_id": "scene0589_01", "origin_question": "In which turn did you see both \"the curtain\" and \"the blue medium-sized folding chair made of metal and fabric\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the curtain\" and \"the blue medium-sized folding chair made of metal and fabric\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0589_01/00600.jpg", "scene0589_01/00640.jpg", "scene0589_01/00680.jpg", "scene0589_01/00720.jpg", "scene0589_01/00760.jpg" ], "turn_id": 4, "scan_id": "scene0589_01", "origin_question": "Is the cylindrical metal or plastic light emitting yellow glow to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the cylindrical metal or plastic light emitting yellow glow to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0589_01/00800.jpg", "scene0589_01/00840.jpg", "scene0589_01/00880.jpg", "scene0589_01/00920.jpg", "scene0589_01/00960.jpg" ], "turn_id": 5, "scan_id": "scene0589_01", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the slender, red guitar made of wood\" from your position?", "option": [], "answer": "30.51", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the slender, red guitar made of wood\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0589_01/01000.jpg", "scene0589_01/01040.jpg", "scene0589_01/01080.jpg", "scene0589_01/01120.jpg", "scene0589_01/01160.jpg" ], "turn_id": 6, "scan_id": "scene0589_01", "origin_question": "Please recall \"the simple rectangular wooden door with a smooth surface\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.58", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the simple rectangular wooden door with a smooth surface\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "scene0589_01/01200.jpg", "scene0589_01/01240.jpg", "scene0589_01/01280.jpg", "scene0589_01/01320.jpg", "scene0589_01/01360.jpg" ], "turn_id": 7, "scan_id": "scene0589_01", "origin_question": "Among these three objects(\"the modern, conical yellow or gold metal lamp\",\"the white vertical metal radiator with horizontal partitions\",\"the simple rectangular wooden door with a smooth surface\" ), which one is the closest to you?", "option": [ "the modern, conical yellow or gold metal lamp", "the white vertical metal radiator with horizontal partitions", "the simple rectangular wooden door with a smooth surface" ], "answer": "the simple rectangular wooden door with a smooth surface", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the modern, conical yellow or gold metal lamp\",\"the white vertical metal radiator with horizontal partitions\",\"the simple rectangular wooden door with a smooth surface\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the modern, conical yellow or gold metal lamp', 'the white vertical metal radiator with horizontal partitions', 'the simple rectangular wooden door with a smooth surface'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0589_01/01400.jpg", "scene0589_01/01440.jpg", "scene0589_01/01480.jpg", "scene0589_01/01520.jpg", "scene0589_01/01560.jpg" ], "turn_id": 8, "scan_id": "scene0589_01", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the modern, conical yellow or gold metal lamp\" from your position?", "option": [], "answer": "147.73", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the modern, conical yellow or gold metal lamp\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0696_01/00000.jpg", "scene0696_01/00040.jpg", "scene0696_01/00080.jpg", "scene0696_01/00120.jpg", "scene0696_01/00160.jpg" ], "turn_id": 1, "scan_id": "scene0696_01", "origin_question": "Remember, how many chair(s) have you seen so far? ", "option": [], "answer": "4", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many chair(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0696_01/00200.jpg", "scene0696_01/00240.jpg", "scene0696_01/00280.jpg", "scene0696_01/00320.jpg", "scene0696_01/00360.jpg" ], "turn_id": 2, "scan_id": "scene0696_01", "origin_question": "In which turn were you closest to \"the radiator\"? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the radiator\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0696_01/00400.jpg", "scene0696_01/00440.jpg", "scene0696_01/00480.jpg", "scene0696_01/00520.jpg", "scene0696_01/00560.jpg" ], "turn_id": 3, "scan_id": "scene0696_01", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the radiator\" from your position?", "option": [], "answer": "147.08", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the radiator\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0696_01/00600.jpg", "scene0696_01/00640.jpg", "scene0696_01/00680.jpg", "scene0696_01/00720.jpg", "scene0696_01/00760.jpg" ], "turn_id": 4, "scan_id": "scene0696_01", "origin_question": "Which two objects are on the same side of you now? \"the door\", \"the rectangular, yellow cotton-polyester couch\", \"the radiator\". ", "option": [ "the door, the rectangular, yellow cotton-polyester couch", "the door, the radiator", "the rectangular, yellow cotton-polyester couch, the radiator" ], "answer": "the rectangular, yellow cotton-polyester couch, the radiator", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the door\", \"the rectangular, yellow cotton-polyester couch\", \"the radiator\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the door, the rectangular, yellow cotton-polyester couch', 'the door, the radiator', 'the rectangular, yellow cotton-polyester couch, the radiator'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0696_01/00800.jpg", "scene0696_01/00840.jpg", "scene0696_01/00880.jpg", "scene0696_01/00920.jpg", "scene0696_01/00960.jpg" ], "turn_id": 5, "scan_id": "scene0696_01", "origin_question": "Which direction is the vintage black rectangular speaker with circular knobs in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the vintage black rectangular speaker with circular knobs in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0696_01/01000.jpg", "scene0696_01/01040.jpg", "scene0696_01/01080.jpg", "scene0696_01/01120.jpg", "scene0696_01/01160.jpg" ], "turn_id": 6, "scan_id": "scene0696_01", "origin_question": "Is the door to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the door to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "scene0696_01/01200.jpg", "scene0696_01/01240.jpg", "scene0696_01/01280.jpg", "scene0696_01/01320.jpg", "scene0696_01/01360.jpg" ], "turn_id": 7, "scan_id": "scene0696_01", "origin_question": "Among these three objects(\"the radiator\",\"the medium cylindrical glass pot with a dome lid\",\"the vintage black rectangular speaker with circular knobs\" ), which one is the farthest from you?", "option": [ "the radiator", "the medium cylindrical glass pot with a dome lid", "the vintage black rectangular speaker with circular knobs" ], "answer": "the radiator", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the radiator\",\"the medium cylindrical glass pot with a dome lid\",\"the vintage black rectangular speaker with circular knobs\" ), which one is the farthest from you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the radiator', 'the medium cylindrical glass pot with a dome lid', 'the vintage black rectangular speaker with circular knobs'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0696_01/01400.jpg", "scene0696_01/01440.jpg", "scene0696_01/01480.jpg", "scene0696_01/01520.jpg", "scene0696_01/01560.jpg" ], "turn_id": 8, "scan_id": "scene0696_01", "origin_question": "Please recall \"the medium cylindrical glass pot with a dome lid\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "3.98", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the medium cylindrical glass pot with a dome lid\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "scene0696_01/01600.jpg", "scene0696_01/01640.jpg", "scene0696_01/01680.jpg", "scene0696_01/01720.jpg", "scene0696_01/01760.jpg" ], "turn_id": 9, "scan_id": "scene0696_01", "origin_question": "Assuming the direction you are facing at the end of the turn 7 is forward, did you move a certain distance forward or backward from that position?", "option": [ "Forward", "Backward" ], "answer": "Forward", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 7 is forward, did you move a certain distance forward or backward from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Forward', 'Backward'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0692_03/00000.jpg", "scene0692_03/00040.jpg", "scene0692_03/00080.jpg", "scene0692_03/00120.jpg" ], "turn_id": 1, "scan_id": "scene0692_03", "origin_question": "Remember, have you seen any shelf(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any shelf(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0692_03/00160.jpg", "scene0692_03/00200.jpg", "scene0692_03/00240.jpg", "scene0692_03/00280.jpg" ], "turn_id": 2, "scan_id": "scene0692_03", "origin_question": "Remember, how many chair(s) have you seen so far? ", "option": [], "answer": "3", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many chair(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0692_03/00320.jpg", "scene0692_03/00360.jpg", "scene0692_03/00400.jpg", "scene0692_03/00440.jpg" ], "turn_id": 3, "scan_id": "scene0692_03", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the door\", Y: \"the round light-colored wooden table with a smooth surface\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the door\", Y: \"the round light-colored wooden table with a smooth surface\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0692_03/00480.jpg", "scene0692_03/00520.jpg", "scene0692_03/00560.jpg", "scene0692_03/00600.jpg" ], "turn_id": 4, "scan_id": "scene0692_03", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the round light-colored wooden table with a smooth surface\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the round light-colored wooden table with a smooth surface\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0692_03/00640.jpg", "scene0692_03/00680.jpg", "scene0692_03/00720.jpg", "scene0692_03/00760.jpg" ], "turn_id": 5, "scan_id": "scene0692_03", "origin_question": "Which two objects are on the same side of you now? \"the door\", \"the cylindrical black bin made of plastic or metal\", \"the large rectangular red and gold carpet\". ", "option": [ "the door, the cylindrical black bin made of plastic or metal", "the door, the large rectangular red and gold carpet", "the cylindrical black bin made of plastic or metal, the large rectangular red and gold carpet" ], "answer": "the cylindrical black bin made of plastic or metal, the large rectangular red and gold carpet", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the door\", \"the cylindrical black bin made of plastic or metal\", \"the large rectangular red and gold carpet\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the door, the cylindrical black bin made of plastic or metal', 'the door, the large rectangular red and gold carpet', 'the cylindrical black bin made of plastic or metal, the large rectangular red and gold carpet'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0692_03/00800.jpg", "scene0692_03/00840.jpg", "scene0692_03/00880.jpg", "scene0692_03/00920.jpg" ], "turn_id": 6, "scan_id": "scene0692_03", "origin_question": "Please recall \"the classic lamp with a conical white fabric shade\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.16", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the classic lamp with a conical white fabric shade\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0065_01/00000.jpg", "scene0065_01/00030.jpg", "scene0065_01/00060.jpg", "scene0065_01/00090.jpg" ], "turn_id": 1, "scan_id": "scene0065_01", "origin_question": "Remember, have you seen any wine(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any wine(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0065_01/00120.jpg", "scene0065_01/00150.jpg", "scene0065_01/00180.jpg", "scene0065_01/00210.jpg" ], "turn_id": 2, "scan_id": "scene0065_01", "origin_question": "In which turn did you see both \"the round white ceramic sink with a glossy surface\" and \"the single, white cylindrical roll of toilet paper\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the round white ceramic sink with a glossy surface\" and \"the single, white cylindrical roll of toilet paper\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0065_01/00240.jpg", "scene0065_01/00270.jpg", "scene0065_01/00300.jpg", "scene0065_01/00330.jpg" ], "turn_id": 3, "scan_id": "scene0065_01", "origin_question": "Is the rectangular wooden cabinet with a granite countertop to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the rectangular wooden cabinet with a granite countertop to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0065_01/00360.jpg", "scene0065_01/00390.jpg", "scene0065_01/00420.jpg", "scene0065_01/00450.jpg" ], "turn_id": 4, "scan_id": "scene0065_01", "origin_question": "Is the modern white ceramic standalone toilet to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the modern white ceramic standalone toilet to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0065_01/00480.jpg", "scene0065_01/00510.jpg", "scene0065_01/00540.jpg", "scene0065_01/00570.jpg" ], "turn_id": 5, "scan_id": "scene0065_01", "origin_question": "Which direction is the toilet plunger with a wooden handle and red cup in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the toilet plunger with a wooden handle and red cup in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "scene0065_01/00600.jpg", "scene0065_01/00630.jpg", "scene0065_01/00660.jpg", "scene0065_01/00690.jpg" ], "turn_id": 6, "scan_id": "scene0065_01", "origin_question": "Among these three objects(\"the single, white cylindrical roll of toilet paper\",\"the medium-sized rectangular white ceramic bathtub\",\"the toilet plunger with a wooden handle and red cup\" ), which one is the closest to you?", "option": [ "the single, white cylindrical roll of toilet paper", "the medium-sized rectangular white ceramic bathtub", "the toilet plunger with a wooden handle and red cup" ], "answer": "the single, white cylindrical roll of toilet paper", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects(\"the single, white cylindrical roll of toilet paper\",\"the medium-sized rectangular white ceramic bathtub\",\"the toilet plunger with a wooden handle and red cup\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the single, white cylindrical roll of toilet paper', 'the medium-sized rectangular white ceramic bathtub', 'the toilet plunger with a wooden handle and red cup'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0065_01/00720.jpg", "scene0065_01/00750.jpg", "scene0065_01/00780.jpg", "scene0065_01/00810.jpg" ], "turn_id": 7, "scan_id": "scene0065_01", "origin_question": "Is the medium-sized rectangular white ceramic bathtub to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the medium-sized rectangular white ceramic bathtub to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42898817/Validation_42898817_42898817_167412.311.png", "Validation_42898817/Validation_42898817_42898817_167414.294.png", "Validation_42898817/Validation_42898817_42898817_167416.293.png", "Validation_42898817/Validation_42898817_42898817_167418.292.png" ], "turn_id": 1, "scan_id": "Validation_42898817", "origin_question": "Remember, have you seen any washing machine(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any washing machine(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_42898817/Validation_42898817_42898817_167420.291.png", "Validation_42898817/Validation_42898817_42898817_167422.290.png", "Validation_42898817/Validation_42898817_42898817_167424.290.png", "Validation_42898817/Validation_42898817_42898817_167426.289.png" ], "turn_id": 2, "scan_id": "Validation_42898817", "origin_question": "When did you first discover \"the sink\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the sink\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_42898817/Validation_42898817_42898817_167428.288.png", "Validation_42898817/Validation_42898817_42898817_167430.287.png", "Validation_42898817/Validation_42898817_42898817_167432.303.png", "Validation_42898817/Validation_42898817_42898817_167434.302.png" ], "turn_id": 3, "scan_id": "Validation_42898817", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the vibrant abstract box in green, yellow, and blue\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the vibrant abstract box in green, yellow, and blue\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_42898817/Validation_42898817_42898817_167436.301.png", "Validation_42898817/Validation_42898817_42898817_167438.300.png", "Validation_42898817/Validation_42898817_42898817_167440.299.png", "Validation_42898817/Validation_42898817_42898817_167442.299.png" ], "turn_id": 4, "scan_id": "Validation_42898817", "origin_question": "Which direction is the sink in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the sink in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_42898817/Validation_42898817_42898817_167444.298.png", "Validation_42898817/Validation_42898817_42898817_167446.297.png", "Validation_42898817/Validation_42898817_42898817_167448.296.png", "Validation_42898817/Validation_42898817_42898817_167450.295.png" ], "turn_id": 5, "scan_id": "Validation_42898817", "origin_question": "In which turn were you closest to \"the sleek silver washing machine\"? ", "option": [], "answer": "2", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the sleek silver washing machine\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_42898817/Validation_42898817_42898817_167452.295.png", "Validation_42898817/Validation_42898817_42898817_167454.294.png", "Validation_42898817/Validation_42898817_42898817_167456.293.png", "Validation_42898817/Validation_42898817_42898817_167458.292.png" ], "turn_id": 6, "scan_id": "Validation_42898817", "origin_question": "At the end of which turn were both of \"the vibrant abstract box in green, yellow, and blue\" and \"the sleek silver washing machine\" on your right side?", "option": [], "answer": "4", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the vibrant abstract box in green, yellow, and blue\" and \"the sleek silver washing machine\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_45261575/Validation_45261575_45261575_77469.527.png", "Validation_45261575/Validation_45261575_45261575_77471.009.png", "Validation_45261575/Validation_45261575_45261575_77472.509.png", "Validation_45261575/Validation_45261575_45261575_77474.008.png" ], "turn_id": 1, "scan_id": "Validation_45261575", "origin_question": "Remember, have you seen any monitor(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any monitor(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_45261575/Validation_45261575_45261575_77475.507.png", "Validation_45261575/Validation_45261575_45261575_77477.007.png", "Validation_45261575/Validation_45261575_45261575_77478.506.png", "Validation_45261575/Validation_45261575_45261575_77480.005.png" ], "turn_id": 2, "scan_id": "Validation_45261575", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_45261575/Validation_45261575_45261575_77481.505.png", "Validation_45261575/Validation_45261575_45261575_77483.004.png", "Validation_45261575/Validation_45261575_45261575_77484.504.png", "Validation_45261575/Validation_45261575_45261575_77486.020.png" ], "turn_id": 3, "scan_id": "Validation_45261575", "origin_question": "When did you first discover \"the medium-sized black flat-screen monitor\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the medium-sized black flat-screen monitor\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_45261575/Validation_45261575_45261575_77487.519.png", "Validation_45261575/Validation_45261575_45261575_77489.018.png", "Validation_45261575/Validation_45261575_45261575_77490.518.png", "Validation_45261575/Validation_45261575_45261575_77492.017.png" ], "turn_id": 4, "scan_id": "Validation_45261575", "origin_question": "In which turn were you farthest from \"the soft pink bed with plush bedding\"? ", "option": [], "answer": "1", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the soft pink bed with plush bedding\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_45261575/Validation_45261575_45261575_77493.517.png", "Validation_45261575/Validation_45261575_45261575_77495.016.png", "Validation_45261575/Validation_45261575_45261575_77496.515.png", "Validation_45261575/Validation_45261575_45261575_77498.015.png" ], "turn_id": 5, "scan_id": "Validation_45261575", "origin_question": "How far is your current position from where you were at the end of the turn 1? (in meters)", "option": [], "answer": "1.81", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 1? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_45261575/Validation_45261575_45261575_77499.514.png", "Validation_45261575/Validation_45261575_45261575_77501.014.png", "Validation_45261575/Validation_45261575_45261575_77502.513.png", "Validation_45261575/Validation_45261575_45261575_77504.012.png" ], "turn_id": 6, "scan_id": "Validation_45261575", "origin_question": "Is the medium-sized black flat-screen monitor to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the medium-sized black flat-screen monitor to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47334237/Validation_47334237_47334237_109431.492.png", "Validation_47334237/Validation_47334237_47334237_109433.474.png", "Validation_47334237/Validation_47334237_47334237_109435.474.png", "Validation_47334237/Validation_47334237_47334237_109437.473.png", "Validation_47334237/Validation_47334237_47334237_109439.472.png" ], "turn_id": 1, "scan_id": "Validation_47334237", "origin_question": "Remember, have you seen any cabinet(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any cabinet(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47334237/Validation_47334237_47334237_109441.471.png", "Validation_47334237/Validation_47334237_47334237_109443.470.png", "Validation_47334237/Validation_47334237_47334237_109445.470.png", "Validation_47334237/Validation_47334237_47334237_109447.485.png", "Validation_47334237/Validation_47334237_47334237_109449.485.png" ], "turn_id": 2, "scan_id": "Validation_47334237", "origin_question": "Remember, have you seen any desk(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any desk(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_47334237/Validation_47334237_47334237_109451.484.png", "Validation_47334237/Validation_47334237_47334237_109453.483.png", "Validation_47334237/Validation_47334237_47334237_109455.482.png", "Validation_47334237/Validation_47334237_47334237_109457.481.png", "Validation_47334237/Validation_47334237_47334237_109459.481.png" ], "turn_id": 3, "scan_id": "Validation_47334237", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_47334237/Validation_47334237_47334237_109461.480.png", "Validation_47334237/Validation_47334237_47334237_109463.479.png", "Validation_47334237/Validation_47334237_47334237_109465.478.png", "Validation_47334237/Validation_47334237_47334237_109467.477.png", "Validation_47334237/Validation_47334237_47334237_109469.477.png" ], "turn_id": 4, "scan_id": "Validation_47334237", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the modern chair with the dark wood frame and beige upholstery\", Y: \"the bold red square plush pillow\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the modern chair with the dark wood frame and beige upholstery\", Y: \"the bold red square plush pillow\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_47334237/Validation_47334237_47334237_109471.476.png", "Validation_47334237/Validation_47334237_47334237_109473.475.png", "Validation_47334237/Validation_47334237_47334237_109475.474.png", "Validation_47334237/Validation_47334237_47334237_109477.473.png", "Validation_47334237/Validation_47334237_47334237_109479.473.png" ], "turn_id": 5, "scan_id": "Validation_47334237", "origin_question": "At the end of which turn were both of \"the compact, cylindrical metal stool in muted red\" and \"the rectangular mahogany table with the glass top\" on your right side?", "option": [], "answer": "5", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the compact, cylindrical metal stool in muted red\" and \"the rectangular mahogany table with the glass top\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_47334237/Validation_47334237_47334237_109481.472.png", "Validation_47334237/Validation_47334237_47334237_109483.471.png", "Validation_47334237/Validation_47334237_47334237_109485.470.png", "Validation_47334237/Validation_47334237_47334237_109487.469.png", "Validation_47334237/Validation_47334237_47334237_109489.485.png" ], "turn_id": 6, "scan_id": "Validation_47334237", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the sleek, light-colored door with the metallic handle\" from your position?", "option": [], "answer": "170.29", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the sleek, light-colored door with the metallic handle\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47334237/Validation_47334237_47334237_109491.484.png", "Validation_47334237/Validation_47334237_47334237_109493.484.png", "Validation_47334237/Validation_47334237_47334237_109495.483.png", "Validation_47334237/Validation_47334237_47334237_109497.482.png", "Validation_47334237/Validation_47334237_47334237_109499.481.png" ], "turn_id": 7, "scan_id": "Validation_47334237", "origin_question": "Which direction is the sleek, light-colored door with the metallic handle in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the sleek, light-colored door with the metallic handle in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_47334237/Validation_47334237_47334237_109501.480.png", "Validation_47334237/Validation_47334237_47334237_109503.480.png", "Validation_47334237/Validation_47334237_47334237_109505.479.png", "Validation_47334237/Validation_47334237_47334237_109507.478.png", "Validation_47334237/Validation_47334237_47334237_109509.477.png" ], "turn_id": 8, "scan_id": "Validation_47334237", "origin_question": "Using your orientation at the end of turn 7 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "67.34", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 7 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_47334237/Validation_47334237_47334237_109511.476.png", "Validation_47334237/Validation_47334237_47334237_109513.476.png", "Validation_47334237/Validation_47334237_47334237_109515.475.png", "Validation_47334237/Validation_47334237_47334237_109517.474.png", "Validation_47334237/Validation_47334237_47334237_109519.473.png" ], "turn_id": 9, "scan_id": "Validation_47334237", "origin_question": "Is the sleek, light-colored door with the metallic handle to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the sleek, light-colored door with the metallic handle to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_47334237/Validation_47334237_47334237_109521.472.png", "Validation_47334237/Validation_47334237_47334237_109523.471.png", "Validation_47334237/Validation_47334237_47334237_109525.471.png", "Validation_47334237/Validation_47334237_47334237_109527.470.png", "Validation_47334237/Validation_47334237_47334237_109529.469.png" ], "turn_id": 10, "scan_id": "Validation_47334237", "origin_question": "Using your orientation at the end of turn 8 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "28.46", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 8 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_41069021/Validation_41069021_41069021_305.377.png", "Validation_41069021/Validation_41069021_41069021_307.360.png", "Validation_41069021/Validation_41069021_41069021_309.359.png", "Validation_41069021/Validation_41069021_41069021_311.358.png", "Validation_41069021/Validation_41069021_41069021_313.357.png" ], "turn_id": 1, "scan_id": "Validation_41069021", "origin_question": "Remember, have you seen any screen(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any screen(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_41069021/Validation_41069021_41069021_315.373.png", "Validation_41069021/Validation_41069021_41069021_317.373.png", "Validation_41069021/Validation_41069021_41069021_319.372.png", "Validation_41069021/Validation_41069021_41069021_321.371.png", "Validation_41069021/Validation_41069021_41069021_323.370.png" ], "turn_id": 2, "scan_id": "Validation_41069021", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the framed abstract picture with geometric shapes\" from your position?", "option": [], "answer": "49.42", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the framed abstract picture with geometric shapes\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_41069021/Validation_41069021_41069021_325.369.png", "Validation_41069021/Validation_41069021_41069021_327.368.png", "Validation_41069021/Validation_41069021_41069021_329.368.png", "Validation_41069021/Validation_41069021_41069021_331.367.png", "Validation_41069021/Validation_41069021_41069021_333.366.png" ], "turn_id": 3, "scan_id": "Validation_41069021", "origin_question": "At the end of which turn were both of \"the dark brown leather three-seat couch\" and \"the medium-sized, modern lamp with the cylindrical white shade\" on your left side?", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the dark brown leather three-seat couch\" and \"the medium-sized, modern lamp with the cylindrical white shade\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "Validation_41069021/Validation_41069021_41069021_335.365.png", "Validation_41069021/Validation_41069021_41069021_337.364.png", "Validation_41069021/Validation_41069021_41069021_339.363.png", "Validation_41069021/Validation_41069021_41069021_341.363.png", "Validation_41069021/Validation_41069021_41069021_343.362.png" ], "turn_id": 4, "scan_id": "Validation_41069021", "origin_question": "Among these three objects(\"the framed abstract picture with geometric shapes\",\"the compact, white metal radiator\",\"the medium-sized, modern lamp with the cylindrical white shade\" ), which one is the farthest from you?", "option": [ "the framed abstract picture with geometric shapes", "the compact, white metal radiator", "the medium-sized, modern lamp with the cylindrical white shade" ], "answer": "the framed abstract picture with geometric shapes", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the framed abstract picture with geometric shapes\",\"the compact, white metal radiator\",\"the medium-sized, modern lamp with the cylindrical white shade\" ), which one is the farthest from you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the framed abstract picture with geometric shapes', 'the compact, white metal radiator', 'the medium-sized, modern lamp with the cylindrical white shade'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_41069021/Validation_41069021_41069021_345.361.png", "Validation_41069021/Validation_41069021_41069021_347.360.png", "Validation_41069021/Validation_41069021_41069021_349.359.png", "Validation_41069021/Validation_41069021_41069021_351.358.png", "Validation_41069021/Validation_41069021_41069021_353.358.png" ], "turn_id": 5, "scan_id": "Validation_41069021", "origin_question": "Is the medium-sized, modern lamp with the cylindrical white shade to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the medium-sized, modern lamp with the cylindrical white shade to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_41069021/Validation_41069021_41069021_355.357.png", "Validation_41069021/Validation_41069021_41069021_357.373.png", "Validation_41069021/Validation_41069021_41069021_359.372.png", "Validation_41069021/Validation_41069021_41069021_361.371.png", "Validation_41069021/Validation_41069021_41069021_363.370.png" ], "turn_id": 6, "scan_id": "Validation_41069021", "origin_question": "Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Left", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_41069021/Validation_41069021_41069021_365.370.png", "Validation_41069021/Validation_41069021_41069021_367.369.png", "Validation_41069021/Validation_41069021_41069021_369.368.png", "Validation_41069021/Validation_41069021_41069021_371.367.png", "Validation_41069021/Validation_41069021_41069021_373.366.png" ], "turn_id": 7, "scan_id": "Validation_41069021", "origin_question": "", "option": [], "answer": "", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_41069021/Validation_41069021_41069021_375.365.png", "Validation_41069021/Validation_41069021_41069021_377.365.png", "Validation_41069021/Validation_41069021_41069021_379.364.png", "Validation_41069021/Validation_41069021_41069021_381.363.png", "Validation_41069021/Validation_41069021_41069021_383.362.png" ], "turn_id": 8, "scan_id": "Validation_41069021", "origin_question": "Please recall \"the light wood door with the smooth finish\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "3.23", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the light wood door with the smooth finish\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_41069021/Validation_41069021_41069021_385.361.png", "Validation_41069021/Validation_41069021_41069021_387.360.png", "Validation_41069021/Validation_41069021_41069021_389.360.png", "Validation_41069021/Validation_41069021_41069021_391.359.png", "Validation_41069021/Validation_41069021_41069021_393.358.png" ], "turn_id": 9, "scan_id": "Validation_41069021", "origin_question": "Which direction is the light wood door with the smooth finish in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the light wood door with the smooth finish in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_41069021/Validation_41069021_41069021_395.357.png", "Validation_41069021/Validation_41069021_41069021_397.373.png", "Validation_41069021/Validation_41069021_41069021_399.372.png", "Validation_41069021/Validation_41069021_41069021_401.371.png", "Validation_41069021/Validation_41069021_41069021_403.370.png" ], "turn_id": 10, "scan_id": "Validation_41069021", "origin_question": "Please recall \"the medium-sized, modern lamp with the cylindrical white shade\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.37", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the medium-sized, modern lamp with the cylindrical white shade\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_41254262/Validation_41254262_41254262_101754.312.png", "Validation_41254262/Validation_41254262_41254262_101756.295.png", "Validation_41254262/Validation_41254262_41254262_101758.294.png", "Validation_41254262/Validation_41254262_41254262_101760.293.png", "Validation_41254262/Validation_41254262_41254262_101762.292.png" ], "turn_id": 1, "scan_id": "Validation_41254262", "origin_question": "Remember, have you seen any bed(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any bed(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_41254262/Validation_41254262_41254262_101764.291.png", "Validation_41254262/Validation_41254262_41254262_101766.307.png", "Validation_41254262/Validation_41254262_41254262_101768.306.png", "Validation_41254262/Validation_41254262_41254262_101770.306.png", "Validation_41254262/Validation_41254262_41254262_101772.305.png" ], "turn_id": 2, "scan_id": "Validation_41254262", "origin_question": "In which turn did you see both \"the bed\" and \"the warm brown cylindrical chandelier\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the bed\" and \"the warm brown cylindrical chandelier\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "Validation_41254262/Validation_41254262_41254262_101774.304.png", "Validation_41254262/Validation_41254262_41254262_101776.303.png", "Validation_41254262/Validation_41254262_41254262_101778.302.png", "Validation_41254262/Validation_41254262_41254262_101780.302.png", "Validation_41254262/Validation_41254262_41254262_101782.301.png" ], "turn_id": 3, "scan_id": "Validation_41254262", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the sleek, light wood veneer cabinet\"; \"the bed\"; \"the rectangular, white-framed window with glass and wood\" ", "option": [ "the sleek, light wood veneer cabinet", "the bed", "the rectangular, white-framed window with glass and wood" ], "answer": "the rectangular, white-framed window with glass and wood", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the sleek, light wood veneer cabinet\"; \"the bed\"; \"the rectangular, white-framed window with glass and wood\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the sleek, light wood veneer cabinet', 'the bed', 'the rectangular, white-framed window with glass and wood'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_41254262/Validation_41254262_41254262_101784.300.png", "Validation_41254262/Validation_41254262_41254262_101786.299.png", "Validation_41254262/Validation_41254262_41254262_101788.298.png", "Validation_41254262/Validation_41254262_41254262_101790.297.png", "Validation_41254262/Validation_41254262_41254262_101792.296.png" ], "turn_id": 4, "scan_id": "Validation_41254262", "origin_question": "At the end of which turn were both of \"the warm brown cylindrical chandelier\" and \"the sleek white rectangular door with the brass handle\" on your left side?", "option": [], "answer": "4", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the warm brown cylindrical chandelier\" and \"the sleek white rectangular door with the brass handle\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_41254262/Validation_41254262_41254262_101794.296.png", "Validation_41254262/Validation_41254262_41254262_101796.295.png", "Validation_41254262/Validation_41254262_41254262_101798.294.png", "Validation_41254262/Validation_41254262_41254262_101800.293.png", "Validation_41254262/Validation_41254262_41254262_101802.292.png" ], "turn_id": 5, "scan_id": "Validation_41254262", "origin_question": "Is the warm brown cylindrical chandelier to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the warm brown cylindrical chandelier to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "Validation_41254262/Validation_41254262_41254262_101804.292.png", "Validation_41254262/Validation_41254262_41254262_101806.307.png", "Validation_41254262/Validation_41254262_41254262_101808.307.png", "Validation_41254262/Validation_41254262_41254262_101810.306.png", "Validation_41254262/Validation_41254262_41254262_101812.305.png" ], "turn_id": 6, "scan_id": "Validation_41254262", "origin_question": "Among these three objects(\"the sleek, light wood veneer cabinet\",\"the sleek white rectangular door with the brass handle\",\"the rectangular, white-framed window with glass and wood\" ), which one is the closest to you?", "option": [ "the sleek, light wood veneer cabinet", "the sleek white rectangular door with the brass handle", "the rectangular, white-framed window with glass and wood" ], "answer": "the rectangular, white-framed window with glass and wood", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the sleek, light wood veneer cabinet\",\"the sleek white rectangular door with the brass handle\",\"the rectangular, white-framed window with glass and wood\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the sleek, light wood veneer cabinet', 'the sleek white rectangular door with the brass handle', 'the rectangular, white-framed window with glass and wood'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_41254262/Validation_41254262_41254262_101814.304.png", "Validation_41254262/Validation_41254262_41254262_101816.303.png", "Validation_41254262/Validation_41254262_41254262_101818.302.png", "Validation_41254262/Validation_41254262_41254262_101820.302.png", "Validation_41254262/Validation_41254262_41254262_101822.301.png" ], "turn_id": 7, "scan_id": "Validation_41254262", "origin_question": "In which turn were you farthest from \"the warm brown cylindrical chandelier\"? ", "option": [], "answer": "1", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the warm brown cylindrical chandelier\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_41254262/Validation_41254262_41254262_101824.300.png", "Validation_41254262/Validation_41254262_41254262_101826.299.png", "Validation_41254262/Validation_41254262_41254262_101828.298.png", "Validation_41254262/Validation_41254262_41254262_101830.297.png", "Validation_41254262/Validation_41254262_41254262_101832.297.png" ], "turn_id": 8, "scan_id": "Validation_41254262", "origin_question": "What will be the first-time appearance order of the following categories: chandelier, door, radiator?", "option": [ "radiator, door, chandelier", "door, chandelier, radiator", "chandelier, door, radiator", "radiator, chandelier, door" ], "answer": "door, chandelier, radiator", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: chandelier, door, radiator? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['radiator, door, chandelier', 'door, chandelier, radiator', 'chandelier, door, radiator', 'radiator, chandelier, door'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_41254262/Validation_41254262_41254262_101834.296.png", "Validation_41254262/Validation_41254262_41254262_101836.295.png", "Validation_41254262/Validation_41254262_41254262_101838.294.png", "Validation_41254262/Validation_41254262_41254262_101840.293.png", "Validation_41254262/Validation_41254262_41254262_101842.292.png" ], "turn_id": 9, "scan_id": "Validation_41254262", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the bed\" from your position?", "option": [], "answer": "164.8", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the bed\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_41254262/Validation_41254262_41254262_101844.292.png", "Validation_41254262/Validation_41254262_41254262_101846.291.png", "Validation_41254262/Validation_41254262_41254262_101848.307.png", "Validation_41254262/Validation_41254262_41254262_101850.306.png", "Validation_41254262/Validation_41254262_41254262_101852.305.png" ], "turn_id": 10, "scan_id": "Validation_41254262", "origin_question": "Please recall \"the sleek white rectangular door with the brass handle\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.13", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the sleek white rectangular door with the brass handle\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "Validation_44358436/Validation_44358436_44358436_42322.907.png", "Validation_44358436/Validation_44358436_44358436_42324.890.png", "Validation_44358436/Validation_44358436_44358436_42326.889.png", "Validation_44358436/Validation_44358436_44358436_42328.888.png", "Validation_44358436/Validation_44358436_44358436_42330.904.png" ], "turn_id": 1, "scan_id": "Validation_44358436", "origin_question": "Remember, how many chair(s) have you seen so far? ", "option": [], "answer": "1", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many chair(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_44358436/Validation_44358436_44358436_42332.903.png", "Validation_44358436/Validation_44358436_44358436_42334.902.png", "Validation_44358436/Validation_44358436_44358436_42336.901.png", "Validation_44358436/Validation_44358436_44358436_42338.901.png", "Validation_44358436/Validation_44358436_44358436_42340.900.png" ], "turn_id": 2, "scan_id": "Validation_44358436", "origin_question": "At the end of which turn were both of \"the rectangular gray couch with plush cushions\" and \"the practical open shelf in neutral color\" on your left side?", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the rectangular gray couch with plush cushions\" and \"the practical open shelf in neutral color\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_44358436/Validation_44358436_44358436_42342.899.png", "Validation_44358436/Validation_44358436_44358436_42344.898.png", "Validation_44358436/Validation_44358436_44358436_42346.897.png", "Validation_44358436/Validation_44358436_44358436_42348.896.png", "Validation_44358436/Validation_44358436_44358436_42350.895.png" ], "turn_id": 3, "scan_id": "Validation_44358436", "origin_question": "Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance forward or backward from that position?", "option": [ "Forward", "Backward" ], "answer": "Forward", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance forward or backward from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Forward', 'Backward'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_44358436/Validation_44358436_44358436_42352.895.png", "Validation_44358436/Validation_44358436_44358436_42354.894.png", "Validation_44358436/Validation_44358436_44358436_42356.893.png", "Validation_44358436/Validation_44358436_44358436_42358.892.png", "Validation_44358436/Validation_44358436_44358436_42360.891.png" ], "turn_id": 4, "scan_id": "Validation_44358436", "origin_question": "Remember, have you seen any stairs(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any stairs(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_44358436/Validation_44358436_44358436_42362.891.png", "Validation_44358436/Validation_44358436_44358436_42364.890.png", "Validation_44358436/Validation_44358436_44358436_42366.889.png", "Validation_44358436/Validation_44358436_44358436_42368.888.png", "Validation_44358436/Validation_44358436_44358436_42370.904.png" ], "turn_id": 5, "scan_id": "Validation_44358436", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular gray couch with plush cushions\", Y: \"the sleek black rectangular monitor with rounded corners\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to Y and farther away from X", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular gray couch with plush cushions\", Y: \"the sleek black rectangular monitor with rounded corners\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_44358436/Validation_44358436_44358436_42372.903.png", "Validation_44358436/Validation_44358436_44358436_42374.902.png", "Validation_44358436/Validation_44358436_44358436_42376.901.png", "Validation_44358436/Validation_44358436_44358436_42378.901.png", "Validation_44358436/Validation_44358436_44358436_42380.900.png" ], "turn_id": 6, "scan_id": "Validation_44358436", "origin_question": "Which two objects are on the same side of you now? \"the rectangular gray couch with plush cushions\", \"the medium-sized, light gray cabinet\", \"the sleek black rectangular monitor with rounded corners\". ", "option": [ "the rectangular gray couch with plush cushions, the medium-sized, light gray cabinet", "the rectangular gray couch with plush cushions, the sleek black rectangular monitor with rounded corners", "the medium-sized, light gray cabinet, the sleek black rectangular monitor with rounded corners" ], "answer": "the medium-sized, light gray cabinet, the sleek black rectangular monitor with rounded corners", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the rectangular gray couch with plush cushions\", \"the medium-sized, light gray cabinet\", \"the sleek black rectangular monitor with rounded corners\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular gray couch with plush cushions, the medium-sized, light gray cabinet', 'the rectangular gray couch with plush cushions, the sleek black rectangular monitor with rounded corners', 'the medium-sized, light gray cabinet, the sleek black rectangular monitor with rounded corners'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_44358436/Validation_44358436_44358436_42382.899.png", "Validation_44358436/Validation_44358436_44358436_42384.898.png", "Validation_44358436/Validation_44358436_44358436_42386.897.png", "Validation_44358436/Validation_44358436_44358436_42388.897.png", "Validation_44358436/Validation_44358436_44358436_42390.896.png" ], "turn_id": 7, "scan_id": "Validation_44358436", "origin_question": "Which direction is the medium-sized, light gray cabinet in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the medium-sized, light gray cabinet in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_44358436/Validation_44358436_44358436_42392.895.png", "Validation_44358436/Validation_44358436_44358436_42394.894.png", "Validation_44358436/Validation_44358436_44358436_42396.893.png", "Validation_44358436/Validation_44358436_44358436_42398.892.png", "Validation_44358436/Validation_44358436_44358436_42400.891.png" ], "turn_id": 8, "scan_id": "Validation_44358436", "origin_question": "Which two objects are on the same side of you now? \"the rectangular gray couch with plush cushions\", \"the medium-sized, light gray cabinet\", \"the sleek black rectangular monitor with rounded corners\". ", "option": [ "the rectangular gray couch with plush cushions, the medium-sized, light gray cabinet", "the rectangular gray couch with plush cushions, the sleek black rectangular monitor with rounded corners", "the medium-sized, light gray cabinet, the sleek black rectangular monitor with rounded corners" ], "answer": "the rectangular gray couch with plush cushions, the medium-sized, light gray cabinet", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the rectangular gray couch with plush cushions\", \"the medium-sized, light gray cabinet\", \"the sleek black rectangular monitor with rounded corners\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular gray couch with plush cushions, the medium-sized, light gray cabinet', 'the rectangular gray couch with plush cushions, the sleek black rectangular monitor with rounded corners', 'the medium-sized, light gray cabinet, the sleek black rectangular monitor with rounded corners'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_45662926/Validation_45662926_45662926_21639.339.png", "Validation_45662926/Validation_45662926_45662926_21641.322.png", "Validation_45662926/Validation_45662926_45662926_21643.321.png", "Validation_45662926/Validation_45662926_45662926_21645.320.png", "Validation_45662926/Validation_45662926_45662926_21647.319.png" ], "turn_id": 1, "scan_id": "Validation_45662926", "origin_question": "Remember, have you seen any stairs(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any stairs(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_45662926/Validation_45662926_45662926_21649.319.png", "Validation_45662926/Validation_45662926_45662926_21651.318.png", "Validation_45662926/Validation_45662926_45662926_21653.317.png", "Validation_45662926/Validation_45662926_45662926_21655.316.png", "Validation_45662926/Validation_45662926_45662926_21657.315.png" ], "turn_id": 2, "scan_id": "Validation_45662926", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "Validation_45662926/Validation_45662926_45662926_21659.331.png", "Validation_45662926/Validation_45662926_45662926_21661.330.png", "Validation_45662926/Validation_45662926_45662926_21663.329.png", "Validation_45662926/Validation_45662926_45662926_21665.329.png", "Validation_45662926/Validation_45662926_45662926_21667.328.png" ], "turn_id": 3, "scan_id": "Validation_45662926", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the large rectangular light wood table\"; \"the straight wooden stairs with the neutral color\"; \"the lush plant with broad, green leaves in the white planter\" ", "option": [ "the large rectangular light wood table", "the straight wooden stairs with the neutral color", "the lush plant with broad, green leaves in the white planter" ], "answer": "the large rectangular light wood table", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the large rectangular light wood table\"; \"the straight wooden stairs with the neutral color\"; \"the lush plant with broad, green leaves in the white planter\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the large rectangular light wood table', 'the straight wooden stairs with the neutral color', 'the lush plant with broad, green leaves in the white planter'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_45662926/Validation_45662926_45662926_21669.327.png", "Validation_45662926/Validation_45662926_45662926_21671.326.png", "Validation_45662926/Validation_45662926_45662926_21673.325.png", "Validation_45662926/Validation_45662926_45662926_21675.324.png", "Validation_45662926/Validation_45662926_45662926_21677.324.png" ], "turn_id": 4, "scan_id": "Validation_45662926", "origin_question": "Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Right", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_45662926/Validation_45662926_45662926_21679.323.png", "Validation_45662926/Validation_45662926_45662926_21681.322.png", "Validation_45662926/Validation_45662926_45662926_21683.321.png", "Validation_45662926/Validation_45662926_45662926_21685.320.png", "Validation_45662926/Validation_45662926_45662926_21687.319.png" ], "turn_id": 5, "scan_id": "Validation_45662926", "origin_question": "How far is your current position from where you were at the end of the turn 3? (in meters)", "option": [], "answer": "4.84", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 3? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_45662926/Validation_45662926_45662926_21689.319.png", "Validation_45662926/Validation_45662926_45662926_21691.318.png", "Validation_45662926/Validation_45662926_45662926_21693.317.png", "Validation_45662926/Validation_45662926_45662926_21695.316.png", "Validation_45662926/Validation_45662926_45662926_21697.315.png" ], "turn_id": 6, "scan_id": "Validation_45662926", "origin_question": "Which direction is the sleek rectangular stainless steel stove in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the sleek rectangular stainless steel stove in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_45662926/Validation_45662926_45662926_21699.331.png", "Validation_45662926/Validation_45662926_45662926_21701.330.png", "Validation_45662926/Validation_45662926_45662926_21703.329.png", "Validation_45662926/Validation_45662926_45662926_21705.329.png", "Validation_45662926/Validation_45662926_45662926_21707.328.png" ], "turn_id": 7, "scan_id": "Validation_45662926", "origin_question": "Which direction is the straight wooden stairs with the neutral color in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the straight wooden stairs with the neutral color in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_45662926/Validation_45662926_45662926_21709.327.png", "Validation_45662926/Validation_45662926_45662926_21711.326.png", "Validation_45662926/Validation_45662926_45662926_21713.325.png", "Validation_45662926/Validation_45662926_45662926_21715.324.png", "Validation_45662926/Validation_45662926_45662926_21717.324.png" ], "turn_id": 8, "scan_id": "Validation_45662926", "origin_question": "How far is your current position from where you were at the end of the turn 3? (in meters)", "option": [], "answer": "6.41", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 3? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_45662926/Validation_45662926_45662926_21719.323.png", "Validation_45662926/Validation_45662926_45662926_21721.322.png", "Validation_45662926/Validation_45662926_45662926_21723.321.png", "Validation_45662926/Validation_45662926_45662926_21725.320.png", "Validation_45662926/Validation_45662926_45662926_21727.320.png" ], "turn_id": 9, "scan_id": "Validation_45662926", "origin_question": "Which two objects are on the same side of you now? \"the straight wooden stairs with the neutral color\", \"the mirror\", \"the large rectangular light wood table\". ", "option": [ "the straight wooden stairs with the neutral color, the mirror", "the straight wooden stairs with the neutral color, the large rectangular light wood table", "the mirror, the large rectangular light wood table" ], "answer": "the straight wooden stairs with the neutral color, the mirror", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the straight wooden stairs with the neutral color\", \"the mirror\", \"the large rectangular light wood table\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the straight wooden stairs with the neutral color, the mirror', 'the straight wooden stairs with the neutral color, the large rectangular light wood table', 'the mirror, the large rectangular light wood table'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_45662926/Validation_45662926_45662926_21729.319.png", "Validation_45662926/Validation_45662926_45662926_21731.318.png", "Validation_45662926/Validation_45662926_45662926_21733.317.png", "Validation_45662926/Validation_45662926_45662926_21735.316.png", "Validation_45662926/Validation_45662926_45662926_21737.315.png" ], "turn_id": 10, "scan_id": "Validation_45662926", "origin_question": "Which two objects are on the same side of you now? \"the straight wooden stairs with the neutral color\", \"the lush plant with broad, green leaves in the white planter\", \"the large rectangular light wood table\". ", "option": [ "the straight wooden stairs with the neutral color, the lush plant with broad, green leaves in the white planter", "the straight wooden stairs with the neutral color, the large rectangular light wood table", "the lush plant with broad, green leaves in the white planter, the large rectangular light wood table" ], "answer": "the lush plant with broad, green leaves in the white planter, the large rectangular light wood table", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the straight wooden stairs with the neutral color\", \"the lush plant with broad, green leaves in the white planter\", \"the large rectangular light wood table\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the straight wooden stairs with the neutral color, the lush plant with broad, green leaves in the white planter', 'the straight wooden stairs with the neutral color, the large rectangular light wood table', 'the lush plant with broad, green leaves in the white planter, the large rectangular light wood table'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42899459/Validation_42899459_42899459_195916.348.png", "Validation_42899459/Validation_42899459_42899459_195918.330.png", "Validation_42899459/Validation_42899459_42899459_195920.329.png", "Validation_42899459/Validation_42899459_42899459_195922.329.png", "Validation_42899459/Validation_42899459_42899459_195924.328.png" ], "turn_id": 1, "scan_id": "Validation_42899459", "origin_question": "Remember, have you seen any washing machine(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any washing machine(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_42899459/Validation_42899459_42899459_195926.327.png", "Validation_42899459/Validation_42899459_42899459_195928.326.png", "Validation_42899459/Validation_42899459_42899459_195930.325.png", "Validation_42899459/Validation_42899459_42899459_195932.324.png", "Validation_42899459/Validation_42899459_42899459_195934.324.png" ], "turn_id": 2, "scan_id": "Validation_42899459", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the rich chocolate brown rectangular couch with plush fabric\", Y: \"the sleek rectangular monitor with thin black bezels\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the rich chocolate brown rectangular couch with plush fabric\", Y: \"the sleek rectangular monitor with thin black bezels\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_42899459/Validation_42899459_42899459_195936.323.png", "Validation_42899459/Validation_42899459_42899459_195938.322.png", "Validation_42899459/Validation_42899459_42899459_195940.321.png", "Validation_42899459/Validation_42899459_42899459_195942.320.png", "Validation_42899459/Validation_42899459_42899459_195944.320.png" ], "turn_id": 3, "scan_id": "Validation_42899459", "origin_question": "Which two objects are on the same side of you now? \"elegant wooden spiral stairs with broad treads\", \"the sleek, rectangular dark metallic fireplace\", \"the sleek rectangular monitor with thin black bezels\". ", "option": [ "elegant wooden spiral stairs with broad treads, the sleek, rectangular dark metallic fireplace", "elegant wooden spiral stairs with broad treads, the sleek rectangular monitor with thin black bezels", "the sleek, rectangular dark metallic fireplace, the sleek rectangular monitor with thin black bezels" ], "answer": "the sleek, rectangular dark metallic fireplace, the sleek rectangular monitor with thin black bezels", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"elegant wooden spiral stairs with broad treads\", \"the sleek, rectangular dark metallic fireplace\", \"the sleek rectangular monitor with thin black bezels\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['elegant wooden spiral stairs with broad treads, the sleek, rectangular dark metallic fireplace', 'elegant wooden spiral stairs with broad treads, the sleek rectangular monitor with thin black bezels', 'the sleek, rectangular dark metallic fireplace, the sleek rectangular monitor with thin black bezels'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_42899459/Validation_42899459_42899459_195946.319.png", "Validation_42899459/Validation_42899459_42899459_195948.318.png", "Validation_42899459/Validation_42899459_42899459_195950.317.png", "Validation_42899459/Validation_42899459_42899459_195952.333.png", "Validation_42899459/Validation_42899459_42899459_195954.332.png" ], "turn_id": 4, "scan_id": "Validation_42899459", "origin_question": "Please recall \"the sleek, rectangular book with the neutral cover\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "4.81", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the sleek, rectangular book with the neutral cover\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_42899459/Validation_42899459_42899459_195956.331.png", "Validation_42899459/Validation_42899459_42899459_195958.330.png", "Validation_42899459/Validation_42899459_42899459_195960.329.png", "Validation_42899459/Validation_42899459_42899459_195962.329.png", "Validation_42899459/Validation_42899459_42899459_195964.328.png" ], "turn_id": 5, "scan_id": "Validation_42899459", "origin_question": "When did you first discover \"the sleek black microwave with the steel exterior\"? (index of the turn)", "option": [], "answer": "5", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the sleek black microwave with the steel exterior\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_42899459/Validation_42899459_42899459_195966.327.png", "Validation_42899459/Validation_42899459_42899459_195968.326.png", "Validation_42899459/Validation_42899459_42899459_195970.325.png", "Validation_42899459/Validation_42899459_42899459_195972.325.png", "Validation_42899459/Validation_42899459_42899459_195974.324.png" ], "turn_id": 6, "scan_id": "Validation_42899459", "origin_question": "What will be the first-time appearance order of the following categories: microwave, stairs, window?", "option": [ "microwave, stairs, window", "stairs, window, microwave", "window, microwave, stairs", "stairs, microwave, window" ], "answer": "stairs, window, microwave", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: microwave, stairs, window? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['microwave, stairs, window', 'stairs, window, microwave', 'window, microwave, stairs', 'stairs, microwave, window'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "Validation_42899459/Validation_42899459_42899459_195976.323.png", "Validation_42899459/Validation_42899459_42899459_195978.322.png", "Validation_42899459/Validation_42899459_42899459_195980.321.png", "Validation_42899459/Validation_42899459_42899459_195982.320.png", "Validation_42899459/Validation_42899459_42899459_195984.320.png" ], "turn_id": 7, "scan_id": "Validation_42899459", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the sleek rectangular monitor with thin black bezels\"; \"the compact, metallic machine with the glossy design\"; \"elegant wooden spiral stairs with broad treads\" ", "option": [ "the sleek rectangular monitor with thin black bezels", "the compact, metallic machine with the glossy design", "elegant wooden spiral stairs with broad treads" ], "answer": "the compact, metallic machine with the glossy design", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the sleek rectangular monitor with thin black bezels\"; \"the compact, metallic machine with the glossy design\"; \"elegant wooden spiral stairs with broad treads\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the sleek rectangular monitor with thin black bezels', 'the compact, metallic machine with the glossy design', 'elegant wooden spiral stairs with broad treads'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_42899459/Validation_42899459_42899459_195986.319.png", "Validation_42899459/Validation_42899459_42899459_195988.318.png", "Validation_42899459/Validation_42899459_42899459_195990.317.png", "Validation_42899459/Validation_42899459_42899459_195992.333.png", "Validation_42899459/Validation_42899459_42899459_195994.332.png" ], "turn_id": 8, "scan_id": "Validation_42899459", "origin_question": "Assuming the direction you are facing at the end of the turn 6 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Right", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 6 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_42899459/Validation_42899459_42899459_195996.331.png", "Validation_42899459/Validation_42899459_42899459_195998.330.png", "Validation_42899459/Validation_42899459_42899459_196000.330.png", "Validation_42899459/Validation_42899459_42899459_196002.329.png", "Validation_42899459/Validation_42899459_42899459_196004.328.png" ], "turn_id": 9, "scan_id": "Validation_42899459", "origin_question": "In which turn did you see both \"the sleek, rectangular dark metallic fireplace\" and \"the compact, metallic machine with the glossy design\" simultaneously? ", "option": [], "answer": "7", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the sleek, rectangular dark metallic fireplace\" and \"the compact, metallic machine with the glossy design\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_42899459/Validation_42899459_42899459_196006.327.png", "Validation_42899459/Validation_42899459_42899459_196008.326.png", "Validation_42899459/Validation_42899459_42899459_196010.326.png", "Validation_42899459/Validation_42899459_42899459_196012.325.png", "Validation_42899459/Validation_42899459_42899459_196014.324.png" ], "turn_id": 10, "scan_id": "Validation_42899459", "origin_question": "Please recall \"the rich chocolate brown rectangular couch with plush fabric\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.28", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the rich chocolate brown rectangular couch with plush fabric\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47333899/Validation_47333899_47333899_81241.075.png", "Validation_47333899/Validation_47333899_47333899_81243.074.png", "Validation_47333899/Validation_47333899_47333899_81245.074.png", "Validation_47333899/Validation_47333899_47333899_81247.073.png" ], "turn_id": 1, "scan_id": "Validation_47333899", "origin_question": "Remember, have you seen any person(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any person(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_47333899/Validation_47333899_47333899_81249.072.png", "Validation_47333899/Validation_47333899_47333899_81251.071.png", "Validation_47333899/Validation_47333899_47333899_81253.070.png", "Validation_47333899/Validation_47333899_47333899_81255.070.png" ], "turn_id": 2, "scan_id": "Validation_47333899", "origin_question": "In which turn did you see both \"the large white rectangular window with the modern frame\" and \"the white, elongated bulb-shaped light\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the large white rectangular window with the modern frame\" and \"the white, elongated bulb-shaped light\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_47333899/Validation_47333899_47333899_81257.069.png", "Validation_47333899/Validation_47333899_47333899_81259.068.png", "Validation_47333899/Validation_47333899_47333899_81261.067.png", "Validation_47333899/Validation_47333899_47333899_81263.066.png" ], "turn_id": 3, "scan_id": "Validation_47333899", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the sleek white rectangular table on four legs\", Y: \"the white, elongated bulb-shaped light\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the sleek white rectangular table on four legs\", Y: \"the white, elongated bulb-shaped light\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_47333899/Validation_47333899_47333899_81265.065.png", "Validation_47333899/Validation_47333899_47333899_81267.065.png", "Validation_47333899/Validation_47333899_47333899_81269.064.png", "Validation_47333899/Validation_47333899_47333899_81271.063.png" ], "turn_id": 4, "scan_id": "Validation_47333899", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_47333899/Validation_47333899_47333899_81273.079.png", "Validation_47333899/Validation_47333899_47333899_81275.078.png", "Validation_47333899/Validation_47333899_47333899_81277.077.png", "Validation_47333899/Validation_47333899_47333899_81279.076.png" ], "turn_id": 5, "scan_id": "Validation_47333899", "origin_question": "Which two objects are on the same side of you now? \"the sleek white rectangular table on four legs\", \"the large white rectangular window with the modern frame\", \"the white, elongated bulb-shaped light\". ", "option": [ "the sleek white rectangular table on four legs, the large white rectangular window with the modern frame", "the sleek white rectangular table on four legs, the white, elongated bulb-shaped light", "the large white rectangular window with the modern frame, the white, elongated bulb-shaped light" ], "answer": "the sleek white rectangular table on four legs, the large white rectangular window with the modern frame", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the sleek white rectangular table on four legs\", \"the large white rectangular window with the modern frame\", \"the white, elongated bulb-shaped light\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the sleek white rectangular table on four legs, the large white rectangular window with the modern frame', 'the sleek white rectangular table on four legs, the white, elongated bulb-shaped light', 'the large white rectangular window with the modern frame, the white, elongated bulb-shaped light'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "Validation_47333899/Validation_47333899_47333899_81281.076.png", "Validation_47333899/Validation_47333899_47333899_81283.075.png", "Validation_47333899/Validation_47333899_47333899_81285.074.png", "Validation_47333899/Validation_47333899_47333899_81287.073.png" ], "turn_id": 6, "scan_id": "Validation_47333899", "origin_question": "Among these three objects(\"the sleek white rectangular table on four legs\",\"the large white rectangular window with the modern frame\",\"the white, elongated bulb-shaped light\" ), which one is the closest to you?", "option": [ "the sleek white rectangular table on four legs", "the large white rectangular window with the modern frame", "the white, elongated bulb-shaped light" ], "answer": "the white, elongated bulb-shaped light", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects(\"the sleek white rectangular table on four legs\",\"the large white rectangular window with the modern frame\",\"the white, elongated bulb-shaped light\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the sleek white rectangular table on four legs', 'the large white rectangular window with the modern frame', 'the white, elongated bulb-shaped light'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_47333899/Validation_47333899_47333899_81289.072.png", "Validation_47333899/Validation_47333899_47333899_81291.072.png", "Validation_47333899/Validation_47333899_47333899_81293.071.png", "Validation_47333899/Validation_47333899_47333899_81295.070.png" ], "turn_id": 7, "scan_id": "Validation_47333899", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "135.9", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "Validation_47333899/Validation_47333899_47333899_81297.069.png", "Validation_47333899/Validation_47333899_47333899_81299.068.png", "Validation_47333899/Validation_47333899_47333899_81301.068.png", "Validation_47333899/Validation_47333899_47333899_81303.067.png" ], "turn_id": 8, "scan_id": "Validation_47333899", "origin_question": "Among these three objects(\"the sleek white rectangular table on four legs\",\"the white, elongated bulb-shaped light\",\"the white, square washing machine with the top control panel\" ), which one is the farthest from you?", "option": [ "the sleek white rectangular table on four legs", "the white, elongated bulb-shaped light", "the white, square washing machine with the top control panel" ], "answer": "the white, square washing machine with the top control panel", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects(\"the sleek white rectangular table on four legs\",\"the white, elongated bulb-shaped light\",\"the white, square washing machine with the top control panel\" ), which one is the farthest from you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the sleek white rectangular table on four legs', 'the white, elongated bulb-shaped light', 'the white, square washing machine with the top control panel'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_47333899/Validation_47333899_47333899_81305.066.png", "Validation_47333899/Validation_47333899_47333899_81307.065.png", "Validation_47333899/Validation_47333899_47333899_81309.064.png", "Validation_47333899/Validation_47333899_47333899_81311.064.png" ], "turn_id": 9, "scan_id": "Validation_47333899", "origin_question": "Using your orientation at the end of turn 7 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "28.32", "user_message": "For the 9 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 7 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_45260928/Validation_45260928_45260928_2338.236.png", "Validation_45260928/Validation_45260928_45260928_2339.719.png", "Validation_45260928/Validation_45260928_45260928_2341.218.png", "Validation_45260928/Validation_45260928_45260928_2342.718.png" ], "turn_id": 1, "scan_id": "Validation_45260928", "origin_question": "Remember, have you seen any table(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any table(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "Validation_45260928/Validation_45260928_45260928_2344.217.png", "Validation_45260928/Validation_45260928_45260928_2345.717.png", "Validation_45260928/Validation_45260928_45260928_2347.216.png", "Validation_45260928/Validation_45260928_45260928_2348.715.png" ], "turn_id": 2, "scan_id": "Validation_45260928", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the classic wooden chair with the dark frame and red cushion\"; \"the sleek black monitor with the modern design\"; \"the glossy ceramic teapot with the bulbous shape\" ", "option": [ "the classic wooden chair with the dark frame and red cushion", "the sleek black monitor with the modern design", "the glossy ceramic teapot with the bulbous shape" ], "answer": "the glossy ceramic teapot with the bulbous shape", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the classic wooden chair with the dark frame and red cushion\"; \"the sleek black monitor with the modern design\"; \"the glossy ceramic teapot with the bulbous shape\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the classic wooden chair with the dark frame and red cushion', 'the sleek black monitor with the modern design', 'the glossy ceramic teapot with the bulbous shape'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_45260928/Validation_45260928_45260928_2350.215.png", "Validation_45260928/Validation_45260928_45260928_2351.714.png", "Validation_45260928/Validation_45260928_45260928_2353.230.png", "Validation_45260928/Validation_45260928_45260928_2354.730.png" ], "turn_id": 3, "scan_id": "Validation_45260928", "origin_question": "Please recall \"the sleek black monitor with the modern design\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.55", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the sleek black monitor with the modern design\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_45260928/Validation_45260928_45260928_2356.229.png", "Validation_45260928/Validation_45260928_45260928_2357.729.png", "Validation_45260928/Validation_45260928_45260928_2359.228.png", "Validation_45260928/Validation_45260928_45260928_2360.727.png" ], "turn_id": 4, "scan_id": "Validation_45260928", "origin_question": "What will be the first-time appearance order of the following categories: couch, table, teapot?", "option": [ "table, teapot, couch", "table, couch, teapot", "couch, table, teapot", "teapot, table, couch" ], "answer": "table, teapot, couch", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: couch, table, teapot? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['table, teapot, couch', 'table, couch, teapot', 'couch, table, teapot', 'teapot, table, couch'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_45260928/Validation_45260928_45260928_2362.227.png", "Validation_45260928/Validation_45260928_45260928_2363.726.png", "Validation_45260928/Validation_45260928_45260928_2365.226.png", "Validation_45260928/Validation_45260928_45260928_2366.725.png" ], "turn_id": 5, "scan_id": "Validation_45260928", "origin_question": "Is the sleek black monitor with the modern design to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the sleek black monitor with the modern design to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_45260928/Validation_45260928_45260928_2368.224.png", "Validation_45260928/Validation_45260928_45260928_2369.724.png", "Validation_45260928/Validation_45260928_45260928_2371.223.png", "Validation_45260928/Validation_45260928_45260928_2372.723.png" ], "turn_id": 6, "scan_id": "Validation_45260928", "origin_question": "Using your orientation at the end of turn 5 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 5 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_45260928/Validation_45260928_45260928_2374.222.png", "Validation_45260928/Validation_45260928_45260928_2375.722.png", "Validation_45260928/Validation_45260928_45260928_2377.221.png", "Validation_45260928/Validation_45260928_45260928_2378.720.png" ], "turn_id": 7, "scan_id": "Validation_45260928", "origin_question": "Please recall \"the glossy ceramic teapot with the bulbous shape\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.17", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the glossy ceramic teapot with the bulbous shape\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_45260928/Validation_45260928_45260928_2380.220.png", "Validation_45260928/Validation_45260928_45260928_2381.719.png", "Validation_45260928/Validation_45260928_45260928_2383.218.png", "Validation_45260928/Validation_45260928_45260928_2384.718.png" ], "turn_id": 8, "scan_id": "Validation_45260928", "origin_question": "Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the dark rectangular fireplace with an ornate design\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the dark rectangular fireplace with an ornate design\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47333920/Validation_47333920_47333920_82545.269.png", "Validation_47333920/Validation_47333920_47333920_82547.251.png", "Validation_47333920/Validation_47333920_47333920_82549.250.png", "Validation_47333920/Validation_47333920_47333920_82551.266.png", "Validation_47333920/Validation_47333920_47333920_82553.265.png" ], "turn_id": 1, "scan_id": "Validation_47333920", "origin_question": "Remember, have you seen any plate(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any plate(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_47333920/Validation_47333920_47333920_82555.265.png", "Validation_47333920/Validation_47333920_47333920_82557.264.png", "Validation_47333920/Validation_47333920_47333920_82559.263.png", "Validation_47333920/Validation_47333920_47333920_82561.262.png", "Validation_47333920/Validation_47333920_47333920_82563.261.png" ], "turn_id": 2, "scan_id": "Validation_47333920", "origin_question": "At the end of which turn were both of \"the rectangular wood table with the white cloth-covered top\" and \"the light\" on your right side?", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the rectangular wood table with the white cloth-covered top\" and \"the light\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_47333920/Validation_47333920_47333920_82565.260.png", "Validation_47333920/Validation_47333920_47333920_82567.260.png", "Validation_47333920/Validation_47333920_47333920_82569.259.png", "Validation_47333920/Validation_47333920_47333920_82571.258.png", "Validation_47333920/Validation_47333920_47333920_82573.257.png" ], "turn_id": 3, "scan_id": "Validation_47333920", "origin_question": "At the end of which turn were both of \"the rectangular wood table with the white cloth-covered top\" and \"the speaker\" on your right side?", "option": [], "answer": "2", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the rectangular wood table with the white cloth-covered top\" and \"the speaker\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_47333920/Validation_47333920_47333920_82575.256.png", "Validation_47333920/Validation_47333920_47333920_82577.256.png", "Validation_47333920/Validation_47333920_47333920_82579.255.png", "Validation_47333920/Validation_47333920_47333920_82581.254.png", "Validation_47333920/Validation_47333920_47333920_82583.253.png" ], "turn_id": 4, "scan_id": "Validation_47333920", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the speaker\" from your position?", "option": [], "answer": "81.45", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the speaker\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_47333920/Validation_47333920_47333920_82585.252.png", "Validation_47333920/Validation_47333920_47333920_82587.251.png", "Validation_47333920/Validation_47333920_47333920_82589.251.png", "Validation_47333920/Validation_47333920_47333920_82591.250.png", "Validation_47333920/Validation_47333920_47333920_82593.266.png" ], "turn_id": 5, "scan_id": "Validation_47333920", "origin_question": "Which two objects are on the same side of you now? \"the rectangular wood table with the white cloth-covered top\", \"the speaker\", \"the large white pvc-framed window\". ", "option": [ "the rectangular wood table with the white cloth-covered top, the speaker", "the rectangular wood table with the white cloth-covered top, the large white pvc-framed window", "the speaker, the large white pvc-framed window" ], "answer": "the speaker, the large white pvc-framed window", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the rectangular wood table with the white cloth-covered top\", \"the speaker\", \"the large white pvc-framed window\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular wood table with the white cloth-covered top, the speaker', 'the rectangular wood table with the white cloth-covered top, the large white pvc-framed window', 'the speaker, the large white pvc-framed window'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_47333920/Validation_47333920_47333920_82595.265.png", "Validation_47333920/Validation_47333920_47333920_82597.264.png", "Validation_47333920/Validation_47333920_47333920_82599.263.png", "Validation_47333920/Validation_47333920_47333920_82601.262.png", "Validation_47333920/Validation_47333920_47333920_82603.262.png" ], "turn_id": 6, "scan_id": "Validation_47333920", "origin_question": "How far is your current position from where you were at the end of the turn 1? (in meters)", "option": [], "answer": "2.68", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 1? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_47333920/Validation_47333920_47333920_82605.261.png", "Validation_47333920/Validation_47333920_47333920_82607.260.png", "Validation_47333920/Validation_47333920_47333920_82609.259.png", "Validation_47333920/Validation_47333920_47333920_82611.258.png", "Validation_47333920/Validation_47333920_47333920_82613.258.png" ], "turn_id": 7, "scan_id": "Validation_47333920", "origin_question": "By the end of the 4 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular wood table with the white cloth-covered top\", Y: \"the large white pvc-framed window\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 4 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular wood table with the white cloth-covered top\", Y: \"the large white pvc-framed window\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_47333920/Validation_47333920_47333920_82615.257.png", "Validation_47333920/Validation_47333920_47333920_82617.256.png", "Validation_47333920/Validation_47333920_47333920_82619.255.png", "Validation_47333920/Validation_47333920_47333920_82621.254.png", "Validation_47333920/Validation_47333920_47333920_82623.254.png" ], "turn_id": 8, "scan_id": "Validation_47333920", "origin_question": "When did you first discover \"the light\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the light\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_47333920/Validation_47333920_47333920_82625.253.png", "Validation_47333920/Validation_47333920_47333920_82627.252.png", "Validation_47333920/Validation_47333920_47333920_82629.251.png", "Validation_47333920/Validation_47333920_47333920_82631.250.png", "Validation_47333920/Validation_47333920_47333920_82633.249.png" ], "turn_id": 9, "scan_id": "Validation_47333920", "origin_question": "Which two objects are on the same side of you now? \"the rectangular wood table with the white cloth-covered top\", \"the light\", \"the large white pvc-framed window\". ", "option": [ "the rectangular wood table with the white cloth-covered top, the light", "the rectangular wood table with the white cloth-covered top, the large white pvc-framed window", "the light, the large white pvc-framed window" ], "answer": "the rectangular wood table with the white cloth-covered top, the light", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the rectangular wood table with the white cloth-covered top\", \"the light\", \"the large white pvc-framed window\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular wood table with the white cloth-covered top, the light', 'the rectangular wood table with the white cloth-covered top, the large white pvc-framed window', 'the light, the large white pvc-framed window'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_47333920/Validation_47333920_47333920_82635.249.png", "Validation_47333920/Validation_47333920_47333920_82637.264.png", "Validation_47333920/Validation_47333920_47333920_82639.264.png", "Validation_47333920/Validation_47333920_47333920_82641.263.png", "Validation_47333920/Validation_47333920_47333920_82643.262.png" ], "turn_id": 10, "scan_id": "Validation_47333920", "origin_question": "Which two objects are on the same side of you now? \"the light\", \"the dark round stool, possibly wooden or metal\", \"the speaker\". ", "option": [ "the light, the dark round stool, possibly wooden or metal", "the light, the speaker", "the dark round stool, possibly wooden or metal, the speaker" ], "answer": "the light, the dark round stool, possibly wooden or metal", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the light\", \"the dark round stool, possibly wooden or metal\", \"the speaker\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the light, the dark round stool, possibly wooden or metal', 'the light, the speaker', 'the dark round stool, possibly wooden or metal, the speaker'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47334091/Validation_47334091_47334091_92065.241.png", "Validation_47334091/Validation_47334091_47334091_92067.224.png", "Validation_47334091/Validation_47334091_47334091_92069.223.png", "Validation_47334091/Validation_47334091_47334091_92071.239.png", "Validation_47334091/Validation_47334091_47334091_92073.238.png" ], "turn_id": 1, "scan_id": "Validation_47334091", "origin_question": "", "option": [], "answer": "", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47334091/Validation_47334091_47334091_92075.237.png", "Validation_47334091/Validation_47334091_47334091_92077.236.png", "Validation_47334091/Validation_47334091_47334091_92079.236.png", "Validation_47334091/Validation_47334091_47334091_92081.235.png", "Validation_47334091/Validation_47334091_47334091_92083.234.png" ], "turn_id": 2, "scan_id": "Validation_47334091", "origin_question": "Remember, have you seen any stool(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any stool(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_47334091/Validation_47334091_47334091_92085.233.png", "Validation_47334091/Validation_47334091_47334091_92087.232.png", "Validation_47334091/Validation_47334091_47334091_92089.232.png", "Validation_47334091/Validation_47334091_47334091_92091.231.png", "Validation_47334091/Validation_47334091_47334091_92093.230.png" ], "turn_id": 3, "scan_id": "Validation_47334091", "origin_question": "In which turn did you see both \"the modern rectangular stool with tufted dark upholstery\" and \"the sleek rectangular monitor with the slim black bezel\" simultaneously? ", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the modern rectangular stool with tufted dark upholstery\" and \"the sleek rectangular monitor with the slim black bezel\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_47334091/Validation_47334091_47334091_92095.229.png", "Validation_47334091/Validation_47334091_47334091_92097.228.png", "Validation_47334091/Validation_47334091_47334091_92099.228.png", "Validation_47334091/Validation_47334091_47334091_92101.227.png", "Validation_47334091/Validation_47334091_47334091_92103.226.png" ], "turn_id": 4, "scan_id": "Validation_47334091", "origin_question": "In which turn did you see both \"the sleek rectangular monitor with the slim black bezel\" and \"the modern rectangular couch with light gray upholstery\" simultaneously? ", "option": [], "answer": "3", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the sleek rectangular monitor with the slim black bezel\" and \"the modern rectangular couch with light gray upholstery\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47334091/Validation_47334091_47334091_92105.225.png", "Validation_47334091/Validation_47334091_47334091_92107.224.png", "Validation_47334091/Validation_47334091_47334091_92109.224.png", "Validation_47334091/Validation_47334091_47334091_92111.223.png", "Validation_47334091/Validation_47334091_47334091_92113.239.png" ], "turn_id": 5, "scan_id": "Validation_47334091", "origin_question": "In which turn were you closest to \"the modern, rectangular stainless steel oven with glass\"? ", "option": [], "answer": "5", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the modern, rectangular stainless steel oven with glass\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47334091/Validation_47334091_47334091_92115.238.png", "Validation_47334091/Validation_47334091_47334091_92117.237.png", "Validation_47334091/Validation_47334091_47334091_92119.236.png", "Validation_47334091/Validation_47334091_47334091_92121.235.png", "Validation_47334091/Validation_47334091_47334091_92123.235.png" ], "turn_id": 6, "scan_id": "Validation_47334091", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the modern, rectangular stainless steel oven with glass\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the modern, rectangular stainless steel oven with glass\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47334091/Validation_47334091_47334091_92125.234.png", "Validation_47334091/Validation_47334091_47334091_92127.233.png", "Validation_47334091/Validation_47334091_47334091_92129.232.png", "Validation_47334091/Validation_47334091_47334091_92131.231.png", "Validation_47334091/Validation_47334091_47334091_92133.231.png" ], "turn_id": 7, "scan_id": "Validation_47334091", "origin_question": "Which direction is the modern rectangular stool with tufted dark upholstery in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the modern rectangular stool with tufted dark upholstery in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47334091/Validation_47334091_47334091_92135.230.png", "Validation_47334091/Validation_47334091_47334091_92137.229.png", "Validation_47334091/Validation_47334091_47334091_92139.228.png", "Validation_47334091/Validation_47334091_47334091_92141.227.png", "Validation_47334091/Validation_47334091_47334091_92143.226.png" ], "turn_id": 8, "scan_id": "Validation_47334091", "origin_question": "", "option": [], "answer": "", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_47334091/Validation_47334091_47334091_92145.226.png", "Validation_47334091/Validation_47334091_47334091_92147.225.png", "Validation_47334091/Validation_47334091_47334091_92149.224.png", "Validation_47334091/Validation_47334091_47334091_92151.223.png", "Validation_47334091/Validation_47334091_47334091_92153.239.png" ], "turn_id": 9, "scan_id": "Validation_47334091", "origin_question": "Which two objects are on the same side of you now? \"the modern rectangular stool with tufted dark upholstery\", \"the modern rectangular couch with light gray upholstery\", \"the compact stainless steel microwave with the digital display\". ", "option": [ "the modern rectangular stool with tufted dark upholstery, the modern rectangular couch with light gray upholstery", "the modern rectangular stool with tufted dark upholstery, the compact stainless steel microwave with the digital display", "the modern rectangular couch with light gray upholstery, the compact stainless steel microwave with the digital display" ], "answer": "the modern rectangular stool with tufted dark upholstery, the modern rectangular couch with light gray upholstery", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the modern rectangular stool with tufted dark upholstery\", \"the modern rectangular couch with light gray upholstery\", \"the compact stainless steel microwave with the digital display\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the modern rectangular stool with tufted dark upholstery, the modern rectangular couch with light gray upholstery', 'the modern rectangular stool with tufted dark upholstery, the compact stainless steel microwave with the digital display', 'the modern rectangular couch with light gray upholstery, the compact stainless steel microwave with the digital display'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_47334091/Validation_47334091_47334091_92155.238.png", "Validation_47334091/Validation_47334091_47334091_92157.237.png", "Validation_47334091/Validation_47334091_47334091_92159.237.png", "Validation_47334091/Validation_47334091_47334091_92161.236.png", "Validation_47334091/Validation_47334091_47334091_92163.235.png" ], "turn_id": 10, "scan_id": "Validation_47334091", "origin_question": "How far is your current position from where you were at the end of the turn 4? (in meters)", "option": [], "answer": "2.05", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 4? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47895357/Validation_47895357_47895357_98547.354.png", "Validation_47895357/Validation_47895357_47895357_98549.336.png", "Validation_47895357/Validation_47895357_47895357_98551.336.png", "Validation_47895357/Validation_47895357_47895357_98553.335.png", "Validation_47895357/Validation_47895357_47895357_98555.334.png" ], "turn_id": 1, "scan_id": "Validation_47895357", "origin_question": "", "option": [], "answer": "", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47895357/Validation_47895357_47895357_98557.333.png", "Validation_47895357/Validation_47895357_47895357_98559.332.png", "Validation_47895357/Validation_47895357_47895357_98561.332.png", "Validation_47895357/Validation_47895357_47895357_98563.331.png", "Validation_47895357/Validation_47895357_47895357_98565.330.png" ], "turn_id": 2, "scan_id": "Validation_47895357", "origin_question": "Remember, have you seen any chair(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any chair(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47895357/Validation_47895357_47895357_98567.329.png", "Validation_47895357/Validation_47895357_47895357_98569.328.png", "Validation_47895357/Validation_47895357_47895357_98571.327.png", "Validation_47895357/Validation_47895357_47895357_98573.327.png", "Validation_47895357/Validation_47895357_47895357_98575.326.png" ], "turn_id": 3, "scan_id": "Validation_47895357", "origin_question": "In which turn were you closest to \"the rectangular book with the single, subtle color\"? ", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the rectangular book with the single, subtle color\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47895357/Validation_47895357_47895357_98577.342.png", "Validation_47895357/Validation_47895357_47895357_98579.341.png", "Validation_47895357/Validation_47895357_47895357_98581.340.png", "Validation_47895357/Validation_47895357_47895357_98583.339.png", "Validation_47895357/Validation_47895357_47895357_98585.338.png" ], "turn_id": 4, "scan_id": "Validation_47895357", "origin_question": "", "option": [], "answer": "", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_47895357/Validation_47895357_47895357_98587.338.png", "Validation_47895357/Validation_47895357_47895357_98589.337.png", "Validation_47895357/Validation_47895357_47895357_98591.336.png", "Validation_47895357/Validation_47895357_47895357_98593.335.png", "Validation_47895357/Validation_47895357_47895357_98595.334.png" ], "turn_id": 5, "scan_id": "Validation_47895357", "origin_question": "What will be the first-time appearance order of the following categories: bed, chair, oven?", "option": [ "oven, chair, bed", "bed, chair, oven", "chair, oven, bed", "bed, oven, chair" ], "answer": "bed, chair, oven", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bed, chair, oven? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['oven, chair, bed', 'bed, chair, oven', 'chair, oven, bed', 'bed, oven, chair'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_47895357/Validation_47895357_47895357_98597.334.png", "Validation_47895357/Validation_47895357_47895357_98599.333.png", "Validation_47895357/Validation_47895357_47895357_98601.332.png", "Validation_47895357/Validation_47895357_47895357_98603.331.png", "Validation_47895357/Validation_47895357_47895357_98605.330.png" ], "turn_id": 6, "scan_id": "Validation_47895357", "origin_question": "When did you first discover \"the compact white plastic bin with the closed lid\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the compact white plastic bin with the closed lid\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47895357/Validation_47895357_47895357_98607.330.png", "Validation_47895357/Validation_47895357_47895357_98609.329.png", "Validation_47895357/Validation_47895357_47895357_98611.328.png", "Validation_47895357/Validation_47895357_47895357_98613.327.png", "Validation_47895357/Validation_47895357_47895357_98615.326.png" ], "turn_id": 7, "scan_id": "Validation_47895357", "origin_question": "Remember, have you seen any faucet(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any faucet(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "Validation_47895357/Validation_47895357_47895357_98617.342.png", "Validation_47895357/Validation_47895357_47895357_98619.341.png", "Validation_47895357/Validation_47895357_47895357_98621.341.png", "Validation_47895357/Validation_47895357_47895357_98623.340.png", "Validation_47895357/Validation_47895357_47895357_98625.339.png" ], "turn_id": 8, "scan_id": "Validation_47895357", "origin_question": "Among these three objects(\"the compact white plastic bin with the closed lid\",\"the large rectangular wooden-framed window with clear glass\",\"the medium-sized light brown cardboard box\" ), which one is the farthest from you?", "option": [ "the compact white plastic bin with the closed lid", "the large rectangular wooden-framed window with clear glass", "the medium-sized light brown cardboard box" ], "answer": "the compact white plastic bin with the closed lid", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the compact white plastic bin with the closed lid\",\"the large rectangular wooden-framed window with clear glass\",\"the medium-sized light brown cardboard box\" ), which one is the farthest from you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the compact white plastic bin with the closed lid', 'the large rectangular wooden-framed window with clear glass', 'the medium-sized light brown cardboard box'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_47895357/Validation_47895357_47895357_98627.338.png", "Validation_47895357/Validation_47895357_47895357_98629.337.png", "Validation_47895357/Validation_47895357_47895357_98631.337.png", "Validation_47895357/Validation_47895357_47895357_98633.336.png", "Validation_47895357/Validation_47895357_47895357_98635.335.png" ], "turn_id": 9, "scan_id": "Validation_47895357", "origin_question": "Using your orientation at the end of turn 7 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 7 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_47895357/Validation_47895357_47895357_98637.334.png", "Validation_47895357/Validation_47895357_47895357_98639.333.png", "Validation_47895357/Validation_47895357_47895357_98641.333.png", "Validation_47895357/Validation_47895357_47895357_98643.332.png", "Validation_47895357/Validation_47895357_47895357_98645.331.png" ], "turn_id": 10, "scan_id": "Validation_47895357", "origin_question": "Is the rectangular book with the single, subtle color to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the rectangular book with the single, subtle color to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_48458656/Validation_48458656_48458656_7627.603.png", "Validation_48458656/Validation_48458656_48458656_7629.586.png", "Validation_48458656/Validation_48458656_48458656_7631.585.png", "Validation_48458656/Validation_48458656_48458656_7633.584.png" ], "turn_id": 1, "scan_id": "Validation_48458656", "origin_question": "Remember, have you seen any towel(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any towel(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_48458656/Validation_48458656_48458656_7635.583.png", "Validation_48458656/Validation_48458656_48458656_7637.583.png", "Validation_48458656/Validation_48458656_48458656_7639.582.png", "Validation_48458656/Validation_48458656_48458656_7641.581.png" ], "turn_id": 2, "scan_id": "Validation_48458656", "origin_question": "In which turn were you closest to \"the compact multi-tier table with the white design\"? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the compact multi-tier table with the white design\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_48458656/Validation_48458656_48458656_7643.597.png", "Validation_48458656/Validation_48458656_48458656_7645.596.png", "Validation_48458656/Validation_48458656_48458656_7647.595.png", "Validation_48458656/Validation_48458656_48458656_7649.594.png" ], "turn_id": 3, "scan_id": "Validation_48458656", "origin_question": "", "option": [], "answer": "", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_48458656/Validation_48458656_48458656_7651.593.png", "Validation_48458656/Validation_48458656_48458656_7653.593.png", "Validation_48458656/Validation_48458656_48458656_7655.592.png", "Validation_48458656/Validation_48458656_48458656_7657.591.png" ], "turn_id": 4, "scan_id": "Validation_48458656", "origin_question": "Which direction is the soft, pink plush carpet for the bathroom in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the soft, pink plush carpet for the bathroom in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_48458656/Validation_48458656_48458656_7659.590.png", "Validation_48458656/Validation_48458656_48458656_7661.589.png", "Validation_48458656/Validation_48458656_48458656_7663.588.png", "Validation_48458656/Validation_48458656_48458656_7665.588.png" ], "turn_id": 5, "scan_id": "Validation_48458656", "origin_question": "In which turn were you closest to \"the compact, white cabinet with modern design\"? ", "option": [], "answer": "2", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the compact, white cabinet with modern design\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "Validation_48458656/Validation_48458656_48458656_7667.587.png", "Validation_48458656/Validation_48458656_48458656_7669.586.png", "Validation_48458656/Validation_48458656_48458656_7671.585.png", "Validation_48458656/Validation_48458656_48458656_7673.584.png" ], "turn_id": 6, "scan_id": "Validation_48458656", "origin_question": "Among these three objects(\"the glossy white oval toilet with the closed lid\",\"the slim, white rectangular radiator made of metal\",\"the compact, white cabinet with modern design\" ), which one is the farthest from you?", "option": [ "the glossy white oval toilet with the closed lid", "the slim, white rectangular radiator made of metal", "the compact, white cabinet with modern design" ], "answer": "the compact, white cabinet with modern design", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects(\"the glossy white oval toilet with the closed lid\",\"the slim, white rectangular radiator made of metal\",\"the compact, white cabinet with modern design\" ), which one is the farthest from you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the glossy white oval toilet with the closed lid', 'the slim, white rectangular radiator made of metal', 'the compact, white cabinet with modern design'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "Validation_48458656/Validation_48458656_48458656_7675.584.png", "Validation_48458656/Validation_48458656_48458656_7677.583.png", "Validation_48458656/Validation_48458656_48458656_7679.582.png", "Validation_48458656/Validation_48458656_48458656_7681.581.png" ], "turn_id": 7, "scan_id": "Validation_48458656", "origin_question": "Among these three objects(\"the soft, pink plush carpet for the bathroom\",\"the sleek sliding door in the soft neutral tone\",\"the medium-sized white shelf with the smooth finish\" ), which one is the closest to you?", "option": [ "the soft, pink plush carpet for the bathroom", "the sleek sliding door in the soft neutral tone", "the medium-sized white shelf with the smooth finish" ], "answer": "the medium-sized white shelf with the smooth finish", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects(\"the soft, pink plush carpet for the bathroom\",\"the sleek sliding door in the soft neutral tone\",\"the medium-sized white shelf with the smooth finish\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the soft, pink plush carpet for the bathroom', 'the sleek sliding door in the soft neutral tone', 'the medium-sized white shelf with the smooth finish'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_45261182/Validation_45261182_45261182_2914.734.png", "Validation_45261182/Validation_45261182_45261182_2916.733.png", "Validation_45261182/Validation_45261182_45261182_2918.733.png", "Validation_45261182/Validation_45261182_45261182_2920.732.png", "Validation_45261182/Validation_45261182_45261182_2922.731.png" ], "turn_id": 1, "scan_id": "Validation_45261182", "origin_question": "Remember, have you seen any toilet(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any toilet(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_45261182/Validation_45261182_45261182_2924.730.png", "Validation_45261182/Validation_45261182_45261182_2926.729.png", "Validation_45261182/Validation_45261182_45261182_2928.729.png", "Validation_45261182/Validation_45261182_45261182_2930.728.png", "Validation_45261182/Validation_45261182_45261182_2932.727.png" ], "turn_id": 2, "scan_id": "Validation_45261182", "origin_question": "When did you first discover \"the sleek rectangular white sink\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the sleek rectangular white sink\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_45261182/Validation_45261182_45261182_2934.726.png", "Validation_45261182/Validation_45261182_45261182_2936.725.png", "Validation_45261182/Validation_45261182_45261182_2938.725.png", "Validation_45261182/Validation_45261182_45261182_2940.724.png", "Validation_45261182/Validation_45261182_45261182_2942.740.png" ], "turn_id": 3, "scan_id": "Validation_45261182", "origin_question": "At the end of which turn were both of \"the sleek rectangular white sink\" and \"the lightweight, cylindrical bin with the white surface\" on your left side?", "option": [], "answer": "2", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the sleek rectangular white sink\" and \"the lightweight, cylindrical bin with the white surface\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_45261182/Validation_45261182_45261182_2944.739.png", "Validation_45261182/Validation_45261182_45261182_2946.738.png", "Validation_45261182/Validation_45261182_45261182_2948.737.png", "Validation_45261182/Validation_45261182_45261182_2950.736.png", "Validation_45261182/Validation_45261182_45261182_2952.736.png" ], "turn_id": 4, "scan_id": "Validation_45261182", "origin_question": "Is the sleek rectangular white sink to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the sleek rectangular white sink to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_45261182/Validation_45261182_45261182_2954.735.png", "Validation_45261182/Validation_45261182_45261182_2956.734.png", "Validation_45261182/Validation_45261182_45261182_2958.733.png", "Validation_45261182/Validation_45261182_45261182_2960.733.png", "Validation_45261182/Validation_45261182_45261182_2962.732.png" ], "turn_id": 5, "scan_id": "Validation_45261182", "origin_question": "In which turn were you farthest from \"the sleek rectangular white sink\"? ", "option": [], "answer": "1", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the sleek rectangular white sink\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_45261182/Validation_45261182_45261182_2964.731.png", "Validation_45261182/Validation_45261182_45261182_2966.730.png", "Validation_45261182/Validation_45261182_45261182_2968.729.png", "Validation_45261182/Validation_45261182_45261182_2970.729.png", "Validation_45261182/Validation_45261182_45261182_2972.728.png" ], "turn_id": 6, "scan_id": "Validation_45261182", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_45261182/Validation_45261182_45261182_2974.727.png", "Validation_45261182/Validation_45261182_45261182_2976.726.png", "Validation_45261182/Validation_45261182_45261182_2978.725.png", "Validation_45261182/Validation_45261182_45261182_2980.725.png", "Validation_45261182/Validation_45261182_45261182_2982.740.png" ], "turn_id": 7, "scan_id": "Validation_45261182", "origin_question": "", "option": [], "answer": "", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_45261182/Validation_45261182_45261182_2984.740.png", "Validation_45261182/Validation_45261182_45261182_2986.739.png", "Validation_45261182/Validation_45261182_45261182_2988.738.png", "Validation_45261182/Validation_45261182_45261182_2990.737.png", "Validation_45261182/Validation_45261182_45261182_2992.737.png" ], "turn_id": 8, "scan_id": "Validation_45261182", "origin_question": "Is the rectangular bed with the soft pink bedspread to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the rectangular bed with the soft pink bedspread to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_45261182/Validation_45261182_45261182_2994.736.png", "Validation_45261182/Validation_45261182_45261182_2996.735.png", "Validation_45261182/Validation_45261182_45261182_2998.734.png", "Validation_45261182/Validation_45261182_45261182_3000.733.png", "Validation_45261182/Validation_45261182_45261182_3002.733.png" ], "turn_id": 9, "scan_id": "Validation_45261182", "origin_question": "Which direction is the lightweight, cylindrical bin with the white surface in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the lightweight, cylindrical bin with the white surface in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_41159555/Validation_41159555_41159555_33469.269.png", "Validation_41159555/Validation_41159555_41159555_33471.252.png", "Validation_41159555/Validation_41159555_41159555_33473.251.png", "Validation_41159555/Validation_41159555_41159555_33475.250.png", "Validation_41159555/Validation_41159555_41159555_33477.250.png" ], "turn_id": 1, "scan_id": "Validation_41159555", "origin_question": "Remember, have you seen any cabinet(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any cabinet(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_41159555/Validation_41159555_41159555_33479.265.png", "Validation_41159555/Validation_41159555_41159555_33481.265.png", "Validation_41159555/Validation_41159555_41159555_33483.264.png", "Validation_41159555/Validation_41159555_41159555_33485.263.png", "Validation_41159555/Validation_41159555_41159555_33487.262.png" ], "turn_id": 2, "scan_id": "Validation_41159555", "origin_question": "Which direction is dark, soft cotton clothes in standard sizes in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is dark, soft cotton clothes in standard sizes in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_41159555/Validation_41159555_41159555_33489.262.png", "Validation_41159555/Validation_41159555_41159555_33491.261.png", "Validation_41159555/Validation_41159555_41159555_33493.260.png", "Validation_41159555/Validation_41159555_41159555_33495.259.png", "Validation_41159555/Validation_41159555_41159555_33497.258.png" ], "turn_id": 3, "scan_id": "Validation_41159555", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the sleek horizontal white metal radiator\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the sleek horizontal white metal radiator\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_41159555/Validation_41159555_41159555_33499.257.png", "Validation_41159555/Validation_41159555_41159555_33501.257.png", "Validation_41159555/Validation_41159555_41159555_33503.256.png", "Validation_41159555/Validation_41159555_41159555_33505.255.png", "Validation_41159555/Validation_41159555_41159555_33507.254.png" ], "turn_id": 4, "scan_id": "Validation_41159555", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_41159555/Validation_41159555_41159555_33509.253.png", "Validation_41159555/Validation_41159555_41159555_33511.253.png", "Validation_41159555/Validation_41159555_41159555_33513.252.png", "Validation_41159555/Validation_41159555_41159555_33515.251.png", "Validation_41159555/Validation_41159555_41159555_33517.250.png" ], "turn_id": 5, "scan_id": "Validation_41159555", "origin_question": "Which direction is the sleek horizontal white metal radiator in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the sleek horizontal white metal radiator in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_41159555/Validation_41159555_41159555_33519.266.png", "Validation_41159555/Validation_41159555_41159555_33521.266.png", "Validation_41159555/Validation_41159555_41159555_33523.265.png", "Validation_41159555/Validation_41159555_41159555_33525.264.png", "Validation_41159555/Validation_41159555_41159555_33527.263.png" ], "turn_id": 6, "scan_id": "Validation_41159555", "origin_question": "At the end of which turn were both of \"the classic wooden bed with the warm finish\" and \"the light wood cabinet with rounded knobs\" on your right side?", "option": [], "answer": "6", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the classic wooden bed with the warm finish\" and \"the light wood cabinet with rounded knobs\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_41159555/Validation_41159555_41159555_33529.262.png", "Validation_41159555/Validation_41159555_41159555_33531.262.png", "Validation_41159555/Validation_41159555_41159555_33533.261.png", "Validation_41159555/Validation_41159555_41159555_33535.260.png", "Validation_41159555/Validation_41159555_41159555_33537.259.png" ], "turn_id": 7, "scan_id": "Validation_41159555", "origin_question": "In which turn were you closest to \"the rustic, medium-brown wooden table\"? ", "option": [], "answer": "7", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the rustic, medium-brown wooden table\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_41159555/Validation_41159555_41159555_33539.258.png", "Validation_41159555/Validation_41159555_41159555_33541.257.png", "Validation_41159555/Validation_41159555_41159555_33543.257.png", "Validation_41159555/Validation_41159555_41159555_33545.256.png", "Validation_41159555/Validation_41159555_41159555_33547.255.png" ], "turn_id": 8, "scan_id": "Validation_41159555", "origin_question": "Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "124.2", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_41159555/Validation_41159555_41159555_33549.254.png", "Validation_41159555/Validation_41159555_41159555_33551.254.png", "Validation_41159555/Validation_41159555_41159555_33553.253.png", "Validation_41159555/Validation_41159555_41159555_33555.252.png", "Validation_41159555/Validation_41159555_41159555_33557.251.png" ], "turn_id": 9, "scan_id": "Validation_41159555", "origin_question": "Assuming the direction you are facing at the end of the turn 5 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Left", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 5 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "Validation_41159555/Validation_41159555_41159555_33559.250.png", "Validation_41159555/Validation_41159555_41159555_33561.266.png", "Validation_41159555/Validation_41159555_41159555_33563.266.png", "Validation_41159555/Validation_41159555_41159555_33565.265.png", "Validation_41159555/Validation_41159555_41159555_33567.264.png" ], "turn_id": 10, "scan_id": "Validation_41159555", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the sturdy beige shoe with durable leather\"; \"the vibrant geometric-patterned carpet in warm colors\"; \"the sleek horizontal white metal radiator\" ", "option": [ "the sturdy beige shoe with durable leather", "the vibrant geometric-patterned carpet in warm colors", "the sleek horizontal white metal radiator" ], "answer": "the sturdy beige shoe with durable leather", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the sturdy beige shoe with durable leather\"; \"the vibrant geometric-patterned carpet in warm colors\"; \"the sleek horizontal white metal radiator\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the sturdy beige shoe with durable leather', 'the vibrant geometric-patterned carpet in warm colors', 'the sleek horizontal white metal radiator'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_45663113/Validation_45663113_45663113_55304.680.png", "Validation_45663113/Validation_45663113_45663113_55306.162.png", "Validation_45663113/Validation_45663113_45663113_55307.662.png", "Validation_45663113/Validation_45663113_45663113_55309.161.png" ], "turn_id": 1, "scan_id": "Validation_45663113", "origin_question": "Remember, have you seen any towel(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any towel(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_45663113/Validation_45663113_45663113_55310.660.png", "Validation_45663113/Validation_45663113_45663113_55312.160.png", "Validation_45663113/Validation_45663113_45663113_55313.659.png", "Validation_45663113/Validation_45663113_45663113_55315.159.png" ], "turn_id": 2, "scan_id": "Validation_45663113", "origin_question": "In which turn did you see both \"the blue and white whale-patterned toilet\" and \"the modern, compact glossy white sink\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the blue and white whale-patterned toilet\" and \"the modern, compact glossy white sink\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_45663113/Validation_45663113_45663113_55316.658.png", "Validation_45663113/Validation_45663113_45663113_55318.157.png", "Validation_45663113/Validation_45663113_45663113_55319.657.png", "Validation_45663113/Validation_45663113_45663113_55321.156.png" ], "turn_id": 3, "scan_id": "Validation_45663113", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the large, soft, absorbent cotton towel\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the large, soft, absorbent cotton towel\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_45663113/Validation_45663113_45663113_55322.672.png", "Validation_45663113/Validation_45663113_45663113_55324.172.png", "Validation_45663113/Validation_45663113_45663113_55325.671.png", "Validation_45663113/Validation_45663113_45663113_55327.170.png" ], "turn_id": 4, "scan_id": "Validation_45663113", "origin_question": "Which direction is the blue and white whale-patterned toilet in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the blue and white whale-patterned toilet in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_45663113/Validation_45663113_45663113_55328.670.png", "Validation_45663113/Validation_45663113_45663113_55330.169.png", "Validation_45663113/Validation_45663113_45663113_55331.668.png", "Validation_45663113/Validation_45663113_45663113_55333.168.png" ], "turn_id": 5, "scan_id": "Validation_45663113", "origin_question": "In which turn were you closest to \"colorful striped clothes made of cotton\"? ", "option": [], "answer": "2", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"colorful striped clothes made of cotton\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_45663113/Validation_45663113_45663113_55334.667.png", "Validation_45663113/Validation_45663113_45663113_55336.167.png", "Validation_45663113/Validation_45663113_45663113_55337.666.png", "Validation_45663113/Validation_45663113_45663113_55339.165.png" ], "turn_id": 6, "scan_id": "Validation_45663113", "origin_question": "Which two objects are on the same side of you now? \"the blue and white whale-patterned toilet\", \"the compact rectangular window with the white upvc frame\", \"colorful striped clothes made of cotton\". ", "option": [ "the blue and white whale-patterned toilet, the compact rectangular window with the white upvc frame", "the blue and white whale-patterned toilet, colorful striped clothes made of cotton", "the compact rectangular window with the white upvc frame, colorful striped clothes made of cotton" ], "answer": "the blue and white whale-patterned toilet, the compact rectangular window with the white upvc frame", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the blue and white whale-patterned toilet\", \"the compact rectangular window with the white upvc frame\", \"colorful striped clothes made of cotton\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the blue and white whale-patterned toilet, the compact rectangular window with the white upvc frame', 'the blue and white whale-patterned toilet, colorful striped clothes made of cotton', 'the compact rectangular window with the white upvc frame, colorful striped clothes made of cotton'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_41159504/Validation_41159504_41159504_34675.004.png", "Validation_41159504/Validation_41159504_41159504_34676.986.png", "Validation_41159504/Validation_41159504_41159504_34678.986.png", "Validation_41159504/Validation_41159504_41159504_34680.985.png", "Validation_41159504/Validation_41159504_41159504_34682.984.png" ], "turn_id": 1, "scan_id": "Validation_41159504", "origin_question": "Remember, have you seen any towel(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any towel(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_41159504/Validation_41159504_41159504_34684.983.png", "Validation_41159504/Validation_41159504_41159504_34686.982.png", "Validation_41159504/Validation_41159504_41159504_34688.982.png", "Validation_41159504/Validation_41159504_41159504_34690.981.png", "Validation_41159504/Validation_41159504_41159504_34692.980.png" ], "turn_id": 2, "scan_id": "Validation_41159504", "origin_question": "In which turn were you closest to \"the compact oval white porcelain sink\"? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the compact oval white porcelain sink\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_41159504/Validation_41159504_41159504_34694.979.png", "Validation_41159504/Validation_41159504_41159504_34696.978.png", "Validation_41159504/Validation_41159504_41159504_34698.977.png", "Validation_41159504/Validation_41159504_41159504_34700.977.png", "Validation_41159504/Validation_41159504_41159504_34702.976.png" ], "turn_id": 3, "scan_id": "Validation_41159504", "origin_question": "When did you first discover \"the polished chrome faucet with the curved spout\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the polished chrome faucet with the curved spout\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_41159504/Validation_41159504_41159504_34704.992.png", "Validation_41159504/Validation_41159504_41159504_34706.991.png", "Validation_41159504/Validation_41159504_41159504_34708.990.png", "Validation_41159504/Validation_41159504_41159504_34710.989.png", "Validation_41159504/Validation_41159504_41159504_34712.988.png" ], "turn_id": 4, "scan_id": "Validation_41159504", "origin_question": "At the end of which turn were both of \"the compact dark brown table\" and \"the polished chrome faucet with the curved spout\" on your right side?", "option": [], "answer": "1", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the compact dark brown table\" and \"the polished chrome faucet with the curved spout\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_41159504/Validation_41159504_41159504_34714.987.png", "Validation_41159504/Validation_41159504_41159504_34716.987.png", "Validation_41159504/Validation_41159504_41159504_34718.986.png", "Validation_41159504/Validation_41159504_41159504_34720.985.png", "Validation_41159504/Validation_41159504_41159504_34722.984.png" ], "turn_id": 5, "scan_id": "Validation_41159504", "origin_question": "Is the light beige modern cabinet with glass-fronted doors to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the light beige modern cabinet with glass-fronted doors to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_41159504/Validation_41159504_41159504_34724.983.png", "Validation_41159504/Validation_41159504_41159504_34726.982.png", "Validation_41159504/Validation_41159504_41159504_34728.982.png", "Validation_41159504/Validation_41159504_41159504_34730.981.png", "Validation_41159504/Validation_41159504_41159504_34732.980.png" ], "turn_id": 6, "scan_id": "Validation_41159504", "origin_question": "Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "131.49", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_41159504/Validation_41159504_41159504_34734.979.png", "Validation_41159504/Validation_41159504_41159504_34736.978.png", "Validation_41159504/Validation_41159504_41159504_34738.978.png", "Validation_41159504/Validation_41159504_41159504_34740.977.png", "Validation_41159504/Validation_41159504_41159504_34742.976.png" ], "turn_id": 7, "scan_id": "Validation_41159504", "origin_question": "Which direction is the sleek oval white bathtub with the glossy finish in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the sleek oval white bathtub with the glossy finish in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_41159504/Validation_41159504_41159504_34744.975.png", "Validation_41159504/Validation_41159504_41159504_34746.974.png", "Validation_41159504/Validation_41159504_41159504_34748.990.png", "Validation_41159504/Validation_41159504_41159504_34750.989.png", "Validation_41159504/Validation_41159504_41159504_34752.988.png" ], "turn_id": 8, "scan_id": "Validation_41159504", "origin_question": "In which turn were you closest to \"the white elongated ceramic toilet\"? ", "option": [], "answer": "8", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the white elongated ceramic toilet\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "Validation_41159504/Validation_41159504_41159504_34754.988.png", "Validation_41159504/Validation_41159504_41159504_34756.987.png", "Validation_41159504/Validation_41159504_41159504_34758.986.png", "Validation_41159504/Validation_41159504_41159504_34760.985.png", "Validation_41159504/Validation_41159504_41159504_34762.984.png" ], "turn_id": 9, "scan_id": "Validation_41159504", "origin_question": "Among these three objects(\"the white elongated ceramic toilet\",\"the sleek dark wooden bed with the minimalist headboard\",\"the medium rectangular beige woven basket\" ), which one is the farthest from you?", "option": [ "the white elongated ceramic toilet", "the sleek dark wooden bed with the minimalist headboard", "the medium rectangular beige woven basket" ], "answer": "the sleek dark wooden bed with the minimalist headboard", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the white elongated ceramic toilet\",\"the sleek dark wooden bed with the minimalist headboard\",\"the medium rectangular beige woven basket\" ), which one is the farthest from you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the white elongated ceramic toilet', 'the sleek dark wooden bed with the minimalist headboard', 'the medium rectangular beige woven basket'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_41159504/Validation_41159504_41159504_34764.983.png", "Validation_41159504/Validation_41159504_41159504_34766.983.png", "Validation_41159504/Validation_41159504_41159504_34768.982.png", "Validation_41159504/Validation_41159504_41159504_34770.981.png", "Validation_41159504/Validation_41159504_41159504_34772.980.png" ], "turn_id": 10, "scan_id": "Validation_41159504", "origin_question": "Please recall \"the sleek dark wooden bed with the minimalist headboard\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "3.78", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the sleek dark wooden bed with the minimalist headboard\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_44358451/Validation_44358451_44358451_43605.908.png", "Validation_44358451/Validation_44358451_44358451_43607.891.png", "Validation_44358451/Validation_44358451_44358451_43609.890.png", "Validation_44358451/Validation_44358451_44358451_43611.889.png", "Validation_44358451/Validation_44358451_44358451_43613.888.png" ], "turn_id": 1, "scan_id": "Validation_44358451", "origin_question": "Remember, have you seen any oven(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any oven(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_44358451/Validation_44358451_44358451_43615.887.png", "Validation_44358451/Validation_44358451_44358451_43617.886.png", "Validation_44358451/Validation_44358451_44358451_43619.886.png", "Validation_44358451/Validation_44358451_44358451_43621.885.png", "Validation_44358451/Validation_44358451_44358451_43623.884.png" ], "turn_id": 2, "scan_id": "Validation_44358451", "origin_question": "In which turn were you farthest from \"the sleek, dark-colored shoe made of synthetic fabric\"? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the sleek, dark-colored shoe made of synthetic fabric\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_44358451/Validation_44358451_44358451_43625.883.png", "Validation_44358451/Validation_44358451_44358451_43627.882.png", "Validation_44358451/Validation_44358451_44358451_43629.881.png", "Validation_44358451/Validation_44358451_44358451_43631.897.png", "Validation_44358451/Validation_44358451_44358451_43633.896.png" ], "turn_id": 3, "scan_id": "Validation_44358451", "origin_question": "When did you first discover \"the sleek white steel oven with the modern design\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the sleek white steel oven with the modern design\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_44358451/Validation_44358451_44358451_43635.896.png", "Validation_44358451/Validation_44358451_44358451_43637.895.png", "Validation_44358451/Validation_44358451_44358451_43639.894.png", "Validation_44358451/Validation_44358451_44358451_43641.893.png", "Validation_44358451/Validation_44358451_44358451_43643.892.png" ], "turn_id": 4, "scan_id": "Validation_44358451", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "78.21", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_44358451/Validation_44358451_44358451_43645.891.png", "Validation_44358451/Validation_44358451_44358451_43647.891.png", "Validation_44358451/Validation_44358451_44358451_43649.890.png", "Validation_44358451/Validation_44358451_44358451_43651.889.png", "Validation_44358451/Validation_44358451_44358451_43653.888.png" ], "turn_id": 5, "scan_id": "Validation_44358451", "origin_question": "", "option": [], "answer": "", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_44358451/Validation_44358451_44358451_43655.887.png", "Validation_44358451/Validation_44358451_44358451_43657.887.png", "Validation_44358451/Validation_44358451_44358451_43659.886.png", "Validation_44358451/Validation_44358451_44358451_43661.885.png", "Validation_44358451/Validation_44358451_44358451_43663.884.png" ], "turn_id": 6, "scan_id": "Validation_44358451", "origin_question": "In which turn did you see both \"the sleek white steel oven with the modern design\" and \"the rack\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the sleek white steel oven with the modern design\" and \"the rack\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_44358451/Validation_44358451_44358451_43665.883.png", "Validation_44358451/Validation_44358451_44358451_43667.882.png", "Validation_44358451/Validation_44358451_44358451_43669.882.png", "Validation_44358451/Validation_44358451_44358451_43671.897.png", "Validation_44358451/Validation_44358451_44358451_43673.897.png" ], "turn_id": 7, "scan_id": "Validation_44358451", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the sleek, dark-colored shoe made of synthetic fabric\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the sleek, dark-colored shoe made of synthetic fabric\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_44358451/Validation_44358451_44358451_43675.896.png", "Validation_44358451/Validation_44358451_44358451_43677.895.png", "Validation_44358451/Validation_44358451_44358451_43679.894.png", "Validation_44358451/Validation_44358451_44358451_43681.893.png", "Validation_44358451/Validation_44358451_44358451_43683.892.png" ], "turn_id": 8, "scan_id": "Validation_44358451", "origin_question": "What will be the first-time appearance order of the following categories: cabinet, clothes, oven?", "option": [ "oven, cabinet, clothes", "cabinet, oven, clothes", "cabinet, clothes, oven", "oven, clothes, cabinet" ], "answer": "cabinet, oven, clothes", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: cabinet, clothes, oven? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['oven, cabinet, clothes', 'cabinet, oven, clothes', 'cabinet, clothes, oven', 'oven, clothes, cabinet'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_44358451/Validation_44358451_44358451_43685.891.png", "Validation_44358451/Validation_44358451_44358451_43687.891.png", "Validation_44358451/Validation_44358451_44358451_43689.890.png", "Validation_44358451/Validation_44358451_44358451_43691.889.png", "Validation_44358451/Validation_44358451_44358451_43693.888.png" ], "turn_id": 9, "scan_id": "Validation_44358451", "origin_question": "Assuming the direction you are facing at the end of the turn 5 is forward, did you move a certain distance forward or backward from that position?", "option": [ "Forward", "Backward" ], "answer": "Backward", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 5 is forward, did you move a certain distance forward or backward from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Forward', 'Backward'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_44358451/Validation_44358451_44358451_43695.887.png", "Validation_44358451/Validation_44358451_44358451_43697.887.png", "Validation_44358451/Validation_44358451_44358451_43699.886.png", "Validation_44358451/Validation_44358451_44358451_43701.885.png", "Validation_44358451/Validation_44358451_44358451_43703.884.png" ], "turn_id": 10, "scan_id": "Validation_44358451", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the sleek white steel oven with the modern design\" from your position?", "option": [], "answer": "68.22", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the sleek white steel oven with the modern design\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_44358442/Validation_44358442_44358442_42891.117.png", "Validation_44358442/Validation_44358442_44358442_42893.099.png", "Validation_44358442/Validation_44358442_44358442_42895.098.png", "Validation_44358442/Validation_44358442_44358442_42897.098.png" ], "turn_id": 1, "scan_id": "Validation_44358442", "origin_question": "Remember, have you seen any clothes(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any clothes(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_44358442/Validation_44358442_44358442_42899.097.png", "Validation_44358442/Validation_44358442_44358442_42901.096.png", "Validation_44358442/Validation_44358442_44358442_42903.095.png", "Validation_44358442/Validation_44358442_44358442_42905.094.png" ], "turn_id": 2, "scan_id": "Validation_44358442", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_44358442/Validation_44358442_44358442_42907.093.png", "Validation_44358442/Validation_44358442_44358442_42909.093.png", "Validation_44358442/Validation_44358442_44358442_42911.092.png", "Validation_44358442/Validation_44358442_44358442_42913.091.png" ], "turn_id": 3, "scan_id": "Validation_44358442", "origin_question": "Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "88.98", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_44358442/Validation_44358442_44358442_42915.090.png", "Validation_44358442/Validation_44358442_44358442_42917.089.png", "Validation_44358442/Validation_44358442_44358442_42919.105.png", "Validation_44358442/Validation_44358442_44358442_42921.104.png" ], "turn_id": 4, "scan_id": "Validation_44358442", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the sophisticated collection of tailored clothes\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the sophisticated collection of tailored clothes\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_44358442/Validation_44358442_44358442_42923.103.png", "Validation_44358442/Validation_44358442_44358442_42925.103.png", "Validation_44358442/Validation_44358442_44358442_42927.102.png", "Validation_44358442/Validation_44358442_44358442_42929.101.png" ], "turn_id": 5, "scan_id": "Validation_44358442", "origin_question": "Which two objects are on the same side of you now? \"the large floral-patterned couch with light blue and pink roses\", \"the sleek black ergonomic chair\", \"the sophisticated collection of tailored clothes\". ", "option": [ "the large floral-patterned couch with light blue and pink roses, the sleek black ergonomic chair", "the large floral-patterned couch with light blue and pink roses, the sophisticated collection of tailored clothes", "the sleek black ergonomic chair, the sophisticated collection of tailored clothes" ], "answer": "the sleek black ergonomic chair, the sophisticated collection of tailored clothes", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the large floral-patterned couch with light blue and pink roses\", \"the sleek black ergonomic chair\", \"the sophisticated collection of tailored clothes\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the large floral-patterned couch with light blue and pink roses, the sleek black ergonomic chair', 'the large floral-patterned couch with light blue and pink roses, the sophisticated collection of tailored clothes', 'the sleek black ergonomic chair, the sophisticated collection of tailored clothes'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_44358442/Validation_44358442_44358442_42931.100.png", "Validation_44358442/Validation_44358442_44358442_42933.099.png", "Validation_44358442/Validation_44358442_44358442_42935.099.png", "Validation_44358442/Validation_44358442_44358442_42937.098.png" ], "turn_id": 6, "scan_id": "Validation_44358442", "origin_question": "When did you first discover \"the sleek black ergonomic chair\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the sleek black ergonomic chair\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47332908/Validation_47332908_47332908_12123.255.png", "Validation_47332908/Validation_47332908_47332908_12125.238.png", "Validation_47332908/Validation_47332908_47332908_12127.237.png", "Validation_47332908/Validation_47332908_47332908_12129.236.png" ], "turn_id": 1, "scan_id": "Validation_47332908", "origin_question": "Remember, have you seen any cushion(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any cushion(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47332908/Validation_47332908_47332908_12131.235.png", "Validation_47332908/Validation_47332908_47332908_12133.235.png", "Validation_47332908/Validation_47332908_47332908_12135.234.png", "Validation_47332908/Validation_47332908_47332908_12137.233.png" ], "turn_id": 2, "scan_id": "Validation_47332908", "origin_question": "In which turn were you closest to \"the medium-sized white wall-mounted shelf\"? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the medium-sized white wall-mounted shelf\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_47332908/Validation_47332908_47332908_12139.232.png", "Validation_47332908/Validation_47332908_47332908_12141.231.png", "Validation_47332908/Validation_47332908_47332908_12143.231.png", "Validation_47332908/Validation_47332908_47332908_12145.230.png" ], "turn_id": 3, "scan_id": "Validation_47332908", "origin_question": "When did you first discover \"the medium-sized white wall-mounted shelf\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the medium-sized white wall-mounted shelf\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47332908/Validation_47332908_47332908_12147.229.png", "Validation_47332908/Validation_47332908_47332908_12149.228.png", "Validation_47332908/Validation_47332908_47332908_12151.227.png", "Validation_47332908/Validation_47332908_47332908_12153.227.png" ], "turn_id": 4, "scan_id": "Validation_47332908", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the medium-sized white wall-mounted shelf\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the medium-sized white wall-mounted shelf\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_47332908/Validation_47332908_47332908_12155.226.png", "Validation_47332908/Validation_47332908_47332908_12157.225.png", "Validation_47332908/Validation_47332908_47332908_12159.224.png", "Validation_47332908/Validation_47332908_47332908_12161.240.png" ], "turn_id": 5, "scan_id": "Validation_47332908", "origin_question": "When did you first discover \"the medium-sized white wall-mounted shelf\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the medium-sized white wall-mounted shelf\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_47332908/Validation_47332908_47332908_12163.239.png", "Validation_47332908/Validation_47332908_47332908_12165.238.png", "Validation_47332908/Validation_47332908_47332908_12167.237.png", "Validation_47332908/Validation_47332908_47332908_12169.237.png" ], "turn_id": 6, "scan_id": "Validation_47332908", "origin_question": "Is the rectangular, gleaming white enamel bathtub to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the rectangular, gleaming white enamel bathtub to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_47332908/Validation_47332908_47332908_12171.236.png", "Validation_47332908/Validation_47332908_47332908_12173.235.png", "Validation_47332908/Validation_47332908_47332908_12175.234.png", "Validation_47332908/Validation_47332908_47332908_12177.234.png" ], "turn_id": 7, "scan_id": "Validation_47332908", "origin_question": "What will be the first-time appearance order of the following categories: bathtub, shelf, washing machine?", "option": [ "bathtub, shelf, washing machine", "shelf, bathtub, washing machine", "washing machine, shelf, bathtub", "bathtub, washing machine, shelf" ], "answer": "shelf, bathtub, washing machine", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bathtub, shelf, washing machine? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['bathtub, shelf, washing machine', 'shelf, bathtub, washing machine', 'washing machine, shelf, bathtub', 'bathtub, washing machine, shelf'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_48458660/Validation_48458660_48458660_8243.956.png", "Validation_48458660/Validation_48458660_48458660_8245.455.png", "Validation_48458660/Validation_48458660_48458660_8246.954.png", "Validation_48458660/Validation_48458660_48458660_8248.454.png" ], "turn_id": 1, "scan_id": "Validation_48458660", "origin_question": "Remember, have you seen any door knob(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any door knob(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_48458660/Validation_48458660_48458660_8249.953.png", "Validation_48458660/Validation_48458660_48458660_8251.452.png", "Validation_48458660/Validation_48458660_48458660_8252.952.png", "Validation_48458660/Validation_48458660_48458660_8254.451.png" ], "turn_id": 2, "scan_id": "Validation_48458660", "origin_question": "Which direction is the square white mesh bucket in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the square white mesh bucket in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_48458660/Validation_48458660_48458660_8255.951.png", "Validation_48458660/Validation_48458660_48458660_8257.450.png", "Validation_48458660/Validation_48458660_48458660_8258.949.png", "Validation_48458660/Validation_48458660_48458660_8260.449.png" ], "turn_id": 3, "scan_id": "Validation_48458660", "origin_question": "Which direction is the sleek dark rectangular monitor with the slim bezel in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the sleek dark rectangular monitor with the slim bezel in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_48458660/Validation_48458660_48458660_8261.948.png", "Validation_48458660/Validation_48458660_48458660_8263.448.png", "Validation_48458660/Validation_48458660_48458660_8264.964.png", "Validation_48458660/Validation_48458660_48458660_8266.463.png" ], "turn_id": 4, "scan_id": "Validation_48458660", "origin_question": "In which turn did you see both \"the square white mesh bucket\" and \"the compact white steel radiator\" simultaneously? ", "option": [], "answer": "4", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the square white mesh bucket\" and \"the compact white steel radiator\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42897538/Validation_42897538_42897538_275492.837.png", "Validation_42897538/Validation_42897538_42897538_275494.836.png", "Validation_42897538/Validation_42897538_42897538_275496.836.png", "Validation_42897538/Validation_42897538_42897538_275498.835.png", "Validation_42897538/Validation_42897538_42897538_275500.834.png" ], "turn_id": 1, "scan_id": "Validation_42897538", "origin_question": "Remember, have you seen any table(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any table(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_42897538/Validation_42897538_42897538_275502.833.png", "Validation_42897538/Validation_42897538_42897538_275504.832.png", "Validation_42897538/Validation_42897538_42897538_275506.832.png", "Validation_42897538/Validation_42897538_42897538_275508.831.png", "Validation_42897538/Validation_42897538_42897538_275510.830.png" ], "turn_id": 2, "scan_id": "Validation_42897538", "origin_question": "In which turn were you closest to \"an elegant beige lamp with the metal base\"? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"an elegant beige lamp with the metal base\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_42897538/Validation_42897538_42897538_275512.829.png", "Validation_42897538/Validation_42897538_42897538_275514.828.png", "Validation_42897538/Validation_42897538_42897538_275516.828.png", "Validation_42897538/Validation_42897538_42897538_275518.827.png", "Validation_42897538/Validation_42897538_42897538_275520.826.png" ], "turn_id": 3, "scan_id": "Validation_42897538", "origin_question": "When did you first discover \"the sleek two-seater couch with beige fabric and rounded backrest\"? (index of the turn)", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the sleek two-seater couch with beige fabric and rounded backrest\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_42897538/Validation_42897538_42897538_275522.825.png", "Validation_42897538/Validation_42897538_42897538_275524.841.png", "Validation_42897538/Validation_42897538_42897538_275526.840.png", "Validation_42897538/Validation_42897538_42897538_275528.839.png", "Validation_42897538/Validation_42897538_42897538_275530.838.png" ], "turn_id": 4, "scan_id": "Validation_42897538", "origin_question": "Which direction is an elegant beige lamp with the metal base in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is an elegant beige lamp with the metal base in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_42897538/Validation_42897538_42897538_275532.838.png", "Validation_42897538/Validation_42897538_42897538_275534.837.png", "Validation_42897538/Validation_42897538_42897538_275536.836.png", "Validation_42897538/Validation_42897538_42897538_275538.835.png", "Validation_42897538/Validation_42897538_42897538_275540.834.png" ], "turn_id": 5, "scan_id": "Validation_42897538", "origin_question": "Is an elegant beige lamp with the metal base to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Is an elegant beige lamp with the metal base to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_42897538/Validation_42897538_42897538_275542.834.png", "Validation_42897538/Validation_42897538_42897538_275544.833.png", "Validation_42897538/Validation_42897538_42897538_275546.832.png", "Validation_42897538/Validation_42897538_42897538_275548.831.png", "Validation_42897538/Validation_42897538_42897538_275550.830.png" ], "turn_id": 6, "scan_id": "Validation_42897538", "origin_question": "How far is your current position from where you were at the end of the turn 3? (in meters)", "option": [], "answer": "3.05", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 3? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_42897538/Validation_42897538_42897538_275552.829.png", "Validation_42897538/Validation_42897538_42897538_275554.829.png", "Validation_42897538/Validation_42897538_42897538_275556.828.png", "Validation_42897538/Validation_42897538_42897538_275558.827.png", "Validation_42897538/Validation_42897538_42897538_275560.826.png" ], "turn_id": 7, "scan_id": "Validation_42897538", "origin_question": "What will be the first-time appearance order of the following categories: couch, monitor, table?", "option": [ "monitor, table, couch", "couch, table, monitor", "couch, monitor, table", "table, monitor, couch" ], "answer": "monitor, table, couch", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: couch, monitor, table? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['monitor, table, couch', 'couch, table, monitor', 'couch, monitor, table', 'table, monitor, couch'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_42897538/Validation_42897538_42897538_275562.825.png", "Validation_42897538/Validation_42897538_42897538_275564.824.png", "Validation_42897538/Validation_42897538_42897538_275566.824.png", "Validation_42897538/Validation_42897538_42897538_275568.840.png", "Validation_42897538/Validation_42897538_42897538_275570.839.png" ], "turn_id": 8, "scan_id": "Validation_42897538", "origin_question": "Compared to your position at the end of the 7 turn, are you now closer or farther away from \"an elegant beige lamp with the metal base\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 7 turn, are you now closer or farther away from \"an elegant beige lamp with the metal base\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_42897538/Validation_42897538_42897538_275572.838.png", "Validation_42897538/Validation_42897538_42897538_275574.837.png", "Validation_42897538/Validation_42897538_42897538_275576.836.png", "Validation_42897538/Validation_42897538_42897538_275578.835.png", "Validation_42897538/Validation_42897538_42897538_275580.834.png" ], "turn_id": 9, "scan_id": "Validation_42897538", "origin_question": "How far is your current position from where you were at the end of the turn 4? (in meters)", "option": [], "answer": "2.58", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 4? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_42897538/Validation_42897538_42897538_275582.834.png", "Validation_42897538/Validation_42897538_42897538_275584.833.png", "Validation_42897538/Validation_42897538_42897538_275586.832.png", "Validation_42897538/Validation_42897538_42897538_275588.831.png", "Validation_42897538/Validation_42897538_42897538_275590.830.png" ], "turn_id": 10, "scan_id": "Validation_42897538", "origin_question": "", "option": [], "answer": "", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_45260856/Validation_45260856_45260856_626.298.png", "Validation_45260856/Validation_45260856_45260856_628.280.png", "Validation_45260856/Validation_45260856_45260856_630.279.png", "Validation_45260856/Validation_45260856_45260856_632.278.png", "Validation_45260856/Validation_45260856_45260856_634.278.png" ], "turn_id": 1, "scan_id": "Validation_45260856", "origin_question": "Remember, have you seen any bread(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any bread(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_45260856/Validation_45260856_45260856_636.277.png", "Validation_45260856/Validation_45260856_45260856_638.293.png", "Validation_45260856/Validation_45260856_45260856_640.292.png", "Validation_45260856/Validation_45260856_45260856_642.291.png", "Validation_45260856/Validation_45260856_45260856_644.290.png" ], "turn_id": 2, "scan_id": "Validation_45260856", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_45260856/Validation_45260856_45260856_646.289.png", "Validation_45260856/Validation_45260856_45260856_648.288.png", "Validation_45260856/Validation_45260856_45260856_650.288.png", "Validation_45260856/Validation_45260856_45260856_652.287.png", "Validation_45260856/Validation_45260856_45260856_654.286.png" ], "turn_id": 3, "scan_id": "Validation_45260856", "origin_question": "In which turn were you farthest from \"the colorful floral bed with the metal frame\"? ", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the colorful floral bed with the metal frame\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_45260856/Validation_45260856_45260856_656.285.png", "Validation_45260856/Validation_45260856_45260856_658.284.png", "Validation_45260856/Validation_45260856_45260856_660.283.png", "Validation_45260856/Validation_45260856_45260856_662.283.png", "Validation_45260856/Validation_45260856_45260856_664.282.png" ], "turn_id": 4, "scan_id": "Validation_45260856", "origin_question": "", "option": [], "answer": "", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_45260856/Validation_45260856_45260856_666.281.png", "Validation_45260856/Validation_45260856_45260856_668.280.png", "Validation_45260856/Validation_45260856_45260856_670.279.png", "Validation_45260856/Validation_45260856_45260856_672.278.png", "Validation_45260856/Validation_45260856_45260856_674.278.png" ], "turn_id": 5, "scan_id": "Validation_45260856", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the medium-sized red bag made of durable fabric\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the medium-sized red bag made of durable fabric\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_45260856/Validation_45260856_45260856_676.277.png", "Validation_45260856/Validation_45260856_45260856_678.276.png", "Validation_45260856/Validation_45260856_45260856_680.292.png", "Validation_45260856/Validation_45260856_45260856_682.291.png", "Validation_45260856/Validation_45260856_45260856_684.290.png" ], "turn_id": 6, "scan_id": "Validation_45260856", "origin_question": "Using your orientation at the end of turn 4 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "116.95", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 4 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_45260856/Validation_45260856_45260856_686.289.png", "Validation_45260856/Validation_45260856_45260856_688.288.png", "Validation_45260856/Validation_45260856_45260856_690.288.png", "Validation_45260856/Validation_45260856_45260856_692.287.png", "Validation_45260856/Validation_45260856_45260856_694.286.png" ], "turn_id": 7, "scan_id": "Validation_45260856", "origin_question": "Compared to your position at the end of the 5 turn, are you now closer or farther away from \"the fireplace\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 5 turn, are you now closer or farther away from \"the fireplace\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_45260856/Validation_45260856_45260856_696.285.png", "Validation_45260856/Validation_45260856_45260856_698.284.png", "Validation_45260856/Validation_45260856_45260856_700.284.png", "Validation_45260856/Validation_45260856_45260856_702.283.png", "Validation_45260856/Validation_45260856_45260856_704.282.png" ], "turn_id": 8, "scan_id": "Validation_45260856", "origin_question": "Compared to your position at the end of the 5 turn, are you now closer or farther away from \"the fireplace\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 5 turn, are you now closer or farther away from \"the fireplace\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_45260856/Validation_45260856_45260856_706.281.png", "Validation_45260856/Validation_45260856_45260856_708.280.png", "Validation_45260856/Validation_45260856_45260856_710.279.png", "Validation_45260856/Validation_45260856_45260856_712.279.png", "Validation_45260856/Validation_45260856_45260856_714.278.png" ], "turn_id": 9, "scan_id": "Validation_45260856", "origin_question": "Compared to your position at the end of the 7 turn, are you now closer or farther away from \"the fireplace\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 7 turn, are you now closer or farther away from \"the fireplace\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_45260856/Validation_45260856_45260856_716.277.png", "Validation_45260856/Validation_45260856_45260856_718.276.png", "Validation_45260856/Validation_45260856_45260856_720.275.png", "Validation_45260856/Validation_45260856_45260856_722.291.png", "Validation_45260856/Validation_45260856_45260856_724.290.png" ], "turn_id": 10, "scan_id": "Validation_45260856", "origin_question": "Compared to your position at the end of the 8 turn, are you now closer or farther away from \"the colorful floral bed with the metal frame\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 8 turn, are you now closer or farther away from \"the colorful floral bed with the metal frame\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42446100/Validation_42446100_42446100_77907.297.png", "Validation_42446100/Validation_42446100_42446100_77909.280.png", "Validation_42446100/Validation_42446100_42446100_77911.295.png", "Validation_42446100/Validation_42446100_42446100_77913.295.png", "Validation_42446100/Validation_42446100_42446100_77915.294.png" ], "turn_id": 1, "scan_id": "Validation_42446100", "origin_question": "Remember, have you seen any tray(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any tray(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_42446100/Validation_42446100_42446100_77917.293.png", "Validation_42446100/Validation_42446100_42446100_77919.292.png", "Validation_42446100/Validation_42446100_42446100_77921.291.png", "Validation_42446100/Validation_42446100_42446100_77923.291.png", "Validation_42446100/Validation_42446100_42446100_77925.290.png" ], "turn_id": 2, "scan_id": "Validation_42446100", "origin_question": "Which direction is the circular polished wood table with the yellow tablecloth in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the circular polished wood table with the yellow tablecloth in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_42446100/Validation_42446100_42446100_77927.289.png", "Validation_42446100/Validation_42446100_42446100_77929.288.png", "Validation_42446100/Validation_42446100_42446100_77931.287.png", "Validation_42446100/Validation_42446100_42446100_77933.287.png", "Validation_42446100/Validation_42446100_42446100_77935.286.png" ], "turn_id": 3, "scan_id": "Validation_42446100", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the transparent cylindrical glass vase\" from your position?", "option": [], "answer": "113.22", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the transparent cylindrical glass vase\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_42446100/Validation_42446100_42446100_77937.285.png", "Validation_42446100/Validation_42446100_42446100_77939.284.png", "Validation_42446100/Validation_42446100_42446100_77941.284.png", "Validation_42446100/Validation_42446100_42446100_77943.283.png", "Validation_42446100/Validation_42446100_42446100_77945.282.png" ], "turn_id": 4, "scan_id": "Validation_42446100", "origin_question": "", "option": [], "answer": "", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_42446100/Validation_42446100_42446100_77947.281.png", "Validation_42446100/Validation_42446100_42446100_77949.280.png", "Validation_42446100/Validation_42446100_42446100_77951.280.png", "Validation_42446100/Validation_42446100_42446100_77953.295.png", "Validation_42446100/Validation_42446100_42446100_77955.295.png" ], "turn_id": 5, "scan_id": "Validation_42446100", "origin_question": "In which turn were you closest to \"the compact three-tier metal shelf with the polished finish\"? ", "option": [], "answer": "2", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the compact three-tier metal shelf with the polished finish\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_42446100/Validation_42446100_42446100_77957.294.png", "Validation_42446100/Validation_42446100_42446100_77959.293.png", "Validation_42446100/Validation_42446100_42446100_77961.292.png", "Validation_42446100/Validation_42446100_42446100_77963.291.png", "Validation_42446100/Validation_42446100_42446100_77965.291.png" ], "turn_id": 6, "scan_id": "Validation_42446100", "origin_question": "Please recall \"the transparent cylindrical glass vase\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.62", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the transparent cylindrical glass vase\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_42446100/Validation_42446100_42446100_77967.290.png", "Validation_42446100/Validation_42446100_42446100_77969.289.png", "Validation_42446100/Validation_42446100_42446100_77971.288.png", "Validation_42446100/Validation_42446100_42446100_77973.288.png", "Validation_42446100/Validation_42446100_42446100_77975.287.png" ], "turn_id": 7, "scan_id": "Validation_42446100", "origin_question": "Please recall \"the sleek rectangular stainless steel sink\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "4.29", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the sleek rectangular stainless steel sink\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "Validation_42446100/Validation_42446100_42446100_77977.286.png", "Validation_42446100/Validation_42446100_42446100_77979.285.png", "Validation_42446100/Validation_42446100_42446100_77981.284.png", "Validation_42446100/Validation_42446100_42446100_77983.284.png", "Validation_42446100/Validation_42446100_42446100_77985.283.png" ], "turn_id": 8, "scan_id": "Validation_42446100", "origin_question": "Remember, how many oven(s) have you seen so far? ", "option": [], "answer": "2", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many oven(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_42446100/Validation_42446100_42446100_77987.282.png", "Validation_42446100/Validation_42446100_42446100_77989.281.png", "Validation_42446100/Validation_42446100_42446100_77991.280.png", "Validation_42446100/Validation_42446100_42446100_77993.280.png", "Validation_42446100/Validation_42446100_42446100_77995.295.png" ], "turn_id": 9, "scan_id": "Validation_42446100", "origin_question": "In which turn did you see both \"the modern stainless steel stove with the glossy finish\" and \"the small stainless steel bin with the shiny finish\" simultaneously? ", "option": [], "answer": "8", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the modern stainless steel stove with the glossy finish\" and \"the small stainless steel bin with the shiny finish\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_42446100/Validation_42446100_42446100_77997.295.png", "Validation_42446100/Validation_42446100_42446100_77999.294.png", "Validation_42446100/Validation_42446100_42446100_78001.293.png", "Validation_42446100/Validation_42446100_42446100_78003.292.png", "Validation_42446100/Validation_42446100_42446100_78005.292.png" ], "turn_id": 10, "scan_id": "Validation_42446100", "origin_question": "Please recall \"the sleek black monitor with the large display\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "4.35", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the sleek black monitor with the large display\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47430422/Validation_47430422_47430422_696386.068.png", "Validation_47430422/Validation_47430422_47430422_696388.050.png", "Validation_47430422/Validation_47430422_47430422_696390.049.png", "Validation_47430422/Validation_47430422_47430422_696392.049.png", "Validation_47430422/Validation_47430422_47430422_696394.048.png" ], "turn_id": 1, "scan_id": "Validation_47430422", "origin_question": "Remember, have you seen any door(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any door(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47430422/Validation_47430422_47430422_696396.047.png", "Validation_47430422/Validation_47430422_47430422_696398.046.png", "Validation_47430422/Validation_47430422_47430422_696400.045.png", "Validation_47430422/Validation_47430422_47430422_696402.045.png", "Validation_47430422/Validation_47430422_47430422_696404.044.png" ], "turn_id": 2, "scan_id": "Validation_47430422", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the sleek oval white toilet made of porcelain\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the sleek oval white toilet made of porcelain\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_47430422/Validation_47430422_47430422_696406.043.png", "Validation_47430422/Validation_47430422_47430422_696408.042.png", "Validation_47430422/Validation_47430422_47430422_696410.041.png", "Validation_47430422/Validation_47430422_47430422_696412.057.png", "Validation_47430422/Validation_47430422_47430422_696414.056.png" ], "turn_id": 3, "scan_id": "Validation_47430422", "origin_question": "Please recall \"the door\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.24", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the door\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47430422/Validation_47430422_47430422_696416.056.png", "Validation_47430422/Validation_47430422_47430422_696418.055.png", "Validation_47430422/Validation_47430422_47430422_696420.054.png", "Validation_47430422/Validation_47430422_47430422_696422.053.png", "Validation_47430422/Validation_47430422_47430422_696424.052.png" ], "turn_id": 4, "scan_id": "Validation_47430422", "origin_question": "Which direction is the door in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the door in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_47430422/Validation_47430422_47430422_696426.052.png", "Validation_47430422/Validation_47430422_47430422_696428.051.png", "Validation_47430422/Validation_47430422_47430422_696430.050.png", "Validation_47430422/Validation_47430422_47430422_696432.049.png", "Validation_47430422/Validation_47430422_47430422_696434.048.png" ], "turn_id": 5, "scan_id": "Validation_47430422", "origin_question": "By the end of the 3 turn, were you getting closer to or farther away from these two objects: X: \"the white front-loading washing machine with the metal body\", Y: \"the door\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "farther away from both X and Y", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 3 turn, were you getting closer to or farther away from these two objects: X: \"the white front-loading washing machine with the metal body\", Y: \"the door\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47430422/Validation_47430422_47430422_696436.048.png", "Validation_47430422/Validation_47430422_47430422_696438.047.png", "Validation_47430422/Validation_47430422_47430422_696440.046.png", "Validation_47430422/Validation_47430422_47430422_696442.045.png", "Validation_47430422/Validation_47430422_47430422_696444.044.png" ], "turn_id": 6, "scan_id": "Validation_47430422", "origin_question": "Which direction is the white front-loading washing machine with the metal body in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the white front-loading washing machine with the metal body in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47430422/Validation_47430422_47430422_696446.044.png", "Validation_47430422/Validation_47430422_47430422_696448.043.png", "Validation_47430422/Validation_47430422_47430422_696450.042.png", "Validation_47430422/Validation_47430422_47430422_696452.058.png", "Validation_47430422/Validation_47430422_47430422_696454.057.png" ], "turn_id": 7, "scan_id": "Validation_47430422", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the door\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the door\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47430422/Validation_47430422_47430422_696456.056.png", "Validation_47430422/Validation_47430422_47430422_696458.055.png", "Validation_47430422/Validation_47430422_47430422_696460.055.png", "Validation_47430422/Validation_47430422_47430422_696462.054.png", "Validation_47430422/Validation_47430422_47430422_696464.053.png" ], "turn_id": 8, "scan_id": "Validation_47430422", "origin_question": "Which direction is the deep purple rectangular terry cloth towel in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the deep purple rectangular terry cloth towel in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_47430422/Validation_47430422_47430422_696466.052.png", "Validation_47430422/Validation_47430422_47430422_696468.051.png", "Validation_47430422/Validation_47430422_47430422_696470.051.png", "Validation_47430422/Validation_47430422_47430422_696472.050.png", "Validation_47430422/Validation_47430422_47430422_696474.049.png" ], "turn_id": 9, "scan_id": "Validation_47430422", "origin_question": "How far is your current position from where you were at the end of the turn 3? (in meters)", "option": [], "answer": "1.43", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 3? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47895736/Validation_47895736_47895736_384654.647.png", "Validation_47895736/Validation_47895736_47895736_384656.629.png", "Validation_47895736/Validation_47895736_47895736_384658.629.png", "Validation_47895736/Validation_47895736_47895736_384660.628.png" ], "turn_id": 1, "scan_id": "Validation_47895736", "origin_question": "", "option": [], "answer": "", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47895736/Validation_47895736_47895736_384662.627.png", "Validation_47895736/Validation_47895736_47895736_384664.626.png", "Validation_47895736/Validation_47895736_47895736_384666.625.png", "Validation_47895736/Validation_47895736_47895736_384668.624.png" ], "turn_id": 2, "scan_id": "Validation_47895736", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_47895736/Validation_47895736_47895736_384670.624.png", "Validation_47895736/Validation_47895736_47895736_384672.623.png", "Validation_47895736/Validation_47895736_47895736_384674.622.png", "Validation_47895736/Validation_47895736_47895736_384676.621.png" ], "turn_id": 3, "scan_id": "Validation_47895736", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_47895736/Validation_47895736_47895736_384678.620.png", "Validation_47895736/Validation_47895736_47895736_384680.619.png", "Validation_47895736/Validation_47895736_47895736_384682.619.png", "Validation_47895736/Validation_47895736_47895736_384684.618.png" ], "turn_id": 4, "scan_id": "Validation_47895736", "origin_question": "Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_47895736/Validation_47895736_47895736_384686.617.png", "Validation_47895736/Validation_47895736_47895736_384688.616.png", "Validation_47895736/Validation_47895736_47895736_384690.615.png", "Validation_47895736/Validation_47895736_47895736_384692.631.png" ], "turn_id": 5, "scan_id": "Validation_47895736", "origin_question": "What will be the first-time appearance order of the following categories: cabinet, couch, table?", "option": [ "cabinet, couch, table", "couch, cabinet, table", "table, couch, cabinet", "couch, table, cabinet" ], "answer": "couch, table, cabinet", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: cabinet, couch, table? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['cabinet, couch, table', 'couch, cabinet, table', 'table, couch, cabinet', 'couch, table, cabinet'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_47895736/Validation_47895736_47895736_384694.630.png", "Validation_47895736/Validation_47895736_47895736_384696.629.png", "Validation_47895736/Validation_47895736_47895736_384698.629.png", "Validation_47895736/Validation_47895736_47895736_384700.628.png" ], "turn_id": 6, "scan_id": "Validation_47895736", "origin_question": "Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance forward or backward from that position?", "option": [ "Forward", "Backward" ], "answer": "Backward", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance forward or backward from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Forward', 'Backward'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47895736/Validation_47895736_47895736_384702.627.png", "Validation_47895736/Validation_47895736_47895736_384704.626.png", "Validation_47895736/Validation_47895736_47895736_384706.625.png", "Validation_47895736/Validation_47895736_47895736_384708.624.png" ], "turn_id": 7, "scan_id": "Validation_47895736", "origin_question": "Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the large, rectangular green cushion made of soft fabric\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the large, rectangular green cushion made of soft fabric\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42445031/Validation_42445031_42445031_49445.411.png", "Validation_42445031/Validation_42445031_42445031_49447.394.png", "Validation_42445031/Validation_42445031_42445031_49449.393.png", "Validation_42445031/Validation_42445031_42445031_49451.392.png" ], "turn_id": 1, "scan_id": "Validation_42445031", "origin_question": "Remember, have you seen any plant(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any plant(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42445031/Validation_42445031_42445031_49453.391.png", "Validation_42445031/Validation_42445031_42445031_49455.391.png", "Validation_42445031/Validation_42445031_42445031_49457.390.png", "Validation_42445031/Validation_42445031_42445031_49459.389.png" ], "turn_id": 2, "scan_id": "Validation_42445031", "origin_question": "Remember, have you seen any door(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any door(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_42445031/Validation_42445031_42445031_49461.388.png", "Validation_42445031/Validation_42445031_42445031_49463.387.png", "Validation_42445031/Validation_42445031_42445031_49465.387.png", "Validation_42445031/Validation_42445031_42445031_49467.402.png" ], "turn_id": 3, "scan_id": "Validation_42445031", "origin_question": "Which two objects are on the same side of you now? \"the modern medium-sized window with white pvc framing\", \"the compact, white metal radiator with vertical fins\", \"the sleek, white door with the smooth finish\". ", "option": [ "the modern medium-sized window with white pvc framing, the compact, white metal radiator with vertical fins", "the modern medium-sized window with white pvc framing, the sleek, white door with the smooth finish", "the compact, white metal radiator with vertical fins, the sleek, white door with the smooth finish" ], "answer": "the modern medium-sized window with white pvc framing, the compact, white metal radiator with vertical fins", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the modern medium-sized window with white pvc framing\", \"the compact, white metal radiator with vertical fins\", \"the sleek, white door with the smooth finish\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the modern medium-sized window with white pvc framing, the compact, white metal radiator with vertical fins', 'the modern medium-sized window with white pvc framing, the sleek, white door with the smooth finish', 'the compact, white metal radiator with vertical fins, the sleek, white door with the smooth finish'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_42445031/Validation_42445031_42445031_49469.402.png", "Validation_42445031/Validation_42445031_42445031_49471.401.png", "Validation_42445031/Validation_42445031_42445031_49473.400.png", "Validation_42445031/Validation_42445031_42445031_49475.399.png" ], "turn_id": 4, "scan_id": "Validation_42445031", "origin_question": "Which direction is the sleek, white door with the smooth finish in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the sleek, white door with the smooth finish in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_42445031/Validation_42445031_42445031_49477.399.png", "Validation_42445031/Validation_42445031_42445031_49479.398.png", "Validation_42445031/Validation_42445031_42445031_49481.397.png", "Validation_42445031/Validation_42445031_42445031_49483.396.png" ], "turn_id": 5, "scan_id": "Validation_42445031", "origin_question": "Which two objects are on the same side of you now? \"the compact, white metal radiator with vertical fins\", \"the bed\", \"the sleek, white door with the smooth finish\". ", "option": [ "the compact, white metal radiator with vertical fins, the bed", "the compact, white metal radiator with vertical fins, the sleek, white door with the smooth finish", "the bed, the sleek, white door with the smooth finish" ], "answer": "the compact, white metal radiator with vertical fins, the bed", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the compact, white metal radiator with vertical fins\", \"the bed\", \"the sleek, white door with the smooth finish\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the compact, white metal radiator with vertical fins, the bed', 'the compact, white metal radiator with vertical fins, the sleek, white door with the smooth finish', 'the bed, the sleek, white door with the smooth finish'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_42445031/Validation_42445031_42445031_49485.395.png", "Validation_42445031/Validation_42445031_42445031_49487.395.png", "Validation_42445031/Validation_42445031_42445031_49489.394.png", "Validation_42445031/Validation_42445031_42445031_49491.393.png" ], "turn_id": 6, "scan_id": "Validation_42445031", "origin_question": "Which two objects are on the same side of you now? \"the sleek, white door with the smooth finish\", \"the compact, white metal radiator with vertical fins\", \"the sleek black rectangular monitor with the slim bezel\". ", "option": [ "the sleek, white door with the smooth finish, the compact, white metal radiator with vertical fins", "the sleek, white door with the smooth finish, the sleek black rectangular monitor with the slim bezel", "the compact, white metal radiator with vertical fins, the sleek black rectangular monitor with the slim bezel" ], "answer": "the sleek, white door with the smooth finish, the compact, white metal radiator with vertical fins", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the sleek, white door with the smooth finish\", \"the compact, white metal radiator with vertical fins\", \"the sleek black rectangular monitor with the slim bezel\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the sleek, white door with the smooth finish, the compact, white metal radiator with vertical fins', 'the sleek, white door with the smooth finish, the sleek black rectangular monitor with the slim bezel', 'the compact, white metal radiator with vertical fins, the sleek black rectangular monitor with the slim bezel'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_42445031/Validation_42445031_42445031_49493.392.png", "Validation_42445031/Validation_42445031_42445031_49495.391.png", "Validation_42445031/Validation_42445031_42445031_49497.391.png", "Validation_42445031/Validation_42445031_42445031_49499.390.png" ], "turn_id": 7, "scan_id": "Validation_42445031", "origin_question": "In which turn were you farthest from \"the compact plant with vibrant, oval-shaped green leaves\"? ", "option": [], "answer": "1", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the compact plant with vibrant, oval-shaped green leaves\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_42445031/Validation_42445031_42445031_49501.389.png", "Validation_42445031/Validation_42445031_42445031_49503.388.png", "Validation_42445031/Validation_42445031_42445031_49505.387.png", "Validation_42445031/Validation_42445031_42445031_49507.387.png" ], "turn_id": 8, "scan_id": "Validation_42445031", "origin_question": "At the end of which turn were both of \"the bed\" and \"the minimalist rectangular light-colored table made of laminated wood\" on your right side?", "option": [], "answer": "5", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the bed\" and \"the minimalist rectangular light-colored table made of laminated wood\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_42445031/Validation_42445031_42445031_49509.403.png", "Validation_42445031/Validation_42445031_42445031_49511.402.png", "Validation_42445031/Validation_42445031_42445031_49513.401.png", "Validation_42445031/Validation_42445031_42445031_49515.400.png" ], "turn_id": 9, "scan_id": "Validation_42445031", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the modern medium-sized window with white pvc framing\" from your position?", "option": [], "answer": "41.09", "user_message": "For the 9 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the modern medium-sized window with white pvc framing\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47334117/Validation_47334117_47334117_94203.270.png", "Validation_47334117/Validation_47334117_47334117_94205.253.png", "Validation_47334117/Validation_47334117_47334117_94207.252.png", "Validation_47334117/Validation_47334117_47334117_94209.251.png", "Validation_47334117/Validation_47334117_47334117_94211.251.png" ], "turn_id": 1, "scan_id": "Validation_47334117", "origin_question": "Remember, have you seen any light(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any light(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_47334117/Validation_47334117_47334117_94213.250.png", "Validation_47334117/Validation_47334117_47334117_94215.249.png", "Validation_47334117/Validation_47334117_47334117_94217.248.png", "Validation_47334117/Validation_47334117_47334117_94219.247.png", "Validation_47334117/Validation_47334117_47334117_94221.247.png" ], "turn_id": 2, "scan_id": "Validation_47334117", "origin_question": "In which turn did you see both \"the round light fixture with transparent glass pendants\" and \"the door\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the round light fixture with transparent glass pendants\" and \"the door\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_47334117/Validation_47334117_47334117_94223.246.png", "Validation_47334117/Validation_47334117_47334117_94225.245.png", "Validation_47334117/Validation_47334117_47334117_94227.244.png", "Validation_47334117/Validation_47334117_47334117_94229.243.png", "Validation_47334117/Validation_47334117_47334117_94231.242.png" ], "turn_id": 3, "scan_id": "Validation_47334117", "origin_question": "Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "121.02", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_47334117/Validation_47334117_47334117_94233.242.png", "Validation_47334117/Validation_47334117_47334117_94235.258.png", "Validation_47334117/Validation_47334117_47334117_94237.257.png", "Validation_47334117/Validation_47334117_47334117_94239.256.png", "Validation_47334117/Validation_47334117_47334117_94241.255.png" ], "turn_id": 4, "scan_id": "Validation_47334117", "origin_question": "At the end of which turn were both of \"the round light fixture with transparent glass pendants\" and \"the sleek, modern white plastic chair\" on your left side?", "option": [], "answer": "2", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the round light fixture with transparent glass pendants\" and \"the sleek, modern white plastic chair\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_47334117/Validation_47334117_47334117_94243.254.png", "Validation_47334117/Validation_47334117_47334117_94245.254.png", "Validation_47334117/Validation_47334117_47334117_94247.253.png", "Validation_47334117/Validation_47334117_47334117_94249.252.png", "Validation_47334117/Validation_47334117_47334117_94251.251.png" ], "turn_id": 5, "scan_id": "Validation_47334117", "origin_question": "What will be the first-time appearance order of the following categories: bed, light, pillow?", "option": [ "pillow, light, bed", "pillow, bed, light", "light, bed, pillow", "bed, light, pillow" ], "answer": "light, bed, pillow", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bed, light, pillow? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['pillow, light, bed', 'pillow, bed, light', 'light, bed, pillow', 'bed, light, pillow'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_47334117/Validation_47334117_47334117_94253.250.png", "Validation_47334117/Validation_47334117_47334117_94255.250.png", "Validation_47334117/Validation_47334117_47334117_94257.249.png", "Validation_47334117/Validation_47334117_47334117_94259.248.png", "Validation_47334117/Validation_47334117_47334117_94261.247.png" ], "turn_id": 6, "scan_id": "Validation_47334117", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the sleek, modern white plastic chair\" from your position?", "option": [], "answer": "99.65", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the sleek, modern white plastic chair\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47334117/Validation_47334117_47334117_94263.246.png", "Validation_47334117/Validation_47334117_47334117_94265.246.png", "Validation_47334117/Validation_47334117_47334117_94267.245.png", "Validation_47334117/Validation_47334117_47334117_94269.244.png", "Validation_47334117/Validation_47334117_47334117_94271.243.png" ], "turn_id": 7, "scan_id": "Validation_47334117", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"an off-white, sleek, rounded vase\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"an off-white, sleek, rounded vase\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_47334117/Validation_47334117_47334117_94273.242.png", "Validation_47334117/Validation_47334117_47334117_94275.241.png", "Validation_47334117/Validation_47334117_47334117_94277.257.png", "Validation_47334117/Validation_47334117_47334117_94279.257.png", "Validation_47334117/Validation_47334117_47334117_94281.256.png" ], "turn_id": 8, "scan_id": "Validation_47334117", "origin_question": "What will be the first-time appearance order of the following categories: bed, light, vase?", "option": [ "vase, light, bed", "vase, bed, light", "light, bed, vase", "light, vase, bed" ], "answer": "light, bed, vase", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bed, light, vase? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['vase, light, bed', 'vase, bed, light', 'light, bed, vase', 'light, vase, bed'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_47334117/Validation_47334117_47334117_94283.255.png", "Validation_47334117/Validation_47334117_47334117_94285.254.png", "Validation_47334117/Validation_47334117_47334117_94287.253.png", "Validation_47334117/Validation_47334117_47334117_94289.252.png", "Validation_47334117/Validation_47334117_47334117_94291.252.png" ], "turn_id": 9, "scan_id": "Validation_47334117", "origin_question": "Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "Validation_47334117/Validation_47334117_47334117_94293.251.png", "Validation_47334117/Validation_47334117_47334117_94295.250.png", "Validation_47334117/Validation_47334117_47334117_94297.249.png", "Validation_47334117/Validation_47334117_47334117_94299.248.png", "Validation_47334117/Validation_47334117_47334117_94301.248.png" ], "turn_id": 10, "scan_id": "Validation_47334117", "origin_question": "Among these three objects(\"the rectangular white bed with the blue quilt\",\"the compact metal stool with the dark cushioned seat\",\"the colorful framed picture against the blue wall\" ), which one is the farthest from you?", "option": [ "the rectangular white bed with the blue quilt", "the compact metal stool with the dark cushioned seat", "the colorful framed picture against the blue wall" ], "answer": "the compact metal stool with the dark cushioned seat", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the rectangular white bed with the blue quilt\",\"the compact metal stool with the dark cushioned seat\",\"the colorful framed picture against the blue wall\" ), which one is the farthest from you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular white bed with the blue quilt', 'the compact metal stool with the dark cushioned seat', 'the colorful framed picture against the blue wall'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47429992/Validation_47429992_47429992_16761.691.png", "Validation_47429992/Validation_47429992_47429992_16763.690.png", "Validation_47429992/Validation_47429992_47429992_16765.689.png", "Validation_47429992/Validation_47429992_47429992_16767.689.png" ], "turn_id": 1, "scan_id": "Validation_47429992", "origin_question": "Remember, have you seen any dustpan(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any dustpan(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47429992/Validation_47429992_47429992_16769.688.png", "Validation_47429992/Validation_47429992_47429992_16771.687.png", "Validation_47429992/Validation_47429992_47429992_16773.686.png", "Validation_47429992/Validation_47429992_47429992_16775.685.png" ], "turn_id": 2, "scan_id": "Validation_47429992", "origin_question": "In which turn were you closest to \"the cabinet\"? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the cabinet\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47429992/Validation_47429992_47429992_16777.684.png", "Validation_47429992/Validation_47429992_47429992_16779.684.png", "Validation_47429992/Validation_47429992_47429992_16781.683.png", "Validation_47429992/Validation_47429992_47429992_16783.682.png" ], "turn_id": 3, "scan_id": "Validation_47429992", "origin_question": "In which turn were you farthest from \"the door\"? ", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the door\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_47429992/Validation_47429992_47429992_16785.681.png", "Validation_47429992/Validation_47429992_47429992_16787.697.png", "Validation_47429992/Validation_47429992_47429992_16789.696.png", "Validation_47429992/Validation_47429992_47429992_16791.695.png" ], "turn_id": 4, "scan_id": "Validation_47429992", "origin_question": "How far is your current position from where you were at the end of the turn 1? (in meters)", "option": [], "answer": "2.33", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 1? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47429992/Validation_47429992_47429992_16793.695.png", "Validation_47429992/Validation_47429992_47429992_16795.694.png", "Validation_47429992/Validation_47429992_47429992_16797.693.png", "Validation_47429992/Validation_47429992_47429992_16799.692.png" ], "turn_id": 5, "scan_id": "Validation_47429992", "origin_question": "In which turn were you closest to \"the sleek rectangular table with the light-colored surface\"? ", "option": [], "answer": "5", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the sleek rectangular table with the light-colored surface\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_47429992/Validation_47429992_47429992_16801.691.png", "Validation_47429992/Validation_47429992_47429992_16803.690.png", "Validation_47429992/Validation_47429992_47429992_16805.690.png", "Validation_47429992/Validation_47429992_47429992_16807.689.png" ], "turn_id": 6, "scan_id": "Validation_47429992", "origin_question": "At the end of which turn were both of \"the sleek metallic silver and black laptop\" and \"the white steel panel radiator with parallel columns\" on your left side?", "option": [], "answer": "6", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the sleek metallic silver and black laptop\" and \"the white steel panel radiator with parallel columns\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_47429992/Validation_47429992_47429992_16809.688.png", "Validation_47429992/Validation_47429992_47429992_16811.687.png", "Validation_47429992/Validation_47429992_47429992_16813.686.png", "Validation_47429992/Validation_47429992_47429992_16815.685.png" ], "turn_id": 7, "scan_id": "Validation_47429992", "origin_question": "By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the sleek metallic silver and black laptop\", Y: \"the white steel panel radiator with parallel columns\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the sleek metallic silver and black laptop\", Y: \"the white steel panel radiator with parallel columns\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "Validation_47204573/Validation_47204573_47204573_4957.623.png", "Validation_47204573/Validation_47204573_47204573_4959.623.png", "Validation_47204573/Validation_47204573_47204573_4961.622.png", "Validation_47204573/Validation_47204573_47204573_4963.621.png", "Validation_47204573/Validation_47204573_47204573_4965.620.png" ], "turn_id": 1, "scan_id": "Validation_47204573", "origin_question": "Remember, how many toy(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many toy(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47204573/Validation_47204573_47204573_4967.619.png", "Validation_47204573/Validation_47204573_47204573_4969.618.png", "Validation_47204573/Validation_47204573_47204573_4971.618.png", "Validation_47204573/Validation_47204573_47204573_4973.617.png", "Validation_47204573/Validation_47204573_47204573_4975.616.png" ], "turn_id": 2, "scan_id": "Validation_47204573", "origin_question": "In which turn were you farthest from \"the sleek white hexagonal table made of metal\"? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the sleek white hexagonal table made of metal\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_47204573/Validation_47204573_47204573_4977.615.png", "Validation_47204573/Validation_47204573_47204573_4979.614.png", "Validation_47204573/Validation_47204573_47204573_4981.614.png", "Validation_47204573/Validation_47204573_47204573_4983.613.png", "Validation_47204573/Validation_47204573_47204573_4985.612.png" ], "turn_id": 3, "scan_id": "Validation_47204573", "origin_question": "When did you first discover \"the sleek rectangular monitor with the dark casing\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the sleek rectangular monitor with the dark casing\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47204573/Validation_47204573_47204573_4987.611.png", "Validation_47204573/Validation_47204573_47204573_4989.610.png", "Validation_47204573/Validation_47204573_47204573_4991.609.png", "Validation_47204573/Validation_47204573_47204573_4993.609.png", "Validation_47204573/Validation_47204573_47204573_4995.608.png" ], "turn_id": 4, "scan_id": "Validation_47204573", "origin_question": "", "option": [], "answer": "", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "Validation_47204573/Validation_47204573_47204573_4997.607.png", "Validation_47204573/Validation_47204573_47204573_4999.606.png", "Validation_47204573/Validation_47204573_47204573_5001.622.png", "Validation_47204573/Validation_47204573_47204573_5003.621.png", "Validation_47204573/Validation_47204573_47204573_5005.620.png" ], "turn_id": 5, "scan_id": "Validation_47204573", "origin_question": "Among these three objects(\"the sleek rectangular monitor with the dark casing\",\"the deep blue rectangular pillow with plush fabric\",\"the large rectangular carpet with the geometric design\" ), which one is the farthest from you?", "option": [ "the sleek rectangular monitor with the dark casing", "the deep blue rectangular pillow with plush fabric", "the large rectangular carpet with the geometric design" ], "answer": "the deep blue rectangular pillow with plush fabric", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the sleek rectangular monitor with the dark casing\",\"the deep blue rectangular pillow with plush fabric\",\"the large rectangular carpet with the geometric design\" ), which one is the farthest from you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the sleek rectangular monitor with the dark casing', 'the deep blue rectangular pillow with plush fabric', 'the large rectangular carpet with the geometric design'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_47204573/Validation_47204573_47204573_5007.619.png", "Validation_47204573/Validation_47204573_47204573_5009.619.png", "Validation_47204573/Validation_47204573_47204573_5011.618.png", "Validation_47204573/Validation_47204573_47204573_5013.617.png", "Validation_47204573/Validation_47204573_47204573_5015.616.png" ], "turn_id": 6, "scan_id": "Validation_47204573", "origin_question": "Is the deep blue rectangular pillow with plush fabric to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the deep blue rectangular pillow with plush fabric to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_47204573/Validation_47204573_47204573_5017.615.png", "Validation_47204573/Validation_47204573_47204573_5019.614.png", "Validation_47204573/Validation_47204573_47204573_5021.614.png", "Validation_47204573/Validation_47204573_47204573_5023.613.png", "Validation_47204573/Validation_47204573_47204573_5025.612.png" ], "turn_id": 7, "scan_id": "Validation_47204573", "origin_question": "Is the deep blue rectangular pillow with plush fabric to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the deep blue rectangular pillow with plush fabric to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_47204573/Validation_47204573_47204573_5027.611.png", "Validation_47204573/Validation_47204573_47204573_5029.610.png", "Validation_47204573/Validation_47204573_47204573_5031.609.png", "Validation_47204573/Validation_47204573_47204573_5033.609.png", "Validation_47204573/Validation_47204573_47204573_5035.608.png" ], "turn_id": 8, "scan_id": "Validation_47204573", "origin_question": "In which turn did you see both \"the deep blue rectangular pillow with plush fabric\" and \"the medium-sized, clear plastic box\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the deep blue rectangular pillow with plush fabric\" and \"the medium-sized, clear plastic box\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_47204573/Validation_47204573_47204573_5037.607.png", "Validation_47204573/Validation_47204573_47204573_5039.606.png", "Validation_47204573/Validation_47204573_47204573_5041.622.png", "Validation_47204573/Validation_47204573_47204573_5043.621.png", "Validation_47204573/Validation_47204573_47204573_5045.620.png" ], "turn_id": 9, "scan_id": "Validation_47204573", "origin_question": "Which two objects are on the same side of you now? \"the vibrant green toy with painted details\", \"the sleek rectangular monitor with the dark casing\", \"the large rectangular carpet with the geometric design\". ", "option": [ "the vibrant green toy with painted details, the sleek rectangular monitor with the dark casing", "the vibrant green toy with painted details, the large rectangular carpet with the geometric design", "the sleek rectangular monitor with the dark casing, the large rectangular carpet with the geometric design" ], "answer": "the sleek rectangular monitor with the dark casing, the large rectangular carpet with the geometric design", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the vibrant green toy with painted details\", \"the sleek rectangular monitor with the dark casing\", \"the large rectangular carpet with the geometric design\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the vibrant green toy with painted details, the sleek rectangular monitor with the dark casing', 'the vibrant green toy with painted details, the large rectangular carpet with the geometric design', 'the sleek rectangular monitor with the dark casing, the large rectangular carpet with the geometric design'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47204573/Validation_47204573_47204573_5047.619.png", "Validation_47204573/Validation_47204573_47204573_5049.619.png", "Validation_47204573/Validation_47204573_47204573_5051.618.png", "Validation_47204573/Validation_47204573_47204573_5053.617.png", "Validation_47204573/Validation_47204573_47204573_5055.616.png" ], "turn_id": 10, "scan_id": "Validation_47204573", "origin_question": "", "option": [], "answer": "", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42446151/Validation_42446151_42446151_78976.436.png", "Validation_42446151/Validation_42446151_42446151_78978.418.png", "Validation_42446151/Validation_42446151_42446151_78980.418.png", "Validation_42446151/Validation_42446151_42446151_78982.417.png" ], "turn_id": 1, "scan_id": "Validation_42446151", "origin_question": "Remember, have you seen any pitcher(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any pitcher(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_42446151/Validation_42446151_42446151_78984.416.png", "Validation_42446151/Validation_42446151_42446151_78986.415.png", "Validation_42446151/Validation_42446151_42446151_78988.414.png", "Validation_42446151/Validation_42446151_42446151_78990.414.png" ], "turn_id": 2, "scan_id": "Validation_42446151", "origin_question": "When did you first discover \"the medium-sized rectangular fireplace with the brick exterior\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the medium-sized rectangular fireplace with the brick exterior\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_42446151/Validation_42446151_42446151_78992.413.png", "Validation_42446151/Validation_42446151_42446151_78994.412.png", "Validation_42446151/Validation_42446151_42446151_78996.411.png", "Validation_42446151/Validation_42446151_42446151_78998.410.png" ], "turn_id": 3, "scan_id": "Validation_42446151", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_42446151/Validation_42446151_42446151_79000.410.png", "Validation_42446151/Validation_42446151_42446151_79002.409.png", "Validation_42446151/Validation_42446151_42446151_79004.408.png", "Validation_42446151/Validation_42446151_42446151_79006.407.png" ], "turn_id": 4, "scan_id": "Validation_42446151", "origin_question": "Which two objects are on the same side of you now? \"the sleek black monitor made of durable plastic\", \"the sturdy wooden shelf with the deep brown hue\", \"the rectangular white wooden door with glass panes\". ", "option": [ "the sleek black monitor made of durable plastic, the sturdy wooden shelf with the deep brown hue", "the sleek black monitor made of durable plastic, the rectangular white wooden door with glass panes", "the sturdy wooden shelf with the deep brown hue, the rectangular white wooden door with glass panes" ], "answer": "the sturdy wooden shelf with the deep brown hue, the rectangular white wooden door with glass panes", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the sleek black monitor made of durable plastic\", \"the sturdy wooden shelf with the deep brown hue\", \"the rectangular white wooden door with glass panes\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the sleek black monitor made of durable plastic, the sturdy wooden shelf with the deep brown hue', 'the sleek black monitor made of durable plastic, the rectangular white wooden door with glass panes', 'the sturdy wooden shelf with the deep brown hue, the rectangular white wooden door with glass panes'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_42446151/Validation_42446151_42446151_79008.423.png", "Validation_42446151/Validation_42446151_42446151_79010.422.png", "Validation_42446151/Validation_42446151_42446151_79012.422.png", "Validation_42446151/Validation_42446151_42446151_79014.421.png" ], "turn_id": 5, "scan_id": "Validation_42446151", "origin_question": "Please recall \"the light beige rectangular pillow made of soft woven material\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.25", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the light beige rectangular pillow made of soft woven material\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_42446151/Validation_42446151_42446151_79016.420.png", "Validation_42446151/Validation_42446151_42446151_79018.419.png", "Validation_42446151/Validation_42446151_42446151_79020.418.png", "Validation_42446151/Validation_42446151_42446151_79022.418.png" ], "turn_id": 6, "scan_id": "Validation_42446151", "origin_question": "Which direction is the rectangular wicker basket in light brown color in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the rectangular wicker basket in light brown color in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_42446151/Validation_42446151_42446151_79024.417.png", "Validation_42446151/Validation_42446151_42446151_79026.416.png", "Validation_42446151/Validation_42446151_42446151_79028.415.png", "Validation_42446151/Validation_42446151_42446151_79030.414.png" ], "turn_id": 7, "scan_id": "Validation_42446151", "origin_question": "Which two objects are on the same side of you now? \"the vertically standing polished wooden cabinet\", \"the rectangular white wooden door with glass panes\", \"the rectangular window with the white wooden frame\". ", "option": [ "the vertically standing polished wooden cabinet, the rectangular white wooden door with glass panes", "the vertically standing polished wooden cabinet, the rectangular window with the white wooden frame", "the rectangular white wooden door with glass panes, the rectangular window with the white wooden frame" ], "answer": "the rectangular white wooden door with glass panes, the rectangular window with the white wooden frame", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the vertically standing polished wooden cabinet\", \"the rectangular white wooden door with glass panes\", \"the rectangular window with the white wooden frame\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the vertically standing polished wooden cabinet, the rectangular white wooden door with glass panes', 'the vertically standing polished wooden cabinet, the rectangular window with the white wooden frame', 'the rectangular white wooden door with glass panes, the rectangular window with the white wooden frame'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_42446151/Validation_42446151_42446151_79032.414.png", "Validation_42446151/Validation_42446151_42446151_79034.413.png", "Validation_42446151/Validation_42446151_42446151_79036.412.png", "Validation_42446151/Validation_42446151_42446151_79038.411.png" ], "turn_id": 8, "scan_id": "Validation_42446151", "origin_question": "Which two objects are on the same side of you now? \"the sleek black monitor made of durable plastic\", \"the rectangular white wooden door with glass panes\", \"the sturdy wooden shelf with the deep brown hue\". ", "option": [ "the sleek black monitor made of durable plastic, the rectangular white wooden door with glass panes", "the sleek black monitor made of durable plastic, the sturdy wooden shelf with the deep brown hue", "the rectangular white wooden door with glass panes, the sturdy wooden shelf with the deep brown hue" ], "answer": "the sleek black monitor made of durable plastic, the rectangular white wooden door with glass panes", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the sleek black monitor made of durable plastic\", \"the rectangular white wooden door with glass panes\", \"the sturdy wooden shelf with the deep brown hue\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the sleek black monitor made of durable plastic, the rectangular white wooden door with glass panes', 'the sleek black monitor made of durable plastic, the sturdy wooden shelf with the deep brown hue', 'the rectangular white wooden door with glass panes, the sturdy wooden shelf with the deep brown hue'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_48458484/Validation_48458484_48458484_111570.315.png", "Validation_48458484/Validation_48458484_48458484_111571.798.png", "Validation_48458484/Validation_48458484_48458484_111573.298.png", "Validation_48458484/Validation_48458484_48458484_111574.797.png" ], "turn_id": 1, "scan_id": "Validation_48458484", "origin_question": "Remember, have you seen any bicycle(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any bicycle(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_48458484/Validation_48458484_48458484_111576.296.png", "Validation_48458484/Validation_48458484_48458484_111577.813.png", "Validation_48458484/Validation_48458484_48458484_111579.312.png", "Validation_48458484/Validation_48458484_48458484_111580.811.png" ], "turn_id": 2, "scan_id": "Validation_48458484", "origin_question": "When did you first discover \"the compact, neutral-colored plastic table\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the compact, neutral-colored plastic table\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_48458484/Validation_48458484_48458484_111582.311.png", "Validation_48458484/Validation_48458484_48458484_111583.810.png", "Validation_48458484/Validation_48458484_48458484_111585.310.png", "Validation_48458484/Validation_48458484_48458484_111586.809.png" ], "turn_id": 3, "scan_id": "Validation_48458484", "origin_question": "In which turn were you closest to \"the spherical glass light providing soft white light\"? ", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the spherical glass light providing soft white light\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_48458484/Validation_48458484_48458484_111588.308.png", "Validation_48458484/Validation_48458484_48458484_111589.808.png", "Validation_48458484/Validation_48458484_48458484_111591.307.png", "Validation_48458484/Validation_48458484_48458484_111592.807.png" ], "turn_id": 4, "scan_id": "Validation_48458484", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the compact silver stainless steel microwave\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the compact silver stainless steel microwave\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_48458484/Validation_48458484_48458484_111594.306.png", "Validation_48458484/Validation_48458484_48458484_111595.806.png", "Validation_48458484/Validation_48458484_48458484_111597.305.png", "Validation_48458484/Validation_48458484_48458484_111598.804.png" ], "turn_id": 5, "scan_id": "Validation_48458484", "origin_question": "Which direction is the compact silver stainless steel microwave in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the compact silver stainless steel microwave in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42898544/Validation_42898544_42898544_752080.372.png", "Validation_42898544/Validation_42898544_42898544_752082.355.png", "Validation_42898544/Validation_42898544_42898544_752084.354.png", "Validation_42898544/Validation_42898544_42898544_752086.353.png", "Validation_42898544/Validation_42898544_42898544_752088.352.png" ], "turn_id": 1, "scan_id": "Validation_42898544", "origin_question": "Remember, have you seen any pillow(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any pillow(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_42898544/Validation_42898544_42898544_752090.351.png", "Validation_42898544/Validation_42898544_42898544_752092.351.png", "Validation_42898544/Validation_42898544_42898544_752094.350.png", "Validation_42898544/Validation_42898544_42898544_752096.349.png", "Validation_42898544/Validation_42898544_42898544_752098.348.png" ], "turn_id": 2, "scan_id": "Validation_42898544", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "Validation_42898544/Validation_42898544_42898544_752100.347.png", "Validation_42898544/Validation_42898544_42898544_752102.346.png", "Validation_42898544/Validation_42898544_42898544_752104.346.png", "Validation_42898544/Validation_42898544_42898544_752106.345.png", "Validation_42898544/Validation_42898544_42898544_752108.344.png" ], "turn_id": 3, "scan_id": "Validation_42898544", "origin_question": "Among these three objects(\"the rectangular bed with the padded headboard\",\"the sleek cylindrical lamp with the glossy base\",\"the white rectangular pillow with the subtle pattern\" ), which one is the closest to you?", "option": [ "the rectangular bed with the padded headboard", "the sleek cylindrical lamp with the glossy base", "the white rectangular pillow with the subtle pattern" ], "answer": "the rectangular bed with the padded headboard", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the rectangular bed with the padded headboard\",\"the sleek cylindrical lamp with the glossy base\",\"the white rectangular pillow with the subtle pattern\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular bed with the padded headboard', 'the sleek cylindrical lamp with the glossy base', 'the white rectangular pillow with the subtle pattern'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_42898544/Validation_42898544_42898544_752110.343.png", "Validation_42898544/Validation_42898544_42898544_752112.342.png", "Validation_42898544/Validation_42898544_42898544_752114.341.png", "Validation_42898544/Validation_42898544_42898544_752116.341.png", "Validation_42898544/Validation_42898544_42898544_752118.340.png" ], "turn_id": 4, "scan_id": "Validation_42898544", "origin_question": "Which direction is the rectangular bed with the padded headboard in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the rectangular bed with the padded headboard in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_42898544/Validation_42898544_42898544_752120.356.png", "Validation_42898544/Validation_42898544_42898544_752122.355.png", "Validation_42898544/Validation_42898544_42898544_752124.354.png", "Validation_42898544/Validation_42898544_42898544_752126.353.png", "Validation_42898544/Validation_42898544_42898544_752128.352.png" ], "turn_id": 5, "scan_id": "Validation_42898544", "origin_question": "Which two objects are on the same side of you now? \"the rectangular bed with the padded headboard\", \"the lush green plant in the neutral-toned pot\", \"the compact rectangular wooden table with the polished surface\". ", "option": [ "the rectangular bed with the padded headboard, the lush green plant in the neutral-toned pot", "the rectangular bed with the padded headboard, the compact rectangular wooden table with the polished surface", "the lush green plant in the neutral-toned pot, the compact rectangular wooden table with the polished surface" ], "answer": "the lush green plant in the neutral-toned pot, the compact rectangular wooden table with the polished surface", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the rectangular bed with the padded headboard\", \"the lush green plant in the neutral-toned pot\", \"the compact rectangular wooden table with the polished surface\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular bed with the padded headboard, the lush green plant in the neutral-toned pot', 'the rectangular bed with the padded headboard, the compact rectangular wooden table with the polished surface', 'the lush green plant in the neutral-toned pot, the compact rectangular wooden table with the polished surface'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_42898544/Validation_42898544_42898544_752130.351.png", "Validation_42898544/Validation_42898544_42898544_752132.351.png", "Validation_42898544/Validation_42898544_42898544_752134.350.png", "Validation_42898544/Validation_42898544_42898544_752136.349.png", "Validation_42898544/Validation_42898544_42898544_752138.348.png" ], "turn_id": 6, "scan_id": "Validation_42898544", "origin_question": "At the end of which turn were both of \"the compact rectangular wooden table with the polished surface\" and \"the sturdy square stool with the light wood frame\" on your left side?", "option": [], "answer": "6", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the compact rectangular wooden table with the polished surface\" and \"the sturdy square stool with the light wood frame\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_42898544/Validation_42898544_42898544_752140.347.png", "Validation_42898544/Validation_42898544_42898544_752142.347.png", "Validation_42898544/Validation_42898544_42898544_752144.346.png", "Validation_42898544/Validation_42898544_42898544_752146.345.png", "Validation_42898544/Validation_42898544_42898544_752148.344.png" ], "turn_id": 7, "scan_id": "Validation_42898544", "origin_question": "By the end of the 5 turn, were you getting closer to or farther away from these two objects: X: \"the lush green plant in the neutral-toned pot\", Y: \"the compact rectangular wooden table with the polished surface\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "farther away from both X and Y", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 5 turn, were you getting closer to or farther away from these two objects: X: \"the lush green plant in the neutral-toned pot\", Y: \"the compact rectangular wooden table with the polished surface\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_42898544/Validation_42898544_42898544_752150.343.png", "Validation_42898544/Validation_42898544_42898544_752152.342.png", "Validation_42898544/Validation_42898544_42898544_752154.342.png", "Validation_42898544/Validation_42898544_42898544_752156.341.png", "Validation_42898544/Validation_42898544_42898544_752158.340.png" ], "turn_id": 8, "scan_id": "Validation_42898544", "origin_question": "By the end of the 7 turn, were you getting closer to or farther away from these two objects: X: \"the compact rectangular wooden table with the polished surface\", Y: \"the sturdy square stool with the light wood frame\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "farther away from both X and Y", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 7 turn, were you getting closer to or farther away from these two objects: X: \"the compact rectangular wooden table with the polished surface\", Y: \"the sturdy square stool with the light wood frame\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_42898544/Validation_42898544_42898544_752160.356.png", "Validation_42898544/Validation_42898544_42898544_752162.355.png", "Validation_42898544/Validation_42898544_42898544_752164.354.png", "Validation_42898544/Validation_42898544_42898544_752166.353.png", "Validation_42898544/Validation_42898544_42898544_752168.352.png" ], "turn_id": 9, "scan_id": "Validation_42898544", "origin_question": "What will be the first-time appearance order of the following categories: bag, bed, stool?", "option": [ "bed, bag, stool", "stool, bag, bed", "bag, bed, stool", "bed, stool, bag" ], "answer": "bed, stool, bag", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bag, bed, stool? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['bed, bag, stool', 'stool, bag, bed', 'bag, bed, stool', 'bed, stool, bag'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_42898544/Validation_42898544_42898544_752170.352.png", "Validation_42898544/Validation_42898544_42898544_752172.351.png", "Validation_42898544/Validation_42898544_42898544_752174.350.png", "Validation_42898544/Validation_42898544_42898544_752176.349.png", "Validation_42898544/Validation_42898544_42898544_752178.348.png" ], "turn_id": 10, "scan_id": "Validation_42898544", "origin_question": "Please recall \"the sturdy square stool with the light wood frame\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.61", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the sturdy square stool with the light wood frame\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_45261632/Validation_45261632_45261632_55891.209.png", "Validation_45261632/Validation_45261632_45261632_55893.192.png", "Validation_45261632/Validation_45261632_45261632_55895.191.png", "Validation_45261632/Validation_45261632_45261632_55897.190.png" ], "turn_id": 1, "scan_id": "Validation_45261632", "origin_question": "Remember, have you seen any washing machine(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any washing machine(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_45261632/Validation_45261632_45261632_55899.189.png", "Validation_45261632/Validation_45261632_45261632_55901.188.png", "Validation_45261632/Validation_45261632_45261632_55903.188.png", "Validation_45261632/Validation_45261632_45261632_55905.204.png" ], "turn_id": 2, "scan_id": "Validation_45261632", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_45261632/Validation_45261632_45261632_55907.203.png", "Validation_45261632/Validation_45261632_45261632_55909.202.png", "Validation_45261632/Validation_45261632_45261632_55911.201.png", "Validation_45261632/Validation_45261632_45261632_55913.200.png" ], "turn_id": 3, "scan_id": "Validation_45261632", "origin_question": "Please recall \"an elegant cylindrical chandelier with the floral pattern\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.54", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"an elegant cylindrical chandelier with the floral pattern\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_45261632/Validation_45261632_45261632_55915.200.png", "Validation_45261632/Validation_45261632_45261632_55917.199.png", "Validation_45261632/Validation_45261632_45261632_55919.198.png", "Validation_45261632/Validation_45261632_45261632_55921.197.png" ], "turn_id": 4, "scan_id": "Validation_45261632", "origin_question": "Please recall \"the sleek white rectangular radiator made of metal\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.19", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the sleek white rectangular radiator made of metal\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_45261632/Validation_45261632_45261632_55923.196.png", "Validation_45261632/Validation_45261632_45261632_55925.196.png", "Validation_45261632/Validation_45261632_45261632_55927.195.png", "Validation_45261632/Validation_45261632_45261632_55929.194.png" ], "turn_id": 5, "scan_id": "Validation_45261632", "origin_question": "Which direction is an elegant cylindrical chandelier with the floral pattern in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is an elegant cylindrical chandelier with the floral pattern in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_45261632/Validation_45261632_45261632_55931.193.png", "Validation_45261632/Validation_45261632_45261632_55933.192.png", "Validation_45261632/Validation_45261632_45261632_55935.192.png", "Validation_45261632/Validation_45261632_45261632_55937.191.png" ], "turn_id": 6, "scan_id": "Validation_45261632", "origin_question": "Which two objects are on the same side of you now? \"the sleek black rectangular monitor with durable materials\", \"the queen-sized bed with the white comforter and pillows\", \"the window\". ", "option": [ "the sleek black rectangular monitor with durable materials, the queen-sized bed with the white comforter and pillows", "the sleek black rectangular monitor with durable materials, the window", "the queen-sized bed with the white comforter and pillows, the window" ], "answer": "the sleek black rectangular monitor with durable materials, the queen-sized bed with the white comforter and pillows", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the sleek black rectangular monitor with durable materials\", \"the queen-sized bed with the white comforter and pillows\", \"the window\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the sleek black rectangular monitor with durable materials, the queen-sized bed with the white comforter and pillows', 'the sleek black rectangular monitor with durable materials, the window', 'the queen-sized bed with the white comforter and pillows, the window'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42897545/Validation_42897545_42897545_276440.879.png", "Validation_42897545/Validation_42897545_42897545_276442.361.png", "Validation_42897545/Validation_42897545_42897545_276443.861.png", "Validation_42897545/Validation_42897545_42897545_276445.360.png" ], "turn_id": 1, "scan_id": "Validation_42897545", "origin_question": "Remember, have you seen any tablet(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any tablet(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_42897545/Validation_42897545_42897545_276446.860.png", "Validation_42897545/Validation_42897545_42897545_276448.359.png", "Validation_42897545/Validation_42897545_42897545_276449.858.png", "Validation_42897545/Validation_42897545_42897545_276451.374.png" ], "turn_id": 2, "scan_id": "Validation_42897545", "origin_question": "In which turn did you see both \"the warm beige, rectangular door\" and \"the window\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the warm beige, rectangular door\" and \"the window\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_42897545/Validation_42897545_42897545_276452.874.png", "Validation_42897545/Validation_42897545_42897545_276454.373.png", "Validation_42897545/Validation_42897545_42897545_276455.872.png", "Validation_42897545/Validation_42897545_42897545_276457.372.png" ], "turn_id": 3, "scan_id": "Validation_42897545", "origin_question": "At the end of which turn were both of \"the modern oval white acrylic bathtub\" and \"the sink\" on your left side?", "option": [], "answer": "2", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the modern oval white acrylic bathtub\" and \"the sink\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_42897545/Validation_42897545_42897545_276458.871.png", "Validation_42897545/Validation_42897545_42897545_276460.371.png", "Validation_42897545/Validation_42897545_42897545_276461.870.png", "Validation_42897545/Validation_42897545_42897545_276463.369.png" ], "turn_id": 4, "scan_id": "Validation_42897545", "origin_question": "Is the contemporary oval white porcelain toilet to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the contemporary oval white porcelain toilet to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_42897545/Validation_42897545_42897545_276464.869.png", "Validation_42897545/Validation_42897545_42897545_276466.368.png", "Validation_42897545/Validation_42897545_42897545_276467.868.png", "Validation_42897545/Validation_42897545_42897545_276469.367.png" ], "turn_id": 5, "scan_id": "Validation_42897545", "origin_question": "Using your orientation at the end of turn 4 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "20.85", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 4 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47430024/Validation_47430024_47430024_18769.060.png", "Validation_47430024/Validation_47430024_47430024_18770.543.png", "Validation_47430024/Validation_47430024_47430024_18772.042.png", "Validation_47430024/Validation_47430024_47430024_18773.542.png" ], "turn_id": 1, "scan_id": "Validation_47430024", "origin_question": "", "option": [], "answer": "", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47430024/Validation_47430024_47430024_18775.041.png", "Validation_47430024/Validation_47430024_47430024_18776.541.png", "Validation_47430024/Validation_47430024_47430024_18778.040.png", "Validation_47430024/Validation_47430024_47430024_18779.539.png" ], "turn_id": 2, "scan_id": "Validation_47430024", "origin_question": "In which turn were you farthest from \"the rectangular, neutral-toned synthetic carpet\"? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the rectangular, neutral-toned synthetic carpet\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47430024/Validation_47430024_47430024_18781.039.png", "Validation_47430024/Validation_47430024_47430024_18782.538.png", "Validation_47430024/Validation_47430024_47430024_18784.038.png", "Validation_47430024/Validation_47430024_47430024_18785.537.png" ], "turn_id": 3, "scan_id": "Validation_47430024", "origin_question": "In which turn were you closest to \"the window\"? ", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the window\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47430024/Validation_47430024_47430024_18787.036.png", "Validation_47430024/Validation_47430024_47430024_18788.536.png", "Validation_47430024/Validation_47430024_47430024_18790.035.png", "Validation_47430024/Validation_47430024_47430024_18791.535.png" ], "turn_id": 4, "scan_id": "Validation_47430024", "origin_question": "In which turn were you closest to \"the sturdy rectangular table with the light brown hue\"? ", "option": [], "answer": "4", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the sturdy rectangular table with the light brown hue\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_47430024/Validation_47430024_47430024_18793.034.png", "Validation_47430024/Validation_47430024_47430024_18794.533.png", "Validation_47430024/Validation_47430024_47430024_18796.033.png", "Validation_47430024/Validation_47430024_47430024_18797.549.png" ], "turn_id": 5, "scan_id": "Validation_47430024", "origin_question": "What will be the first-time appearance order of the following categories: cabinet, chair, monitor?", "option": [ "cabinet, monitor, chair", "cabinet, chair, monitor", "monitor, chair, cabinet", "chair, monitor, cabinet" ], "answer": "cabinet, chair, monitor", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: cabinet, chair, monitor? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['cabinet, monitor, chair', 'cabinet, chair, monitor', 'monitor, chair, cabinet', 'chair, monitor, cabinet'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_48018367/Validation_48018367_48018367_389445.594.png", "Validation_48018367/Validation_48018367_48018367_389447.077.png", "Validation_48018367/Validation_48018367_48018367_389448.576.png", "Validation_48018367/Validation_48018367_48018367_389450.076.png" ], "turn_id": 1, "scan_id": "Validation_48018367", "origin_question": "", "option": [], "answer": "", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_48018367/Validation_48018367_48018367_389451.575.png", "Validation_48018367/Validation_48018367_48018367_389453.075.png", "Validation_48018367/Validation_48018367_48018367_389454.574.png", "Validation_48018367/Validation_48018367_48018367_389456.073.png" ], "turn_id": 2, "scan_id": "Validation_48018367", "origin_question": "In which turn did you see both \"the compact white refrigerator with metal exterior\" and \"the sink\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the compact white refrigerator with metal exterior\" and \"the sink\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_48018367/Validation_48018367_48018367_389457.573.png", "Validation_48018367/Validation_48018367_48018367_389459.072.png", "Validation_48018367/Validation_48018367_48018367_389460.572.png", "Validation_48018367/Validation_48018367_48018367_389462.071.png" ], "turn_id": 3, "scan_id": "Validation_48018367", "origin_question": "Please recall \"the stove\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.67", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the stove\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_48018367/Validation_48018367_48018367_389463.570.png", "Validation_48018367/Validation_48018367_48018367_389465.070.png", "Validation_48018367/Validation_48018367_48018367_389466.569.png", "Validation_48018367/Validation_48018367_48018367_389468.068.png" ], "turn_id": 4, "scan_id": "Validation_48018367", "origin_question": "How far is your current position from where you were at the end of the turn 3? (in meters)", "option": [], "answer": "1.32", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 3? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_48018367/Validation_48018367_48018367_389469.584.png", "Validation_48018367/Validation_48018367_48018367_389471.084.png", "Validation_48018367/Validation_48018367_48018367_389472.583.png", "Validation_48018367/Validation_48018367_48018367_389474.083.png" ], "turn_id": 5, "scan_id": "Validation_48018367", "origin_question": "Please recall \"the oven\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.7", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the oven\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_48018367/Validation_48018367_48018367_389475.582.png", "Validation_48018367/Validation_48018367_48018367_389477.081.png", "Validation_48018367/Validation_48018367_48018367_389478.581.png", "Validation_48018367/Validation_48018367_48018367_389480.080.png" ], "turn_id": 6, "scan_id": "Validation_48018367", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the stove\" from your position?", "option": [], "answer": "72.0", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the stove\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_48018367/Validation_48018367_48018367_389481.579.png", "Validation_48018367/Validation_48018367_48018367_389483.079.png", "Validation_48018367/Validation_48018367_48018367_389484.578.png", "Validation_48018367/Validation_48018367_48018367_389486.078.png" ], "turn_id": 7, "scan_id": "Validation_48018367", "origin_question": "Which direction is the vibrant red round stool with metallic legs in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the vibrant red round stool with metallic legs in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47430001/Validation_47430001_47430001_17208.359.png", "Validation_47430001/Validation_47430001_47430001_17209.859.png", "Validation_47430001/Validation_47430001_47430001_17211.358.png", "Validation_47430001/Validation_47430001_47430001_17212.858.png" ], "turn_id": 1, "scan_id": "Validation_47430001", "origin_question": "Remember, have you seen any headphones(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any headphones(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47430001/Validation_47430001_47430001_17214.357.png", "Validation_47430001/Validation_47430001_47430001_17215.857.png", "Validation_47430001/Validation_47430001_47430001_17217.356.png", "Validation_47430001/Validation_47430001_47430001_17218.855.png" ], "turn_id": 2, "scan_id": "Validation_47430001", "origin_question": "In which turn were you closest to \"the compact, rectangular dark fabric bed\"? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the compact, rectangular dark fabric bed\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_47430001/Validation_47430001_47430001_17220.355.png", "Validation_47430001/Validation_47430001_47430001_17221.854.png", "Validation_47430001/Validation_47430001_47430001_17223.353.png", "Validation_47430001/Validation_47430001_47430001_17224.853.png" ], "turn_id": 3, "scan_id": "Validation_47430001", "origin_question": "What will be the first-time appearance order of the following categories: cabinet, door, window?", "option": [ "window, cabinet, door", "cabinet, window, door", "cabinet, door, window", "window, door, cabinet" ], "answer": "window, door, cabinet", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: cabinet, door, window? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['window, cabinet, door', 'cabinet, window, door', 'cabinet, door, window', 'window, door, cabinet'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_47430001/Validation_47430001_47430001_17226.352.png", "Validation_47430001/Validation_47430001_47430001_17227.852.png", "Validation_47430001/Validation_47430001_47430001_17229.351.png", "Validation_47430001/Validation_47430001_47430001_17230.850.png" ], "turn_id": 4, "scan_id": "Validation_47430001", "origin_question": "What will be the first-time appearance order of the following categories: flowerpot, shelf, window?", "option": [ "flowerpot, window, shelf", "shelf, flowerpot, window", "window, flowerpot, shelf", "window, shelf, flowerpot" ], "answer": "window, flowerpot, shelf", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: flowerpot, shelf, window? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['flowerpot, window, shelf', 'shelf, flowerpot, window', 'window, flowerpot, shelf', 'window, shelf, flowerpot'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_47430001/Validation_47430001_47430001_17232.350.png", "Validation_47430001/Validation_47430001_47430001_17233.849.png", "Validation_47430001/Validation_47430001_47430001_17235.349.png", "Validation_47430001/Validation_47430001_47430001_17236.848.png" ], "turn_id": 5, "scan_id": "Validation_47430001", "origin_question": "When did you first discover \"the compact, rectangular dark fabric bed\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the compact, rectangular dark fabric bed\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47333456/Validation_47333456_47333456_55008.133.png", "Validation_47333456/Validation_47333456_47333456_55010.116.png", "Validation_47333456/Validation_47333456_47333456_55012.115.png", "Validation_47333456/Validation_47333456_47333456_55014.114.png" ], "turn_id": 1, "scan_id": "Validation_47333456", "origin_question": "Remember, have you seen any refrigerator(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any refrigerator(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_47333456/Validation_47333456_47333456_55016.113.png", "Validation_47333456/Validation_47333456_47333456_55018.112.png", "Validation_47333456/Validation_47333456_47333456_55020.111.png", "Validation_47333456/Validation_47333456_47333456_55022.111.png" ], "turn_id": 2, "scan_id": "Validation_47333456", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the sleek stainless steel and black glass stove\" from your position?", "option": [], "answer": "170.51", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the sleek stainless steel and black glass stove\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47333456/Validation_47333456_47333456_55024.110.png", "Validation_47333456/Validation_47333456_47333456_55026.109.png", "Validation_47333456/Validation_47333456_47333456_55028.108.png", "Validation_47333456/Validation_47333456_47333456_55030.107.png" ], "turn_id": 3, "scan_id": "Validation_47333456", "origin_question": "In which turn were you closest to \"the bed\"? ", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the bed\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47333456/Validation_47333456_47333456_55032.123.png", "Validation_47333456/Validation_47333456_47333456_55034.122.png", "Validation_47333456/Validation_47333456_47333456_55036.122.png", "Validation_47333456/Validation_47333456_47333456_55038.121.png" ], "turn_id": 4, "scan_id": "Validation_47333456", "origin_question": "Which direction is the sleek stainless steel and black glass stove in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the sleek stainless steel and black glass stove in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_47333456/Validation_47333456_47333456_55040.120.png", "Validation_47333456/Validation_47333456_47333456_55042.119.png", "Validation_47333456/Validation_47333456_47333456_55044.118.png", "Validation_47333456/Validation_47333456_47333456_55046.117.png" ], "turn_id": 5, "scan_id": "Validation_47333456", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the sleek, black glossy monitor with the thin bezel\", Y: \"the medium-sized cream ceramic flowerpot\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the sleek, black glossy monitor with the thin bezel\", Y: \"the medium-sized cream ceramic flowerpot\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_47333456/Validation_47333456_47333456_55048.116.png", "Validation_47333456/Validation_47333456_47333456_55050.116.png", "Validation_47333456/Validation_47333456_47333456_55052.115.png", "Validation_47333456/Validation_47333456_47333456_55054.114.png" ], "turn_id": 6, "scan_id": "Validation_47333456", "origin_question": "Please recall \"the sleek, black glossy monitor with the thin bezel\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.13", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the sleek, black glossy monitor with the thin bezel\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_47333456/Validation_47333456_47333456_55056.113.png", "Validation_47333456/Validation_47333456_47333456_55058.112.png", "Validation_47333456/Validation_47333456_47333456_55060.112.png", "Validation_47333456/Validation_47333456_47333456_55062.111.png" ], "turn_id": 7, "scan_id": "Validation_47333456", "origin_question": "Using your orientation at the end of turn 4 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "154.84", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 4 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_47333456/Validation_47333456_47333456_55064.110.png", "Validation_47333456/Validation_47333456_47333456_55066.109.png", "Validation_47333456/Validation_47333456_47333456_55068.108.png", "Validation_47333456/Validation_47333456_47333456_55070.107.png" ], "turn_id": 8, "scan_id": "Validation_47333456", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the large cylindrical vase\" from your position?", "option": [], "answer": "163.88", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the large cylindrical vase\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_47333456/Validation_47333456_47333456_55072.107.png", "Validation_47333456/Validation_47333456_47333456_55074.122.png", "Validation_47333456/Validation_47333456_47333456_55076.122.png", "Validation_47333456/Validation_47333456_47333456_55078.121.png" ], "turn_id": 9, "scan_id": "Validation_47333456", "origin_question": "Please recall \"the medium-sized natural wood shelf with the modern design\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.42", "user_message": "For the 9 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the medium-sized natural wood shelf with the modern design\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47333916/Validation_47333916_47333916_82368.557.png", "Validation_47333916/Validation_47333916_47333916_82370.556.png", "Validation_47333916/Validation_47333916_47333916_82372.555.png", "Validation_47333916/Validation_47333916_47333916_82374.555.png", "Validation_47333916/Validation_47333916_47333916_82376.554.png" ], "turn_id": 1, "scan_id": "Validation_47333916", "origin_question": "Remember, have you seen any flowerpot(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any flowerpot(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47333916/Validation_47333916_47333916_82378.553.png", "Validation_47333916/Validation_47333916_47333916_82380.552.png", "Validation_47333916/Validation_47333916_47333916_82382.551.png", "Validation_47333916/Validation_47333916_47333916_82384.567.png", "Validation_47333916/Validation_47333916_47333916_82386.566.png" ], "turn_id": 2, "scan_id": "Validation_47333916", "origin_question": "In which turn were you closest to \"the square, soft beige pillow\"? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the square, soft beige pillow\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_47333916/Validation_47333916_47333916_82388.566.png", "Validation_47333916/Validation_47333916_47333916_82390.565.png", "Validation_47333916/Validation_47333916_47333916_82392.564.png", "Validation_47333916/Validation_47333916_47333916_82394.563.png", "Validation_47333916/Validation_47333916_47333916_82396.562.png" ], "turn_id": 3, "scan_id": "Validation_47333916", "origin_question": "In which turn did you see both \"the rectangular table covered in light cloth\" and \"the speaker\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the rectangular table covered in light cloth\" and \"the speaker\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_47333916/Validation_47333916_47333916_82398.562.png", "Validation_47333916/Validation_47333916_47333916_82400.561.png", "Validation_47333916/Validation_47333916_47333916_82402.560.png", "Validation_47333916/Validation_47333916_47333916_82404.559.png", "Validation_47333916/Validation_47333916_47333916_82406.558.png" ], "turn_id": 4, "scan_id": "Validation_47333916", "origin_question": "Is the sleek square wooden stool with straight legs to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the sleek square wooden stool with straight legs to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47333916/Validation_47333916_47333916_82408.557.png", "Validation_47333916/Validation_47333916_47333916_82410.557.png", "Validation_47333916/Validation_47333916_47333916_82412.556.png", "Validation_47333916/Validation_47333916_47333916_82414.555.png", "Validation_47333916/Validation_47333916_47333916_82416.554.png" ], "turn_id": 5, "scan_id": "Validation_47333916", "origin_question": "Which direction is the light-colored rectangular door with frosted glass in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the light-colored rectangular door with frosted glass in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_47333916/Validation_47333916_47333916_82418.553.png", "Validation_47333916/Validation_47333916_47333916_82420.553.png", "Validation_47333916/Validation_47333916_47333916_82422.552.png", "Validation_47333916/Validation_47333916_47333916_82424.551.png", "Validation_47333916/Validation_47333916_47333916_82426.550.png" ], "turn_id": 6, "scan_id": "Validation_47333916", "origin_question": "Is the square, soft beige pillow to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the square, soft beige pillow to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "Validation_47333916/Validation_47333916_47333916_82428.566.png", "Validation_47333916/Validation_47333916_47333916_82430.565.png", "Validation_47333916/Validation_47333916_47333916_82432.564.png", "Validation_47333916/Validation_47333916_47333916_82434.564.png", "Validation_47333916/Validation_47333916_47333916_82436.563.png" ], "turn_id": 7, "scan_id": "Validation_47333916", "origin_question": "Among these three objects(\"the sleek square wooden stool with straight legs\",\"the classic dome-shaped frosted glass light\",\"the square, soft beige pillow\" ), which one is the closest to you?", "option": [ "the sleek square wooden stool with straight legs", "the classic dome-shaped frosted glass light", "the square, soft beige pillow" ], "answer": "the classic dome-shaped frosted glass light", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the sleek square wooden stool with straight legs\",\"the classic dome-shaped frosted glass light\",\"the square, soft beige pillow\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the sleek square wooden stool with straight legs', 'the classic dome-shaped frosted glass light', 'the square, soft beige pillow'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47333916/Validation_47333916_47333916_82438.562.png", "Validation_47333916/Validation_47333916_47333916_82440.561.png", "Validation_47333916/Validation_47333916_47333916_82442.560.png", "Validation_47333916/Validation_47333916_47333916_82444.560.png", "Validation_47333916/Validation_47333916_47333916_82446.559.png" ], "turn_id": 8, "scan_id": "Validation_47333916", "origin_question": "Compared to your position at the end of the 7 turn, are you now closer or farther away from \"the sleek square wooden stool with straight legs\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 7 turn, are you now closer or farther away from \"the sleek square wooden stool with straight legs\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42446167/Validation_42446167_42446167_79643.743.png", "Validation_42446167/Validation_42446167_42446167_79645.726.png", "Validation_42446167/Validation_42446167_42446167_79647.725.png", "Validation_42446167/Validation_42446167_42446167_79649.724.png" ], "turn_id": 1, "scan_id": "Validation_42446167", "origin_question": "Remember, have you seen any door(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any door(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_42446167/Validation_42446167_42446167_79651.723.png", "Validation_42446167/Validation_42446167_42446167_79653.723.png", "Validation_42446167/Validation_42446167_42446167_79655.722.png", "Validation_42446167/Validation_42446167_42446167_79657.721.png" ], "turn_id": 2, "scan_id": "Validation_42446167", "origin_question": "In which turn were you closest to \"an off-white interior door with the sleek design\"? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"an off-white interior door with the sleek design\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_42446167/Validation_42446167_42446167_79659.720.png", "Validation_42446167/Validation_42446167_42446167_79661.719.png", "Validation_42446167/Validation_42446167_42446167_79663.719.png", "Validation_42446167/Validation_42446167_42446167_79665.734.png" ], "turn_id": 3, "scan_id": "Validation_42446167", "origin_question": "Which two objects are on the same side of you now? \"the metal bed with an arched headboard and footboard\", \"the sleek, cylindrical gray fabric bin\", \"the large, sleek white rectangular radiator\". ", "option": [ "the metal bed with an arched headboard and footboard, the sleek, cylindrical gray fabric bin", "the metal bed with an arched headboard and footboard, the large, sleek white rectangular radiator", "the sleek, cylindrical gray fabric bin, the large, sleek white rectangular radiator" ], "answer": "the metal bed with an arched headboard and footboard, the sleek, cylindrical gray fabric bin", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the metal bed with an arched headboard and footboard\", \"the sleek, cylindrical gray fabric bin\", \"the large, sleek white rectangular radiator\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the metal bed with an arched headboard and footboard, the sleek, cylindrical gray fabric bin', 'the metal bed with an arched headboard and footboard, the large, sleek white rectangular radiator', 'the sleek, cylindrical gray fabric bin, the large, sleek white rectangular radiator'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_42446167/Validation_42446167_42446167_79667.734.png", "Validation_42446167/Validation_42446167_42446167_79669.733.png", "Validation_42446167/Validation_42446167_42446167_79671.732.png", "Validation_42446167/Validation_42446167_42446167_79673.731.png" ], "turn_id": 4, "scan_id": "Validation_42446167", "origin_question": "Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_42446167/Validation_42446167_42446167_79675.730.png", "Validation_42446167/Validation_42446167_42446167_79677.730.png", "Validation_42446167/Validation_42446167_42446167_79679.729.png", "Validation_42446167/Validation_42446167_42446167_79681.728.png" ], "turn_id": 5, "scan_id": "Validation_42446167", "origin_question": "At the end of which turn were both of \"the sleek, white ceramic toilet with the standard shape\" and \"the rectangular light-colored soft pillow\" on your right side?", "option": [], "answer": "1", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the sleek, white ceramic toilet with the standard shape\" and \"the rectangular light-colored soft pillow\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_42446167/Validation_42446167_42446167_79683.727.png", "Validation_42446167/Validation_42446167_42446167_79685.727.png", "Validation_42446167/Validation_42446167_42446167_79687.726.png", "Validation_42446167/Validation_42446167_42446167_79689.725.png" ], "turn_id": 6, "scan_id": "Validation_42446167", "origin_question": "Compared to your position at the end of the 5 turn, are you now closer or farther away from \"an off-white interior door with the sleek design\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 5 turn, are you now closer or farther away from \"an off-white interior door with the sleek design\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_42446167/Validation_42446167_42446167_79691.724.png", "Validation_42446167/Validation_42446167_42446167_79693.723.png", "Validation_42446167/Validation_42446167_42446167_79695.723.png", "Validation_42446167/Validation_42446167_42446167_79697.722.png" ], "turn_id": 7, "scan_id": "Validation_42446167", "origin_question": "Please recall \"the rectangular light-colored soft pillow\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.57", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the rectangular light-colored soft pillow\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_42446167/Validation_42446167_42446167_79699.721.png", "Validation_42446167/Validation_42446167_42446167_79701.720.png", "Validation_42446167/Validation_42446167_42446167_79703.719.png", "Validation_42446167/Validation_42446167_42446167_79705.719.png" ], "turn_id": 8, "scan_id": "Validation_42446167", "origin_question": "Which direction is the sleek, modern white rectangular table in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the sleek, modern white rectangular table in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_48018358/Validation_48018358_48018358_389049.965.png", "Validation_48018358/Validation_48018358_48018358_389051.948.png", "Validation_48018358/Validation_48018358_48018358_389053.947.png", "Validation_48018358/Validation_48018358_48018358_389055.946.png" ], "turn_id": 1, "scan_id": "Validation_48018358", "origin_question": "Remember, have you seen any telephone(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any telephone(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_48018358/Validation_48018358_48018358_389057.945.png", "Validation_48018358/Validation_48018358_48018358_389059.944.png", "Validation_48018358/Validation_48018358_48018358_389061.944.png", "Validation_48018358/Validation_48018358_48018358_389063.943.png" ], "turn_id": 2, "scan_id": "Validation_48018358", "origin_question": "At the end of which turn were both of \"the contemporary geometric light with metal and glass\" and \"the window\" on your right side?", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the contemporary geometric light with metal and glass\" and \"the window\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_48018358/Validation_48018358_48018358_389065.942.png", "Validation_48018358/Validation_48018358_48018358_389067.958.png", "Validation_48018358/Validation_48018358_48018358_389069.957.png", "Validation_48018358/Validation_48018358_48018358_389071.956.png" ], "turn_id": 3, "scan_id": "Validation_48018358", "origin_question": "At the end of which turn were both of \"the contemporary geometric light with metal and glass\" and \"the sleek rectangular table with light wood finish and black legs\" on your left side?", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the contemporary geometric light with metal and glass\" and \"the sleek rectangular table with light wood finish and black legs\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_48018358/Validation_48018358_48018358_389073.955.png", "Validation_48018358/Validation_48018358_48018358_389075.954.png", "Validation_48018358/Validation_48018358_48018358_389077.954.png", "Validation_48018358/Validation_48018358_48018358_389079.953.png" ], "turn_id": 4, "scan_id": "Validation_48018358", "origin_question": "Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "56.86", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_48018358/Validation_48018358_48018358_389081.952.png", "Validation_48018358/Validation_48018358_48018358_389083.951.png", "Validation_48018358/Validation_48018358_48018358_389085.950.png", "Validation_48018358/Validation_48018358_48018358_389087.949.png" ], "turn_id": 5, "scan_id": "Validation_48018358", "origin_question": "When did you first discover \"the window\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the window\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_48018358/Validation_48018358_48018358_389089.949.png", "Validation_48018358/Validation_48018358_48018358_389091.948.png", "Validation_48018358/Validation_48018358_48018358_389093.947.png", "Validation_48018358/Validation_48018358_48018358_389095.946.png" ], "turn_id": 6, "scan_id": "Validation_48018358", "origin_question": "Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the vibrant red, compact bed\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the vibrant red, compact bed\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "Validation_42899717/Validation_42899717_42899717_796800.113.png", "Validation_42899717/Validation_42899717_42899717_796802.095.png", "Validation_42899717/Validation_42899717_42899717_796804.094.png", "Validation_42899717/Validation_42899717_42899717_796806.094.png", "Validation_42899717/Validation_42899717_42899717_796808.109.png" ], "turn_id": 1, "scan_id": "Validation_42899717", "origin_question": "Remember, how many pillow(s) have you seen so far? ", "option": [], "answer": "2", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many pillow(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_42899717/Validation_42899717_42899717_796810.109.png", "Validation_42899717/Validation_42899717_42899717_796812.108.png", "Validation_42899717/Validation_42899717_42899717_796814.107.png", "Validation_42899717/Validation_42899717_42899717_796816.106.png", "Validation_42899717/Validation_42899717_42899717_796818.105.png" ], "turn_id": 2, "scan_id": "Validation_42899717", "origin_question": "In which turn did you see both \"the tall, slender white cabinet with the smooth finish\" and \"the modern, glossy black rectangular table\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the tall, slender white cabinet with the smooth finish\" and \"the modern, glossy black rectangular table\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_42899717/Validation_42899717_42899717_796820.105.png", "Validation_42899717/Validation_42899717_42899717_796822.104.png", "Validation_42899717/Validation_42899717_42899717_796824.103.png", "Validation_42899717/Validation_42899717_42899717_796826.102.png", "Validation_42899717/Validation_42899717_42899717_796828.101.png" ], "turn_id": 3, "scan_id": "Validation_42899717", "origin_question": "What will be the first-time appearance order of the following categories: chair, microwave, pillow?", "option": [ "microwave, chair, pillow", "microwave, pillow, chair", "chair, microwave, pillow", "pillow, chair, microwave" ], "answer": "pillow, chair, microwave", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: chair, microwave, pillow? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['microwave, chair, pillow', 'microwave, pillow, chair', 'chair, microwave, pillow', 'pillow, chair, microwave'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_42899717/Validation_42899717_42899717_796830.100.png", "Validation_42899717/Validation_42899717_42899717_796832.100.png", "Validation_42899717/Validation_42899717_42899717_796834.099.png", "Validation_42899717/Validation_42899717_42899717_796836.098.png", "Validation_42899717/Validation_42899717_42899717_796838.097.png" ], "turn_id": 4, "scan_id": "Validation_42899717", "origin_question": "In which turn were you closest to \"the gray, rectangular plastic bin\"? ", "option": [], "answer": "3", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the gray, rectangular plastic bin\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_42899717/Validation_42899717_42899717_796840.096.png", "Validation_42899717/Validation_42899717_42899717_796842.095.png", "Validation_42899717/Validation_42899717_42899717_796844.095.png", "Validation_42899717/Validation_42899717_42899717_796846.094.png", "Validation_42899717/Validation_42899717_42899717_796848.093.png" ], "turn_id": 5, "scan_id": "Validation_42899717", "origin_question": "Please recall \"the gray, rectangular plastic bin\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.76", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the gray, rectangular plastic bin\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_42899717/Validation_42899717_42899717_796850.092.png", "Validation_42899717/Validation_42899717_42899717_796852.108.png", "Validation_42899717/Validation_42899717_42899717_796854.107.png", "Validation_42899717/Validation_42899717_42899717_796856.106.png", "Validation_42899717/Validation_42899717_42899717_796858.105.png" ], "turn_id": 6, "scan_id": "Validation_42899717", "origin_question": "Which two objects are on the same side of you now? \"the tall stainless steel oven with the sleek design\", \"the gray, rectangular plastic bin\", \"the compact stainless steel microwave in metallic gray\". ", "option": [ "the tall stainless steel oven with the sleek design, the gray, rectangular plastic bin", "the tall stainless steel oven with the sleek design, the compact stainless steel microwave in metallic gray", "the gray, rectangular plastic bin, the compact stainless steel microwave in metallic gray" ], "answer": "the gray, rectangular plastic bin, the compact stainless steel microwave in metallic gray", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the tall stainless steel oven with the sleek design\", \"the gray, rectangular plastic bin\", \"the compact stainless steel microwave in metallic gray\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the tall stainless steel oven with the sleek design, the gray, rectangular plastic bin', 'the tall stainless steel oven with the sleek design, the compact stainless steel microwave in metallic gray', 'the gray, rectangular plastic bin, the compact stainless steel microwave in metallic gray'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_42899717/Validation_42899717_42899717_796860.105.png", "Validation_42899717/Validation_42899717_42899717_796862.104.png", "Validation_42899717/Validation_42899717_42899717_796864.103.png", "Validation_42899717/Validation_42899717_42899717_796866.102.png", "Validation_42899717/Validation_42899717_42899717_796868.101.png" ], "turn_id": 7, "scan_id": "Validation_42899717", "origin_question": "Please recall \"the large, rectangular stainless steel sink\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "3.82", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the large, rectangular stainless steel sink\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_42899717/Validation_42899717_42899717_796870.100.png", "Validation_42899717/Validation_42899717_42899717_796872.100.png", "Validation_42899717/Validation_42899717_42899717_796874.099.png", "Validation_42899717/Validation_42899717_42899717_796876.098.png", "Validation_42899717/Validation_42899717_42899717_796878.097.png" ], "turn_id": 8, "scan_id": "Validation_42899717", "origin_question": "In which turn did you see both \"the gray, rectangular plastic bin\" and \"the compact stainless steel microwave in metallic gray\" simultaneously? ", "option": [], "answer": "3", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the gray, rectangular plastic bin\" and \"the compact stainless steel microwave in metallic gray\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_42899717/Validation_42899717_42899717_796880.096.png", "Validation_42899717/Validation_42899717_42899717_796882.096.png", "Validation_42899717/Validation_42899717_42899717_796884.095.png", "Validation_42899717/Validation_42899717_42899717_796886.094.png", "Validation_42899717/Validation_42899717_42899717_796888.093.png" ], "turn_id": 9, "scan_id": "Validation_42899717", "origin_question": "Which two objects are on the same side of you now? \"the gray, rectangular plastic bin\", \"the large, rectangular stainless steel sink\", \"the sleek black flat-screen monitor with the rectangular shape\". ", "option": [ "the gray, rectangular plastic bin, the large, rectangular stainless steel sink", "the gray, rectangular plastic bin, the sleek black flat-screen monitor with the rectangular shape", "the large, rectangular stainless steel sink, the sleek black flat-screen monitor with the rectangular shape" ], "answer": "the gray, rectangular plastic bin, the large, rectangular stainless steel sink", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the gray, rectangular plastic bin\", \"the large, rectangular stainless steel sink\", \"the sleek black flat-screen monitor with the rectangular shape\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the gray, rectangular plastic bin, the large, rectangular stainless steel sink', 'the gray, rectangular plastic bin, the sleek black flat-screen monitor with the rectangular shape', 'the large, rectangular stainless steel sink, the sleek black flat-screen monitor with the rectangular shape'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_42899717/Validation_42899717_42899717_796890.092.png", "Validation_42899717/Validation_42899717_42899717_796892.108.png", "Validation_42899717/Validation_42899717_42899717_796894.107.png", "Validation_42899717/Validation_42899717_42899717_796896.106.png", "Validation_42899717/Validation_42899717_42899717_796898.106.png" ], "turn_id": 10, "scan_id": "Validation_42899717", "origin_question": "Which direction is the compact stainless steel microwave in metallic gray in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the compact stainless steel microwave in metallic gray in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "Validation_42899691/Validation_42899691_42899691_792261.692.png", "Validation_42899691/Validation_42899691_42899691_792263.675.png", "Validation_42899691/Validation_42899691_42899691_792265.674.png", "Validation_42899691/Validation_42899691_42899691_792267.673.png", "Validation_42899691/Validation_42899691_42899691_792269.689.png" ], "turn_id": 1, "scan_id": "Validation_42899691", "origin_question": "Remember, how many table(s) have you seen so far? ", "option": [], "answer": "2", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many table(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_42899691/Validation_42899691_42899691_792271.688.png", "Validation_42899691/Validation_42899691_42899691_792273.687.png", "Validation_42899691/Validation_42899691_42899691_792275.687.png", "Validation_42899691/Validation_42899691_42899691_792277.686.png", "Validation_42899691/Validation_42899691_42899691_792279.685.png" ], "turn_id": 2, "scan_id": "Validation_42899691", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the rectangular bed with the dark upholstered headboard\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the rectangular bed with the dark upholstered headboard\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_42899691/Validation_42899691_42899691_792281.684.png", "Validation_42899691/Validation_42899691_42899691_792283.683.png", "Validation_42899691/Validation_42899691_42899691_792285.682.png", "Validation_42899691/Validation_42899691_42899691_792287.682.png", "Validation_42899691/Validation_42899691_42899691_792289.681.png" ], "turn_id": 3, "scan_id": "Validation_42899691", "origin_question": "", "option": [], "answer": "", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_42899691/Validation_42899691_42899691_792291.680.png", "Validation_42899691/Validation_42899691_42899691_792293.679.png", "Validation_42899691/Validation_42899691_42899691_792295.678.png", "Validation_42899691/Validation_42899691_42899691_792297.677.png", "Validation_42899691/Validation_42899691_42899691_792299.677.png" ], "turn_id": 4, "scan_id": "Validation_42899691", "origin_question": "Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "159.08", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_42899691/Validation_42899691_42899691_792301.676.png", "Validation_42899691/Validation_42899691_42899691_792303.675.png", "Validation_42899691/Validation_42899691_42899691_792305.674.png", "Validation_42899691/Validation_42899691_42899691_792307.673.png", "Validation_42899691/Validation_42899691_42899691_792309.689.png" ], "turn_id": 5, "scan_id": "Validation_42899691", "origin_question": "Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the sleek, black rectangular monitor with the plastic frame\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the sleek, black rectangular monitor with the plastic frame\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_42899691/Validation_42899691_42899691_792311.688.png", "Validation_42899691/Validation_42899691_42899691_792313.688.png", "Validation_42899691/Validation_42899691_42899691_792315.687.png", "Validation_42899691/Validation_42899691_42899691_792317.686.png", "Validation_42899691/Validation_42899691_42899691_792319.685.png" ], "turn_id": 6, "scan_id": "Validation_42899691", "origin_question": "How far is your current position from where you were at the end of the turn 3? (in meters)", "option": [], "answer": "3.37", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 3? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_42899691/Validation_42899691_42899691_792321.684.png", "Validation_42899691/Validation_42899691_42899691_792323.684.png", "Validation_42899691/Validation_42899691_42899691_792325.683.png", "Validation_42899691/Validation_42899691_42899691_792327.682.png", "Validation_42899691/Validation_42899691_42899691_792329.681.png" ], "turn_id": 7, "scan_id": "Validation_42899691", "origin_question": "What will be the first-time appearance order of the following categories: bed, bin, cabinet?", "option": [ "cabinet, bin, bed", "bin, bed, cabinet", "bed, bin, cabinet", "bed, cabinet, bin" ], "answer": "bed, cabinet, bin", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bed, bin, cabinet? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['cabinet, bin, bed', 'bin, bed, cabinet', 'bed, bin, cabinet', 'bed, cabinet, bin'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_42899691/Validation_42899691_42899691_792331.680.png", "Validation_42899691/Validation_42899691_42899691_792333.679.png", "Validation_42899691/Validation_42899691_42899691_792335.679.png", "Validation_42899691/Validation_42899691_42899691_792337.678.png", "Validation_42899691/Validation_42899691_42899691_792339.677.png" ], "turn_id": 8, "scan_id": "Validation_42899691", "origin_question": "At the end of which turn were both of \"the rectangular bed with the dark upholstered headboard\" and \"the large frameless rectangular mirror with clear glass\" on your left side?", "option": [], "answer": "3", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the rectangular bed with the dark upholstered headboard\" and \"the large frameless rectangular mirror with clear glass\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42899698/Validation_42899698_42899698_793201.164.png", "Validation_42899698/Validation_42899698_42899698_793202.647.png", "Validation_42899698/Validation_42899698_42899698_793204.146.png", "Validation_42899698/Validation_42899698_42899698_793205.645.png" ], "turn_id": 1, "scan_id": "Validation_42899698", "origin_question": "Remember, have you seen any bathtub(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any bathtub(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_42899698/Validation_42899698_42899698_793207.161.png", "Validation_42899698/Validation_42899698_42899698_793208.661.png", "Validation_42899698/Validation_42899698_42899698_793210.160.png", "Validation_42899698/Validation_42899698_42899698_793211.659.png" ], "turn_id": 2, "scan_id": "Validation_42899698", "origin_question": "When did you first discover \"the modern rectangular glossy white ceramic sink\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the modern rectangular glossy white ceramic sink\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_42899698/Validation_42899698_42899698_793213.159.png", "Validation_42899698/Validation_42899698_42899698_793214.658.png", "Validation_42899698/Validation_42899698_42899698_793216.158.png", "Validation_42899698/Validation_42899698_42899698_793217.657.png" ], "turn_id": 3, "scan_id": "Validation_42899698", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "81.72", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_42899698/Validation_42899698_42899698_793219.156.png", "Validation_42899698/Validation_42899698_42899698_793220.656.png", "Validation_42899698/Validation_42899698_42899698_793222.155.png", "Validation_42899698/Validation_42899698_42899698_793223.655.png" ], "turn_id": 4, "scan_id": "Validation_42899698", "origin_question": "Please recall \"the sleek, rectangular white acrylic bathtub\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.29", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the sleek, rectangular white acrylic bathtub\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_42899698/Validation_42899698_42899698_793225.154.png", "Validation_42899698/Validation_42899698_42899698_793226.653.png", "Validation_42899698/Validation_42899698_42899698_793228.153.png", "Validation_42899698/Validation_42899698_42899698_793229.652.png" ], "turn_id": 5, "scan_id": "Validation_42899698", "origin_question": "Which direction is the sleek, rectangular white acrylic bathtub in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the sleek, rectangular white acrylic bathtub in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_42899698/Validation_42899698_42899698_793231.152.png", "Validation_42899698/Validation_42899698_42899698_793232.651.png", "Validation_42899698/Validation_42899698_42899698_793234.150.png", "Validation_42899698/Validation_42899698_42899698_793235.650.png" ], "turn_id": 6, "scan_id": "Validation_42899698", "origin_question": "Is the standard white roll of soft toilet paper to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the standard white roll of soft toilet paper to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_42899698/Validation_42899698_42899698_793237.149.png", "Validation_42899698/Validation_42899698_42899698_793238.648.png", "Validation_42899698/Validation_42899698_42899698_793240.148.png", "Validation_42899698/Validation_42899698_42899698_793241.647.png" ], "turn_id": 7, "scan_id": "Validation_42899698", "origin_question": "Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47331971/Validation_47331971_47331971_168966.397.png", "Validation_47331971/Validation_47331971_47331971_168968.379.png", "Validation_47331971/Validation_47331971_47331971_168970.378.png", "Validation_47331971/Validation_47331971_47331971_168972.378.png", "Validation_47331971/Validation_47331971_47331971_168974.377.png" ], "turn_id": 1, "scan_id": "Validation_47331971", "origin_question": "", "option": [], "answer": "", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_47331971/Validation_47331971_47331971_168976.376.png", "Validation_47331971/Validation_47331971_47331971_168978.375.png", "Validation_47331971/Validation_47331971_47331971_168980.375.png", "Validation_47331971/Validation_47331971_47331971_168982.374.png", "Validation_47331971/Validation_47331971_47331971_168984.373.png" ], "turn_id": 2, "scan_id": "Validation_47331971", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the medium-sized cylindrical flowerpot with the glossy beige surface\", Y: \"the sleek white rectangular wooden table\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the medium-sized cylindrical flowerpot with the glossy beige surface\", Y: \"the sleek white rectangular wooden table\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_47331971/Validation_47331971_47331971_168986.372.png", "Validation_47331971/Validation_47331971_47331971_168988.371.png", "Validation_47331971/Validation_47331971_47331971_168990.370.png", "Validation_47331971/Validation_47331971_47331971_168992.370.png", "Validation_47331971/Validation_47331971_47331971_168994.369.png" ], "turn_id": 3, "scan_id": "Validation_47331971", "origin_question": "Is the sleek white rectangular wooden table to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the sleek white rectangular wooden table to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_47331971/Validation_47331971_47331971_168996.368.png", "Validation_47331971/Validation_47331971_47331971_168998.367.png", "Validation_47331971/Validation_47331971_47331971_169000.366.png", "Validation_47331971/Validation_47331971_47331971_169002.366.png", "Validation_47331971/Validation_47331971_47331971_169004.365.png" ], "turn_id": 4, "scan_id": "Validation_47331971", "origin_question": "What will be the first-time appearance order of the following categories: chair, couch, window?", "option": [ "window, chair, couch", "couch, window, chair", "window, couch, chair", "chair, couch, window" ], "answer": "chair, couch, window", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: chair, couch, window? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['window, chair, couch', 'couch, window, chair', 'window, couch, chair', 'chair, couch, window'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47331971/Validation_47331971_47331971_169006.364.png", "Validation_47331971/Validation_47331971_47331971_169008.380.png", "Validation_47331971/Validation_47331971_47331971_169010.379.png", "Validation_47331971/Validation_47331971_47331971_169012.378.png", "Validation_47331971/Validation_47331971_47331971_169014.377.png" ], "turn_id": 5, "scan_id": "Validation_47331971", "origin_question": "Remember, have you seen any body loofah(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any body loofah(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_47331971/Validation_47331971_47331971_169016.377.png", "Validation_47331971/Validation_47331971_47331971_169018.376.png", "Validation_47331971/Validation_47331971_47331971_169020.375.png", "Validation_47331971/Validation_47331971_47331971_169022.374.png", "Validation_47331971/Validation_47331971_47331971_169024.373.png" ], "turn_id": 6, "scan_id": "Validation_47331971", "origin_question": "How far is your current position from where you were at the end of the turn 3? (in meters)", "option": [], "answer": "2.08", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 3? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_47331971/Validation_47331971_47331971_169026.372.png", "Validation_47331971/Validation_47331971_47331971_169028.372.png", "Validation_47331971/Validation_47331971_47331971_169030.371.png", "Validation_47331971/Validation_47331971_47331971_169032.370.png", "Validation_47331971/Validation_47331971_47331971_169034.369.png" ], "turn_id": 7, "scan_id": "Validation_47331971", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47331971/Validation_47331971_47331971_169036.368.png", "Validation_47331971/Validation_47331971_47331971_169038.368.png", "Validation_47331971/Validation_47331971_47331971_169040.367.png", "Validation_47331971/Validation_47331971_47331971_169042.366.png", "Validation_47331971/Validation_47331971_47331971_169044.365.png" ], "turn_id": 8, "scan_id": "Validation_47331971", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the geometric dark green ceramic lamp\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the geometric dark green ceramic lamp\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_47331971/Validation_47331971_47331971_169046.364.png", "Validation_47331971/Validation_47331971_47331971_169048.380.png", "Validation_47331971/Validation_47331971_47331971_169050.379.png", "Validation_47331971/Validation_47331971_47331971_169052.379.png", "Validation_47331971/Validation_47331971_47331971_169054.378.png" ], "turn_id": 9, "scan_id": "Validation_47331971", "origin_question": "Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_47331971/Validation_47331971_47331971_169056.377.png", "Validation_47331971/Validation_47331971_47331971_169058.376.png", "Validation_47331971/Validation_47331971_47331971_169060.375.png", "Validation_47331971/Validation_47331971_47331971_169062.375.png", "Validation_47331971/Validation_47331971_47331971_169064.374.png" ], "turn_id": 10, "scan_id": "Validation_47331971", "origin_question": "When did you first discover \"the window\"? (index of the turn)", "option": [], "answer": "4", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the window\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_48018730/Validation_48018730_48018730_1668.491.png", "Validation_48018730/Validation_48018730_48018730_1670.473.png", "Validation_48018730/Validation_48018730_48018730_1672.473.png", "Validation_48018730/Validation_48018730_48018730_1674.472.png", "Validation_48018730/Validation_48018730_48018730_1676.471.png" ], "turn_id": 1, "scan_id": "Validation_48018730", "origin_question": "Remember, have you seen any cabinet(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any cabinet(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_48018730/Validation_48018730_48018730_1678.470.png", "Validation_48018730/Validation_48018730_48018730_1680.469.png", "Validation_48018730/Validation_48018730_48018730_1682.469.png", "Validation_48018730/Validation_48018730_48018730_1684.468.png", "Validation_48018730/Validation_48018730_48018730_1686.467.png" ], "turn_id": 2, "scan_id": "Validation_48018730", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the large rectangular white-framed window\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the large rectangular white-framed window\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_48018730/Validation_48018730_48018730_1688.466.png", "Validation_48018730/Validation_48018730_48018730_1690.465.png", "Validation_48018730/Validation_48018730_48018730_1692.465.png", "Validation_48018730/Validation_48018730_48018730_1694.480.png", "Validation_48018730/Validation_48018730_48018730_1696.480.png" ], "turn_id": 3, "scan_id": "Validation_48018730", "origin_question": "Is the large rectangular white-framed window to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the large rectangular white-framed window to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_48018730/Validation_48018730_48018730_1698.479.png", "Validation_48018730/Validation_48018730_48018730_1700.478.png", "Validation_48018730/Validation_48018730_48018730_1702.477.png", "Validation_48018730/Validation_48018730_48018730_1704.476.png", "Validation_48018730/Validation_48018730_48018730_1706.476.png" ], "turn_id": 4, "scan_id": "Validation_48018730", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the sleek, cube-shaped frosted glass light\" from your position?", "option": [], "answer": "51.29", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the sleek, cube-shaped frosted glass light\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_48018730/Validation_48018730_48018730_1708.475.png", "Validation_48018730/Validation_48018730_48018730_1710.474.png", "Validation_48018730/Validation_48018730_48018730_1712.473.png", "Validation_48018730/Validation_48018730_48018730_1714.472.png", "Validation_48018730/Validation_48018730_48018730_1716.472.png" ], "turn_id": 5, "scan_id": "Validation_48018730", "origin_question": "At the end of which turn were both of \"the large rectangular white-framed window\" and \"the full-sized wooden bed with the mosaic-patterned bedspread\" on your left side?", "option": [], "answer": "3", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the large rectangular white-framed window\" and \"the full-sized wooden bed with the mosaic-patterned bedspread\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_48018730/Validation_48018730_48018730_1718.471.png", "Validation_48018730/Validation_48018730_48018730_1720.470.png", "Validation_48018730/Validation_48018730_48018730_1722.469.png", "Validation_48018730/Validation_48018730_48018730_1724.468.png", "Validation_48018730/Validation_48018730_48018730_1726.467.png" ], "turn_id": 6, "scan_id": "Validation_48018730", "origin_question": "When did you first discover \"the sleek flat-screen monitor with the matte black frame\"? (index of the turn)", "option": [], "answer": "6", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the sleek flat-screen monitor with the matte black frame\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_48018730/Validation_48018730_48018730_1728.467.png", "Validation_48018730/Validation_48018730_48018730_1730.466.png", "Validation_48018730/Validation_48018730_48018730_1732.465.png", "Validation_48018730/Validation_48018730_48018730_1734.464.png", "Validation_48018730/Validation_48018730_48018730_1736.480.png" ], "turn_id": 7, "scan_id": "Validation_48018730", "origin_question": "Please recall \"the full-sized wooden bed with the mosaic-patterned bedspread\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.34", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the full-sized wooden bed with the mosaic-patterned bedspread\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_48018730/Validation_48018730_48018730_1738.479.png", "Validation_48018730/Validation_48018730_48018730_1740.478.png", "Validation_48018730/Validation_48018730_48018730_1742.478.png", "Validation_48018730/Validation_48018730_48018730_1744.477.png", "Validation_48018730/Validation_48018730_48018730_1746.476.png" ], "turn_id": 8, "scan_id": "Validation_48018730", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "35.24", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_48018730/Validation_48018730_48018730_1748.475.png", "Validation_48018730/Validation_48018730_48018730_1750.474.png", "Validation_48018730/Validation_48018730_48018730_1752.474.png", "Validation_48018730/Validation_48018730_48018730_1754.473.png", "Validation_48018730/Validation_48018730_48018730_1756.472.png" ], "turn_id": 9, "scan_id": "Validation_48018730", "origin_question": "At the end of which turn were both of \"the large rectangular white-framed window\" and \"the dark brown faux leather cushion stool with straight legs\" on your left side?", "option": [], "answer": "4", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the large rectangular white-framed window\" and \"the dark brown faux leather cushion stool with straight legs\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_48018730/Validation_48018730_48018730_1758.471.png", "Validation_48018730/Validation_48018730_48018730_1760.470.png", "Validation_48018730/Validation_48018730_48018730_1762.470.png", "Validation_48018730/Validation_48018730_48018730_1764.469.png", "Validation_48018730/Validation_48018730_48018730_1766.468.png" ], "turn_id": 10, "scan_id": "Validation_48018730", "origin_question": "Is the rich brown upholstered chair with curved armrests to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the rich brown upholstered chair with curved armrests to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_41069046/Validation_41069046_41069046_2911.344.png", "Validation_41069046/Validation_41069046_41069046_2913.326.png", "Validation_41069046/Validation_41069046_41069046_2915.325.png", "Validation_41069046/Validation_41069046_41069046_2917.325.png", "Validation_41069046/Validation_41069046_41069046_2919.324.png" ], "turn_id": 1, "scan_id": "Validation_41069046", "origin_question": "Remember, have you seen any bread(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any bread(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_41069046/Validation_41069046_41069046_2921.323.png", "Validation_41069046/Validation_41069046_41069046_2923.339.png", "Validation_41069046/Validation_41069046_41069046_2925.338.png", "Validation_41069046/Validation_41069046_41069046_2927.337.png", "Validation_41069046/Validation_41069046_41069046_2929.336.png" ], "turn_id": 2, "scan_id": "Validation_41069046", "origin_question": "When did you first discover \"the bed\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the bed\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_41069046/Validation_41069046_41069046_2931.335.png", "Validation_41069046/Validation_41069046_41069046_2933.334.png", "Validation_41069046/Validation_41069046_41069046_2935.334.png", "Validation_41069046/Validation_41069046_41069046_2937.333.png", "Validation_41069046/Validation_41069046_41069046_2939.332.png" ], "turn_id": 3, "scan_id": "Validation_41069046", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "57.12", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_41069046/Validation_41069046_41069046_2941.331.png", "Validation_41069046/Validation_41069046_41069046_2943.330.png", "Validation_41069046/Validation_41069046_41069046_2945.329.png", "Validation_41069046/Validation_41069046_41069046_2947.329.png", "Validation_41069046/Validation_41069046_41069046_2949.328.png" ], "turn_id": 4, "scan_id": "Validation_41069046", "origin_question": "Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "121.13", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_41069046/Validation_41069046_41069046_2951.327.png", "Validation_41069046/Validation_41069046_41069046_2953.326.png", "Validation_41069046/Validation_41069046_41069046_2955.325.png", "Validation_41069046/Validation_41069046_41069046_2957.325.png", "Validation_41069046/Validation_41069046_41069046_2959.324.png" ], "turn_id": 5, "scan_id": "Validation_41069046", "origin_question": "Using your orientation at the end of turn 4 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "57.67", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 4 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_41069046/Validation_41069046_41069046_2961.323.png", "Validation_41069046/Validation_41069046_41069046_2963.322.png", "Validation_41069046/Validation_41069046_41069046_2965.338.png", "Validation_41069046/Validation_41069046_41069046_2967.337.png", "Validation_41069046/Validation_41069046_41069046_2969.336.png" ], "turn_id": 6, "scan_id": "Validation_41069046", "origin_question": "What will be the first-time appearance order of the following categories: door, light, pillow?", "option": [ "door, light, pillow", "door, pillow, light", "pillow, light, door", "pillow, door, light" ], "answer": "door, light, pillow", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: door, light, pillow? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['door, light, pillow', 'door, pillow, light', 'pillow, light, door', 'pillow, door, light'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_41069046/Validation_41069046_41069046_2971.336.png", "Validation_41069046/Validation_41069046_41069046_2973.335.png", "Validation_41069046/Validation_41069046_41069046_2975.334.png", "Validation_41069046/Validation_41069046_41069046_2977.333.png", "Validation_41069046/Validation_41069046_41069046_2979.332.png" ], "turn_id": 7, "scan_id": "Validation_41069046", "origin_question": "Compared to your position at the end of the 5 turn, are you now closer or farther away from \"the medium-sized square picture with the colorful geometric design\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 5 turn, are you now closer or farther away from \"the medium-sized square picture with the colorful geometric design\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_41069046/Validation_41069046_41069046_2981.331.png", "Validation_41069046/Validation_41069046_41069046_2983.331.png", "Validation_41069046/Validation_41069046_41069046_2985.330.png", "Validation_41069046/Validation_41069046_41069046_2987.329.png", "Validation_41069046/Validation_41069046_41069046_2989.328.png" ], "turn_id": 8, "scan_id": "Validation_41069046", "origin_question": "Compared to your position at the end of the 6 turn, are you now closer or farther away from \"the bed\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 6 turn, are you now closer or farther away from \"the bed\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_41069046/Validation_41069046_41069046_2991.327.png", "Validation_41069046/Validation_41069046_41069046_2993.326.png", "Validation_41069046/Validation_41069046_41069046_2995.325.png", "Validation_41069046/Validation_41069046_41069046_2997.325.png", "Validation_41069046/Validation_41069046_41069046_2999.324.png" ], "turn_id": 9, "scan_id": "Validation_41069046", "origin_question": "Assuming the direction you are facing at the end of the turn 5 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Right", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 5 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_41069046/Validation_41069046_41069046_3001.323.png", "Validation_41069046/Validation_41069046_41069046_3003.322.png", "Validation_41069046/Validation_41069046_41069046_3005.321.png", "Validation_41069046/Validation_41069046_41069046_3007.337.png", "Validation_41069046/Validation_41069046_41069046_3009.336.png" ], "turn_id": 10, "scan_id": "Validation_41069046", "origin_question": "In which turn were you closest to \"the medium-sized square picture with the colorful geometric design\"? ", "option": [], "answer": "7", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the medium-sized square picture with the colorful geometric design\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_48458732/Validation_48458732_48458732_139096.306.png", "Validation_48458732/Validation_48458732_48458732_139098.289.png", "Validation_48458732/Validation_48458732_48458732_139100.288.png", "Validation_48458732/Validation_48458732_48458732_139102.287.png" ], "turn_id": 1, "scan_id": "Validation_48458732", "origin_question": "Remember, have you seen any chair(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any chair(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_48458732/Validation_48458732_48458732_139104.286.png", "Validation_48458732/Validation_48458732_48458732_139106.286.png", "Validation_48458732/Validation_48458732_48458732_139108.285.png", "Validation_48458732/Validation_48458732_48458732_139110.284.png" ], "turn_id": 2, "scan_id": "Validation_48458732", "origin_question": "In which turn did you see both \"the sleek rectangular black monitor with the slim bezel\" and \"the sleek, compact microwave with the metal and plastic design\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the sleek rectangular black monitor with the slim bezel\" and \"the sleek, compact microwave with the metal and plastic design\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_48458732/Validation_48458732_48458732_139112.283.png", "Validation_48458732/Validation_48458732_48458732_139114.283.png", "Validation_48458732/Validation_48458732_48458732_139116.298.png", "Validation_48458732/Validation_48458732_48458732_139118.298.png" ], "turn_id": 3, "scan_id": "Validation_48458732", "origin_question": "Which direction is the oven in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the oven in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_48458732/Validation_48458732_48458732_139120.297.png", "Validation_48458732/Validation_48458732_48458732_139122.296.png", "Validation_48458732/Validation_48458732_48458732_139124.295.png", "Validation_48458732/Validation_48458732_48458732_139126.294.png" ], "turn_id": 4, "scan_id": "Validation_48458732", "origin_question": "Is the oven to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the oven to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_48458732/Validation_48458732_48458732_139128.294.png", "Validation_48458732/Validation_48458732_48458732_139130.293.png", "Validation_48458732/Validation_48458732_48458732_139132.292.png", "Validation_48458732/Validation_48458732_48458732_139134.291.png" ], "turn_id": 5, "scan_id": "Validation_48458732", "origin_question": "In which turn did you see both \"the oven\" and \"the large rectangular window framed in white\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the oven\" and \"the large rectangular window framed in white\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_48458732/Validation_48458732_48458732_139136.291.png", "Validation_48458732/Validation_48458732_48458732_139138.290.png", "Validation_48458732/Validation_48458732_48458732_139140.289.png", "Validation_48458732/Validation_48458732_48458732_139142.288.png" ], "turn_id": 6, "scan_id": "Validation_48458732", "origin_question": "Please recall \"the sleek, compact microwave with the metal and plastic design\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.12", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the sleek, compact microwave with the metal and plastic design\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_48458732/Validation_48458732_48458732_139144.287.png", "Validation_48458732/Validation_48458732_48458732_139146.286.png", "Validation_48458732/Validation_48458732_48458732_139148.286.png", "Validation_48458732/Validation_48458732_48458732_139150.285.png" ], "turn_id": 7, "scan_id": "Validation_48458732", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_48458732/Validation_48458732_48458732_139152.284.png", "Validation_48458732/Validation_48458732_48458732_139154.283.png", "Validation_48458732/Validation_48458732_48458732_139156.299.png", "Validation_48458732/Validation_48458732_48458732_139158.298.png" ], "turn_id": 8, "scan_id": "Validation_48458732", "origin_question": "How far is your current position from where you were at the end of the turn 4? (in meters)", "option": [], "answer": "2.05", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 4? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47332005/Validation_47332005_47332005_172063.764.png", "Validation_47332005/Validation_47332005_47332005_172065.730.png", "Validation_47332005/Validation_47332005_47332005_172067.729.png", "Validation_47332005/Validation_47332005_47332005_172069.728.png", "Validation_47332005/Validation_47332005_47332005_172071.728.png" ], "turn_id": 1, "scan_id": "Validation_47332005", "origin_question": "", "option": [], "answer": "", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47332005/Validation_47332005_47332005_172073.744.png", "Validation_47332005/Validation_47332005_47332005_172075.743.png", "Validation_47332005/Validation_47332005_47332005_172077.742.png", "Validation_47332005/Validation_47332005_47332005_172079.741.png", "Validation_47332005/Validation_47332005_47332005_172081.740.png" ], "turn_id": 2, "scan_id": "Validation_47332005", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_47332005/Validation_47332005_47332005_172083.739.png", "Validation_47332005/Validation_47332005_47332005_172085.739.png", "Validation_47332005/Validation_47332005_47332005_172087.738.png", "Validation_47332005/Validation_47332005_47332005_172089.737.png", "Validation_47332005/Validation_47332005_47332005_172091.736.png" ], "turn_id": 3, "scan_id": "Validation_47332005", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the sleek gray cabinet with the glossy finish\", Y: \"the small, cylindrical ceramic flowerpot in an earthy tone\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the sleek gray cabinet with the glossy finish\", Y: \"the small, cylindrical ceramic flowerpot in an earthy tone\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47332005/Validation_47332005_47332005_172093.735.png", "Validation_47332005/Validation_47332005_47332005_172095.735.png", "Validation_47332005/Validation_47332005_47332005_172097.734.png", "Validation_47332005/Validation_47332005_47332005_172099.733.png", "Validation_47332005/Validation_47332005_47332005_172101.732.png" ], "turn_id": 4, "scan_id": "Validation_47332005", "origin_question": "", "option": [], "answer": "", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_47332005/Validation_47332005_47332005_172103.731.png", "Validation_47332005/Validation_47332005_47332005_172105.731.png", "Validation_47332005/Validation_47332005_47332005_172107.730.png", "Validation_47332005/Validation_47332005_47332005_172109.729.png", "Validation_47332005/Validation_47332005_47332005_172111.728.png" ], "turn_id": 5, "scan_id": "Validation_47332005", "origin_question": "Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_47332005/Validation_47332005_47332005_172113.727.png", "Validation_47332005/Validation_47332005_47332005_172115.743.png", "Validation_47332005/Validation_47332005_47332005_172117.742.png", "Validation_47332005/Validation_47332005_47332005_172119.742.png", "Validation_47332005/Validation_47332005_47332005_172121.741.png" ], "turn_id": 6, "scan_id": "Validation_47332005", "origin_question": "How far is your current position from where you were at the end of the turn 3? (in meters)", "option": [], "answer": "1.13", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 3? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47332005/Validation_47332005_47332005_172123.740.png", "Validation_47332005/Validation_47332005_47332005_172125.739.png", "Validation_47332005/Validation_47332005_47332005_172127.738.png", "Validation_47332005/Validation_47332005_47332005_172129.737.png", "Validation_47332005/Validation_47332005_47332005_172131.737.png" ], "turn_id": 7, "scan_id": "Validation_47332005", "origin_question": "", "option": [], "answer": "", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_47332005/Validation_47332005_47332005_172133.736.png", "Validation_47332005/Validation_47332005_47332005_172135.735.png", "Validation_47332005/Validation_47332005_47332005_172137.734.png", "Validation_47332005/Validation_47332005_47332005_172139.733.png", "Validation_47332005/Validation_47332005_47332005_172141.733.png" ], "turn_id": 8, "scan_id": "Validation_47332005", "origin_question": "How far is your current position from where you were at the end of the turn 3? (in meters)", "option": [], "answer": "1.27", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 3? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47332005/Validation_47332005_47332005_172143.732.png", "Validation_47332005/Validation_47332005_47332005_172145.731.png", "Validation_47332005/Validation_47332005_47332005_172147.730.png", "Validation_47332005/Validation_47332005_47332005_172149.729.png", "Validation_47332005/Validation_47332005_47332005_172151.728.png" ], "turn_id": 9, "scan_id": "Validation_47332005", "origin_question": "Which direction is the sleek white rectangular metal radiator in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the sleek white rectangular metal radiator in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47332005/Validation_47332005_47332005_172153.728.png", "Validation_47332005/Validation_47332005_47332005_172155.744.png", "Validation_47332005/Validation_47332005_47332005_172157.743.png", "Validation_47332005/Validation_47332005_47332005_172159.742.png", "Validation_47332005/Validation_47332005_47332005_172161.741.png" ], "turn_id": 10, "scan_id": "Validation_47332005", "origin_question": "Which direction is the sleek, glossy white ceramic sink with the rectangular basin in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the sleek, glossy white ceramic sink with the rectangular basin in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47331989/Validation_47331989_47331989_170414.880.png", "Validation_47331989/Validation_47331989_47331989_170416.862.png", "Validation_47331989/Validation_47331989_47331989_170418.861.png", "Validation_47331989/Validation_47331989_47331989_170420.861.png" ], "turn_id": 1, "scan_id": "Validation_47331989", "origin_question": "", "option": [], "answer": "", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47331989/Validation_47331989_47331989_170422.860.png", "Validation_47331989/Validation_47331989_47331989_170424.859.png", "Validation_47331989/Validation_47331989_47331989_170426.858.png", "Validation_47331989/Validation_47331989_47331989_170428.857.png" ], "turn_id": 2, "scan_id": "Validation_47331989", "origin_question": "In which turn were you closest to \"the slim, black glossy monitor with the rectangular shape\"? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the slim, black glossy monitor with the rectangular shape\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47331989/Validation_47331989_47331989_170430.857.png", "Validation_47331989/Validation_47331989_47331989_170432.856.png", "Validation_47331989/Validation_47331989_47331989_170434.855.png", "Validation_47331989/Validation_47331989_47331989_170436.854.png" ], "turn_id": 3, "scan_id": "Validation_47331989", "origin_question": "", "option": [], "answer": "", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_47331989/Validation_47331989_47331989_170438.853.png", "Validation_47331989/Validation_47331989_47331989_170440.853.png", "Validation_47331989/Validation_47331989_47331989_170442.852.png", "Validation_47331989/Validation_47331989_47331989_170444.851.png" ], "turn_id": 4, "scan_id": "Validation_47331989", "origin_question": "Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "99.99", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_47331989/Validation_47331989_47331989_170446.850.png", "Validation_47331989/Validation_47331989_47331989_170448.849.png", "Validation_47331989/Validation_47331989_47331989_170450.865.png", "Validation_47331989/Validation_47331989_47331989_170452.864.png" ], "turn_id": 5, "scan_id": "Validation_47331989", "origin_question": "Is the compact, cozy couch in the muted color to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the compact, cozy couch in the muted color to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_47331989/Validation_47331989_47331989_170454.864.png", "Validation_47331989/Validation_47331989_47331989_170456.863.png", "Validation_47331989/Validation_47331989_47331989_170458.862.png", "Validation_47331989/Validation_47331989_47331989_170460.861.png" ], "turn_id": 6, "scan_id": "Validation_47331989", "origin_question": "Using your orientation at the end of turn 4 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "44.58", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 4 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_47331989/Validation_47331989_47331989_170462.860.png", "Validation_47331989/Validation_47331989_47331989_170464.860.png", "Validation_47331989/Validation_47331989_47331989_170466.859.png", "Validation_47331989/Validation_47331989_47331989_170468.858.png" ], "turn_id": 7, "scan_id": "Validation_47331989", "origin_question": "How far is your current position from where you were at the end of the turn 3? (in meters)", "option": [], "answer": "1.07", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 3? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47332901/Validation_47332901_47332901_11017.617.png", "Validation_47332901/Validation_47332901_47332901_11019.599.png", "Validation_47332901/Validation_47332901_47332901_11021.598.png", "Validation_47332901/Validation_47332901_47332901_11023.597.png" ], "turn_id": 1, "scan_id": "Validation_47332901", "origin_question": "Remember, have you seen any kitchen island(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any kitchen island(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47332901/Validation_47332901_47332901_11025.597.png", "Validation_47332901/Validation_47332901_47332901_11027.596.png", "Validation_47332901/Validation_47332901_47332901_11029.595.png", "Validation_47332901/Validation_47332901_47332901_11031.594.png" ], "turn_id": 2, "scan_id": "Validation_47332901", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47332901/Validation_47332901_47332901_11033.593.png", "Validation_47332901/Validation_47332901_47332901_11035.593.png", "Validation_47332901/Validation_47332901_47332901_11037.592.png", "Validation_47332901/Validation_47332901_47332901_11039.591.png" ], "turn_id": 3, "scan_id": "Validation_47332901", "origin_question": "In which turn were you closest to \"the sleek black stainless steel stove with multiple burners\"? ", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the sleek black stainless steel stove with multiple burners\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_47332901/Validation_47332901_47332901_11041.590.png", "Validation_47332901/Validation_47332901_47332901_11043.589.png", "Validation_47332901/Validation_47332901_47332901_11045.588.png", "Validation_47332901/Validation_47332901_47332901_11047.588.png" ], "turn_id": 4, "scan_id": "Validation_47332901", "origin_question": "Please recall \"the sleek rectangular microwave with the metal and plastic body\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.69", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the sleek rectangular microwave with the metal and plastic body\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_47332901/Validation_47332901_47332901_11049.587.png", "Validation_47332901/Validation_47332901_47332901_11051.586.png", "Validation_47332901/Validation_47332901_47332901_11053.585.png", "Validation_47332901/Validation_47332901_47332901_11055.601.png" ], "turn_id": 5, "scan_id": "Validation_47332901", "origin_question": "When did you first discover \"the sleek black stainless steel stove with multiple burners\"? (index of the turn)", "option": [], "answer": "3", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the sleek black stainless steel stove with multiple burners\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "Validation_47332901/Validation_47332901_47332901_11057.600.png", "Validation_47332901/Validation_47332901_47332901_11059.600.png", "Validation_47332901/Validation_47332901_47332901_11061.599.png", "Validation_47332901/Validation_47332901_47332901_11063.598.png" ], "turn_id": 6, "scan_id": "Validation_47332901", "origin_question": "Among these three objects(\"the modern gray stainless steel refrigerator\",\"the sleek rectangular microwave with the metal and plastic body\",\"the sleek black stainless steel stove with multiple burners\" ), which one is the farthest from you?", "option": [ "the modern gray stainless steel refrigerator", "the sleek rectangular microwave with the metal and plastic body", "the sleek black stainless steel stove with multiple burners" ], "answer": "the sleek black stainless steel stove with multiple burners", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects(\"the modern gray stainless steel refrigerator\",\"the sleek rectangular microwave with the metal and plastic body\",\"the sleek black stainless steel stove with multiple burners\" ), which one is the farthest from you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the modern gray stainless steel refrigerator', 'the sleek rectangular microwave with the metal and plastic body', 'the sleek black stainless steel stove with multiple burners'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_47332901/Validation_47332901_47332901_11065.597.png", "Validation_47332901/Validation_47332901_47332901_11067.596.png", "Validation_47332901/Validation_47332901_47332901_11069.595.png", "Validation_47332901/Validation_47332901_47332901_11071.595.png" ], "turn_id": 7, "scan_id": "Validation_47332901", "origin_question": "By the end of the 6 turn, were you getting closer to or farther away from these two objects: X: \"the compact rectangular metal oven with the white exterior\", Y: \"the sleek black stainless steel stove with multiple burners\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 6 turn, were you getting closer to or farther away from these two objects: X: \"the compact rectangular metal oven with the white exterior\", Y: \"the sleek black stainless steel stove with multiple burners\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_48018739/Validation_48018739_48018739_3431.543.png", "Validation_48018739/Validation_48018739_48018739_3433.026.png", "Validation_48018739/Validation_48018739_48018739_3434.525.png", "Validation_48018739/Validation_48018739_48018739_3436.024.png" ], "turn_id": 1, "scan_id": "Validation_48018739", "origin_question": "Remember, have you seen any kettle(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any kettle(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_48018739/Validation_48018739_48018739_3437.524.png", "Validation_48018739/Validation_48018739_48018739_3439.023.png", "Validation_48018739/Validation_48018739_48018739_3440.522.png", "Validation_48018739/Validation_48018739_48018739_3442.022.png" ], "turn_id": 2, "scan_id": "Validation_48018739", "origin_question": "In which turn did you see both \"the bathtub\" and \"the medium rectangular beige box with brown accents\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the bathtub\" and \"the medium rectangular beige box with brown accents\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_48018739/Validation_48018739_48018739_3443.521.png", "Validation_48018739/Validation_48018739_48018739_3445.021.png", "Validation_48018739/Validation_48018739_48018739_3446.520.png", "Validation_48018739/Validation_48018739_48018739_3448.019.png" ], "turn_id": 3, "scan_id": "Validation_48018739", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"an oval white ceramic sink with the chrome faucet\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"an oval white ceramic sink with the chrome faucet\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_48018739/Validation_48018739_48018739_3449.519.png", "Validation_48018739/Validation_48018739_48018739_3451.018.png", "Validation_48018739/Validation_48018739_48018739_3452.518.png", "Validation_48018739/Validation_48018739_48018739_3454.017.png" ], "turn_id": 4, "scan_id": "Validation_48018739", "origin_question": "When did you first discover \"the bathtub\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the bathtub\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_48018739/Validation_48018739_48018739_3455.516.png", "Validation_48018739/Validation_48018739_48018739_3457.016.png", "Validation_48018739/Validation_48018739_48018739_3458.515.png", "Validation_48018739/Validation_48018739_48018739_3460.015.png" ], "turn_id": 5, "scan_id": "Validation_48018739", "origin_question": "In which turn were you closest to \"the sleek, circular light with the matte finish\"? ", "option": [], "answer": "4", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the sleek, circular light with the matte finish\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47429904/Validation_47429904_47429904_135188.972.png", "Validation_47429904/Validation_47429904_47429904_135190.455.png", "Validation_47429904/Validation_47429904_47429904_135191.954.png", "Validation_47429904/Validation_47429904_47429904_135193.454.png" ], "turn_id": 1, "scan_id": "Validation_47429904", "origin_question": "", "option": [], "answer": "", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47429904/Validation_47429904_47429904_135194.953.png", "Validation_47429904/Validation_47429904_47429904_135196.453.png", "Validation_47429904/Validation_47429904_47429904_135197.952.png", "Validation_47429904/Validation_47429904_47429904_135199.451.png" ], "turn_id": 2, "scan_id": "Validation_47429904", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_47429904/Validation_47429904_47429904_135200.951.png", "Validation_47429904/Validation_47429904_47429904_135202.450.png", "Validation_47429904/Validation_47429904_47429904_135203.949.png", "Validation_47429904/Validation_47429904_47429904_135205.449.png" ], "turn_id": 3, "scan_id": "Validation_47429904", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "1.98", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_47429904/Validation_47429904_47429904_135206.948.png", "Validation_47429904/Validation_47429904_47429904_135208.448.png", "Validation_47429904/Validation_47429904_47429904_135209.947.png", "Validation_47429904/Validation_47429904_47429904_135211.446.png" ], "turn_id": 4, "scan_id": "Validation_47429904", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_47429904/Validation_47429904_47429904_135212.946.png", "Validation_47429904/Validation_47429904_47429904_135214.445.png", "Validation_47429904/Validation_47429904_47429904_135215.961.png", "Validation_47429904/Validation_47429904_47429904_135217.461.png" ], "turn_id": 5, "scan_id": "Validation_47429904", "origin_question": "When did you first discover \"the sleek black rectangular speaker with the mesh front\"? (index of the turn)", "option": [], "answer": "3", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the sleek black rectangular speaker with the mesh front\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_47429904/Validation_47429904_47429904_135218.960.png", "Validation_47429904/Validation_47429904_47429904_135220.459.png", "Validation_47429904/Validation_47429904_47429904_135221.959.png", "Validation_47429904/Validation_47429904_47429904_135223.458.png" ], "turn_id": 6, "scan_id": "Validation_47429904", "origin_question": "When did you first discover \"the window\"? (index of the turn)", "option": [], "answer": "5", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the window\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_47429904/Validation_47429904_47429904_135224.958.png", "Validation_47429904/Validation_47429904_47429904_135226.457.png", "Validation_47429904/Validation_47429904_47429904_135227.956.png", "Validation_47429904/Validation_47429904_47429904_135229.456.png" ], "turn_id": 7, "scan_id": "Validation_47429904", "origin_question": "Please recall \"the modern stainless steel oven with the glass front\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "3.55", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the modern stainless steel oven with the glass front\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_47429904/Validation_47429904_47429904_135230.955.png", "Validation_47429904/Validation_47429904_47429904_135232.455.png", "Validation_47429904/Validation_47429904_47429904_135233.954.png", "Validation_47429904/Validation_47429904_47429904_135235.453.png" ], "turn_id": 8, "scan_id": "Validation_47429904", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the modern stainless steel oven with the glass front\" from your position?", "option": [], "answer": "55.52", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the modern stainless steel oven with the glass front\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47430043/Validation_47430043_47430043_20578.300.png", "Validation_47430043/Validation_47430043_47430043_20579.783.png", "Validation_47430043/Validation_47430043_47430043_20581.282.png", "Validation_47430043/Validation_47430043_47430043_20582.782.png" ], "turn_id": 1, "scan_id": "Validation_47430043", "origin_question": "Remember, have you seen any furniture(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any furniture(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47430043/Validation_47430043_47430043_20584.281.png", "Validation_47430043/Validation_47430043_47430043_20585.797.png", "Validation_47430043/Validation_47430043_47430043_20587.296.png", "Validation_47430043/Validation_47430043_47430043_20588.796.png" ], "turn_id": 2, "scan_id": "Validation_47430043", "origin_question": "Remember, have you seen any treadmill(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any treadmill(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_47430043/Validation_47430043_47430043_20590.295.png", "Validation_47430043/Validation_47430043_47430043_20591.794.png", "Validation_47430043/Validation_47430043_47430043_20593.294.png", "Validation_47430043/Validation_47430043_47430043_20594.793.png" ], "turn_id": 3, "scan_id": "Validation_47430043", "origin_question": "When did you first discover \"the vibrant green plant with elongated oval leaves\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the vibrant green plant with elongated oval leaves\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_47430043/Validation_47430043_47430043_20596.293.png", "Validation_47430043/Validation_47430043_47430043_20597.792.png", "Validation_47430043/Validation_47430043_47430043_20599.291.png", "Validation_47430043/Validation_47430043_47430043_20600.791.png" ], "turn_id": 4, "scan_id": "Validation_47430043", "origin_question": "In which turn did you see both \"the rectangular window with the clear glass pane\" and \"the small rectangular table with the light wood tone\" simultaneously? ", "option": [], "answer": "4", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the rectangular window with the clear glass pane\" and \"the small rectangular table with the light wood tone\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_47430043/Validation_47430043_47430043_20602.290.png", "Validation_47430043/Validation_47430043_47430043_20603.789.png", "Validation_47430043/Validation_47430043_47430043_20605.289.png", "Validation_47430043/Validation_47430043_47430043_20606.788.png" ], "turn_id": 5, "scan_id": "Validation_47430043", "origin_question": "In which turn did you see both \"the round light-colored ceramic plate\" and \"the sleek circular stainless steel sink\" simultaneously? ", "option": [], "answer": "5", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the round light-colored ceramic plate\" and \"the sleek circular stainless steel sink\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_47430043/Validation_47430043_47430043_20608.288.png", "Validation_47430043/Validation_47430043_47430043_20609.787.png", "Validation_47430043/Validation_47430043_47430043_20611.286.png", "Validation_47430043/Validation_47430043_47430043_20612.786.png" ], "turn_id": 6, "scan_id": "Validation_47430043", "origin_question": "When did you first discover \"the medium-sized cylindrical glass flowerpot\"? (index of the turn)", "option": [], "answer": "6", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the medium-sized cylindrical glass flowerpot\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "Validation_47204556/Validation_47204556_47204556_3286.405.png", "Validation_47204556/Validation_47204556_47204556_3288.387.png", "Validation_47204556/Validation_47204556_47204556_3290.403.png", "Validation_47204556/Validation_47204556_47204556_3292.402.png" ], "turn_id": 1, "scan_id": "Validation_47204556", "origin_question": "Remember, how many light(s) have you seen so far? ", "option": [], "answer": "3", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many light(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_47204556/Validation_47204556_47204556_3294.401.png", "Validation_47204556/Validation_47204556_47204556_3296.401.png", "Validation_47204556/Validation_47204556_47204556_3298.400.png", "Validation_47204556/Validation_47204556_47204556_3300.399.png" ], "turn_id": 2, "scan_id": "Validation_47204556", "origin_question": "In which turn did you see both \"the sleek rectangular stainless steel sink\" and \"the natural wood-tone rectangular door with simple surface\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the sleek rectangular stainless steel sink\" and \"the natural wood-tone rectangular door with simple surface\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_47204556/Validation_47204556_47204556_3302.398.png", "Validation_47204556/Validation_47204556_47204556_3304.397.png", "Validation_47204556/Validation_47204556_47204556_3306.397.png", "Validation_47204556/Validation_47204556_47204556_3308.396.png" ], "turn_id": 3, "scan_id": "Validation_47204556", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the sleek rectangular stainless steel sink\", Y: \"the large rectangular window with the modern white frame\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to X and farther away from Y", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the sleek rectangular stainless steel sink\", Y: \"the large rectangular window with the modern white frame\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_47204556/Validation_47204556_47204556_3310.395.png", "Validation_47204556/Validation_47204556_47204556_3312.394.png", "Validation_47204556/Validation_47204556_47204556_3314.393.png", "Validation_47204556/Validation_47204556_47204556_3316.392.png" ], "turn_id": 4, "scan_id": "Validation_47204556", "origin_question": "At the end of which turn were both of \"the sleek, transparent table with smooth edges and slim legs\" and \"the natural wood-tone rectangular door with simple surface\" on your left side?", "option": [], "answer": "4", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the sleek, transparent table with smooth edges and slim legs\" and \"the natural wood-tone rectangular door with simple surface\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_47204556/Validation_47204556_47204556_3318.392.png", "Validation_47204556/Validation_47204556_47204556_3320.391.png", "Validation_47204556/Validation_47204556_47204556_3322.390.png", "Validation_47204556/Validation_47204556_47204556_3324.389.png" ], "turn_id": 5, "scan_id": "Validation_47204556", "origin_question": "At the end of which turn were both of \"the large rectangular window with the modern white frame\" and \"the sleek rectangular monitor with the neutral-colored frame\" on your left side?", "option": [], "answer": "5", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the large rectangular window with the modern white frame\" and \"the sleek rectangular monitor with the neutral-colored frame\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_47204556/Validation_47204556_47204556_3326.388.png", "Validation_47204556/Validation_47204556_47204556_3328.387.png", "Validation_47204556/Validation_47204556_47204556_3330.403.png", "Validation_47204556/Validation_47204556_47204556_3332.403.png" ], "turn_id": 6, "scan_id": "Validation_47204556", "origin_question": "Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "91.9", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47204556/Validation_47204556_47204556_3334.402.png", "Validation_47204556/Validation_47204556_47204556_3336.401.png", "Validation_47204556/Validation_47204556_47204556_3338.400.png", "Validation_47204556/Validation_47204556_47204556_3340.399.png" ], "turn_id": 7, "scan_id": "Validation_47204556", "origin_question": "Which direction is the sleek rectangular stainless steel sink in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the sleek rectangular stainless steel sink in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42445022/Validation_42445022_42445022_49008.276.png", "Validation_42445022/Validation_42445022_42445022_49010.258.png", "Validation_42445022/Validation_42445022_42445022_49012.257.png", "Validation_42445022/Validation_42445022_42445022_49014.257.png" ], "turn_id": 1, "scan_id": "Validation_42445022", "origin_question": "Remember, have you seen any cabinet(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any cabinet(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_42445022/Validation_42445022_42445022_49016.256.png", "Validation_42445022/Validation_42445022_42445022_49018.255.png", "Validation_42445022/Validation_42445022_42445022_49020.271.png", "Validation_42445022/Validation_42445022_42445022_49022.270.png" ], "turn_id": 2, "scan_id": "Validation_42445022", "origin_question": "In which turn were you closest to \"the rectangular pillow with the black flower pattern\"? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the rectangular pillow with the black flower pattern\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "Validation_42445022/Validation_42445022_42445022_49024.269.png", "Validation_42445022/Validation_42445022_42445022_49026.268.png", "Validation_42445022/Validation_42445022_42445022_49028.268.png", "Validation_42445022/Validation_42445022_42445022_49030.267.png" ], "turn_id": 3, "scan_id": "Validation_42445022", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the table\"; \"the sleek, gray upholstered couch with rounded edges\"; \"the rectangular pillow with the black flower pattern\" ", "option": [ "the table", "the sleek, gray upholstered couch with rounded edges", "the rectangular pillow with the black flower pattern" ], "answer": "the table", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the table\"; \"the sleek, gray upholstered couch with rounded edges\"; \"the rectangular pillow with the black flower pattern\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the table', 'the sleek, gray upholstered couch with rounded edges', 'the rectangular pillow with the black flower pattern'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_42445022/Validation_42445022_42445022_49032.266.png", "Validation_42445022/Validation_42445022_42445022_49034.265.png", "Validation_42445022/Validation_42445022_42445022_49036.264.png", "Validation_42445022/Validation_42445022_42445022_49038.264.png" ], "turn_id": 4, "scan_id": "Validation_42445022", "origin_question": "In which turn were you farthest from \"the table\"? ", "option": [], "answer": "1", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the table\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_42445022/Validation_42445022_42445022_49040.263.png", "Validation_42445022/Validation_42445022_42445022_49042.262.png", "Validation_42445022/Validation_42445022_42445022_49044.261.png", "Validation_42445022/Validation_42445022_42445022_49046.260.png" ], "turn_id": 5, "scan_id": "Validation_42445022", "origin_question": "Is the large rectangular window with the white pvc frame to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the large rectangular window with the white pvc frame to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_42445022/Validation_42445022_42445022_49048.260.png", "Validation_42445022/Validation_42445022_42445022_49050.259.png", "Validation_42445022/Validation_42445022_42445022_49052.258.png", "Validation_42445022/Validation_42445022_42445022_49054.257.png" ], "turn_id": 6, "scan_id": "Validation_42445022", "origin_question": "Which direction is the sleek, white plastic chair with no armrests in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the sleek, white plastic chair with no armrests in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_42445022/Validation_42445022_42445022_49056.257.png", "Validation_42445022/Validation_42445022_42445022_49058.256.png", "Validation_42445022/Validation_42445022_42445022_49060.255.png", "Validation_42445022/Validation_42445022_42445022_49062.271.png" ], "turn_id": 7, "scan_id": "Validation_42445022", "origin_question": "Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_42445022/Validation_42445022_42445022_49064.270.png", "Validation_42445022/Validation_42445022_42445022_49066.269.png", "Validation_42445022/Validation_42445022_42445022_49068.268.png", "Validation_42445022/Validation_42445022_42445022_49070.268.png" ], "turn_id": 8, "scan_id": "Validation_42445022", "origin_question": "Which direction is the table in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the table in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_42445022/Validation_42445022_42445022_49072.267.png", "Validation_42445022/Validation_42445022_42445022_49074.266.png", "Validation_42445022/Validation_42445022_42445022_49076.265.png", "Validation_42445022/Validation_42445022_42445022_49078.264.png" ], "turn_id": 9, "scan_id": "Validation_42445022", "origin_question": "Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "85.21", "user_message": "For the 9 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_48018966/Validation_48018966_48018966_556759.073.png", "Validation_48018966/Validation_48018966_48018966_556760.555.png", "Validation_48018966/Validation_48018966_48018966_556762.055.png", "Validation_48018966/Validation_48018966_48018966_556763.554.png" ], "turn_id": 1, "scan_id": "Validation_48018966", "origin_question": "Remember, have you seen any roof(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any roof(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_48018966/Validation_48018966_48018966_556765.053.png", "Validation_48018966/Validation_48018966_48018966_556766.553.png", "Validation_48018966/Validation_48018966_48018966_556768.052.png", "Validation_48018966/Validation_48018966_48018966_556769.551.png" ], "turn_id": 2, "scan_id": "Validation_48018966", "origin_question": "In which turn did you see both \"the door\" and \"the rectangular window in the contemporary white frame\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the door\" and \"the rectangular window in the contemporary white frame\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "Validation_48018966/Validation_48018966_48018966_556771.051.png", "Validation_48018966/Validation_48018966_48018966_556772.550.png", "Validation_48018966/Validation_48018966_48018966_556774.050.png", "Validation_48018966/Validation_48018966_48018966_556775.549.png" ], "turn_id": 3, "scan_id": "Validation_48018966", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the shelf\"; \"the rectangular window in the contemporary white frame\"; \"the sleek rectangular light wood table\" ", "option": [ "the shelf", "the rectangular window in the contemporary white frame", "the sleek rectangular light wood table" ], "answer": "the sleek rectangular light wood table", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the shelf\"; \"the rectangular window in the contemporary white frame\"; \"the sleek rectangular light wood table\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the shelf', 'the rectangular window in the contemporary white frame', 'the sleek rectangular light wood table'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_48018966/Validation_48018966_48018966_556777.048.png", "Validation_48018966/Validation_48018966_48018966_556778.548.png", "Validation_48018966/Validation_48018966_48018966_556780.047.png", "Validation_48018966/Validation_48018966_48018966_556781.547.png" ], "turn_id": 4, "scan_id": "Validation_48018966", "origin_question": "What will be the first-time appearance order of the following categories: blanket, cabinet, chair?", "option": [ "blanket, cabinet, chair", "chair, blanket, cabinet", "cabinet, chair, blanket", "blanket, chair, cabinet" ], "answer": "blanket, cabinet, chair", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: blanket, cabinet, chair? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['blanket, cabinet, chair', 'chair, blanket, cabinet', 'cabinet, chair, blanket', 'blanket, chair, cabinet'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_48018966/Validation_48018966_48018966_556783.062.png", "Validation_48018966/Validation_48018966_48018966_556784.562.png", "Validation_48018966/Validation_48018966_48018966_556786.061.png", "Validation_48018966/Validation_48018966_48018966_556787.561.png" ], "turn_id": 5, "scan_id": "Validation_48018966", "origin_question": "Assuming the direction you are facing at the end of the turn 4 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Left", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 4 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47333431/Validation_47333431_47333431_52247.332.png", "Validation_47333431/Validation_47333431_47333431_52249.314.png", "Validation_47333431/Validation_47333431_47333431_52251.314.png", "Validation_47333431/Validation_47333431_47333431_52253.313.png" ], "turn_id": 1, "scan_id": "Validation_47333431", "origin_question": "Remember, have you seen any cabinet(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any cabinet(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47333431/Validation_47333431_47333431_52255.312.png", "Validation_47333431/Validation_47333431_47333431_52257.311.png", "Validation_47333431/Validation_47333431_47333431_52259.310.png", "Validation_47333431/Validation_47333431_47333431_52261.310.png" ], "turn_id": 2, "scan_id": "Validation_47333431", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the sleek, large flat-screen tv with the black frame\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the sleek, large flat-screen tv with the black frame\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_47333431/Validation_47333431_47333431_52263.309.png", "Validation_47333431/Validation_47333431_47333431_52265.308.png", "Validation_47333431/Validation_47333431_47333431_52267.307.png", "Validation_47333431/Validation_47333431_47333431_52269.306.png" ], "turn_id": 3, "scan_id": "Validation_47333431", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_47333431/Validation_47333431_47333431_52271.305.png", "Validation_47333431/Validation_47333431_47333431_52273.305.png", "Validation_47333431/Validation_47333431_47333431_52275.304.png", "Validation_47333431/Validation_47333431_47333431_52277.303.png" ], "turn_id": 4, "scan_id": "Validation_47333431", "origin_question": "When did you first discover \"the bed\"? (index of the turn)", "option": [], "answer": "3", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the bed\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_47333431/Validation_47333431_47333431_52279.302.png", "Validation_47333431/Validation_47333431_47333431_52281.301.png", "Validation_47333431/Validation_47333431_47333431_52283.300.png", "Validation_47333431/Validation_47333431_47333431_52285.316.png" ], "turn_id": 5, "scan_id": "Validation_47333431", "origin_question": "When did you first discover \"the sleek, large flat-screen tv with the black frame\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the sleek, large flat-screen tv with the black frame\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_47333431/Validation_47333431_47333431_52287.315.png", "Validation_47333431/Validation_47333431_47333431_52289.315.png", "Validation_47333431/Validation_47333431_47333431_52291.314.png", "Validation_47333431/Validation_47333431_47333431_52293.313.png" ], "turn_id": 6, "scan_id": "Validation_47333431", "origin_question": "Please recall \"the sleek, large flat-screen tv with the black frame\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.03", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the sleek, large flat-screen tv with the black frame\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_45260898/Validation_45260898_45260898_765.008.png", "Validation_45260898/Validation_45260898_45260898_766.491.png", "Validation_45260898/Validation_45260898_45260898_767.990.png", "Validation_45260898/Validation_45260898_45260898_769.490.png" ], "turn_id": 1, "scan_id": "Validation_45260898", "origin_question": "Remember, have you seen any sink(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any sink(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_45260898/Validation_45260898_45260898_770.989.png", "Validation_45260898/Validation_45260898_45260898_772.489.png", "Validation_45260898/Validation_45260898_45260898_773.988.png", "Validation_45260898/Validation_45260898_45260898_775.487.png" ], "turn_id": 2, "scan_id": "Validation_45260898", "origin_question": "In which turn were you closest to \"the cylindrical silver bin with the sleek design\"? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the cylindrical silver bin with the sleek design\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_45260898/Validation_45260898_45260898_776.987.png", "Validation_45260898/Validation_45260898_45260898_778.486.png", "Validation_45260898/Validation_45260898_45260898_779.986.png", "Validation_45260898/Validation_45260898_45260898_781.485.png" ], "turn_id": 3, "scan_id": "Validation_45260898", "origin_question": "Which direction is the cylindrical silver bin with the sleek design in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the cylindrical silver bin with the sleek design in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_45260898/Validation_45260898_45260898_782.984.png", "Validation_45260898/Validation_45260898_45260898_784.484.png", "Validation_45260898/Validation_45260898_45260898_785.983.png", "Validation_45260898/Validation_45260898_45260898_787.483.png" ], "turn_id": 4, "scan_id": "Validation_45260898", "origin_question": "Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "79.54", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_45260898/Validation_45260898_45260898_788.982.png", "Validation_45260898/Validation_45260898_45260898_790.481.png", "Validation_45260898/Validation_45260898_45260898_791.981.png", "Validation_45260898/Validation_45260898_45260898_793.480.png" ], "turn_id": 5, "scan_id": "Validation_45260898", "origin_question": "Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance forward or backward from that position?", "option": [ "Forward", "Backward" ], "answer": "Forward", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance forward or backward from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Forward', 'Backward'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "Validation_48458424/Validation_48458424_48458424_115103.660.png", "Validation_48458424/Validation_48458424_48458424_115105.659.png", "Validation_48458424/Validation_48458424_48458424_115107.658.png", "Validation_48458424/Validation_48458424_48458424_115109.657.png" ], "turn_id": 1, "scan_id": "Validation_48458424", "origin_question": "Remember, how many door(s) have you seen so far? ", "option": [], "answer": "3", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many door(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_48458424/Validation_48458424_48458424_115111.657.png", "Validation_48458424/Validation_48458424_48458424_115113.656.png", "Validation_48458424/Validation_48458424_48458424_115115.655.png", "Validation_48458424/Validation_48458424_48458424_115117.654.png" ], "turn_id": 2, "scan_id": "Validation_48458424", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_48458424/Validation_48458424_48458424_115119.653.png", "Validation_48458424/Validation_48458424_48458424_115121.652.png", "Validation_48458424/Validation_48458424_48458424_115123.652.png", "Validation_48458424/Validation_48458424_48458424_115125.651.png" ], "turn_id": 3, "scan_id": "Validation_48458424", "origin_question": "", "option": [], "answer": "", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_48458424/Validation_48458424_48458424_115127.650.png", "Validation_48458424/Validation_48458424_48458424_115129.649.png", "Validation_48458424/Validation_48458424_48458424_115131.648.png", "Validation_48458424/Validation_48458424_48458424_115133.648.png" ], "turn_id": 4, "scan_id": "Validation_48458424", "origin_question": "Please recall \"an oversized rectangular window with the white frame\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.69", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"an oversized rectangular window with the white frame\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_48458424/Validation_48458424_48458424_115135.647.png", "Validation_48458424/Validation_48458424_48458424_115137.646.png", "Validation_48458424/Validation_48458424_48458424_115139.662.png", "Validation_48458424/Validation_48458424_48458424_115141.661.png" ], "turn_id": 5, "scan_id": "Validation_48458424", "origin_question": "Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Left", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_48458424/Validation_48458424_48458424_115143.660.png", "Validation_48458424/Validation_48458424_48458424_115145.659.png", "Validation_48458424/Validation_48458424_48458424_115147.658.png", "Validation_48458424/Validation_48458424_48458424_115149.658.png" ], "turn_id": 6, "scan_id": "Validation_48458424", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the vibrant yellow fabric stool with metal legs\" from your position?", "option": [], "answer": "120.45", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the vibrant yellow fabric stool with metal legs\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_48458424/Validation_48458424_48458424_115151.657.png", "Validation_48458424/Validation_48458424_48458424_115153.656.png", "Validation_48458424/Validation_48458424_48458424_115155.655.png", "Validation_48458424/Validation_48458424_48458424_115157.654.png" ], "turn_id": 7, "scan_id": "Validation_48458424", "origin_question": "Which direction is the glossy paper wall calendar with bold black numbers in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the glossy paper wall calendar with bold black numbers in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_48458424/Validation_48458424_48458424_115159.653.png", "Validation_48458424/Validation_48458424_48458424_115161.653.png", "Validation_48458424/Validation_48458424_48458424_115163.652.png", "Validation_48458424/Validation_48458424_48458424_115165.651.png" ], "turn_id": 8, "scan_id": "Validation_48458424", "origin_question": "Which direction is the sleek, black rectangular monitor made of plastic and glass in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the sleek, black rectangular monitor made of plastic and glass in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_41254405/Validation_41254405_41254405_95575.915.png", "Validation_41254405/Validation_41254405_41254405_95577.381.png", "Validation_41254405/Validation_41254405_41254405_95578.881.png", "Validation_41254405/Validation_41254405_41254405_95580.380.png" ], "turn_id": 1, "scan_id": "Validation_41254405", "origin_question": "Remember, have you seen any computer(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any computer(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 1 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_41125763/Validation_41125763_41125763_5429.015.png", "Validation_41125763/Validation_41125763_41125763_5430.998.png", "Validation_41125763/Validation_41125763_41125763_5432.997.png", "Validation_41125763/Validation_41125763_41125763_5434.996.png" ], "turn_id": 1, "scan_id": "Validation_41125763", "origin_question": "Remember, have you seen any door(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any door(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_41125763/Validation_41125763_41125763_5436.995.png", "Validation_41125763/Validation_41125763_41125763_5439.011.png", "Validation_41125763/Validation_41125763_41125763_5441.010.png", "Validation_41125763/Validation_41125763_41125763_5443.010.png" ], "turn_id": 2, "scan_id": "Validation_41125763", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_41125763/Validation_41125763_41125763_5445.009.png", "Validation_41125763/Validation_41125763_41125763_5447.008.png", "Validation_41125763/Validation_41125763_41125763_5449.007.png", "Validation_41125763/Validation_41125763_41125763_5451.006.png" ], "turn_id": 3, "scan_id": "Validation_41125763", "origin_question": "What will be the first-time appearance order of the following categories: cabinet, door, mirror?", "option": [ "cabinet, mirror, door", "door, mirror, cabinet", "door, cabinet, mirror", "cabinet, door, mirror" ], "answer": "door, cabinet, mirror", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: cabinet, door, mirror? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['cabinet, mirror, door', 'door, mirror, cabinet', 'door, cabinet, mirror', 'cabinet, door, mirror'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_41125763/Validation_41125763_41125763_5453.006.png", "Validation_41125763/Validation_41125763_41125763_5455.005.png", "Validation_41125763/Validation_41125763_41125763_5457.004.png", "Validation_41125763/Validation_41125763_41125763_5459.003.png" ], "turn_id": 4, "scan_id": "Validation_41125763", "origin_question": "Is the sleek, light-colored modern door to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the sleek, light-colored modern door to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_41125763/Validation_41125763_41125763_5461.002.png", "Validation_41125763/Validation_41125763_41125763_5463.001.png", "Validation_41125763/Validation_41125763_41125763_5465.001.png", "Validation_41125763/Validation_41125763_41125763_5467.000.png" ], "turn_id": 5, "scan_id": "Validation_41125763", "origin_question": "Which direction is the sleek white ceramic toilet with the contoured seat in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the sleek white ceramic toilet with the contoured seat in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_41125763/Validation_41125763_41125763_5468.999.png", "Validation_41125763/Validation_41125763_41125763_5470.998.png", "Validation_41125763/Validation_41125763_41125763_5472.997.png", "Validation_41125763/Validation_41125763_41125763_5474.997.png" ], "turn_id": 6, "scan_id": "Validation_41125763", "origin_question": "Which direction is the sink in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the sink in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_41125763/Validation_41125763_41125763_5476.996.png", "Validation_41125763/Validation_41125763_41125763_5478.995.png", "Validation_41125763/Validation_41125763_41125763_5481.011.png", "Validation_41125763/Validation_41125763_41125763_5483.010.png" ], "turn_id": 7, "scan_id": "Validation_41125763", "origin_question": "Which direction is the rectangular white plastic bin with the sleek design in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the rectangular white plastic bin with the sleek design in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47430003/Validation_47430003_47430003_17587.982.png", "Validation_47430003/Validation_47430003_47430003_17589.464.png", "Validation_47430003/Validation_47430003_47430003_17590.964.png", "Validation_47430003/Validation_47430003_47430003_17592.463.png" ], "turn_id": 1, "scan_id": "Validation_47430003", "origin_question": "Remember, have you seen any kitchenware(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any kitchenware(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 2 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_47430003/Validation_47430003_47430003_17593.962.png", "Validation_47430003/Validation_47430003_47430003_17595.462.png", "Validation_47430003/Validation_47430003_47430003_17596.961.png", "Validation_47430003/Validation_47430003_47430003_17598.461.png" ], "turn_id": 2, "scan_id": "Validation_47430003", "origin_question": "In which turn did you see both \"the rectangular, white ceramic sink with rounded edges\" and \"the sleek, modern rectangular white bathtub\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the rectangular, white ceramic sink with rounded edges\" and \"the sleek, modern rectangular white bathtub\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 2 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42444954/Validation_42444954_42444954_224879.784.png", "Validation_42444954/Validation_42444954_42444954_224881.767.png", "Validation_42444954/Validation_42444954_42444954_224883.766.png", "Validation_42444954/Validation_42444954_42444954_224885.765.png" ], "turn_id": 1, "scan_id": "Validation_42444954", "origin_question": "Remember, have you seen any bathtub(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any bathtub(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_42444954/Validation_42444954_42444954_224887.764.png", "Validation_42444954/Validation_42444954_42444954_224889.764.png", "Validation_42444954/Validation_42444954_42444954_224891.763.png", "Validation_42444954/Validation_42444954_42444954_224893.762.png" ], "turn_id": 2, "scan_id": "Validation_42444954", "origin_question": "In which turn were you closest to \"the glossy white ceramic toilet with an elongated oval seat\"? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the glossy white ceramic toilet with an elongated oval seat\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_42444954/Validation_42444954_42444954_224895.761.png", "Validation_42444954/Validation_42444954_42444954_224897.760.png", "Validation_42444954/Validation_42444954_42444954_224899.759.png", "Validation_42444954/Validation_42444954_42444954_224901.759.png" ], "turn_id": 3, "scan_id": "Validation_42444954", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the modern white ceramic sink with the round basin\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the modern white ceramic sink with the round basin\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_42444954/Validation_42444954_42444954_224903.758.png", "Validation_42444954/Validation_42444954_42444954_224905.757.png", "Validation_42444954/Validation_42444954_42444954_224907.756.png", "Validation_42444954/Validation_42444954_42444954_224909.755.png" ], "turn_id": 4, "scan_id": "Validation_42444954", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_42444954/Validation_42444954_42444954_224911.755.png", "Validation_42444954/Validation_42444954_42444954_224913.770.png", "Validation_42444954/Validation_42444954_42444954_224915.770.png", "Validation_42444954/Validation_42444954_42444954_224917.769.png" ], "turn_id": 5, "scan_id": "Validation_42444954", "origin_question": "Is the modern white ceramic sink with the round basin to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the modern white ceramic sink with the round basin to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_42444954/Validation_42444954_42444954_224919.768.png", "Validation_42444954/Validation_42444954_42444954_224921.767.png", "Validation_42444954/Validation_42444954_42444954_224923.766.png", "Validation_42444954/Validation_42444954_42444954_224925.766.png" ], "turn_id": 6, "scan_id": "Validation_42444954", "origin_question": "Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "139.26", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_42444954/Validation_42444954_42444954_224927.765.png", "Validation_42444954/Validation_42444954_42444954_224929.764.png", "Validation_42444954/Validation_42444954_42444954_224931.763.png", "Validation_42444954/Validation_42444954_42444954_224933.762.png" ], "turn_id": 7, "scan_id": "Validation_42444954", "origin_question": "", "option": [], "answer": "", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_42444954/Validation_42444954_42444954_224935.761.png", "Validation_42444954/Validation_42444954_42444954_224937.761.png", "Validation_42444954/Validation_42444954_42444954_224939.760.png", "Validation_42444954/Validation_42444954_42444954_224941.759.png" ], "turn_id": 8, "scan_id": "Validation_42444954", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the glossy white ceramic toilet with an elongated oval seat\" from your position?", "option": [], "answer": "165.09", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the glossy white ceramic toilet with an elongated oval seat\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_42444954/Validation_42444954_42444954_224943.758.png", "Validation_42444954/Validation_42444954_42444954_224945.757.png", "Validation_42444954/Validation_42444954_42444954_224947.756.png", "Validation_42444954/Validation_42444954_42444954_224949.756.png" ], "turn_id": 9, "scan_id": "Validation_42444954", "origin_question": "Which direction is the classic rectangular white bathtub with the sleek design in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 9 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the classic rectangular white bathtub with the sleek design in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "Validation_42446546/Validation_42446546_42446546_201330.169.png", "Validation_42446546/Validation_42446546_42446546_201332.152.png", "Validation_42446546/Validation_42446546_42446546_201334.151.png", "Validation_42446546/Validation_42446546_42446546_201336.150.png", "Validation_42446546/Validation_42446546_42446546_201338.149.png" ], "turn_id": 1, "scan_id": "Validation_42446546", "origin_question": "Remember, how many sink(s) have you seen so far? ", "option": [], "answer": "2", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many sink(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_42446546/Validation_42446546_42446546_201340.148.png", "Validation_42446546/Validation_42446546_42446546_201342.147.png", "Validation_42446546/Validation_42446546_42446546_201344.147.png", "Validation_42446546/Validation_42446546_42446546_201346.146.png", "Validation_42446546/Validation_42446546_42446546_201348.145.png" ], "turn_id": 2, "scan_id": "Validation_42446546", "origin_question": "Is the round dark polished table with the floral tablecloth to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the round dark polished table with the floral tablecloth to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_42446546/Validation_42446546_42446546_201350.144.png", "Validation_42446546/Validation_42446546_42446546_201352.143.png", "Validation_42446546/Validation_42446546_42446546_201354.143.png", "Validation_42446546/Validation_42446546_42446546_201356.142.png", "Validation_42446546/Validation_42446546_42446546_201358.141.png" ], "turn_id": 3, "scan_id": "Validation_42446546", "origin_question": "Which two objects are on the same side of you now? \"the round dark polished table with the floral tablecloth\", \"the polished stainless steel pyramid-shaped range hood\", \"the rectangular stainless steel oven with the black finish\". ", "option": [ "the round dark polished table with the floral tablecloth, the polished stainless steel pyramid-shaped range hood", "the round dark polished table with the floral tablecloth, the rectangular stainless steel oven with the black finish", "the polished stainless steel pyramid-shaped range hood, the rectangular stainless steel oven with the black finish" ], "answer": "the polished stainless steel pyramid-shaped range hood, the rectangular stainless steel oven with the black finish", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the round dark polished table with the floral tablecloth\", \"the polished stainless steel pyramid-shaped range hood\", \"the rectangular stainless steel oven with the black finish\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the round dark polished table with the floral tablecloth, the polished stainless steel pyramid-shaped range hood', 'the round dark polished table with the floral tablecloth, the rectangular stainless steel oven with the black finish', 'the polished stainless steel pyramid-shaped range hood, the rectangular stainless steel oven with the black finish'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "Validation_42446546/Validation_42446546_42446546_201360.140.png", "Validation_42446546/Validation_42446546_42446546_201362.139.png", "Validation_42446546/Validation_42446546_42446546_201364.139.png", "Validation_42446546/Validation_42446546_42446546_201366.138.png", "Validation_42446546/Validation_42446546_42446546_201368.137.png" ], "turn_id": 4, "scan_id": "Validation_42446546", "origin_question": "Among these three objects(\"the tall stainless steel refrigerator with adjustable shelving\",\"the rectangular stainless steel oven with the black finish\",\"the sleek stainless steel microwave with the glossy black window\" ), which one is the closest to you?", "option": [ "the tall stainless steel refrigerator with adjustable shelving", "the rectangular stainless steel oven with the black finish", "the sleek stainless steel microwave with the glossy black window" ], "answer": "the tall stainless steel refrigerator with adjustable shelving", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the tall stainless steel refrigerator with adjustable shelving\",\"the rectangular stainless steel oven with the black finish\",\"the sleek stainless steel microwave with the glossy black window\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the tall stainless steel refrigerator with adjustable shelving', 'the rectangular stainless steel oven with the black finish', 'the sleek stainless steel microwave with the glossy black window'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_42446546/Validation_42446546_42446546_201370.153.png", "Validation_42446546/Validation_42446546_42446546_201372.152.png", "Validation_42446546/Validation_42446546_42446546_201374.151.png", "Validation_42446546/Validation_42446546_42446546_201376.150.png", "Validation_42446546/Validation_42446546_42446546_201378.149.png" ], "turn_id": 5, "scan_id": "Validation_42446546", "origin_question": "Which two objects are on the same side of you now? \"the polished stainless steel pyramid-shaped range hood\", \"the sleek square stool with the light wood frame\", \"the vibrant green plant\". ", "option": [ "the polished stainless steel pyramid-shaped range hood, the sleek square stool with the light wood frame", "the polished stainless steel pyramid-shaped range hood, the vibrant green plant", "the sleek square stool with the light wood frame, the vibrant green plant" ], "answer": "the sleek square stool with the light wood frame, the vibrant green plant", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the polished stainless steel pyramid-shaped range hood\", \"the sleek square stool with the light wood frame\", \"the vibrant green plant\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the polished stainless steel pyramid-shaped range hood, the sleek square stool with the light wood frame', 'the polished stainless steel pyramid-shaped range hood, the vibrant green plant', 'the sleek square stool with the light wood frame, the vibrant green plant'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_42446546/Validation_42446546_42446546_201380.149.png", "Validation_42446546/Validation_42446546_42446546_201382.148.png", "Validation_42446546/Validation_42446546_42446546_201384.147.png", "Validation_42446546/Validation_42446546_42446546_201386.146.png", "Validation_42446546/Validation_42446546_42446546_201388.145.png" ], "turn_id": 6, "scan_id": "Validation_42446546", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_42446546/Validation_42446546_42446546_201390.144.png", "Validation_42446546/Validation_42446546_42446546_201392.144.png", "Validation_42446546/Validation_42446546_42446546_201394.143.png", "Validation_42446546/Validation_42446546_42446546_201396.142.png", "Validation_42446546/Validation_42446546_42446546_201398.141.png" ], "turn_id": 7, "scan_id": "Validation_42446546", "origin_question": "Which two objects are on the same side of you now? \"the transparent rectangular wire mesh basket with the metallic color\", \"the polished stainless steel pyramid-shaped range hood\", \"the vibrant green plant\". ", "option": [ "the transparent rectangular wire mesh basket with the metallic color, the polished stainless steel pyramid-shaped range hood", "the transparent rectangular wire mesh basket with the metallic color, the vibrant green plant", "the polished stainless steel pyramid-shaped range hood, the vibrant green plant" ], "answer": "the transparent rectangular wire mesh basket with the metallic color, the polished stainless steel pyramid-shaped range hood", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the transparent rectangular wire mesh basket with the metallic color\", \"the polished stainless steel pyramid-shaped range hood\", \"the vibrant green plant\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the transparent rectangular wire mesh basket with the metallic color, the polished stainless steel pyramid-shaped range hood', 'the transparent rectangular wire mesh basket with the metallic color, the vibrant green plant', 'the polished stainless steel pyramid-shaped range hood, the vibrant green plant'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_42446546/Validation_42446546_42446546_201400.140.png", "Validation_42446546/Validation_42446546_42446546_201402.139.png", "Validation_42446546/Validation_42446546_42446546_201404.139.png", "Validation_42446546/Validation_42446546_42446546_201406.138.png", "Validation_42446546/Validation_42446546_42446546_201408.137.png" ], "turn_id": 8, "scan_id": "Validation_42446546", "origin_question": "How far is your current position from where you were at the end of the turn 5? (in meters)", "option": [], "answer": "2.79", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 5? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_42446546/Validation_42446546_42446546_201410.136.png", "Validation_42446546/Validation_42446546_42446546_201412.152.png", "Validation_42446546/Validation_42446546_42446546_201414.151.png", "Validation_42446546/Validation_42446546_42446546_201416.150.png", "Validation_42446546/Validation_42446546_42446546_201418.150.png" ], "turn_id": 9, "scan_id": "Validation_42446546", "origin_question": "At the end of which turn were both of \"the tall stainless steel refrigerator with adjustable shelving\" and \"the sleek black rectangular monitor with the glossy finish\" on your left side?", "option": [], "answer": "8", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the tall stainless steel refrigerator with adjustable shelving\" and \"the sleek black rectangular monitor with the glossy finish\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_42446546/Validation_42446546_42446546_201420.149.png", "Validation_42446546/Validation_42446546_42446546_201422.148.png", "Validation_42446546/Validation_42446546_42446546_201424.147.png", "Validation_42446546/Validation_42446546_42446546_201426.146.png", "Validation_42446546/Validation_42446546_42446546_201428.146.png" ], "turn_id": 10, "scan_id": "Validation_42446546", "origin_question": "Which direction is the sleek, dark rectangular washing machine made of smooth material in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the sleek, dark rectangular washing machine made of smooth material in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_48458654/Validation_48458654_48458654_7548.203.png", "Validation_48458654/Validation_48458654_48458654_7549.686.png", "Validation_48458654/Validation_48458654_48458654_7551.185.png", "Validation_48458654/Validation_48458654_48458654_7552.685.png" ], "turn_id": 1, "scan_id": "Validation_48458654", "origin_question": "Remember, have you seen any pool(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any pool(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_48458654/Validation_48458654_48458654_7554.184.png", "Validation_48458654/Validation_48458654_48458654_7555.683.png", "Validation_48458654/Validation_48458654_48458654_7557.183.png", "Validation_48458654/Validation_48458654_48458654_7558.682.png" ], "turn_id": 2, "scan_id": "Validation_48458654", "origin_question": "At the end of which turn were both of \"the glossy wall-mounted shelf with the rectangular shape\" and \"the sink\" on your left side?", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the glossy wall-mounted shelf with the rectangular shape\" and \"the sink\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_48458654/Validation_48458654_48458654_7560.198.png", "Validation_48458654/Validation_48458654_48458654_7561.697.png", "Validation_48458654/Validation_48458654_48458654_7563.197.png", "Validation_48458654/Validation_48458654_48458654_7564.696.png" ], "turn_id": 3, "scan_id": "Validation_48458654", "origin_question": "In which turn did you see both \"the rectangular, light tan door\" and \"the soft light gray rectangular blanket\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the rectangular, light tan door\" and \"the soft light gray rectangular blanket\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_48458654/Validation_48458654_48458654_7566.195.png", "Validation_48458654/Validation_48458654_48458654_7567.695.png", "Validation_48458654/Validation_48458654_48458654_7569.194.png", "Validation_48458654/Validation_48458654_48458654_7570.694.png" ], "turn_id": 4, "scan_id": "Validation_48458654", "origin_question": "Which direction is the rectangular soft pink chenille carpet in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the rectangular soft pink chenille carpet in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_45261615/Validation_45261615_45261615_54998.696.png", "Validation_45261615/Validation_45261615_45261615_55000.678.png", "Validation_45261615/Validation_45261615_45261615_55002.694.png", "Validation_45261615/Validation_45261615_45261615_55004.693.png" ], "turn_id": 1, "scan_id": "Validation_45261615", "origin_question": "Remember, have you seen any monitor(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any monitor(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_45261615/Validation_45261615_45261615_55006.692.png", "Validation_45261615/Validation_45261615_45261615_55008.692.png", "Validation_45261615/Validation_45261615_45261615_55010.691.png", "Validation_45261615/Validation_45261615_45261615_55012.690.png" ], "turn_id": 2, "scan_id": "Validation_45261615", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the light-colored wooden door with refined panels\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the light-colored wooden door with refined panels\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_45261615/Validation_45261615_45261615_55014.689.png", "Validation_45261615/Validation_45261615_45261615_55016.688.png", "Validation_45261615/Validation_45261615_45261615_55018.688.png", "Validation_45261615/Validation_45261615_45261615_55020.687.png" ], "turn_id": 3, "scan_id": "Validation_45261615", "origin_question": "What will be the first-time appearance order of the following categories: curtain, picture, plant?", "option": [ "plant, picture, curtain", "curtain, picture, plant", "picture, plant, curtain", "plant, curtain, picture" ], "answer": "plant, picture, curtain", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: curtain, picture, plant? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['plant, picture, curtain', 'curtain, picture, plant', 'picture, plant, curtain', 'plant, curtain, picture'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_45261615/Validation_45261615_45261615_55022.686.png", "Validation_45261615/Validation_45261615_45261615_55024.685.png", "Validation_45261615/Validation_45261615_45261615_55026.685.png", "Validation_45261615/Validation_45261615_45261615_55028.684.png" ], "turn_id": 4, "scan_id": "Validation_45261615", "origin_question": "In which turn did you see both \"the light-colored wooden door with refined panels\" and \"the sleek rectangular white steel radiator\" simultaneously? ", "option": [], "answer": "3", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the light-colored wooden door with refined panels\" and \"the sleek rectangular white steel radiator\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_45261615/Validation_45261615_45261615_55030.683.png", "Validation_45261615/Validation_45261615_45261615_55032.682.png", "Validation_45261615/Validation_45261615_45261615_55034.681.png", "Validation_45261615/Validation_45261615_45261615_55036.681.png" ], "turn_id": 5, "scan_id": "Validation_45261615", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular picture with an ornate metallic frame\", Y: \"the sleek rectangular white steel radiator\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular picture with an ornate metallic frame\", Y: \"the sleek rectangular white steel radiator\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_45261615/Validation_45261615_45261615_55038.680.png", "Validation_45261615/Validation_45261615_45261615_55040.679.png", "Validation_45261615/Validation_45261615_45261615_55042.678.png", "Validation_45261615/Validation_45261615_45261615_55044.694.png" ], "turn_id": 6, "scan_id": "Validation_45261615", "origin_question": "Which two objects are on the same side of you now? \"the dark vase with vibrant yellow plant\", \"the light-colored wooden door with refined panels\", \"the sleek rectangular white steel radiator\". ", "option": [ "the dark vase with vibrant yellow plant, the light-colored wooden door with refined panels", "the dark vase with vibrant yellow plant, the sleek rectangular white steel radiator", "the light-colored wooden door with refined panels, the sleek rectangular white steel radiator" ], "answer": "the dark vase with vibrant yellow plant, the light-colored wooden door with refined panels", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the dark vase with vibrant yellow plant\", \"the light-colored wooden door with refined panels\", \"the sleek rectangular white steel radiator\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the dark vase with vibrant yellow plant, the light-colored wooden door with refined panels', 'the dark vase with vibrant yellow plant, the sleek rectangular white steel radiator', 'the light-colored wooden door with refined panels, the sleek rectangular white steel radiator'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_45261615/Validation_45261615_45261615_55046.693.png", "Validation_45261615/Validation_45261615_45261615_55048.692.png", "Validation_45261615/Validation_45261615_45261615_55050.692.png", "Validation_45261615/Validation_45261615_45261615_55052.691.png" ], "turn_id": 7, "scan_id": "Validation_45261615", "origin_question": "Please recall \"the rectangular picture with an ornate metallic frame\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.55", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the rectangular picture with an ornate metallic frame\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47334115/Validation_47334115_47334115_94120.371.png", "Validation_47334115/Validation_47334115_47334115_94122.353.png", "Validation_47334115/Validation_47334115_47334115_94124.352.png", "Validation_47334115/Validation_47334115_47334115_94126.352.png", "Validation_47334115/Validation_47334115_47334115_94128.351.png" ], "turn_id": 1, "scan_id": "Validation_47334115", "origin_question": "Remember, have you seen any coffee maker(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any coffee maker(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_47334115/Validation_47334115_47334115_94130.350.png", "Validation_47334115/Validation_47334115_47334115_94132.349.png", "Validation_47334115/Validation_47334115_47334115_94134.348.png", "Validation_47334115/Validation_47334115_47334115_94136.348.png", "Validation_47334115/Validation_47334115_47334115_94138.347.png" ], "turn_id": 2, "scan_id": "Validation_47334115", "origin_question": "When did you first discover \"the sleek, off-white modern cabinet with multiple drawers\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the sleek, off-white modern cabinet with multiple drawers\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_47334115/Validation_47334115_47334115_94140.346.png", "Validation_47334115/Validation_47334115_47334115_94142.345.png", "Validation_47334115/Validation_47334115_47334115_94144.344.png", "Validation_47334115/Validation_47334115_47334115_94146.343.png", "Validation_47334115/Validation_47334115_47334115_94148.343.png" ], "turn_id": 3, "scan_id": "Validation_47334115", "origin_question": "Which two objects are on the same side of you now? \"the large white rectangular glossy frame\", \"the modern, warm brown rectangular door\", \"the large, white-framed rectangular sliding window\". ", "option": [ "the large white rectangular glossy frame, the modern, warm brown rectangular door", "the large white rectangular glossy frame, the large, white-framed rectangular sliding window", "the modern, warm brown rectangular door, the large, white-framed rectangular sliding window" ], "answer": "the large white rectangular glossy frame, the modern, warm brown rectangular door", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the large white rectangular glossy frame\", \"the modern, warm brown rectangular door\", \"the large, white-framed rectangular sliding window\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the large white rectangular glossy frame, the modern, warm brown rectangular door', 'the large white rectangular glossy frame, the large, white-framed rectangular sliding window', 'the modern, warm brown rectangular door, the large, white-framed rectangular sliding window'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_47334115/Validation_47334115_47334115_94150.342.png", "Validation_47334115/Validation_47334115_47334115_94152.358.png", "Validation_47334115/Validation_47334115_47334115_94154.357.png", "Validation_47334115/Validation_47334115_47334115_94156.356.png", "Validation_47334115/Validation_47334115_47334115_94158.355.png" ], "turn_id": 4, "scan_id": "Validation_47334115", "origin_question": "When did you first discover \"the large white rectangular glossy frame\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the large white rectangular glossy frame\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47334115/Validation_47334115_47334115_94160.354.png", "Validation_47334115/Validation_47334115_47334115_94162.354.png", "Validation_47334115/Validation_47334115_47334115_94164.353.png", "Validation_47334115/Validation_47334115_47334115_94166.352.png", "Validation_47334115/Validation_47334115_47334115_94168.351.png" ], "turn_id": 5, "scan_id": "Validation_47334115", "origin_question": "", "option": [], "answer": "", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_47334115/Validation_47334115_47334115_94170.350.png", "Validation_47334115/Validation_47334115_47334115_94172.350.png", "Validation_47334115/Validation_47334115_47334115_94174.349.png", "Validation_47334115/Validation_47334115_47334115_94176.348.png", "Validation_47334115/Validation_47334115_47334115_94178.347.png" ], "turn_id": 6, "scan_id": "Validation_47334115", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_47334115/Validation_47334115_47334115_94180.346.png", "Validation_47334115/Validation_47334115_47334115_94182.346.png", "Validation_47334115/Validation_47334115_47334115_94184.345.png", "Validation_47334115/Validation_47334115_47334115_94186.344.png", "Validation_47334115/Validation_47334115_47334115_94188.343.png" ], "turn_id": 7, "scan_id": "Validation_47334115", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_47334115/Validation_47334115_47334115_94190.342.png", "Validation_47334115/Validation_47334115_47334115_94192.341.png", "Validation_47334115/Validation_47334115_47334115_94194.357.png", "Validation_47334115/Validation_47334115_47334115_94196.357.png", "Validation_47334115/Validation_47334115_47334115_94198.356.png" ], "turn_id": 8, "scan_id": "Validation_47334115", "origin_question": "Please recall \"the sleek white upholstered chair with dark legs\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.69", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the sleek white upholstered chair with dark legs\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_45261133/Validation_45261133_45261133_4007.282.png", "Validation_45261133/Validation_45261133_45261133_4009.264.png", "Validation_45261133/Validation_45261133_45261133_4011.263.png", "Validation_45261133/Validation_45261133_45261133_4013.262.png", "Validation_45261133/Validation_45261133_45261133_4015.262.png" ], "turn_id": 1, "scan_id": "Validation_45261133", "origin_question": "Remember, have you seen any monitor(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any monitor(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_45261133/Validation_45261133_45261133_4017.261.png", "Validation_45261133/Validation_45261133_45261133_4019.260.png", "Validation_45261133/Validation_45261133_45261133_4021.259.png", "Validation_45261133/Validation_45261133_45261133_4023.258.png", "Validation_45261133/Validation_45261133_45261133_4025.274.png" ], "turn_id": 2, "scan_id": "Validation_45261133", "origin_question": "Remember, have you seen any couch(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any couch(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_45261133/Validation_45261133_45261133_4027.273.png", "Validation_45261133/Validation_45261133_45261133_4029.273.png", "Validation_45261133/Validation_45261133_45261133_4031.272.png", "Validation_45261133/Validation_45261133_45261133_4033.271.png", "Validation_45261133/Validation_45261133_45261133_4035.270.png" ], "turn_id": 3, "scan_id": "Validation_45261133", "origin_question": "In which turn did you see both \"the modern, sleek washing machine with the white exterior\" and \"the sleek black modern monitor with plastic and metal\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the modern, sleek washing machine with the white exterior\" and \"the sleek black modern monitor with plastic and metal\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_45261133/Validation_45261133_45261133_4037.269.png", "Validation_45261133/Validation_45261133_45261133_4039.268.png", "Validation_45261133/Validation_45261133_45261133_4041.268.png", "Validation_45261133/Validation_45261133_45261133_4043.267.png", "Validation_45261133/Validation_45261133_45261133_4045.266.png" ], "turn_id": 4, "scan_id": "Validation_45261133", "origin_question": "By the end of the 3 turn, were you getting closer to or farther away from these two objects: X: \"the modern, sleek washing machine with the white exterior\", Y: \"the sleek black modern monitor with plastic and metal\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to Y and farther away from X", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 3 turn, were you getting closer to or farther away from these two objects: X: \"the modern, sleek washing machine with the white exterior\", Y: \"the sleek black modern monitor with plastic and metal\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_45261133/Validation_45261133_45261133_4047.265.png", "Validation_45261133/Validation_45261133_45261133_4049.264.png", "Validation_45261133/Validation_45261133_45261133_4051.263.png", "Validation_45261133/Validation_45261133_45261133_4053.263.png", "Validation_45261133/Validation_45261133_45261133_4055.262.png" ], "turn_id": 5, "scan_id": "Validation_45261133", "origin_question": "Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance forward or backward from that position?", "option": [ "Forward", "Backward" ], "answer": "Backward", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance forward or backward from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Forward', 'Backward'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_45261133/Validation_45261133_45261133_4057.261.png", "Validation_45261133/Validation_45261133_45261133_4059.260.png", "Validation_45261133/Validation_45261133_45261133_4061.259.png", "Validation_45261133/Validation_45261133_45261133_4063.258.png", "Validation_45261133/Validation_45261133_45261133_4065.274.png" ], "turn_id": 6, "scan_id": "Validation_45261133", "origin_question": "Assuming the direction you are facing at the end of the turn 5 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Right", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 5 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_45261133/Validation_45261133_45261133_4067.273.png", "Validation_45261133/Validation_45261133_45261133_4069.273.png", "Validation_45261133/Validation_45261133_45261133_4071.272.png", "Validation_45261133/Validation_45261133_45261133_4073.271.png", "Validation_45261133/Validation_45261133_45261133_4075.270.png" ], "turn_id": 7, "scan_id": "Validation_45261133", "origin_question": "How far is your current position from where you were at the end of the turn 4? (in meters)", "option": [], "answer": "1.54", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 4? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_45261133/Validation_45261133_45261133_4077.269.png", "Validation_45261133/Validation_45261133_45261133_4079.268.png", "Validation_45261133/Validation_45261133_45261133_4081.268.png", "Validation_45261133/Validation_45261133_45261133_4083.267.png", "Validation_45261133/Validation_45261133_45261133_4085.266.png" ], "turn_id": 8, "scan_id": "Validation_45261133", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_45261133/Validation_45261133_45261133_4087.265.png", "Validation_45261133/Validation_45261133_45261133_4089.264.png", "Validation_45261133/Validation_45261133_45261133_4091.263.png", "Validation_45261133/Validation_45261133_45261133_4093.263.png", "Validation_45261133/Validation_45261133_45261133_4095.262.png" ], "turn_id": 9, "scan_id": "Validation_45261133", "origin_question": "In which turn were you closest to \"the modern, sleek washing machine with the white exterior\"? ", "option": [], "answer": "7", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the modern, sleek washing machine with the white exterior\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_45261133/Validation_45261133_45261133_4097.261.png", "Validation_45261133/Validation_45261133_45261133_4099.260.png", "Validation_45261133/Validation_45261133_45261133_4101.259.png", "Validation_45261133/Validation_45261133_45261133_4103.258.png", "Validation_45261133/Validation_45261133_45261133_4105.258.png" ], "turn_id": 10, "scan_id": "Validation_45261133", "origin_question": "", "option": [], "answer": "", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42445029/Validation_42445029_42445029_49525.613.png", "Validation_42445029/Validation_42445029_42445029_49527.595.png", "Validation_42445029/Validation_42445029_42445029_49529.595.png", "Validation_42445029/Validation_42445029_42445029_49531.594.png" ], "turn_id": 1, "scan_id": "Validation_42445029", "origin_question": "Remember, have you seen any cabinet(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any cabinet(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_42445029/Validation_42445029_42445029_49533.593.png", "Validation_42445029/Validation_42445029_42445029_49535.592.png", "Validation_42445029/Validation_42445029_42445029_49537.591.png", "Validation_42445029/Validation_42445029_42445029_49539.591.png" ], "turn_id": 2, "scan_id": "Validation_42445029", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "Validation_42445029/Validation_42445029_42445029_49541.590.png", "Validation_42445029/Validation_42445029_42445029_49543.589.png", "Validation_42445029/Validation_42445029_42445029_49545.588.png", "Validation_42445029/Validation_42445029_42445029_49547.587.png" ], "turn_id": 3, "scan_id": "Validation_42445029", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the stylish, sizable black monitor with the glossy finish\"; \"the shelf\"; \"the sleek, white rectangular door\" ", "option": [ "the stylish, sizable black monitor with the glossy finish", "the shelf", "the sleek, white rectangular door" ], "answer": "the shelf", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the stylish, sizable black monitor with the glossy finish\"; \"the shelf\"; \"the sleek, white rectangular door\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the stylish, sizable black monitor with the glossy finish', 'the shelf', 'the sleek, white rectangular door'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_42445029/Validation_42445029_42445029_49549.603.png", "Validation_42445029/Validation_42445029_42445029_49551.603.png", "Validation_42445029/Validation_42445029_42445029_49553.602.png", "Validation_42445029/Validation_42445029_42445029_49555.601.png" ], "turn_id": 4, "scan_id": "Validation_42445029", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the stylish, sizable black monitor with the glossy finish\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the stylish, sizable black monitor with the glossy finish\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_42445029/Validation_42445029_42445029_49557.600.png", "Validation_42445029/Validation_42445029_42445029_49559.599.png", "Validation_42445029/Validation_42445029_42445029_49561.599.png", "Validation_42445029/Validation_42445029_42445029_49563.598.png" ], "turn_id": 5, "scan_id": "Validation_42445029", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "63.02", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_42445029/Validation_42445029_42445029_49565.597.png", "Validation_42445029/Validation_42445029_42445029_49567.596.png", "Validation_42445029/Validation_42445029_42445029_49569.595.png", "Validation_42445029/Validation_42445029_42445029_49571.595.png" ], "turn_id": 6, "scan_id": "Validation_42445029", "origin_question": "When did you first discover \"the shelf\"? (index of the turn)", "option": [], "answer": "3", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the shelf\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_42445029/Validation_42445029_42445029_49573.594.png", "Validation_42445029/Validation_42445029_42445029_49575.593.png", "Validation_42445029/Validation_42445029_42445029_49577.592.png", "Validation_42445029/Validation_42445029_42445029_49579.591.png" ], "turn_id": 7, "scan_id": "Validation_42445029", "origin_question": "Compared to your position at the end of the 6 turn, are you now closer or farther away from \"the stylish, sizable black monitor with the glossy finish\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 6 turn, are you now closer or farther away from \"the stylish, sizable black monitor with the glossy finish\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_42445029/Validation_42445029_42445029_49581.591.png", "Validation_42445029/Validation_42445029_42445029_49583.590.png", "Validation_42445029/Validation_42445029_42445029_49585.589.png", "Validation_42445029/Validation_42445029_42445029_49587.588.png" ], "turn_id": 8, "scan_id": "Validation_42445029", "origin_question": "Please recall \"the bed\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.97", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the bed\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_41159572/Validation_41159572_41159572_35059.131.png", "Validation_41159572/Validation_41159572_41159572_35061.114.png", "Validation_41159572/Validation_41159572_41159572_35063.113.png", "Validation_41159572/Validation_41159572_41159572_35065.112.png", "Validation_41159572/Validation_41159572_41159572_35067.111.png" ], "turn_id": 1, "scan_id": "Validation_41159572", "origin_question": "", "option": [], "answer": "", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_41159572/Validation_41159572_41159572_35069.111.png", "Validation_41159572/Validation_41159572_41159572_35071.110.png", "Validation_41159572/Validation_41159572_41159572_35073.109.png", "Validation_41159572/Validation_41159572_41159572_35075.125.png", "Validation_41159572/Validation_41159572_41159572_35077.124.png" ], "turn_id": 2, "scan_id": "Validation_41159572", "origin_question": "When did you first discover \"the straight wooden stairs with white risers\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the straight wooden stairs with white risers\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_41159572/Validation_41159572_41159572_35079.123.png", "Validation_41159572/Validation_41159572_41159572_35081.123.png", "Validation_41159572/Validation_41159572_41159572_35083.122.png", "Validation_41159572/Validation_41159572_41159572_35085.121.png", "Validation_41159572/Validation_41159572_41159572_35087.120.png" ], "turn_id": 3, "scan_id": "Validation_41159572", "origin_question": "At the end of which turn were both of \"the straight wooden stairs with white risers\" and \"the rustic brick fireplace in earthy tones\" on your right side?", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the straight wooden stairs with white risers\" and \"the rustic brick fireplace in earthy tones\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_41159572/Validation_41159572_41159572_35089.120.png", "Validation_41159572/Validation_41159572_41159572_35091.119.png", "Validation_41159572/Validation_41159572_41159572_35093.118.png", "Validation_41159572/Validation_41159572_41159572_35095.117.png", "Validation_41159572/Validation_41159572_41159572_35097.116.png" ], "turn_id": 4, "scan_id": "Validation_41159572", "origin_question": "Which two objects are on the same side of you now? \"the straight wooden stairs with white risers\", \"the rectangular table with the rich natural wood finish\", \"an off-white wooden door with vertical paneling\". ", "option": [ "the straight wooden stairs with white risers, the rectangular table with the rich natural wood finish", "the straight wooden stairs with white risers, an off-white wooden door with vertical paneling", "the rectangular table with the rich natural wood finish, an off-white wooden door with vertical paneling" ], "answer": "the rectangular table with the rich natural wood finish, an off-white wooden door with vertical paneling", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the straight wooden stairs with white risers\", \"the rectangular table with the rich natural wood finish\", \"an off-white wooden door with vertical paneling\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the straight wooden stairs with white risers, the rectangular table with the rich natural wood finish', 'the straight wooden stairs with white risers, an off-white wooden door with vertical paneling', 'the rectangular table with the rich natural wood finish, an off-white wooden door with vertical paneling'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_41159572/Validation_41159572_41159572_35099.115.png", "Validation_41159572/Validation_41159572_41159572_35101.115.png", "Validation_41159572/Validation_41159572_41159572_35103.114.png", "Validation_41159572/Validation_41159572_41159572_35105.113.png", "Validation_41159572/Validation_41159572_41159572_35107.112.png" ], "turn_id": 5, "scan_id": "Validation_41159572", "origin_question": "", "option": [], "answer": "", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_41159572/Validation_41159572_41159572_35109.112.png", "Validation_41159572/Validation_41159572_41159572_35111.111.png", "Validation_41159572/Validation_41159572_41159572_35113.110.png", "Validation_41159572/Validation_41159572_41159572_35115.109.png", "Validation_41159572/Validation_41159572_41159572_35117.125.png" ], "turn_id": 6, "scan_id": "Validation_41159572", "origin_question": "In which turn were you closest to \"the rectangular silver-framed picture on the blue wall\"? ", "option": [], "answer": "6", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the rectangular silver-framed picture on the blue wall\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "Validation_41159572/Validation_41159572_41159572_35119.124.png", "Validation_41159572/Validation_41159572_41159572_35121.124.png", "Validation_41159572/Validation_41159572_41159572_35123.123.png", "Validation_41159572/Validation_41159572_41159572_35125.122.png", "Validation_41159572/Validation_41159572_41159572_35127.121.png" ], "turn_id": 7, "scan_id": "Validation_41159572", "origin_question": "Remember, how many table(s) have you seen so far? ", "option": [], "answer": "2", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many table(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_41159572/Validation_41159572_41159572_35129.120.png", "Validation_41159572/Validation_41159572_41159572_35131.120.png", "Validation_41159572/Validation_41159572_41159572_35133.119.png", "Validation_41159572/Validation_41159572_41159572_35135.118.png", "Validation_41159572/Validation_41159572_41159572_35137.117.png" ], "turn_id": 8, "scan_id": "Validation_41159572", "origin_question": "Is the sleek, glossy dark lamp with the light lampshade to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the sleek, glossy dark lamp with the light lampshade to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "Validation_41159572/Validation_41159572_41159572_35139.116.png", "Validation_41159572/Validation_41159572_41159572_35141.116.png", "Validation_41159572/Validation_41159572_41159572_35143.115.png", "Validation_41159572/Validation_41159572_41159572_35145.114.png", "Validation_41159572/Validation_41159572_41159572_35147.113.png" ], "turn_id": 9, "scan_id": "Validation_41159572", "origin_question": "Among these three objects(\"the straight wooden stairs with white risers\",\"the compact cube-shaped stool with grey upholstery\",\"the rectangular silver-framed picture on the blue wall\" ), which one is the farthest from you?", "option": [ "the straight wooden stairs with white risers", "the compact cube-shaped stool with grey upholstery", "the rectangular silver-framed picture on the blue wall" ], "answer": "the straight wooden stairs with white risers", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the straight wooden stairs with white risers\",\"the compact cube-shaped stool with grey upholstery\",\"the rectangular silver-framed picture on the blue wall\" ), which one is the farthest from you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the straight wooden stairs with white risers', 'the compact cube-shaped stool with grey upholstery', 'the rectangular silver-framed picture on the blue wall'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_41159572/Validation_41159572_41159572_35149.112.png", "Validation_41159572/Validation_41159572_41159572_35151.112.png", "Validation_41159572/Validation_41159572_41159572_35153.111.png", "Validation_41159572/Validation_41159572_41159572_35155.110.png", "Validation_41159572/Validation_41159572_41159572_35157.109.png" ], "turn_id": 10, "scan_id": "Validation_41159572", "origin_question": "Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "53.34", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42898570/Validation_42898570_42898570_752817.597.png", "Validation_42898570/Validation_42898570_42898570_752819.579.png", "Validation_42898570/Validation_42898570_42898570_752821.578.png", "Validation_42898570/Validation_42898570_42898570_752823.578.png", "Validation_42898570/Validation_42898570_42898570_752825.577.png" ], "turn_id": 1, "scan_id": "Validation_42898570", "origin_question": "Remember, have you seen any bed(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any bed(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_42898570/Validation_42898570_42898570_752827.576.png", "Validation_42898570/Validation_42898570_42898570_752829.575.png", "Validation_42898570/Validation_42898570_42898570_752831.574.png", "Validation_42898570/Validation_42898570_42898570_752833.574.png", "Validation_42898570/Validation_42898570_42898570_752835.573.png" ], "turn_id": 2, "scan_id": "Validation_42898570", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the sleek, glossy black monitor with slim bezels\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the sleek, glossy black monitor with slim bezels\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_42898570/Validation_42898570_42898570_752837.572.png", "Validation_42898570/Validation_42898570_42898570_752839.571.png", "Validation_42898570/Validation_42898570_42898570_752841.587.png", "Validation_42898570/Validation_42898570_42898570_752843.586.png", "Validation_42898570/Validation_42898570_42898570_752845.585.png" ], "turn_id": 3, "scan_id": "Validation_42898570", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the white floral-printed bed with the medium-thick mattress\", Y: \"the sleek, glossy black monitor with slim bezels\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the white floral-printed bed with the medium-thick mattress\", Y: \"the sleek, glossy black monitor with slim bezels\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_42898570/Validation_42898570_42898570_752847.584.png", "Validation_42898570/Validation_42898570_42898570_752849.584.png", "Validation_42898570/Validation_42898570_42898570_752851.583.png", "Validation_42898570/Validation_42898570_42898570_752853.582.png", "Validation_42898570/Validation_42898570_42898570_752855.581.png" ], "turn_id": 4, "scan_id": "Validation_42898570", "origin_question": "Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance forward or backward from that position?", "option": [ "Forward", "Backward" ], "answer": "Forward", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance forward or backward from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Forward', 'Backward'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_42898570/Validation_42898570_42898570_752857.580.png", "Validation_42898570/Validation_42898570_42898570_752859.579.png", "Validation_42898570/Validation_42898570_42898570_752861.579.png", "Validation_42898570/Validation_42898570_42898570_752863.578.png", "Validation_42898570/Validation_42898570_42898570_752865.577.png" ], "turn_id": 5, "scan_id": "Validation_42898570", "origin_question": "By the end of the 3 turn, were you getting closer to or farther away from these two objects: X: \"the white floral-printed bed with the medium-thick mattress\", Y: \"the medium-sized lamp with the cream square lampshade\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "farther away from both X and Y", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 3 turn, were you getting closer to or farther away from these two objects: X: \"the white floral-printed bed with the medium-thick mattress\", Y: \"the medium-sized lamp with the cream square lampshade\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_42898570/Validation_42898570_42898570_752867.576.png", "Validation_42898570/Validation_42898570_42898570_752869.575.png", "Validation_42898570/Validation_42898570_42898570_752871.575.png", "Validation_42898570/Validation_42898570_42898570_752873.574.png", "Validation_42898570/Validation_42898570_42898570_752875.573.png" ], "turn_id": 6, "scan_id": "Validation_42898570", "origin_question": "What will be the first-time appearance order of the following categories: cabinet, light, monitor?", "option": [ "cabinet, monitor, light", "cabinet, light, monitor", "monitor, cabinet, light", "light, monitor, cabinet" ], "answer": "cabinet, monitor, light", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: cabinet, light, monitor? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['cabinet, monitor, light', 'cabinet, light, monitor', 'monitor, cabinet, light', 'light, monitor, cabinet'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_42898570/Validation_42898570_42898570_752877.572.png", "Validation_42898570/Validation_42898570_42898570_752879.571.png", "Validation_42898570/Validation_42898570_42898570_752881.587.png", "Validation_42898570/Validation_42898570_42898570_752883.586.png", "Validation_42898570/Validation_42898570_42898570_752885.586.png" ], "turn_id": 7, "scan_id": "Validation_42898570", "origin_question": "Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the medium-sized lamp with the cream square lampshade\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the medium-sized lamp with the cream square lampshade\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_42898570/Validation_42898570_42898570_752887.585.png", "Validation_42898570/Validation_42898570_42898570_752889.584.png", "Validation_42898570/Validation_42898570_42898570_752891.583.png", "Validation_42898570/Validation_42898570_42898570_752893.582.png", "Validation_42898570/Validation_42898570_42898570_752895.581.png" ], "turn_id": 8, "scan_id": "Validation_42898570", "origin_question": "What will be the first-time appearance order of the following categories: cabinet, monitor, table?", "option": [ "table, cabinet, monitor", "table, monitor, cabinet", "cabinet, monitor, table", "monitor, cabinet, table" ], "answer": "cabinet, monitor, table", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: cabinet, monitor, table? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['table, cabinet, monitor', 'table, monitor, cabinet', 'cabinet, monitor, table', 'monitor, cabinet, table'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_42898570/Validation_42898570_42898570_752897.581.png", "Validation_42898570/Validation_42898570_42898570_752899.580.png", "Validation_42898570/Validation_42898570_42898570_752901.579.png", "Validation_42898570/Validation_42898570_42898570_752903.578.png", "Validation_42898570/Validation_42898570_42898570_752905.577.png" ], "turn_id": 9, "scan_id": "Validation_42898570", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the medium-sized lamp with the cream square lampshade\", Y: \"the rectangular rich brown wooden table with drawers\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the medium-sized lamp with the cream square lampshade\", Y: \"the rectangular rich brown wooden table with drawers\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_42898570/Validation_42898570_42898570_752907.576.png", "Validation_42898570/Validation_42898570_42898570_752909.576.png", "Validation_42898570/Validation_42898570_42898570_752911.575.png", "Validation_42898570/Validation_42898570_42898570_752913.574.png", "Validation_42898570/Validation_42898570_42898570_752915.573.png" ], "turn_id": 10, "scan_id": "Validation_42898570", "origin_question": "Which direction is the rectangular rich brown wooden table with drawers in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the rectangular rich brown wooden table with drawers in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42899620/Validation_42899620_42899620_228847.689.png", "Validation_42899620/Validation_42899620_42899620_228849.172.png", "Validation_42899620/Validation_42899620_42899620_228850.671.png", "Validation_42899620/Validation_42899620_42899620_228852.171.png" ], "turn_id": 1, "scan_id": "Validation_42899620", "origin_question": "Remember, have you seen any piano(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any piano(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "Validation_42899620/Validation_42899620_42899620_228853.670.png", "Validation_42899620/Validation_42899620_42899620_228855.170.png", "Validation_42899620/Validation_42899620_42899620_228856.669.png", "Validation_42899620/Validation_42899620_42899620_228858.168.png" ], "turn_id": 2, "scan_id": "Validation_42899620", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the classic rectangular white porcelain sink\"; \"the vertical ladder with metallic silver bars\"; \"the bathtub\" ", "option": [ "the classic rectangular white porcelain sink", "the vertical ladder with metallic silver bars", "the bathtub" ], "answer": "the bathtub", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the classic rectangular white porcelain sink\"; \"the vertical ladder with metallic silver bars\"; \"the bathtub\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the classic rectangular white porcelain sink', 'the vertical ladder with metallic silver bars', 'the bathtub'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_42899620/Validation_42899620_42899620_228859.668.png", "Validation_42899620/Validation_42899620_42899620_228861.167.png", "Validation_42899620/Validation_42899620_42899620_228862.667.png", "Validation_42899620/Validation_42899620_42899620_228864.166.png" ], "turn_id": 3, "scan_id": "Validation_42899620", "origin_question": "Which two objects are on the same side of you now? \"the vertical ladder with metallic silver bars\", \"the sleek white wooden door with the rectangular shape\", \"the bathtub\". ", "option": [ "the vertical ladder with metallic silver bars, the sleek white wooden door with the rectangular shape", "the vertical ladder with metallic silver bars, the bathtub", "the sleek white wooden door with the rectangular shape, the bathtub" ], "answer": "the vertical ladder with metallic silver bars, the sleek white wooden door with the rectangular shape", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the vertical ladder with metallic silver bars\", \"the sleek white wooden door with the rectangular shape\", \"the bathtub\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the vertical ladder with metallic silver bars, the sleek white wooden door with the rectangular shape', 'the vertical ladder with metallic silver bars, the bathtub', 'the sleek white wooden door with the rectangular shape, the bathtub'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_42899620/Validation_42899620_42899620_228865.666.png", "Validation_42899620/Validation_42899620_42899620_228867.165.png", "Validation_42899620/Validation_42899620_42899620_228868.664.png", "Validation_42899620/Validation_42899620_42899620_228870.164.png" ], "turn_id": 4, "scan_id": "Validation_42899620", "origin_question": "Which direction is the vertical ladder with metallic silver bars in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the vertical ladder with metallic silver bars in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_42899620/Validation_42899620_42899620_228871.663.png", "Validation_42899620/Validation_42899620_42899620_228873.163.png", "Validation_42899620/Validation_42899620_42899620_228874.662.png", "Validation_42899620/Validation_42899620_42899620_228876.161.png" ], "turn_id": 5, "scan_id": "Validation_42899620", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47333931/Validation_47333931_47333931_83913.716.png", "Validation_47333931/Validation_47333931_47333931_83915.698.png", "Validation_47333931/Validation_47333931_47333931_83917.697.png", "Validation_47333931/Validation_47333931_47333931_83919.713.png" ], "turn_id": 1, "scan_id": "Validation_47333931", "origin_question": "Remember, have you seen any toiletry(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any toiletry(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47333931/Validation_47333931_47333931_83921.712.png", "Validation_47333931/Validation_47333931_47333931_83923.712.png", "Validation_47333931/Validation_47333931_47333931_83925.711.png", "Validation_47333931/Validation_47333931_47333931_83927.710.png" ], "turn_id": 2, "scan_id": "Validation_47333931", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_47333931/Validation_47333931_47333931_83929.709.png", "Validation_47333931/Validation_47333931_47333931_83931.708.png", "Validation_47333931/Validation_47333931_47333931_83933.708.png", "Validation_47333931/Validation_47333931_47333931_83935.707.png" ], "turn_id": 3, "scan_id": "Validation_47333931", "origin_question": "Is the warm-toned wooden table with the polished finish to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the warm-toned wooden table with the polished finish to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_47333931/Validation_47333931_47333931_83937.706.png", "Validation_47333931/Validation_47333931_47333931_83939.705.png", "Validation_47333931/Validation_47333931_47333931_83941.704.png", "Validation_47333931/Validation_47333931_47333931_83943.704.png" ], "turn_id": 4, "scan_id": "Validation_47333931", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "1.39", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_47333931/Validation_47333931_47333931_83945.703.png", "Validation_47333931/Validation_47333931_47333931_83947.702.png", "Validation_47333931/Validation_47333931_47333931_83949.701.png", "Validation_47333931/Validation_47333931_47333931_83951.700.png" ], "turn_id": 5, "scan_id": "Validation_47333931", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_47333931/Validation_47333931_47333931_83953.700.png", "Validation_47333931/Validation_47333931_47333931_83955.699.png", "Validation_47333931/Validation_47333931_47333931_83957.698.png", "Validation_47333931/Validation_47333931_47333931_83959.697.png" ], "turn_id": 6, "scan_id": "Validation_47333931", "origin_question": "Using your orientation at the end of turn 4 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "103.4", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 4 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_47333931/Validation_47333931_47333931_83961.713.png", "Validation_47333931/Validation_47333931_47333931_83963.712.png", "Validation_47333931/Validation_47333931_47333931_83965.711.png", "Validation_47333931/Validation_47333931_47333931_83967.711.png" ], "turn_id": 7, "scan_id": "Validation_47333931", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the compact, white vertical radiator with metal panels\" from your position?", "option": [], "answer": "168.68", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the compact, white vertical radiator with metal panels\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_47333931/Validation_47333931_47333931_83969.710.png", "Validation_47333931/Validation_47333931_47333931_83971.709.png", "Validation_47333931/Validation_47333931_47333931_83973.708.png", "Validation_47333931/Validation_47333931_47333931_83975.707.png" ], "turn_id": 8, "scan_id": "Validation_47333931", "origin_question": "Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "153.07", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_47333931/Validation_47333931_47333931_83977.706.png", "Validation_47333931/Validation_47333931_47333931_83979.706.png", "Validation_47333931/Validation_47333931_47333931_83981.705.png", "Validation_47333931/Validation_47333931_47333931_83983.704.png" ], "turn_id": 9, "scan_id": "Validation_47333931", "origin_question": "At the end of which turn were both of \"the warm-toned wooden table with the polished finish\" and \"the window\" on your right side?", "option": [], "answer": "3", "user_message": "For the 9 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the warm-toned wooden table with the polished finish\" and \"the window\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_41142281/Validation_41142281_41142281_3918.399.png", "Validation_41142281/Validation_41142281_41142281_3920.381.png", "Validation_41142281/Validation_41142281_41142281_3922.381.png", "Validation_41142281/Validation_41142281_41142281_3924.380.png" ], "turn_id": 1, "scan_id": "Validation_41142281", "origin_question": "Remember, have you seen any picture(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any picture(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_41142281/Validation_41142281_41142281_3926.379.png", "Validation_41142281/Validation_41142281_41142281_3928.378.png", "Validation_41142281/Validation_41142281_41142281_3930.377.png", "Validation_41142281/Validation_41142281_41142281_3932.376.png" ], "turn_id": 2, "scan_id": "Validation_41142281", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the cylindrical stainless steel bin with the black lid\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the cylindrical stainless steel bin with the black lid\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_41142281/Validation_41142281_41142281_3934.376.png", "Validation_41142281/Validation_41142281_41142281_3936.375.png", "Validation_41142281/Validation_41142281_41142281_3938.374.png", "Validation_41142281/Validation_41142281_41142281_3940.373.png" ], "turn_id": 3, "scan_id": "Validation_41142281", "origin_question": "Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "75.8", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_41142281/Validation_41142281_41142281_3942.372.png", "Validation_41142281/Validation_41142281_41142281_3944.372.png", "Validation_41142281/Validation_41142281_41142281_3946.371.png", "Validation_41142281/Validation_41142281_41142281_3948.370.png" ], "turn_id": 4, "scan_id": "Validation_41142281", "origin_question": "In which turn were you closest to \"the picture with brown and beige cover\"? ", "option": [], "answer": "1", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the picture with brown and beige cover\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_41142281/Validation_41142281_41142281_3950.369.png", "Validation_41142281/Validation_41142281_41142281_3952.385.png", "Validation_41142281/Validation_41142281_41142281_3954.384.png", "Validation_41142281/Validation_41142281_41142281_3956.383.png" ], "turn_id": 5, "scan_id": "Validation_41142281", "origin_question": "Please recall \"the picture with brown and beige cover\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.73", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the picture with brown and beige cover\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "Validation_41142281/Validation_41142281_41142281_3958.382.png", "Validation_41142281/Validation_41142281_41142281_3960.382.png", "Validation_41142281/Validation_41142281_41142281_3962.381.png", "Validation_41142281/Validation_41142281_41142281_3964.380.png" ], "turn_id": 6, "scan_id": "Validation_41142281", "origin_question": "Among these three objects(\"the rounded white toilet with the smooth finish\",\"the light-toned natural wood door with visible hinges\",\"the sleek, white moisturizing soap bar\" ), which one is the farthest from you?", "option": [ "the rounded white toilet with the smooth finish", "the light-toned natural wood door with visible hinges", "the sleek, white moisturizing soap bar" ], "answer": "the light-toned natural wood door with visible hinges", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects(\"the rounded white toilet with the smooth finish\",\"the light-toned natural wood door with visible hinges\",\"the sleek, white moisturizing soap bar\" ), which one is the farthest from you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rounded white toilet with the smooth finish', 'the light-toned natural wood door with visible hinges', 'the sleek, white moisturizing soap bar'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_41142281/Validation_41142281_41142281_3966.379.png", "Validation_41142281/Validation_41142281_41142281_3968.378.png", "Validation_41142281/Validation_41142281_41142281_3970.378.png", "Validation_41142281/Validation_41142281_41142281_3972.377.png" ], "turn_id": 7, "scan_id": "Validation_41142281", "origin_question": "What will be the first-time appearance order of the following categories: picture, soap, towel?", "option": [ "towel, picture, soap", "picture, towel, soap", "picture, soap, towel", "soap, towel, picture" ], "answer": "picture, towel, soap", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: picture, soap, towel? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['towel, picture, soap', 'picture, towel, soap', 'picture, soap, towel', 'soap, towel, picture'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_41142281/Validation_41142281_41142281_3974.376.png", "Validation_41142281/Validation_41142281_41142281_3976.375.png", "Validation_41142281/Validation_41142281_41142281_3978.374.png", "Validation_41142281/Validation_41142281_41142281_3980.373.png" ], "turn_id": 8, "scan_id": "Validation_41142281", "origin_question": "Compared to your position at the end of the 6 turn, are you now closer or farther away from \"the rounded white toilet with the smooth finish\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 6 turn, are you now closer or farther away from \"the rounded white toilet with the smooth finish\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "Validation_41142281/Validation_41142281_41142281_3982.372.png", "Validation_41142281/Validation_41142281_41142281_3984.372.png", "Validation_41142281/Validation_41142281_41142281_3986.371.png", "Validation_41142281/Validation_41142281_41142281_3988.370.png" ], "turn_id": 9, "scan_id": "Validation_41142281", "origin_question": "Among these three objects(\"the cylindrical stainless steel bin with the black lid\",\"the light-toned natural wood door with visible hinges\",\"the sleek, white moisturizing soap bar\" ), which one is the farthest from you?", "option": [ "the cylindrical stainless steel bin with the black lid", "the light-toned natural wood door with visible hinges", "the sleek, white moisturizing soap bar" ], "answer": "the cylindrical stainless steel bin with the black lid", "user_message": "For the 9 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects(\"the cylindrical stainless steel bin with the black lid\",\"the light-toned natural wood door with visible hinges\",\"the sleek, white moisturizing soap bar\" ), which one is the farthest from you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the cylindrical stainless steel bin with the black lid', 'the light-toned natural wood door with visible hinges', 'the sleek, white moisturizing soap bar'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42899680/Validation_42899680_42899680_790592.924.png", "Validation_42899680/Validation_42899680_42899680_790594.907.png", "Validation_42899680/Validation_42899680_42899680_790596.906.png", "Validation_42899680/Validation_42899680_42899680_790598.905.png", "Validation_42899680/Validation_42899680_42899680_790600.904.png" ], "turn_id": 1, "scan_id": "Validation_42899680", "origin_question": "Remember, have you seen any drum(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any drum(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_42899680/Validation_42899680_42899680_790602.903.png", "Validation_42899680/Validation_42899680_42899680_790604.919.png", "Validation_42899680/Validation_42899680_42899680_790606.918.png", "Validation_42899680/Validation_42899680_42899680_790608.917.png", "Validation_42899680/Validation_42899680_42899680_790610.917.png" ], "turn_id": 2, "scan_id": "Validation_42899680", "origin_question": "Is the clean white rectangular bed with the sturdy frame to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the clean white rectangular bed with the sturdy frame to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "Validation_42899680/Validation_42899680_42899680_790612.916.png", "Validation_42899680/Validation_42899680_42899680_790614.915.png", "Validation_42899680/Validation_42899680_42899680_790616.914.png", "Validation_42899680/Validation_42899680_42899680_790618.913.png", "Validation_42899680/Validation_42899680_42899680_790620.913.png" ], "turn_id": 3, "scan_id": "Validation_42899680", "origin_question": "Among these three objects(\"the medium to large cotton towel with vibrant stripes\",\"the light cream cylindrical metal bin\",\"the clean white rectangular bed with the sturdy frame\" ), which one is the closest to you?", "option": [ "the medium to large cotton towel with vibrant stripes", "the light cream cylindrical metal bin", "the clean white rectangular bed with the sturdy frame" ], "answer": "the light cream cylindrical metal bin", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the medium to large cotton towel with vibrant stripes\",\"the light cream cylindrical metal bin\",\"the clean white rectangular bed with the sturdy frame\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the medium to large cotton towel with vibrant stripes', 'the light cream cylindrical metal bin', 'the clean white rectangular bed with the sturdy frame'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_42899680/Validation_42899680_42899680_790622.912.png", "Validation_42899680/Validation_42899680_42899680_790624.911.png", "Validation_42899680/Validation_42899680_42899680_790626.910.png", "Validation_42899680/Validation_42899680_42899680_790628.909.png", "Validation_42899680/Validation_42899680_42899680_790630.908.png" ], "turn_id": 4, "scan_id": "Validation_42899680", "origin_question": "Which two objects are on the same side of you now? \"the clean white rectangular bed with the sturdy frame\", \"the sleek white rectangular ceramic sink\", \"the compact plant with lush green leaves\". ", "option": [ "the clean white rectangular bed with the sturdy frame, the sleek white rectangular ceramic sink", "the clean white rectangular bed with the sturdy frame, the compact plant with lush green leaves", "the sleek white rectangular ceramic sink, the compact plant with lush green leaves" ], "answer": "the sleek white rectangular ceramic sink, the compact plant with lush green leaves", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the clean white rectangular bed with the sturdy frame\", \"the sleek white rectangular ceramic sink\", \"the compact plant with lush green leaves\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the clean white rectangular bed with the sturdy frame, the sleek white rectangular ceramic sink', 'the clean white rectangular bed with the sturdy frame, the compact plant with lush green leaves', 'the sleek white rectangular ceramic sink, the compact plant with lush green leaves'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_42899680/Validation_42899680_42899680_790632.908.png", "Validation_42899680/Validation_42899680_42899680_790634.907.png", "Validation_42899680/Validation_42899680_42899680_790636.906.png", "Validation_42899680/Validation_42899680_42899680_790638.905.png", "Validation_42899680/Validation_42899680_42899680_790640.904.png" ], "turn_id": 5, "scan_id": "Validation_42899680", "origin_question": "", "option": [], "answer": "", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "Validation_42899680/Validation_42899680_42899680_790642.904.png", "Validation_42899680/Validation_42899680_42899680_790644.919.png", "Validation_42899680/Validation_42899680_42899680_790646.919.png", "Validation_42899680/Validation_42899680_42899680_790648.918.png", "Validation_42899680/Validation_42899680_42899680_790650.917.png" ], "turn_id": 6, "scan_id": "Validation_42899680", "origin_question": "Among these three objects(\"the sleek white dishwasher with the metal and plastic exterior\",\"the machine\",\"the sleek, cylindrical stainless steel kettle\" ), which one is the closest to you?", "option": [ "the sleek white dishwasher with the metal and plastic exterior", "the machine", "the sleek, cylindrical stainless steel kettle" ], "answer": "the sleek white dishwasher with the metal and plastic exterior", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the sleek white dishwasher with the metal and plastic exterior\",\"the machine\",\"the sleek, cylindrical stainless steel kettle\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the sleek white dishwasher with the metal and plastic exterior', 'the machine', 'the sleek, cylindrical stainless steel kettle'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_42899680/Validation_42899680_42899680_790652.916.png", "Validation_42899680/Validation_42899680_42899680_790654.915.png", "Validation_42899680/Validation_42899680_42899680_790656.914.png", "Validation_42899680/Validation_42899680_42899680_790658.914.png", "Validation_42899680/Validation_42899680_42899680_790660.913.png" ], "turn_id": 7, "scan_id": "Validation_42899680", "origin_question": "In which turn were you closest to \"the medium-sized, cube-shaped box\"? ", "option": [], "answer": "7", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the medium-sized, cube-shaped box\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_42899680/Validation_42899680_42899680_790662.912.png", "Validation_42899680/Validation_42899680_42899680_790664.911.png", "Validation_42899680/Validation_42899680_42899680_790666.910.png", "Validation_42899680/Validation_42899680_42899680_790668.909.png", "Validation_42899680/Validation_42899680_42899680_790670.909.png" ], "turn_id": 8, "scan_id": "Validation_42899680", "origin_question": "Which two objects are on the same side of you now? \"the machine\", \"the black stainless-steel oven with the modern design\", \"the clear cylindrical bottle made of transparent material\". ", "option": [ "the machine, the black stainless-steel oven with the modern design", "the machine, the clear cylindrical bottle made of transparent material", "the black stainless-steel oven with the modern design, the clear cylindrical bottle made of transparent material" ], "answer": "the machine, the black stainless-steel oven with the modern design", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the machine\", \"the black stainless-steel oven with the modern design\", \"the clear cylindrical bottle made of transparent material\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the machine, the black stainless-steel oven with the modern design', 'the machine, the clear cylindrical bottle made of transparent material', 'the black stainless-steel oven with the modern design, the clear cylindrical bottle made of transparent material'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_42899680/Validation_42899680_42899680_790672.908.png", "Validation_42899680/Validation_42899680_42899680_790674.907.png", "Validation_42899680/Validation_42899680_42899680_790676.906.png", "Validation_42899680/Validation_42899680_42899680_790678.905.png", "Validation_42899680/Validation_42899680_42899680_790680.904.png" ], "turn_id": 9, "scan_id": "Validation_42899680", "origin_question": "Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the sleek, cylindrical stainless steel kettle\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the sleek, cylindrical stainless steel kettle\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_42899680/Validation_42899680_42899680_790682.904.png", "Validation_42899680/Validation_42899680_42899680_790684.903.png", "Validation_42899680/Validation_42899680_42899680_790686.919.png", "Validation_42899680/Validation_42899680_42899680_790688.918.png", "Validation_42899680/Validation_42899680_42899680_790690.917.png" ], "turn_id": 10, "scan_id": "Validation_42899680", "origin_question": "How far is your current position from where you were at the end of the turn 4? (in meters)", "option": [], "answer": "5.2", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 4? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47895740/Validation_47895740_47895740_384929.349.png", "Validation_47895740/Validation_47895740_47895740_384930.849.png", "Validation_47895740/Validation_47895740_47895740_384932.348.png", "Validation_47895740/Validation_47895740_47895740_384933.847.png" ], "turn_id": 1, "scan_id": "Validation_47895740", "origin_question": "Remember, have you seen any wine(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any wine(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47895740/Validation_47895740_47895740_384935.347.png", "Validation_47895740/Validation_47895740_47895740_384936.846.png", "Validation_47895740/Validation_47895740_47895740_384938.345.png", "Validation_47895740/Validation_47895740_47895740_384939.845.png" ], "turn_id": 2, "scan_id": "Validation_47895740", "origin_question": "In which turn were you farthest from \"the classic white rectangular bathtub\"? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the classic white rectangular bathtub\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_47895740/Validation_47895740_47895740_384941.344.png", "Validation_47895740/Validation_47895740_47895740_384942.844.png", "Validation_47895740/Validation_47895740_47895740_384944.343.png", "Validation_47895740/Validation_47895740_47895740_384945.842.png" ], "turn_id": 3, "scan_id": "Validation_47895740", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_47895740/Validation_47895740_47895740_384947.342.png", "Validation_47895740/Validation_47895740_47895740_384948.841.png", "Validation_47895740/Validation_47895740_47895740_384950.340.png", "Validation_47895740/Validation_47895740_47895740_384951.840.png" ], "turn_id": 4, "scan_id": "Validation_47895740", "origin_question": "What will be the first-time appearance order of the following categories: bathtub, cabinet, washing machine?", "option": [ "cabinet, bathtub, washing machine", "washing machine, bathtub, cabinet", "bathtub, washing machine, cabinet", "bathtub, cabinet, washing machine" ], "answer": "cabinet, bathtub, washing machine", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bathtub, cabinet, washing machine? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['cabinet, bathtub, washing machine', 'washing machine, bathtub, cabinet', 'bathtub, washing machine, cabinet', 'bathtub, cabinet, washing machine'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47331064/Validation_47331064_47331064_129663.173.png", "Validation_47331064/Validation_47331064_47331064_129665.156.png", "Validation_47331064/Validation_47331064_47331064_129667.155.png", "Validation_47331064/Validation_47331064_47331064_129669.154.png", "Validation_47331064/Validation_47331064_47331064_129671.153.png" ], "turn_id": 1, "scan_id": "Validation_47331064", "origin_question": "Remember, have you seen any coil(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any coil(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_47331064/Validation_47331064_47331064_129673.152.png", "Validation_47331064/Validation_47331064_47331064_129675.152.png", "Validation_47331064/Validation_47331064_47331064_129677.151.png", "Validation_47331064/Validation_47331064_47331064_129679.150.png", "Validation_47331064/Validation_47331064_47331064_129681.149.png" ], "turn_id": 2, "scan_id": "Validation_47331064", "origin_question": "When did you first discover \"the compact, curved chair with dark upholstery\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the compact, curved chair with dark upholstery\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_47331064/Validation_47331064_47331064_129683.148.png", "Validation_47331064/Validation_47331064_47331064_129685.148.png", "Validation_47331064/Validation_47331064_47331064_129687.147.png", "Validation_47331064/Validation_47331064_47331064_129689.146.png", "Validation_47331064/Validation_47331064_47331064_129691.145.png" ], "turn_id": 3, "scan_id": "Validation_47331064", "origin_question": "When did you first discover \"the sleek, white upright refrigerator with the metal exterior\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the sleek, white upright refrigerator with the metal exterior\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_47331064/Validation_47331064_47331064_129693.144.png", "Validation_47331064/Validation_47331064_47331064_129695.144.png", "Validation_47331064/Validation_47331064_47331064_129697.143.png", "Validation_47331064/Validation_47331064_47331064_129699.142.png", "Validation_47331064/Validation_47331064_47331064_129701.141.png" ], "turn_id": 4, "scan_id": "Validation_47331064", "origin_question": "Is the modern black ceramic stove with the glossy surface to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the modern black ceramic stove with the glossy surface to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_47331064/Validation_47331064_47331064_129703.140.png", "Validation_47331064/Validation_47331064_47331064_129705.156.png", "Validation_47331064/Validation_47331064_47331064_129707.155.png", "Validation_47331064/Validation_47331064_47331064_129709.155.png", "Validation_47331064/Validation_47331064_47331064_129711.154.png" ], "turn_id": 5, "scan_id": "Validation_47331064", "origin_question": "In which turn did you see both \"the rectangular ceramic or stainless steel sink\" and \"the sleek stainless steel built-in oven\" simultaneously? ", "option": [], "answer": "5", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the rectangular ceramic or stainless steel sink\" and \"the sleek stainless steel built-in oven\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47331064/Validation_47331064_47331064_129713.153.png", "Validation_47331064/Validation_47331064_47331064_129715.152.png", "Validation_47331064/Validation_47331064_47331064_129717.151.png", "Validation_47331064/Validation_47331064_47331064_129719.150.png", "Validation_47331064/Validation_47331064_47331064_129721.150.png" ], "turn_id": 6, "scan_id": "Validation_47331064", "origin_question": "", "option": [], "answer": "", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_47331064/Validation_47331064_47331064_129723.149.png", "Validation_47331064/Validation_47331064_47331064_129725.148.png", "Validation_47331064/Validation_47331064_47331064_129727.147.png", "Validation_47331064/Validation_47331064_47331064_129729.146.png", "Validation_47331064/Validation_47331064_47331064_129731.146.png" ], "turn_id": 7, "scan_id": "Validation_47331064", "origin_question": "Please recall \"the light cream-colored microwave made of metal\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.81", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the light cream-colored microwave made of metal\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47331064/Validation_47331064_47331064_129733.145.png", "Validation_47331064/Validation_47331064_47331064_129735.144.png", "Validation_47331064/Validation_47331064_47331064_129737.143.png", "Validation_47331064/Validation_47331064_47331064_129739.142.png", "Validation_47331064/Validation_47331064_47331064_129741.141.png" ], "turn_id": 8, "scan_id": "Validation_47331064", "origin_question": "Which direction is the sleek white plastic kettle with an ergonomic handle in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the sleek white plastic kettle with an ergonomic handle in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47331064/Validation_47331064_47331064_129743.141.png", "Validation_47331064/Validation_47331064_47331064_129745.156.png", "Validation_47331064/Validation_47331064_47331064_129747.156.png", "Validation_47331064/Validation_47331064_47331064_129749.155.png", "Validation_47331064/Validation_47331064_47331064_129751.154.png" ], "turn_id": 9, "scan_id": "Validation_47331064", "origin_question": "Which direction is the light cream-colored microwave made of metal in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the light cream-colored microwave made of metal in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_47331064/Validation_47331064_47331064_129753.153.png", "Validation_47331064/Validation_47331064_47331064_129755.152.png", "Validation_47331064/Validation_47331064_47331064_129757.152.png", "Validation_47331064/Validation_47331064_47331064_129759.151.png", "Validation_47331064/Validation_47331064_47331064_129761.150.png" ], "turn_id": 10, "scan_id": "Validation_47331064", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "1.99", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42897554/Validation_42897554_42897554_277314.999.png", "Validation_42897554/Validation_42897554_42897554_277316.498.png", "Validation_42897554/Validation_42897554_42897554_277317.998.png", "Validation_42897554/Validation_42897554_42897554_277319.497.png" ], "turn_id": 1, "scan_id": "Validation_42897554", "origin_question": "Remember, have you seen any window(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any window(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_42897554/Validation_42897554_42897554_277320.997.png", "Validation_42897554/Validation_42897554_42897554_277322.496.png", "Validation_42897554/Validation_42897554_42897554_277323.995.png", "Validation_42897554/Validation_42897554_42897554_277325.495.png" ], "turn_id": 2, "scan_id": "Validation_42897554", "origin_question": "When did you first discover \"the bed\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the bed\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_42897554/Validation_42897554_42897554_277326.994.png", "Validation_42897554/Validation_42897554_42897554_277328.494.png", "Validation_42897554/Validation_42897554_42897554_277330.010.png", "Validation_42897554/Validation_42897554_42897554_277331.509.png" ], "turn_id": 3, "scan_id": "Validation_42897554", "origin_question": "When did you first discover \"the large rectangular window with the sturdy frame\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the large rectangular window with the sturdy frame\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_42897554/Validation_42897554_42897554_277333.008.png", "Validation_42897554/Validation_42897554_42897554_277334.508.png", "Validation_42897554/Validation_42897554_42897554_277336.007.png", "Validation_42897554/Validation_42897554_42897554_277337.507.png" ], "turn_id": 4, "scan_id": "Validation_42897554", "origin_question": "", "option": [], "answer": "", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_42897554/Validation_42897554_42897554_277339.006.png", "Validation_42897554/Validation_42897554_42897554_277340.505.png", "Validation_42897554/Validation_42897554_42897554_277342.005.png", "Validation_42897554/Validation_42897554_42897554_277343.504.png" ], "turn_id": 5, "scan_id": "Validation_42897554", "origin_question": "In which turn did you see both \"the bed\" and \"the vertically positioned wooden door in the light tone\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the bed\" and \"the vertically positioned wooden door in the light tone\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_41159530/Validation_41159530_41159530_39160.424.png", "Validation_41159530/Validation_41159530_41159530_39162.406.png", "Validation_41159530/Validation_41159530_41159530_39164.406.png", "Validation_41159530/Validation_41159530_41159530_39166.405.png", "Validation_41159530/Validation_41159530_41159530_39168.404.png" ], "turn_id": 1, "scan_id": "Validation_41159530", "origin_question": "Remember, have you seen any door(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any door(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_41159530/Validation_41159530_41159530_39170.403.png", "Validation_41159530/Validation_41159530_41159530_39172.419.png", "Validation_41159530/Validation_41159530_41159530_39174.418.png", "Validation_41159530/Validation_41159530_41159530_39176.417.png", "Validation_41159530/Validation_41159530_41159530_39178.416.png" ], "turn_id": 2, "scan_id": "Validation_41159530", "origin_question": "In which turn did you see both \"the sleek light gray rectangular table with closed drawers\" and \"the cap\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the sleek light gray rectangular table with closed drawers\" and \"the cap\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_41159530/Validation_41159530_41159530_39180.416.png", "Validation_41159530/Validation_41159530_41159530_39182.415.png", "Validation_41159530/Validation_41159530_41159530_39184.414.png", "Validation_41159530/Validation_41159530_41159530_39186.413.png", "Validation_41159530/Validation_41159530_41159530_39188.412.png" ], "turn_id": 3, "scan_id": "Validation_41159530", "origin_question": "What will be the first-time appearance order of the following categories: door, hat, kettle?", "option": [ "door, kettle, hat", "kettle, hat, door", "kettle, door, hat", "door, hat, kettle" ], "answer": "door, hat, kettle", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: door, hat, kettle? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['door, kettle, hat', 'kettle, hat, door', 'kettle, door, hat', 'door, hat, kettle'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_41159530/Validation_41159530_41159530_39190.412.png", "Validation_41159530/Validation_41159530_41159530_39192.411.png", "Validation_41159530/Validation_41159530_41159530_39194.410.png", "Validation_41159530/Validation_41159530_41159530_39196.409.png", "Validation_41159530/Validation_41159530_41159530_39198.408.png" ], "turn_id": 4, "scan_id": "Validation_41159530", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the warm cylindrical light fixture with the fabric shade\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the warm cylindrical light fixture with the fabric shade\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_41159530/Validation_41159530_41159530_39200.407.png", "Validation_41159530/Validation_41159530_41159530_39202.406.png", "Validation_41159530/Validation_41159530_41159530_39204.406.png", "Validation_41159530/Validation_41159530_41159530_39206.405.png", "Validation_41159530/Validation_41159530_41159530_39208.404.png" ], "turn_id": 5, "scan_id": "Validation_41159530", "origin_question": "What will be the first-time appearance order of the following categories: hat, light, shelf?", "option": [ "hat, light, shelf", "light, shelf, hat", "light, hat, shelf", "hat, shelf, light" ], "answer": "hat, light, shelf", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: hat, light, shelf? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['hat, light, shelf', 'light, shelf, hat', 'light, hat, shelf', 'hat, shelf, light'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_41159530/Validation_41159530_41159530_39210.403.png", "Validation_41159530/Validation_41159530_41159530_39212.419.png", "Validation_41159530/Validation_41159530_41159530_39214.418.png", "Validation_41159530/Validation_41159530_41159530_39216.417.png", "Validation_41159530/Validation_41159530_41159530_39218.416.png" ], "turn_id": 6, "scan_id": "Validation_41159530", "origin_question": "Using your orientation at the end of turn 4 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "54.67", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 4 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_41159530/Validation_41159530_41159530_39220.416.png", "Validation_41159530/Validation_41159530_41159530_39222.415.png", "Validation_41159530/Validation_41159530_41159530_39224.414.png", "Validation_41159530/Validation_41159530_41159530_39226.413.png", "Validation_41159530/Validation_41159530_41159530_39228.412.png" ], "turn_id": 7, "scan_id": "Validation_41159530", "origin_question": "What will be the first-time appearance order of the following categories: door, hat, shelf?", "option": [ "shelf, hat, door", "door, hat, shelf", "hat, door, shelf", "hat, shelf, door" ], "answer": "door, hat, shelf", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: door, hat, shelf? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['shelf, hat, door', 'door, hat, shelf', 'hat, door, shelf', 'hat, shelf, door'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_41159530/Validation_41159530_41159530_39230.412.png", "Validation_41159530/Validation_41159530_41159530_39232.411.png", "Validation_41159530/Validation_41159530_41159530_39234.410.png", "Validation_41159530/Validation_41159530_41159530_39236.409.png", "Validation_41159530/Validation_41159530_41159530_39238.408.png" ], "turn_id": 8, "scan_id": "Validation_41159530", "origin_question": "By the end of the 4 turn, were you getting closer to or farther away from these two objects: X: \"the cap\", Y: \"the medium-sized green metal kettle with the rounded shape\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "farther away from both X and Y", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 4 turn, were you getting closer to or farther away from these two objects: X: \"the cap\", Y: \"the medium-sized green metal kettle with the rounded shape\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_45662943/Validation_45662943_45662943_23019.268.png", "Validation_45662943/Validation_45662943_45662943_23021.267.png", "Validation_45662943/Validation_45662943_45662943_23023.266.png", "Validation_45662943/Validation_45662943_45662943_23025.266.png", "Validation_45662943/Validation_45662943_45662943_23027.265.png" ], "turn_id": 1, "scan_id": "Validation_45662943", "origin_question": "Remember, have you seen any car(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any car(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "Validation_45662943/Validation_45662943_45662943_23029.264.png", "Validation_45662943/Validation_45662943_45662943_23031.263.png", "Validation_45662943/Validation_45662943_45662943_23033.262.png", "Validation_45662943/Validation_45662943_45662943_23035.262.png", "Validation_45662943/Validation_45662943_45662943_23037.261.png" ], "turn_id": 2, "scan_id": "Validation_45662943", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the sleek white modern-design table with light wood surface\"; \"an elegant golden chandelier with the circular shape\"; \"the striking dark blue medium-weight curtain\" ", "option": [ "the sleek white modern-design table with light wood surface", "an elegant golden chandelier with the circular shape", "the striking dark blue medium-weight curtain" ], "answer": "an elegant golden chandelier with the circular shape", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the sleek white modern-design table with light wood surface\"; \"an elegant golden chandelier with the circular shape\"; \"the striking dark blue medium-weight curtain\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the sleek white modern-design table with light wood surface', 'an elegant golden chandelier with the circular shape', 'the striking dark blue medium-weight curtain'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_45662943/Validation_45662943_45662943_23039.260.png", "Validation_45662943/Validation_45662943_45662943_23041.259.png", "Validation_45662943/Validation_45662943_45662943_23043.258.png", "Validation_45662943/Validation_45662943_45662943_23045.257.png", "Validation_45662943/Validation_45662943_45662943_23047.256.png" ], "turn_id": 3, "scan_id": "Validation_45662943", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the modern grey upholstered bed with the padded headboard\", Y: \"the white rectangular interior door with the smooth finish\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to X and farther away from Y", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the modern grey upholstered bed with the padded headboard\", Y: \"the white rectangular interior door with the smooth finish\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_45662943/Validation_45662943_45662943_23049.256.png", "Validation_45662943/Validation_45662943_45662943_23051.255.png", "Validation_45662943/Validation_45662943_45662943_23053.254.png", "Validation_45662943/Validation_45662943_45662943_23055.253.png", "Validation_45662943/Validation_45662943_45662943_23057.269.png" ], "turn_id": 4, "scan_id": "Validation_45662943", "origin_question": "Which two objects are on the same side of you now? \"the striking dark blue medium-weight curtain\", \"the sleek black monitor with matte and glossy finishes\", \"the white rectangular interior door with the smooth finish\". ", "option": [ "the striking dark blue medium-weight curtain, the sleek black monitor with matte and glossy finishes", "the striking dark blue medium-weight curtain, the white rectangular interior door with the smooth finish", "the sleek black monitor with matte and glossy finishes, the white rectangular interior door with the smooth finish" ], "answer": "the striking dark blue medium-weight curtain, the sleek black monitor with matte and glossy finishes", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the striking dark blue medium-weight curtain\", \"the sleek black monitor with matte and glossy finishes\", \"the white rectangular interior door with the smooth finish\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the striking dark blue medium-weight curtain, the sleek black monitor with matte and glossy finishes', 'the striking dark blue medium-weight curtain, the white rectangular interior door with the smooth finish', 'the sleek black monitor with matte and glossy finishes, the white rectangular interior door with the smooth finish'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_45662943/Validation_45662943_45662943_23059.268.png", "Validation_45662943/Validation_45662943_45662943_23061.267.png", "Validation_45662943/Validation_45662943_45662943_23063.267.png", "Validation_45662943/Validation_45662943_45662943_23065.266.png", "Validation_45662943/Validation_45662943_45662943_23067.265.png" ], "turn_id": 5, "scan_id": "Validation_45662943", "origin_question": "By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular classic fireplace with the white exterior\", Y: \"the white rectangular interior door with the smooth finish\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to X and farther away from Y", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular classic fireplace with the white exterior\", Y: \"the white rectangular interior door with the smooth finish\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_45662943/Validation_45662943_45662943_23069.264.png", "Validation_45662943/Validation_45662943_45662943_23071.263.png", "Validation_45662943/Validation_45662943_45662943_23073.262.png", "Validation_45662943/Validation_45662943_45662943_23075.262.png", "Validation_45662943/Validation_45662943_45662943_23077.261.png" ], "turn_id": 6, "scan_id": "Validation_45662943", "origin_question": "Please recall \"the white rectangular interior door with the smooth finish\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.1", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the white rectangular interior door with the smooth finish\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_45662943/Validation_45662943_45662943_23079.260.png", "Validation_45662943/Validation_45662943_45662943_23081.259.png", "Validation_45662943/Validation_45662943_45662943_23083.258.png", "Validation_45662943/Validation_45662943_45662943_23085.257.png", "Validation_45662943/Validation_45662943_45662943_23087.256.png" ], "turn_id": 7, "scan_id": "Validation_45662943", "origin_question": "Which direction is an elegant golden chandelier with the circular shape in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is an elegant golden chandelier with the circular shape in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_45662943/Validation_45662943_45662943_23089.256.png", "Validation_45662943/Validation_45662943_45662943_23091.255.png", "Validation_45662943/Validation_45662943_45662943_23093.254.png", "Validation_45662943/Validation_45662943_45662943_23095.253.png", "Validation_45662943/Validation_45662943_45662943_23097.269.png" ], "turn_id": 8, "scan_id": "Validation_45662943", "origin_question": "How far is your current position from where you were at the end of the turn 1? (in meters)", "option": [], "answer": "3.05", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 1? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_45662943/Validation_45662943_45662943_23099.268.png", "Validation_45662943/Validation_45662943_45662943_23101.267.png", "Validation_45662943/Validation_45662943_45662943_23103.267.png", "Validation_45662943/Validation_45662943_45662943_23105.266.png", "Validation_45662943/Validation_45662943_45662943_23107.265.png" ], "turn_id": 9, "scan_id": "Validation_45662943", "origin_question": "Which direction is the striking dark blue medium-weight curtain in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the striking dark blue medium-weight curtain in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_45662943/Validation_45662943_45662943_23109.264.png", "Validation_45662943/Validation_45662943_45662943_23111.263.png", "Validation_45662943/Validation_45662943_45662943_23113.262.png", "Validation_45662943/Validation_45662943_45662943_23115.262.png", "Validation_45662943/Validation_45662943_45662943_23117.261.png" ], "turn_id": 10, "scan_id": "Validation_45662943", "origin_question": "Please recall \"the white rectangular interior door with the smooth finish\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "5.29", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the white rectangular interior door with the smooth finish\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47895776/Validation_47895776_47895776_386555.692.png", "Validation_47895776/Validation_47895776_47895776_386557.175.png", "Validation_47895776/Validation_47895776_47895776_386558.674.png", "Validation_47895776/Validation_47895776_47895776_386560.173.png" ], "turn_id": 1, "scan_id": "Validation_47895776", "origin_question": "Remember, have you seen any wardrobe(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any wardrobe(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_47895776/Validation_47895776_47895776_386561.673.png", "Validation_47895776/Validation_47895776_47895776_386563.172.png", "Validation_47895776/Validation_47895776_47895776_386564.672.png", "Validation_47895776/Validation_47895776_47895776_386566.171.png" ], "turn_id": 2, "scan_id": "Validation_47895776", "origin_question": "In which turn did you see both \"the clothes\" and \"the sleek, black wall-mounted monitor with the thin bezel\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the clothes\" and \"the sleek, black wall-mounted monitor with the thin bezel\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_47895776/Validation_47895776_47895776_386567.671.png", "Validation_47895776/Validation_47895776_47895776_386569.170.png", "Validation_47895776/Validation_47895776_47895776_386570.669.png", "Validation_47895776/Validation_47895776_47895776_386572.169.png" ], "turn_id": 3, "scan_id": "Validation_47895776", "origin_question": "What will be the first-time appearance order of the following categories: bed, clothes, monitor?", "option": [ "monitor, bed, clothes", "clothes, monitor, bed", "bed, monitor, clothes", "clothes, bed, monitor" ], "answer": "clothes, monitor, bed", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bed, clothes, monitor? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['monitor, bed, clothes', 'clothes, monitor, bed', 'bed, monitor, clothes', 'clothes, bed, monitor'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47895776/Validation_47895776_47895776_386573.668.png", "Validation_47895776/Validation_47895776_47895776_386575.167.png", "Validation_47895776/Validation_47895776_47895776_386576.667.png", "Validation_47895776/Validation_47895776_47895776_386578.166.png" ], "turn_id": 4, "scan_id": "Validation_47895776", "origin_question": "Which direction is the rectangular white cotton pillow with the simple design in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the rectangular white cotton pillow with the simple design in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_47895776/Validation_47895776_47895776_386579.666.png", "Validation_47895776/Validation_47895776_47895776_386581.165.png", "Validation_47895776/Validation_47895776_47895776_386582.664.png", "Validation_47895776/Validation_47895776_47895776_386584.164.png" ], "turn_id": 5, "scan_id": "Validation_47895776", "origin_question": "Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_47895776/Validation_47895776_47895776_386585.663.png", "Validation_47895776/Validation_47895776_47895776_386587.162.png", "Validation_47895776/Validation_47895776_47895776_386588.662.png", "Validation_47895776/Validation_47895776_47895776_386590.161.png" ], "turn_id": 6, "scan_id": "Validation_47895776", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "57.01", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_47895776/Validation_47895776_47895776_386591.677.png", "Validation_47895776/Validation_47895776_47895776_386593.177.png", "Validation_47895776/Validation_47895776_47895776_386594.676.png", "Validation_47895776/Validation_47895776_47895776_386596.176.png" ], "turn_id": 7, "scan_id": "Validation_47895776", "origin_question": "Which two objects are on the same side of you now? \"the bed\", \"the rectangular white cotton pillow with the simple design\", \"the book\". ", "option": [ "the bed, the rectangular white cotton pillow with the simple design", "the bed, the book", "the rectangular white cotton pillow with the simple design, the book" ], "answer": "the bed, the rectangular white cotton pillow with the simple design", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the bed\", \"the rectangular white cotton pillow with the simple design\", \"the book\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the bed, the rectangular white cotton pillow with the simple design', 'the bed, the book', 'the rectangular white cotton pillow with the simple design, the book'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "Validation_47895776/Validation_47895776_47895776_386597.675.png", "Validation_47895776/Validation_47895776_47895776_386599.174.png", "Validation_47895776/Validation_47895776_47895776_386600.674.png", "Validation_47895776/Validation_47895776_47895776_386602.173.png" ], "turn_id": 8, "scan_id": "Validation_47895776", "origin_question": "Among these three objects(\"the bed\",\"the rectangular white cotton pillow with the simple design\",\"the microwave\" ), which one is the farthest from you?", "option": [ "the bed", "the rectangular white cotton pillow with the simple design", "the microwave" ], "answer": "the microwave", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects(\"the bed\",\"the rectangular white cotton pillow with the simple design\",\"the microwave\" ), which one is the farthest from you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the bed', 'the rectangular white cotton pillow with the simple design', 'the microwave'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42446532/Validation_42446532_42446532_198808.960.png", "Validation_42446532/Validation_42446532_42446532_198810.942.png", "Validation_42446532/Validation_42446532_42446532_198812.941.png", "Validation_42446532/Validation_42446532_42446532_198814.940.png", "Validation_42446532/Validation_42446532_42446532_198816.940.png" ], "turn_id": 1, "scan_id": "Validation_42446532", "origin_question": "Remember, have you seen any shelf(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any shelf(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_42446532/Validation_42446532_42446532_198818.939.png", "Validation_42446532/Validation_42446532_42446532_198820.938.png", "Validation_42446532/Validation_42446532_42446532_198822.937.png", "Validation_42446532/Validation_42446532_42446532_198824.936.png", "Validation_42446532/Validation_42446532_42446532_198826.935.png" ], "turn_id": 2, "scan_id": "Validation_42446532", "origin_question": "At the end of which turn were both of \"the rectangular light-colored pillow with gray-blue patterns\" and \"the light-colored wooden door with the sleek design\" on your right side?", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the rectangular light-colored pillow with gray-blue patterns\" and \"the light-colored wooden door with the sleek design\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_42446532/Validation_42446532_42446532_198828.935.png", "Validation_42446532/Validation_42446532_42446532_198830.934.png", "Validation_42446532/Validation_42446532_42446532_198832.950.png", "Validation_42446532/Validation_42446532_42446532_198834.949.png", "Validation_42446532/Validation_42446532_42446532_198836.948.png" ], "turn_id": 3, "scan_id": "Validation_42446532", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "75.02", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_42446532/Validation_42446532_42446532_198838.947.png", "Validation_42446532/Validation_42446532_42446532_198840.946.png", "Validation_42446532/Validation_42446532_42446532_198842.946.png", "Validation_42446532/Validation_42446532_42446532_198844.945.png", "Validation_42446532/Validation_42446532_42446532_198846.944.png" ], "turn_id": 4, "scan_id": "Validation_42446532", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the light-colored wooden door with the sleek design\" from your position?", "option": [], "answer": "111.9", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the light-colored wooden door with the sleek design\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_42446532/Validation_42446532_42446532_198848.943.png", "Validation_42446532/Validation_42446532_42446532_198850.942.png", "Validation_42446532/Validation_42446532_42446532_198852.941.png", "Validation_42446532/Validation_42446532_42446532_198854.941.png", "Validation_42446532/Validation_42446532_42446532_198856.940.png" ], "turn_id": 5, "scan_id": "Validation_42446532", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the sleek white steel radiator with moderate size\" from your position?", "option": [], "answer": "92.03", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the sleek white steel radiator with moderate size\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "Validation_42446532/Validation_42446532_42446532_198858.939.png", "Validation_42446532/Validation_42446532_42446532_198860.938.png", "Validation_42446532/Validation_42446532_42446532_198862.937.png", "Validation_42446532/Validation_42446532_42446532_198864.937.png", "Validation_42446532/Validation_42446532_42446532_198866.936.png" ], "turn_id": 6, "scan_id": "Validation_42446532", "origin_question": "Among these three objects(\"the modern chair with the circular seat and low backrest\",\"the rectangular window with the white frame and clear glass\",\"the sleek white steel radiator with moderate size\" ), which one is the closest to you?", "option": [ "the modern chair with the circular seat and low backrest", "the rectangular window with the white frame and clear glass", "the sleek white steel radiator with moderate size" ], "answer": "the modern chair with the circular seat and low backrest", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the modern chair with the circular seat and low backrest\",\"the rectangular window with the white frame and clear glass\",\"the sleek white steel radiator with moderate size\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the modern chair with the circular seat and low backrest', 'the rectangular window with the white frame and clear glass', 'the sleek white steel radiator with moderate size'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_42446532/Validation_42446532_42446532_198868.935.png", "Validation_42446532/Validation_42446532_42446532_198870.934.png", "Validation_42446532/Validation_42446532_42446532_198872.950.png", "Validation_42446532/Validation_42446532_42446532_198874.949.png", "Validation_42446532/Validation_42446532_42446532_198876.948.png" ], "turn_id": 7, "scan_id": "Validation_42446532", "origin_question": "Which direction is the compact rectangular light wood table in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the compact rectangular light wood table in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "Validation_42446532/Validation_42446532_42446532_198878.947.png", "Validation_42446532/Validation_42446532_42446532_198880.947.png", "Validation_42446532/Validation_42446532_42446532_198882.946.png", "Validation_42446532/Validation_42446532_42446532_198884.945.png", "Validation_42446532/Validation_42446532_42446532_198886.944.png" ], "turn_id": 8, "scan_id": "Validation_42446532", "origin_question": "Among these three objects(\"the shelf\",\"the modern chair with the circular seat and low backrest\",\"the light-colored wooden door with the sleek design\" ), which one is the farthest from you?", "option": [ "the shelf", "the modern chair with the circular seat and low backrest", "the light-colored wooden door with the sleek design" ], "answer": "the modern chair with the circular seat and low backrest", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the shelf\",\"the modern chair with the circular seat and low backrest\",\"the light-colored wooden door with the sleek design\" ), which one is the farthest from you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the shelf', 'the modern chair with the circular seat and low backrest', 'the light-colored wooden door with the sleek design'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "Validation_42446532/Validation_42446532_42446532_198888.943.png", "Validation_42446532/Validation_42446532_42446532_198890.942.png", "Validation_42446532/Validation_42446532_42446532_198892.942.png", "Validation_42446532/Validation_42446532_42446532_198894.941.png", "Validation_42446532/Validation_42446532_42446532_198896.940.png" ], "turn_id": 9, "scan_id": "Validation_42446532", "origin_question": "Among these three objects(\"the modern chair with the circular seat and low backrest\",\"the light-colored wooden door with the sleek design\",\"the foldable metallic rack with vertical bars\" ), which one is the closest to you?", "option": [ "the modern chair with the circular seat and low backrest", "the light-colored wooden door with the sleek design", "the foldable metallic rack with vertical bars" ], "answer": "the foldable metallic rack with vertical bars", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the modern chair with the circular seat and low backrest\",\"the light-colored wooden door with the sleek design\",\"the foldable metallic rack with vertical bars\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the modern chair with the circular seat and low backrest', 'the light-colored wooden door with the sleek design', 'the foldable metallic rack with vertical bars'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_42446532/Validation_42446532_42446532_198898.939.png", "Validation_42446532/Validation_42446532_42446532_198900.938.png", "Validation_42446532/Validation_42446532_42446532_198902.938.png", "Validation_42446532/Validation_42446532_42446532_198904.937.png", "Validation_42446532/Validation_42446532_42446532_198906.936.png" ], "turn_id": 10, "scan_id": "Validation_42446532", "origin_question": "Please recall \"the modern chair with the circular seat and low backrest\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "4.22", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the modern chair with the circular seat and low backrest\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47332910/Validation_47332910_47332910_12187.329.png", "Validation_47332910/Validation_47332910_47332910_12189.328.png", "Validation_47332910/Validation_47332910_47332910_12191.328.png", "Validation_47332910/Validation_47332910_47332910_12193.327.png" ], "turn_id": 1, "scan_id": "Validation_47332910", "origin_question": "Remember, have you seen any door(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any door(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_47332910/Validation_47332910_47332910_12195.326.png", "Validation_47332910/Validation_47332910_47332910_12197.325.png", "Validation_47332910/Validation_47332910_47332910_12199.325.png", "Validation_47332910/Validation_47332910_47332910_12201.324.png" ], "turn_id": 2, "scan_id": "Validation_47332910", "origin_question": "When did you first discover \"the sleek white top-loading washing machine\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the sleek white top-loading washing machine\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47332910/Validation_47332910_47332910_12203.323.png", "Validation_47332910/Validation_47332910_47332910_12205.339.png", "Validation_47332910/Validation_47332910_47332910_12207.338.png", "Validation_47332910/Validation_47332910_47332910_12209.337.png" ], "turn_id": 3, "scan_id": "Validation_47332910", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the sleek white top-loading washing machine\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the sleek white top-loading washing machine\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_47332910/Validation_47332910_47332910_12211.336.png", "Validation_47332910/Validation_47332910_47332910_12213.335.png", "Validation_47332910/Validation_47332910_47332910_12215.335.png", "Validation_47332910/Validation_47332910_47332910_12217.334.png" ], "turn_id": 4, "scan_id": "Validation_47332910", "origin_question": "In which turn did you see both \"the bathtub\" and \"the colorful elongated cylindrical body loofah with abstract patterns\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the bathtub\" and \"the colorful elongated cylindrical body loofah with abstract patterns\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47332910/Validation_47332910_47332910_12219.333.png", "Validation_47332910/Validation_47332910_47332910_12221.332.png", "Validation_47332910/Validation_47332910_47332910_12223.331.png", "Validation_47332910/Validation_47332910_47332910_12225.331.png" ], "turn_id": 5, "scan_id": "Validation_47332910", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the sleek white top-loading washing machine\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the sleek white top-loading washing machine\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47332910/Validation_47332910_47332910_12227.330.png", "Validation_47332910/Validation_47332910_47332910_12229.329.png", "Validation_47332910/Validation_47332910_47332910_12231.328.png", "Validation_47332910/Validation_47332910_47332910_12233.327.png" ], "turn_id": 6, "scan_id": "Validation_47332910", "origin_question": "", "option": [], "answer": "", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_47332910/Validation_47332910_47332910_12235.326.png", "Validation_47332910/Validation_47332910_47332910_12237.326.png", "Validation_47332910/Validation_47332910_47332910_12239.325.png", "Validation_47332910/Validation_47332910_47332910_12241.324.png" ], "turn_id": 7, "scan_id": "Validation_47332910", "origin_question": "By the end of the 6 turn, were you getting closer to or farther away from these two objects: X: \"the bathtub\", Y: \"the sleek white top-loading washing machine\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "farther away from both X and Y", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 6 turn, were you getting closer to or farther away from these two objects: X: \"the bathtub\", Y: \"the sleek white top-loading washing machine\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47332910/Validation_47332910_47332910_12243.323.png", "Validation_47332910/Validation_47332910_47332910_12245.339.png", "Validation_47332910/Validation_47332910_47332910_12247.338.png", "Validation_47332910/Validation_47332910_47332910_12249.337.png" ], "turn_id": 8, "scan_id": "Validation_47332910", "origin_question": "Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the sleek white top-loading washing machine\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the sleek white top-loading washing machine\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47115525/Validation_47115525_47115525_14480.812.png", "Validation_47115525/Validation_47115525_47115525_14482.295.png", "Validation_47115525/Validation_47115525_47115525_14483.794.png", "Validation_47115525/Validation_47115525_47115525_14485.294.png" ], "turn_id": 1, "scan_id": "Validation_47115525", "origin_question": "Remember, have you seen any fireplace(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any fireplace(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47115525/Validation_47115525_47115525_14486.793.png", "Validation_47115525/Validation_47115525_47115525_14488.292.png", "Validation_47115525/Validation_47115525_47115525_14489.792.png", "Validation_47115525/Validation_47115525_47115525_14491.291.png" ], "turn_id": 2, "scan_id": "Validation_47115525", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_47115525/Validation_47115525_47115525_14492.791.png", "Validation_47115525/Validation_47115525_47115525_14494.290.png", "Validation_47115525/Validation_47115525_47115525_14495.790.png", "Validation_47115525/Validation_47115525_47115525_14497.289.png" ], "turn_id": 3, "scan_id": "Validation_47115525", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the rustic medium-brown wooden table with thick legs\", Y: \"the sleek black glossy monitor with the narrow bezel\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the rustic medium-brown wooden table with thick legs\", Y: \"the sleek black glossy monitor with the narrow bezel\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_47115525/Validation_47115525_47115525_14498.788.png", "Validation_47115525/Validation_47115525_47115525_14500.288.png", "Validation_47115525/Validation_47115525_47115525_14501.787.png", "Validation_47115525/Validation_47115525_47115525_14503.287.png" ], "turn_id": 4, "scan_id": "Validation_47115525", "origin_question": "At the end of which turn were both of \"the rectangular metal and stone fireplace\" and \"the rustic medium-brown wooden table with thick legs\" on your left side?", "option": [], "answer": "2", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the rectangular metal and stone fireplace\" and \"the rustic medium-brown wooden table with thick legs\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_47115525/Validation_47115525_47115525_14504.786.png", "Validation_47115525/Validation_47115525_47115525_14506.285.png", "Validation_47115525/Validation_47115525_47115525_14507.785.png", "Validation_47115525/Validation_47115525_47115525_14509.284.png" ], "turn_id": 5, "scan_id": "Validation_47115525", "origin_question": "Is the rectangular metal and stone fireplace to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the rectangular metal and stone fireplace to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_47115525/Validation_47115525_47115525_14510.783.png", "Validation_47115525/Validation_47115525_47115525_14512.283.png", "Validation_47115525/Validation_47115525_47115525_14513.782.png", "Validation_47115525/Validation_47115525_47115525_14515.282.png" ], "turn_id": 6, "scan_id": "Validation_47115525", "origin_question": "Which two objects are on the same side of you now? \"the rectangular metal and stone fireplace\", \"the medium-sized green ceramic flowerpot with the glossy finish\", \"the rustic medium-brown wooden table with thick legs\". ", "option": [ "the rectangular metal and stone fireplace, the medium-sized green ceramic flowerpot with the glossy finish", "the rectangular metal and stone fireplace, the rustic medium-brown wooden table with thick legs", "the medium-sized green ceramic flowerpot with the glossy finish, the rustic medium-brown wooden table with thick legs" ], "answer": "the rectangular metal and stone fireplace, the medium-sized green ceramic flowerpot with the glossy finish", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the rectangular metal and stone fireplace\", \"the medium-sized green ceramic flowerpot with the glossy finish\", \"the rustic medium-brown wooden table with thick legs\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular metal and stone fireplace, the medium-sized green ceramic flowerpot with the glossy finish', 'the rectangular metal and stone fireplace, the rustic medium-brown wooden table with thick legs', 'the medium-sized green ceramic flowerpot with the glossy finish, the rustic medium-brown wooden table with thick legs'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47115525/Validation_47115525_47115525_14516.798.png", "Validation_47115525/Validation_47115525_47115525_14518.297.png", "Validation_47115525/Validation_47115525_47115525_14519.797.png", "Validation_47115525/Validation_47115525_47115525_14521.296.png" ], "turn_id": 7, "scan_id": "Validation_47115525", "origin_question": "In which turn were you closest to \"the sleek black glossy monitor with the narrow bezel\"? ", "option": [], "answer": "7", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the sleek black glossy monitor with the narrow bezel\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "Validation_42898868/Validation_42898868_42898868_197646.853.png", "Validation_42898868/Validation_42898868_42898868_197648.835.png", "Validation_42898868/Validation_42898868_42898868_197650.835.png", "Validation_42898868/Validation_42898868_42898868_197652.834.png" ], "turn_id": 1, "scan_id": "Validation_42898868", "origin_question": "Remember, how many computer(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many computer(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_42898868/Validation_42898868_42898868_197654.833.png", "Validation_42898868/Validation_42898868_42898868_197656.832.png", "Validation_42898868/Validation_42898868_42898868_197658.832.png", "Validation_42898868/Validation_42898868_42898868_197660.831.png" ], "turn_id": 2, "scan_id": "Validation_42898868", "origin_question": "In which turn were you closest to \"the spherical glass light emitting warm white light\"? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the spherical glass light emitting warm white light\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_42898868/Validation_42898868_42898868_197662.830.png", "Validation_42898868/Validation_42898868_42898868_197664.829.png", "Validation_42898868/Validation_42898868_42898868_197666.828.png", "Validation_42898868/Validation_42898868_42898868_197668.828.png" ], "turn_id": 3, "scan_id": "Validation_42898868", "origin_question": "When did you first discover \"the medium to large sleek monitor with the thin black frame\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the medium to large sleek monitor with the thin black frame\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_42898868/Validation_42898868_42898868_197670.827.png", "Validation_42898868/Validation_42898868_42898868_197672.826.png", "Validation_42898868/Validation_42898868_42898868_197674.825.png", "Validation_42898868/Validation_42898868_42898868_197676.824.png" ], "turn_id": 4, "scan_id": "Validation_42898868", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_42898868/Validation_42898868_42898868_197678.824.png", "Validation_42898868/Validation_42898868_42898868_197680.839.png", "Validation_42898868/Validation_42898868_42898868_197682.839.png", "Validation_42898868/Validation_42898868_42898868_197684.838.png" ], "turn_id": 5, "scan_id": "Validation_42898868", "origin_question": "Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "128.19", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_42898868/Validation_42898868_42898868_197686.837.png", "Validation_42898868/Validation_42898868_42898868_197688.836.png", "Validation_42898868/Validation_42898868_42898868_197690.835.png", "Validation_42898868/Validation_42898868_42898868_197692.835.png" ], "turn_id": 6, "scan_id": "Validation_42898868", "origin_question": "Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "30.41", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47334093/Validation_47334093_47334093_91840.632.png", "Validation_47334093/Validation_47334093_47334093_91842.631.png", "Validation_47334093/Validation_47334093_47334093_91844.630.png", "Validation_47334093/Validation_47334093_47334093_91846.630.png", "Validation_47334093/Validation_47334093_47334093_91848.629.png" ], "turn_id": 1, "scan_id": "Validation_47334093", "origin_question": "", "option": [], "answer": "", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_47334093/Validation_47334093_47334093_91850.628.png", "Validation_47334093/Validation_47334093_47334093_91852.627.png", "Validation_47334093/Validation_47334093_47334093_91854.626.png", "Validation_47334093/Validation_47334093_47334093_91856.626.png", "Validation_47334093/Validation_47334093_47334093_91858.625.png" ], "turn_id": 2, "scan_id": "Validation_47334093", "origin_question": "When did you first discover \"the rectangular, tufted brown leather stool\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the rectangular, tufted brown leather stool\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47334093/Validation_47334093_47334093_91860.624.png", "Validation_47334093/Validation_47334093_47334093_91862.640.png", "Validation_47334093/Validation_47334093_47334093_91864.639.png", "Validation_47334093/Validation_47334093_47334093_91866.638.png", "Validation_47334093/Validation_47334093_47334093_91868.637.png" ], "turn_id": 3, "scan_id": "Validation_47334093", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the stylish stainless steel and glass oven\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the stylish stainless steel and glass oven\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_47334093/Validation_47334093_47334093_91870.637.png", "Validation_47334093/Validation_47334093_47334093_91872.636.png", "Validation_47334093/Validation_47334093_47334093_91874.635.png", "Validation_47334093/Validation_47334093_47334093_91876.634.png", "Validation_47334093/Validation_47334093_47334093_91878.633.png" ], "turn_id": 4, "scan_id": "Validation_47334093", "origin_question": "Please recall \"the matte black, sleek rectangular monitor\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "4.01", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the matte black, sleek rectangular monitor\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47334093/Validation_47334093_47334093_91880.633.png", "Validation_47334093/Validation_47334093_47334093_91882.632.png", "Validation_47334093/Validation_47334093_47334093_91884.631.png", "Validation_47334093/Validation_47334093_47334093_91886.630.png", "Validation_47334093/Validation_47334093_47334093_91888.629.png" ], "turn_id": 5, "scan_id": "Validation_47334093", "origin_question": "", "option": [], "answer": "", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47334093/Validation_47334093_47334093_91890.629.png", "Validation_47334093/Validation_47334093_47334093_91892.628.png", "Validation_47334093/Validation_47334093_47334093_91894.627.png", "Validation_47334093/Validation_47334093_47334093_91896.626.png", "Validation_47334093/Validation_47334093_47334093_91898.625.png" ], "turn_id": 6, "scan_id": "Validation_47334093", "origin_question": "Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the stylish stainless steel and glass oven\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the stylish stainless steel and glass oven\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "Validation_47334093/Validation_47334093_47334093_91900.625.png", "Validation_47334093/Validation_47334093_47334093_91902.624.png", "Validation_47334093/Validation_47334093_47334093_91904.640.png", "Validation_47334093/Validation_47334093_47334093_91906.639.png", "Validation_47334093/Validation_47334093_47334093_91908.638.png" ], "turn_id": 7, "scan_id": "Validation_47334093", "origin_question": "Among these three objects(\"the rectangular, tufted brown leather stool\",\"the stylish stainless steel and glass oven\",\"the modern, compact couch with the neutral color palette\" ), which one is the closest to you?", "option": [ "the rectangular, tufted brown leather stool", "the stylish stainless steel and glass oven", "the modern, compact couch with the neutral color palette" ], "answer": "the stylish stainless steel and glass oven", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the rectangular, tufted brown leather stool\",\"the stylish stainless steel and glass oven\",\"the modern, compact couch with the neutral color palette\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular, tufted brown leather stool', 'the stylish stainless steel and glass oven', 'the modern, compact couch with the neutral color palette'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_47334093/Validation_47334093_47334093_91910.637.png", "Validation_47334093/Validation_47334093_47334093_91912.636.png", "Validation_47334093/Validation_47334093_47334093_91914.636.png", "Validation_47334093/Validation_47334093_47334093_91916.635.png", "Validation_47334093/Validation_47334093_47334093_91918.634.png" ], "turn_id": 8, "scan_id": "Validation_47334093", "origin_question": "Please recall \"the rectangular window with the neutral white frame\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.06", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the rectangular window with the neutral white frame\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_47334093/Validation_47334093_47334093_91920.633.png", "Validation_47334093/Validation_47334093_47334093_91922.632.png", "Validation_47334093/Validation_47334093_47334093_91924.632.png", "Validation_47334093/Validation_47334093_47334093_91926.631.png", "Validation_47334093/Validation_47334093_47334093_91928.630.png" ], "turn_id": 9, "scan_id": "Validation_47334093", "origin_question": "By the end of the 5 turn, were you getting closer to or farther away from these two objects: X: \"the stylish stainless steel and glass oven\", Y: \"the rectangular window with the neutral white frame\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to X and farther away from Y", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 5 turn, were you getting closer to or farther away from these two objects: X: \"the stylish stainless steel and glass oven\", Y: \"the rectangular window with the neutral white frame\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_47334093/Validation_47334093_47334093_91930.629.png", "Validation_47334093/Validation_47334093_47334093_91932.628.png", "Validation_47334093/Validation_47334093_47334093_91934.628.png", "Validation_47334093/Validation_47334093_47334093_91936.627.png", "Validation_47334093/Validation_47334093_47334093_91938.626.png" ], "turn_id": 10, "scan_id": "Validation_47334093", "origin_question": "When did you first discover \"the rectangular window with the neutral white frame\"? (index of the turn)", "option": [], "answer": "5", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the rectangular window with the neutral white frame\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42898581/Validation_42898581_42898581_753062.380.png", "Validation_42898581/Validation_42898581_42898581_753064.379.png", "Validation_42898581/Validation_42898581_42898581_753066.378.png", "Validation_42898581/Validation_42898581_42898581_753068.377.png", "Validation_42898581/Validation_42898581_42898581_753070.376.png" ], "turn_id": 1, "scan_id": "Validation_42898581", "origin_question": "Remember, have you seen any bed(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any bed(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_42898581/Validation_42898581_42898581_753072.375.png", "Validation_42898581/Validation_42898581_42898581_753074.375.png", "Validation_42898581/Validation_42898581_42898581_753076.374.png", "Validation_42898581/Validation_42898581_42898581_753078.373.png", "Validation_42898581/Validation_42898581_42898581_753080.372.png" ], "turn_id": 2, "scan_id": "Validation_42898581", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_42898581/Validation_42898581_42898581_753082.371.png", "Validation_42898581/Validation_42898581_42898581_753084.370.png", "Validation_42898581/Validation_42898581_42898581_753086.370.png", "Validation_42898581/Validation_42898581_42898581_753088.369.png", "Validation_42898581/Validation_42898581_42898581_753090.385.png" ], "turn_id": 3, "scan_id": "Validation_42898581", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_42898581/Validation_42898581_42898581_753092.384.png", "Validation_42898581/Validation_42898581_42898581_753094.383.png", "Validation_42898581/Validation_42898581_42898581_753096.382.png", "Validation_42898581/Validation_42898581_42898581_753098.381.png", "Validation_42898581/Validation_42898581_42898581_753100.381.png" ], "turn_id": 4, "scan_id": "Validation_42898581", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "2.14", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_42898581/Validation_42898581_42898581_753102.380.png", "Validation_42898581/Validation_42898581_42898581_753104.379.png", "Validation_42898581/Validation_42898581_42898581_753106.378.png", "Validation_42898581/Validation_42898581_42898581_753108.377.png", "Validation_42898581/Validation_42898581_42898581_753110.377.png" ], "turn_id": 5, "scan_id": "Validation_42898581", "origin_question": "Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Left", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_42898581/Validation_42898581_42898581_753112.376.png", "Validation_42898581/Validation_42898581_42898581_753114.375.png", "Validation_42898581/Validation_42898581_42898581_753116.374.png", "Validation_42898581/Validation_42898581_42898581_753118.373.png", "Validation_42898581/Validation_42898581_42898581_753120.372.png" ], "turn_id": 6, "scan_id": "Validation_42898581", "origin_question": "When did you first discover \"the classic wooden bed with the patterned bedspread\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the classic wooden bed with the patterned bedspread\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_42898581/Validation_42898581_42898581_753122.372.png", "Validation_42898581/Validation_42898581_42898581_753124.371.png", "Validation_42898581/Validation_42898581_42898581_753126.370.png", "Validation_42898581/Validation_42898581_42898581_753128.369.png", "Validation_42898581/Validation_42898581_42898581_753130.368.png" ], "turn_id": 7, "scan_id": "Validation_42898581", "origin_question": "Using your orientation at the end of turn 6 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 6 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_42898581/Validation_42898581_42898581_753132.384.png", "Validation_42898581/Validation_42898581_42898581_753134.383.png", "Validation_42898581/Validation_42898581_42898581_753136.382.png", "Validation_42898581/Validation_42898581_42898581_753138.382.png", "Validation_42898581/Validation_42898581_42898581_753140.381.png" ], "turn_id": 8, "scan_id": "Validation_42898581", "origin_question": "In which turn were you closest to \"the classic rectangular table with the light wood finish\"? ", "option": [], "answer": "8", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the classic rectangular table with the light wood finish\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_42898581/Validation_42898581_42898581_753142.380.png", "Validation_42898581/Validation_42898581_42898581_753144.379.png", "Validation_42898581/Validation_42898581_42898581_753146.378.png", "Validation_42898581/Validation_42898581_42898581_753148.377.png", "Validation_42898581/Validation_42898581_42898581_753150.377.png" ], "turn_id": 9, "scan_id": "Validation_42898581", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the classic rectangular table with the light wood finish\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the classic rectangular table with the light wood finish\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_42898581/Validation_42898581_42898581_753152.376.png", "Validation_42898581/Validation_42898581_42898581_753154.375.png", "Validation_42898581/Validation_42898581_42898581_753156.374.png", "Validation_42898581/Validation_42898581_42898581_753158.373.png", "Validation_42898581/Validation_42898581_42898581_753160.373.png" ], "turn_id": 10, "scan_id": "Validation_42898581", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the wooden stool with tapered legs and the black leather cushion\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the wooden stool with tapered legs and the black leather cushion\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42899723/Validation_42899723_42899723_797672.699.png", "Validation_42899723/Validation_42899723_42899723_797674.682.png", "Validation_42899723/Validation_42899723_42899723_797676.681.png", "Validation_42899723/Validation_42899723_42899723_797678.680.png" ], "turn_id": 1, "scan_id": "Validation_42899723", "origin_question": "Remember, have you seen any window(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any window(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_42899723/Validation_42899723_42899723_797680.679.png", "Validation_42899723/Validation_42899723_42899723_797682.678.png", "Validation_42899723/Validation_42899723_42899723_797684.677.png", "Validation_42899723/Validation_42899723_42899723_797686.677.png" ], "turn_id": 2, "scan_id": "Validation_42899723", "origin_question": "When did you first discover \"the modern oval white toilet made of ceramic\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the modern oval white toilet made of ceramic\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_42899723/Validation_42899723_42899723_797688.676.png", "Validation_42899723/Validation_42899723_42899723_797690.692.png", "Validation_42899723/Validation_42899723_42899723_797692.691.png", "Validation_42899723/Validation_42899723_42899723_797694.690.png" ], "turn_id": 3, "scan_id": "Validation_42899723", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_42899723/Validation_42899723_42899723_797696.689.png", "Validation_42899723/Validation_42899723_42899723_797698.688.png", "Validation_42899723/Validation_42899723_42899723_797700.688.png", "Validation_42899723/Validation_42899723_42899723_797702.687.png" ], "turn_id": 4, "scan_id": "Validation_42899723", "origin_question": "At the end of which turn were both of \"the compact silver metallic bin with the foot pedal\" and \"the modern rectangular white ceramic sink\" on your right side?", "option": [], "answer": "1", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the compact silver metallic bin with the foot pedal\" and \"the modern rectangular white ceramic sink\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_42899723/Validation_42899723_42899723_797704.686.png", "Validation_42899723/Validation_42899723_42899723_797706.685.png", "Validation_42899723/Validation_42899723_42899723_797708.684.png", "Validation_42899723/Validation_42899723_42899723_797710.683.png" ], "turn_id": 5, "scan_id": "Validation_42899723", "origin_question": "What will be the first-time appearance order of the following categories: bin, door, sink?", "option": [ "sink, door, bin", "door, sink, bin", "door, bin, sink", "bin, sink, door" ], "answer": "bin, sink, door", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bin, door, sink? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['sink, door, bin', 'door, sink, bin', 'door, bin, sink', 'bin, sink, door'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_42899723/Validation_42899723_42899723_797712.683.png", "Validation_42899723/Validation_42899723_42899723_797714.682.png", "Validation_42899723/Validation_42899723_42899723_797716.681.png", "Validation_42899723/Validation_42899723_42899723_797718.680.png" ], "turn_id": 6, "scan_id": "Validation_42899723", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the modern rectangular white ceramic sink\" from your position?", "option": [], "answer": "109.09", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the modern rectangular white ceramic sink\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_42899723/Validation_42899723_42899723_797720.679.png", "Validation_42899723/Validation_42899723_42899723_797722.679.png", "Validation_42899723/Validation_42899723_42899723_797724.678.png", "Validation_42899723/Validation_42899723_42899723_797726.677.png" ], "turn_id": 7, "scan_id": "Validation_42899723", "origin_question": "When did you first discover \"the modern rectangular white ceramic sink\"? (index of the turn)", "option": [], "answer": "4", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the modern rectangular white ceramic sink\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_42899723/Validation_42899723_42899723_797728.676.png", "Validation_42899723/Validation_42899723_42899723_797730.675.png", "Validation_42899723/Validation_42899723_42899723_797732.674.png", "Validation_42899723/Validation_42899723_42899723_797734.690.png" ], "turn_id": 8, "scan_id": "Validation_42899723", "origin_question": "When did you first discover \"the compact silver metallic bin with the foot pedal\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the compact silver metallic bin with the foot pedal\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42445984/Validation_42445984_42445984_75145.654.png", "Validation_42445984/Validation_42445984_42445984_75147.636.png", "Validation_42445984/Validation_42445984_42445984_75149.636.png", "Validation_42445984/Validation_42445984_42445984_75151.635.png", "Validation_42445984/Validation_42445984_42445984_75153.634.png" ], "turn_id": 1, "scan_id": "Validation_42445984", "origin_question": "Remember, have you seen any cabinet(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any cabinet(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_42445984/Validation_42445984_42445984_75155.633.png", "Validation_42445984/Validation_42445984_42445984_75157.632.png", "Validation_42445984/Validation_42445984_42445984_75159.631.png", "Validation_42445984/Validation_42445984_42445984_75161.630.png", "Validation_42445984/Validation_42445984_42445984_75163.630.png" ], "turn_id": 2, "scan_id": "Validation_42445984", "origin_question": "Please recall \"the sturdy dark fireplace made of brick or stone\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.17", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the sturdy dark fireplace made of brick or stone\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_42445984/Validation_42445984_42445984_75165.629.png", "Validation_42445984/Validation_42445984_42445984_75167.628.png", "Validation_42445984/Validation_42445984_42445984_75169.627.png", "Validation_42445984/Validation_42445984_42445984_75171.626.png", "Validation_42445984/Validation_42445984_42445984_75173.625.png" ], "turn_id": 3, "scan_id": "Validation_42445984", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_42445984/Validation_42445984_42445984_75175.625.png", "Validation_42445984/Validation_42445984_42445984_75177.624.png", "Validation_42445984/Validation_42445984_42445984_75179.623.png", "Validation_42445984/Validation_42445984_42445984_75181.639.png", "Validation_42445984/Validation_42445984_42445984_75183.638.png" ], "turn_id": 4, "scan_id": "Validation_42445984", "origin_question": "Which direction is the window in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the window in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "Validation_42445984/Validation_42445984_42445984_75185.637.png", "Validation_42445984/Validation_42445984_42445984_75187.636.png", "Validation_42445984/Validation_42445984_42445984_75189.636.png", "Validation_42445984/Validation_42445984_42445984_75191.635.png", "Validation_42445984/Validation_42445984_42445984_75193.634.png" ], "turn_id": 5, "scan_id": "Validation_42445984", "origin_question": "Among these three objects(\"the sturdy dark fireplace made of brick or stone\",\"the vivid yellow book with colorful illustrations\",\"the window\" ), which one is the closest to you?", "option": [ "the sturdy dark fireplace made of brick or stone", "the vivid yellow book with colorful illustrations", "the window" ], "answer": "the vivid yellow book with colorful illustrations", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the sturdy dark fireplace made of brick or stone\",\"the vivid yellow book with colorful illustrations\",\"the window\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the sturdy dark fireplace made of brick or stone', 'the vivid yellow book with colorful illustrations', 'the window'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_42445984/Validation_42445984_42445984_75195.633.png", "Validation_42445984/Validation_42445984_42445984_75197.632.png", "Validation_42445984/Validation_42445984_42445984_75199.631.png", "Validation_42445984/Validation_42445984_42445984_75201.631.png", "Validation_42445984/Validation_42445984_42445984_75203.630.png" ], "turn_id": 6, "scan_id": "Validation_42445984", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the window\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the window\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_42445984/Validation_42445984_42445984_75205.629.png", "Validation_42445984/Validation_42445984_42445984_75207.628.png", "Validation_42445984/Validation_42445984_42445984_75209.627.png", "Validation_42445984/Validation_42445984_42445984_75211.626.png", "Validation_42445984/Validation_42445984_42445984_75213.626.png" ], "turn_id": 7, "scan_id": "Validation_42445984", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "74.93", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_42445984/Validation_42445984_42445984_75215.625.png", "Validation_42445984/Validation_42445984_42445984_75217.624.png", "Validation_42445984/Validation_42445984_42445984_75219.623.png", "Validation_42445984/Validation_42445984_42445984_75221.622.png", "Validation_42445984/Validation_42445984_42445984_75223.638.png" ], "turn_id": 8, "scan_id": "Validation_42445984", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_42445984/Validation_42445984_42445984_75225.637.png", "Validation_42445984/Validation_42445984_42445984_75227.636.png", "Validation_42445984/Validation_42445984_42445984_75229.636.png", "Validation_42445984/Validation_42445984_42445984_75231.635.png", "Validation_42445984/Validation_42445984_42445984_75233.634.png" ], "turn_id": 9, "scan_id": "Validation_42445984", "origin_question": "Assuming the direction you are facing at the end of the turn 8 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Right", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 8 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47429906/Validation_47429906_47429906_135312.972.png", "Validation_47429906/Validation_47429906_47429906_135314.954.png", "Validation_47429906/Validation_47429906_47429906_135316.953.png", "Validation_47429906/Validation_47429906_47429906_135318.953.png", "Validation_47429906/Validation_47429906_47429906_135320.952.png" ], "turn_id": 1, "scan_id": "Validation_47429906", "origin_question": "Remember, have you seen any cabinet(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any cabinet(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_47429906/Validation_47429906_47429906_135322.951.png", "Validation_47429906/Validation_47429906_47429906_135324.950.png", "Validation_47429906/Validation_47429906_47429906_135326.949.png", "Validation_47429906/Validation_47429906_47429906_135328.948.png", "Validation_47429906/Validation_47429906_47429906_135330.948.png" ], "turn_id": 2, "scan_id": "Validation_47429906", "origin_question": "In which turn did you see both \"the dark leather couch with the sleek rectangular shape\" and \"the large rectangular wooden-framed window\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the dark leather couch with the sleek rectangular shape\" and \"the large rectangular wooden-framed window\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47429906/Validation_47429906_47429906_135332.947.png", "Validation_47429906/Validation_47429906_47429906_135334.946.png", "Validation_47429906/Validation_47429906_47429906_135336.945.png", "Validation_47429906/Validation_47429906_47429906_135338.944.png", "Validation_47429906/Validation_47429906_47429906_135340.960.png" ], "turn_id": 3, "scan_id": "Validation_47429906", "origin_question": "", "option": [], "answer": "", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_47429906/Validation_47429906_47429906_135342.959.png", "Validation_47429906/Validation_47429906_47429906_135344.959.png", "Validation_47429906/Validation_47429906_47429906_135346.958.png", "Validation_47429906/Validation_47429906_47429906_135348.957.png", "Validation_47429906/Validation_47429906_47429906_135350.956.png" ], "turn_id": 4, "scan_id": "Validation_47429906", "origin_question": "Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Left", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_47429906/Validation_47429906_47429906_135352.955.png", "Validation_47429906/Validation_47429906_47429906_135354.955.png", "Validation_47429906/Validation_47429906_47429906_135356.954.png", "Validation_47429906/Validation_47429906_47429906_135358.953.png", "Validation_47429906/Validation_47429906_47429906_135360.952.png" ], "turn_id": 5, "scan_id": "Validation_47429906", "origin_question": "What will be the first-time appearance order of the following categories: plate, sink, window?", "option": [ "plate, window, sink", "window, plate, sink", "window, sink, plate", "sink, plate, window" ], "answer": "window, plate, sink", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: plate, sink, window? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['plate, window, sink', 'window, plate, sink', 'window, sink, plate', 'sink, plate, window'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_47429906/Validation_47429906_47429906_135362.951.png", "Validation_47429906/Validation_47429906_47429906_135364.951.png", "Validation_47429906/Validation_47429906_47429906_135366.950.png", "Validation_47429906/Validation_47429906_47429906_135368.949.png", "Validation_47429906/Validation_47429906_47429906_135370.948.png" ], "turn_id": 6, "scan_id": "Validation_47429906", "origin_question": "Please recall \"the dark leather couch with the sleek rectangular shape\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.96", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the dark leather couch with the sleek rectangular shape\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_47429906/Validation_47429906_47429906_135372.947.png", "Validation_47429906/Validation_47429906_47429906_135374.946.png", "Validation_47429906/Validation_47429906_47429906_135376.946.png", "Validation_47429906/Validation_47429906_47429906_135378.945.png", "Validation_47429906/Validation_47429906_47429906_135380.961.png" ], "turn_id": 7, "scan_id": "Validation_47429906", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the dark leather couch with the sleek rectangular shape\" from your position?", "option": [], "answer": "153.9", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the dark leather couch with the sleek rectangular shape\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_47429906/Validation_47429906_47429906_135382.960.png", "Validation_47429906/Validation_47429906_47429906_135384.959.png", "Validation_47429906/Validation_47429906_47429906_135386.958.png", "Validation_47429906/Validation_47429906_47429906_135388.957.png", "Validation_47429906/Validation_47429906_47429906_135390.957.png" ], "turn_id": 8, "scan_id": "Validation_47429906", "origin_question": "In which turn did you see both \"the monitor\" and \"the dark leather couch with the sleek rectangular shape\" simultaneously? ", "option": [], "answer": "7", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the monitor\" and \"the dark leather couch with the sleek rectangular shape\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_44358584/Validation_44358584_44358584_55637.049.png", "Validation_44358584/Validation_44358584_44358584_55639.031.png", "Validation_44358584/Validation_44358584_44358584_55641.030.png", "Validation_44358584/Validation_44358584_44358584_55643.029.png" ], "turn_id": 1, "scan_id": "Validation_44358584", "origin_question": "Remember, have you seen any bed(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any bed(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_44358584/Validation_44358584_44358584_55645.029.png", "Validation_44358584/Validation_44358584_44358584_55647.028.png", "Validation_44358584/Validation_44358584_44358584_55649.027.png", "Validation_44358584/Validation_44358584_44358584_55651.026.png" ], "turn_id": 2, "scan_id": "Validation_44358584", "origin_question": "When did you first discover \"the light-colored rectangular bed with cotton bedding\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the light-colored rectangular bed with cotton bedding\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_44358584/Validation_44358584_44358584_55653.026.png", "Validation_44358584/Validation_44358584_44358584_55655.025.png", "Validation_44358584/Validation_44358584_44358584_55657.024.png", "Validation_44358584/Validation_44358584_44358584_55659.023.png" ], "turn_id": 3, "scan_id": "Validation_44358584", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_44358584/Validation_44358584_44358584_55661.022.png", "Validation_44358584/Validation_44358584_44358584_55663.022.png", "Validation_44358584/Validation_44358584_44358584_55665.021.png", "Validation_44358584/Validation_44358584_44358584_55667.020.png" ], "turn_id": 4, "scan_id": "Validation_44358584", "origin_question": "In which turn were you closest to \"the light-colored rectangular bed with cotton bedding\"? ", "option": [], "answer": "4", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the light-colored rectangular bed with cotton bedding\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_44358584/Validation_44358584_44358584_55669.019.png", "Validation_44358584/Validation_44358584_44358584_55671.035.png", "Validation_44358584/Validation_44358584_44358584_55673.034.png", "Validation_44358584/Validation_44358584_44358584_55675.033.png" ], "turn_id": 5, "scan_id": "Validation_44358584", "origin_question": "", "option": [], "answer": "", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_44358584/Validation_44358584_44358584_55677.033.png", "Validation_44358584/Validation_44358584_44358584_55679.032.png", "Validation_44358584/Validation_44358584_44358584_55681.031.png", "Validation_44358584/Validation_44358584_44358584_55683.030.png" ], "turn_id": 6, "scan_id": "Validation_44358584", "origin_question": "How far is your current position from where you were at the end of the turn 3? (in meters)", "option": [], "answer": "1.85", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 3? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_44358584/Validation_44358584_44358584_55685.029.png", "Validation_44358584/Validation_44358584_44358584_55687.029.png", "Validation_44358584/Validation_44358584_44358584_55689.028.png", "Validation_44358584/Validation_44358584_44358584_55691.027.png" ], "turn_id": 7, "scan_id": "Validation_44358584", "origin_question": "Assuming the direction you are facing at the end of the turn 4 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Right", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 4 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_44358584/Validation_44358584_44358584_55693.026.png", "Validation_44358584/Validation_44358584_44358584_55695.026.png", "Validation_44358584/Validation_44358584_44358584_55697.025.png", "Validation_44358584/Validation_44358584_44358584_55699.024.png" ], "turn_id": 8, "scan_id": "Validation_44358584", "origin_question": "Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42446165/Validation_42446165_42446165_79715.748.png", "Validation_42446165/Validation_42446165_42446165_79717.731.png", "Validation_42446165/Validation_42446165_42446165_79719.730.png", "Validation_42446165/Validation_42446165_42446165_79721.729.png", "Validation_42446165/Validation_42446165_42446165_79723.728.png" ], "turn_id": 1, "scan_id": "Validation_42446165", "origin_question": "Remember, have you seen any door(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any door(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_42446165/Validation_42446165_42446165_79725.727.png", "Validation_42446165/Validation_42446165_42446165_79727.727.png", "Validation_42446165/Validation_42446165_42446165_79729.726.png", "Validation_42446165/Validation_42446165_42446165_79731.725.png", "Validation_42446165/Validation_42446165_42446165_79733.724.png" ], "turn_id": 2, "scan_id": "Validation_42446165", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the modern bed with the dark bronze metal frame\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the modern bed with the dark bronze metal frame\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_42446165/Validation_42446165_42446165_79735.723.png", "Validation_42446165/Validation_42446165_42446165_79737.723.png", "Validation_42446165/Validation_42446165_42446165_79739.722.png", "Validation_42446165/Validation_42446165_42446165_79741.721.png", "Validation_42446165/Validation_42446165_42446165_79743.720.png" ], "turn_id": 3, "scan_id": "Validation_42446165", "origin_question": "Please recall \"the round white bin with the red lid\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.5", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the round white bin with the red lid\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_42446165/Validation_42446165_42446165_79745.719.png", "Validation_42446165/Validation_42446165_42446165_79747.719.png", "Validation_42446165/Validation_42446165_42446165_79749.735.png", "Validation_42446165/Validation_42446165_42446165_79751.734.png", "Validation_42446165/Validation_42446165_42446165_79753.733.png" ], "turn_id": 4, "scan_id": "Validation_42446165", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the modern bed with the dark bronze metal frame\", Y: \"the sleek, minimalist door with the neutral color\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to X and farther away from Y", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the modern bed with the dark bronze metal frame\", Y: \"the sleek, minimalist door with the neutral color\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "Validation_42446165/Validation_42446165_42446165_79755.732.png", "Validation_42446165/Validation_42446165_42446165_79757.731.png", "Validation_42446165/Validation_42446165_42446165_79759.731.png", "Validation_42446165/Validation_42446165_42446165_79761.730.png", "Validation_42446165/Validation_42446165_42446165_79763.729.png" ], "turn_id": 5, "scan_id": "Validation_42446165", "origin_question": "Among these three objects(\"the sleek, minimalist door with the neutral color\",\"the modern, sleek monitor with the thin black bezel\",\"the modern, sleek steel radiator with horizontal ridges\" ), which one is the farthest from you?", "option": [ "the sleek, minimalist door with the neutral color", "the modern, sleek monitor with the thin black bezel", "the modern, sleek steel radiator with horizontal ridges" ], "answer": "the sleek, minimalist door with the neutral color", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the sleek, minimalist door with the neutral color\",\"the modern, sleek monitor with the thin black bezel\",\"the modern, sleek steel radiator with horizontal ridges\" ), which one is the farthest from you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the sleek, minimalist door with the neutral color', 'the modern, sleek monitor with the thin black bezel', 'the modern, sleek steel radiator with horizontal ridges'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_42446165/Validation_42446165_42446165_79765.728.png", "Validation_42446165/Validation_42446165_42446165_79767.727.png", "Validation_42446165/Validation_42446165_42446165_79769.727.png", "Validation_42446165/Validation_42446165_42446165_79771.726.png", "Validation_42446165/Validation_42446165_42446165_79773.725.png" ], "turn_id": 6, "scan_id": "Validation_42446165", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the modern, sleek monitor with the thin black bezel\" from your position?", "option": [], "answer": "113.6", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the modern, sleek monitor with the thin black bezel\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_42446165/Validation_42446165_42446165_79775.724.png", "Validation_42446165/Validation_42446165_42446165_79777.723.png", "Validation_42446165/Validation_42446165_42446165_79779.723.png", "Validation_42446165/Validation_42446165_42446165_79781.722.png", "Validation_42446165/Validation_42446165_42446165_79783.721.png" ], "turn_id": 7, "scan_id": "Validation_42446165", "origin_question": "What will be the first-time appearance order of the following categories: bed, monitor, table?", "option": [ "table, bed, monitor", "bed, table, monitor", "bed, monitor, table", "monitor, bed, table" ], "answer": "bed, monitor, table", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bed, monitor, table? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['table, bed, monitor', 'bed, table, monitor', 'bed, monitor, table', 'monitor, bed, table'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_42446165/Validation_42446165_42446165_79785.720.png", "Validation_42446165/Validation_42446165_42446165_79787.719.png", "Validation_42446165/Validation_42446165_42446165_79789.735.png", "Validation_42446165/Validation_42446165_42446165_79791.735.png", "Validation_42446165/Validation_42446165_42446165_79793.734.png" ], "turn_id": 8, "scan_id": "Validation_42446165", "origin_question": "At the end of which turn were both of \"the modern bed with the dark bronze metal frame\" and \"the window\" on your left side?", "option": [], "answer": "4", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the modern bed with the dark bronze metal frame\" and \"the window\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "Validation_47204578/Validation_47204578_47204578_4894.233.png", "Validation_47204578/Validation_47204578_47204578_4896.216.png", "Validation_47204578/Validation_47204578_47204578_4898.215.png", "Validation_47204578/Validation_47204578_47204578_4900.214.png" ], "turn_id": 1, "scan_id": "Validation_47204578", "origin_question": "Remember, how many dishwasher(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many dishwasher(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_47204578/Validation_47204578_47204578_4902.213.png", "Validation_47204578/Validation_47204578_47204578_4904.212.png", "Validation_47204578/Validation_47204578_47204578_4906.211.png", "Validation_47204578/Validation_47204578_47204578_4908.211.png" ], "turn_id": 2, "scan_id": "Validation_47204578", "origin_question": "Which two objects are on the same side of you now? \"the hexagonal glass top table with the metallic frame\", \"the large, soft carpet with the geometric pattern\", \"the light wooden door with the clean design\". ", "option": [ "the hexagonal glass top table with the metallic frame, the large, soft carpet with the geometric pattern", "the hexagonal glass top table with the metallic frame, the light wooden door with the clean design", "the large, soft carpet with the geometric pattern, the light wooden door with the clean design" ], "answer": "the hexagonal glass top table with the metallic frame, the large, soft carpet with the geometric pattern", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the hexagonal glass top table with the metallic frame\", \"the large, soft carpet with the geometric pattern\", \"the light wooden door with the clean design\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the hexagonal glass top table with the metallic frame, the large, soft carpet with the geometric pattern', 'the hexagonal glass top table with the metallic frame, the light wooden door with the clean design', 'the large, soft carpet with the geometric pattern, the light wooden door with the clean design'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_47204578/Validation_47204578_47204578_4910.210.png", "Validation_47204578/Validation_47204578_47204578_4912.209.png", "Validation_47204578/Validation_47204578_47204578_4914.208.png", "Validation_47204578/Validation_47204578_47204578_4916.207.png" ], "turn_id": 3, "scan_id": "Validation_47204578", "origin_question": "By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the light wooden door with the clean design\", Y: \"the large, soft carpet with the geometric pattern\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to Y and farther away from X", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the light wooden door with the clean design\", Y: \"the large, soft carpet with the geometric pattern\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_47204578/Validation_47204578_47204578_4918.223.png", "Validation_47204578/Validation_47204578_47204578_4920.222.png", "Validation_47204578/Validation_47204578_47204578_4922.221.png", "Validation_47204578/Validation_47204578_47204578_4924.221.png" ], "turn_id": 4, "scan_id": "Validation_47204578", "origin_question": "Is the large monitor with the slender black frame to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the large monitor with the slender black frame to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_47204578/Validation_47204578_47204578_4926.220.png", "Validation_47204578/Validation_47204578_47204578_4928.219.png", "Validation_47204578/Validation_47204578_47204578_4930.218.png", "Validation_47204578/Validation_47204578_47204578_4932.217.png" ], "turn_id": 5, "scan_id": "Validation_47204578", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the hexagonal glass top table with the metallic frame\" from your position?", "option": [], "answer": "107.16", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the hexagonal glass top table with the metallic frame\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_47204578/Validation_47204578_47204578_4934.217.png", "Validation_47204578/Validation_47204578_47204578_4936.216.png", "Validation_47204578/Validation_47204578_47204578_4938.215.png", "Validation_47204578/Validation_47204578_47204578_4940.214.png" ], "turn_id": 6, "scan_id": "Validation_47204578", "origin_question": "Please recall \"the compact, minimalist stool made of durable material\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.93", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the compact, minimalist stool made of durable material\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_45662924/Validation_45662924_45662924_21555.941.png", "Validation_45662924/Validation_45662924_45662924_21557.923.png", "Validation_45662924/Validation_45662924_45662924_21559.922.png", "Validation_45662924/Validation_45662924_45662924_21561.921.png" ], "turn_id": 1, "scan_id": "Validation_45662924", "origin_question": "Remember, have you seen any coffee maker(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any coffee maker(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_45662924/Validation_45662924_45662924_21563.921.png", "Validation_45662924/Validation_45662924_45662924_21565.920.png", "Validation_45662924/Validation_45662924_45662924_21567.919.png", "Validation_45662924/Validation_45662924_45662924_21569.918.png" ], "turn_id": 2, "scan_id": "Validation_45662924", "origin_question": "In which turn were you closest to \"the sleek, dark modern fireplace made of metal or glass\"? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the sleek, dark modern fireplace made of metal or glass\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_45662924/Validation_45662924_45662924_21571.917.png", "Validation_45662924/Validation_45662924_45662924_21573.916.png", "Validation_45662924/Validation_45662924_45662924_21575.916.png", "Validation_45662924/Validation_45662924_45662924_21577.931.png" ], "turn_id": 3, "scan_id": "Validation_45662924", "origin_question": "When did you first discover \"the modern stairs with white wooden steps\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the modern stairs with white wooden steps\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_45662924/Validation_45662924_45662924_21579.931.png", "Validation_45662924/Validation_45662924_45662924_21581.930.png", "Validation_45662924/Validation_45662924_45662924_21583.929.png", "Validation_45662924/Validation_45662924_45662924_21585.928.png" ], "turn_id": 4, "scan_id": "Validation_45662924", "origin_question": "At the end of which turn were both of \"the modern stairs with white wooden steps\" and \"the modern, spacious light-colored wooden table\" on your left side?", "option": [], "answer": "4", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the modern stairs with white wooden steps\" and \"the modern, spacious light-colored wooden table\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_45662924/Validation_45662924_45662924_21587.927.png", "Validation_45662924/Validation_45662924_45662924_21589.927.png", "Validation_45662924/Validation_45662924_45662924_21591.926.png", "Validation_45662924/Validation_45662924_45662924_21593.925.png" ], "turn_id": 5, "scan_id": "Validation_45662924", "origin_question": "Which direction is the medium-sized, vibrant green plant with glossy leaves in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the medium-sized, vibrant green plant with glossy leaves in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_45662924/Validation_45662924_45662924_21595.924.png", "Validation_45662924/Validation_45662924_45662924_21597.923.png", "Validation_45662924/Validation_45662924_45662924_21599.922.png", "Validation_45662924/Validation_45662924_45662924_21601.922.png" ], "turn_id": 6, "scan_id": "Validation_45662924", "origin_question": "By the end of the 3 turn, were you getting closer to or farther away from these two objects: X: \"the sleek, dark modern fireplace made of metal or glass\", Y: \"the modern, spacious light-colored wooden table\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to Y and farther away from X", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 3 turn, were you getting closer to or farther away from these two objects: X: \"the sleek, dark modern fireplace made of metal or glass\", Y: \"the modern, spacious light-colored wooden table\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_45662924/Validation_45662924_45662924_21603.921.png", "Validation_45662924/Validation_45662924_45662924_21605.920.png", "Validation_45662924/Validation_45662924_45662924_21607.919.png", "Validation_45662924/Validation_45662924_45662924_21609.918.png" ], "turn_id": 7, "scan_id": "Validation_45662924", "origin_question": "In which turn were you closest to \"the modern, spacious light-colored wooden table\"? ", "option": [], "answer": "7", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the modern, spacious light-colored wooden table\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_45662924/Validation_45662924_45662924_21611.917.png", "Validation_45662924/Validation_45662924_45662924_21613.916.png", "Validation_45662924/Validation_45662924_45662924_21615.916.png", "Validation_45662924/Validation_45662924_45662924_21617.931.png" ], "turn_id": 8, "scan_id": "Validation_45662924", "origin_question": "Using your orientation at the end of turn 5 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 5 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_45662924/Validation_45662924_45662924_21619.931.png", "Validation_45662924/Validation_45662924_45662924_21621.930.png", "Validation_45662924/Validation_45662924_45662924_21623.929.png", "Validation_45662924/Validation_45662924_45662924_21625.928.png" ], "turn_id": 9, "scan_id": "Validation_45662924", "origin_question": "Is the sleek, dark modern fireplace made of metal or glass to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 9 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the sleek, dark modern fireplace made of metal or glass to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_45662975/Validation_45662975_45662975_26065.135.png", "Validation_45662975/Validation_45662975_45662975_26066.634.png", "Validation_45662975/Validation_45662975_45662975_26068.134.png", "Validation_45662975/Validation_45662975_45662975_26069.633.png" ], "turn_id": 1, "scan_id": "Validation_45662975", "origin_question": "Remember, have you seen any basket(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any basket(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_45662975/Validation_45662975_45662975_26071.132.png", "Validation_45662975/Validation_45662975_45662975_26072.632.png", "Validation_45662975/Validation_45662975_45662975_26074.131.png", "Validation_45662975/Validation_45662975_45662975_26075.631.png" ], "turn_id": 2, "scan_id": "Validation_45662975", "origin_question": "In which turn were you closest to \"the vibrant green plant with broad, healthy leaves\"? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the vibrant green plant with broad, healthy leaves\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_45662975/Validation_45662975_45662975_26077.130.png", "Validation_45662975/Validation_45662975_45662975_26078.630.png", "Validation_45662975/Validation_45662975_45662975_26080.129.png", "Validation_45662975/Validation_45662975_45662975_26081.628.png" ], "turn_id": 3, "scan_id": "Validation_45662975", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_45662975/Validation_45662975_45662975_26083.128.png", "Validation_45662975/Validation_45662975_45662975_26084.627.png", "Validation_45662975/Validation_45662975_45662975_26086.127.png", "Validation_45662975/Validation_45662975_45662975_26087.626.png" ], "turn_id": 4, "scan_id": "Validation_45662975", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the sink\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the sink\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_45662975/Validation_45662975_45662975_26089.125.png", "Validation_45662975/Validation_45662975_45662975_26090.625.png", "Validation_45662975/Validation_45662975_45662975_26092.124.png", "Validation_45662975/Validation_45662975_45662975_26093.624.png" ], "turn_id": 5, "scan_id": "Validation_45662975", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42897696/Validation_42897696_42897696_480414.911.png", "Validation_42897696/Validation_42897696_42897696_480416.894.png", "Validation_42897696/Validation_42897696_42897696_480418.893.png", "Validation_42897696/Validation_42897696_42897696_480420.892.png", "Validation_42897696/Validation_42897696_42897696_480422.891.png" ], "turn_id": 1, "scan_id": "Validation_42897696", "origin_question": "Remember, have you seen any table(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any table(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "Validation_42897696/Validation_42897696_42897696_480424.890.png", "Validation_42897696/Validation_42897696_42897696_480426.890.png", "Validation_42897696/Validation_42897696_42897696_480428.889.png", "Validation_42897696/Validation_42897696_42897696_480430.888.png", "Validation_42897696/Validation_42897696_42897696_480432.904.png" ], "turn_id": 2, "scan_id": "Validation_42897696", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the soft, pale rectangular bed with the crumpled texture\"; \"the soft blue cylindrical lamp with the ceramic base\"; \"the medium-sized black and white landscape picture\" ", "option": [ "the soft, pale rectangular bed with the crumpled texture", "the soft blue cylindrical lamp with the ceramic base", "the medium-sized black and white landscape picture" ], "answer": "the medium-sized black and white landscape picture", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the soft, pale rectangular bed with the crumpled texture\"; \"the soft blue cylindrical lamp with the ceramic base\"; \"the medium-sized black and white landscape picture\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the soft, pale rectangular bed with the crumpled texture', 'the soft blue cylindrical lamp with the ceramic base', 'the medium-sized black and white landscape picture'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_42897696/Validation_42897696_42897696_480434.903.png", "Validation_42897696/Validation_42897696_42897696_480436.902.png", "Validation_42897696/Validation_42897696_42897696_480438.901.png", "Validation_42897696/Validation_42897696_42897696_480440.900.png", "Validation_42897696/Validation_42897696_42897696_480442.900.png" ], "turn_id": 3, "scan_id": "Validation_42897696", "origin_question": "In which turn were you closest to \"the soft, pale rectangular bed with the crumpled texture\"? ", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the soft, pale rectangular bed with the crumpled texture\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_42897696/Validation_42897696_42897696_480444.899.png", "Validation_42897696/Validation_42897696_42897696_480446.898.png", "Validation_42897696/Validation_42897696_42897696_480448.897.png", "Validation_42897696/Validation_42897696_42897696_480450.896.png", "Validation_42897696/Validation_42897696_42897696_480452.895.png" ], "turn_id": 4, "scan_id": "Validation_42897696", "origin_question": "Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Left", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_42897696/Validation_42897696_42897696_480454.895.png", "Validation_42897696/Validation_42897696_42897696_480456.894.png", "Validation_42897696/Validation_42897696_42897696_480458.893.png", "Validation_42897696/Validation_42897696_42897696_480460.892.png", "Validation_42897696/Validation_42897696_42897696_480462.891.png" ], "turn_id": 5, "scan_id": "Validation_42897696", "origin_question": "Which direction is the tall, narrow wooden shelf with the light brown finish in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the tall, narrow wooden shelf with the light brown finish in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_42897696/Validation_42897696_42897696_480464.890.png", "Validation_42897696/Validation_42897696_42897696_480466.890.png", "Validation_42897696/Validation_42897696_42897696_480468.889.png", "Validation_42897696/Validation_42897696_42897696_480470.888.png", "Validation_42897696/Validation_42897696_42897696_480472.904.png" ], "turn_id": 6, "scan_id": "Validation_42897696", "origin_question": "Is the warm, light brown wooden table to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the warm, light brown wooden table to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_42897696/Validation_42897696_42897696_480474.903.png", "Validation_42897696/Validation_42897696_42897696_480476.902.png", "Validation_42897696/Validation_42897696_42897696_480478.901.png", "Validation_42897696/Validation_42897696_42897696_480480.901.png", "Validation_42897696/Validation_42897696_42897696_480482.900.png" ], "turn_id": 7, "scan_id": "Validation_42897696", "origin_question": "Assuming the direction you are facing at the end of the turn 3 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Left", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 3 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_42897696/Validation_42897696_42897696_480484.899.png", "Validation_42897696/Validation_42897696_42897696_480486.898.png", "Validation_42897696/Validation_42897696_42897696_480488.897.png", "Validation_42897696/Validation_42897696_42897696_480490.896.png", "Validation_42897696/Validation_42897696_42897696_480492.896.png" ], "turn_id": 8, "scan_id": "Validation_42897696", "origin_question": "What will be the first-time appearance order of the following categories: radiator, shelf, table?", "option": [ "radiator, shelf, table", "table, shelf, radiator", "table, radiator, shelf", "shelf, table, radiator" ], "answer": "table, shelf, radiator", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: radiator, shelf, table? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['radiator, shelf, table', 'table, shelf, radiator', 'table, radiator, shelf', 'shelf, table, radiator'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_42897696/Validation_42897696_42897696_480494.895.png", "Validation_42897696/Validation_42897696_42897696_480496.894.png", "Validation_42897696/Validation_42897696_42897696_480498.893.png", "Validation_42897696/Validation_42897696_42897696_480500.892.png", "Validation_42897696/Validation_42897696_42897696_480502.892.png" ], "turn_id": 9, "scan_id": "Validation_42897696", "origin_question": "Compared to your position at the end of the 5 turn, are you now closer or farther away from \"the window\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 5 turn, are you now closer or farther away from \"the window\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_42897696/Validation_42897696_42897696_480504.891.png", "Validation_42897696/Validation_42897696_42897696_480506.890.png", "Validation_42897696/Validation_42897696_42897696_480508.889.png", "Validation_42897696/Validation_42897696_42897696_480510.888.png", "Validation_42897696/Validation_42897696_42897696_480512.887.png" ], "turn_id": 10, "scan_id": "Validation_42897696", "origin_question": "At the end of which turn were both of \"the soft blue cylindrical lamp with the ceramic base\" and \"the slim wooden cabinet with the warm natural finish\" on your right side?", "option": [], "answer": "7", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the soft blue cylindrical lamp with the ceramic base\" and \"the slim wooden cabinet with the warm natural finish\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_45261129/Validation_45261129_45261129_3345.372.png", "Validation_45261129/Validation_45261129_45261129_3347.371.png", "Validation_45261129/Validation_45261129_45261129_3349.370.png", "Validation_45261129/Validation_45261129_45261129_3351.369.png" ], "turn_id": 1, "scan_id": "Validation_45261129", "origin_question": "Remember, have you seen any shelf(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any shelf(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_45261129/Validation_45261129_45261129_3353.369.png", "Validation_45261129/Validation_45261129_45261129_3355.368.png", "Validation_45261129/Validation_45261129_45261129_3357.367.png", "Validation_45261129/Validation_45261129_45261129_3359.366.png" ], "turn_id": 2, "scan_id": "Validation_45261129", "origin_question": "At the end of which turn were both of \"the modern silver stainless steel oven with ergonomic design\" and \"the sleek black gas stove with stainless steel finish\" on your left side?", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the modern silver stainless steel oven with ergonomic design\" and \"the sleek black gas stove with stainless steel finish\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_45261129/Validation_45261129_45261129_3361.365.png", "Validation_45261129/Validation_45261129_45261129_3363.364.png", "Validation_45261129/Validation_45261129_45261129_3365.363.png", "Validation_45261129/Validation_45261129_45261129_3367.363.png" ], "turn_id": 3, "scan_id": "Validation_45261129", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_45261129/Validation_45261129_45261129_3369.362.png", "Validation_45261129/Validation_45261129_45261129_3371.361.png", "Validation_45261129/Validation_45261129_45261129_3373.360.png", "Validation_45261129/Validation_45261129_45261129_3375.359.png" ], "turn_id": 4, "scan_id": "Validation_45261129", "origin_question": "What will be the first-time appearance order of the following categories: refrigerator, sink, table?", "option": [ "sink, refrigerator, table", "table, sink, refrigerator", "table, refrigerator, sink", "refrigerator, sink, table" ], "answer": "refrigerator, sink, table", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: refrigerator, sink, table? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['sink, refrigerator, table', 'table, sink, refrigerator', 'table, refrigerator, sink', 'refrigerator, sink, table'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_45261129/Validation_45261129_45261129_3377.358.png", "Validation_45261129/Validation_45261129_45261129_3379.358.png", "Validation_45261129/Validation_45261129_45261129_3381.357.png", "Validation_45261129/Validation_45261129_45261129_3383.356.png" ], "turn_id": 5, "scan_id": "Validation_45261129", "origin_question": "Which two objects are on the same side of you now? \"the microwave\", \"the polished wooden table with the white-painted base\", \"the dark wood stool with the circular seat\". ", "option": [ "the microwave, the polished wooden table with the white-painted base", "the microwave, the dark wood stool with the circular seat", "the polished wooden table with the white-painted base, the dark wood stool with the circular seat" ], "answer": "the polished wooden table with the white-painted base, the dark wood stool with the circular seat", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the microwave\", \"the polished wooden table with the white-painted base\", \"the dark wood stool with the circular seat\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the microwave, the polished wooden table with the white-painted base', 'the microwave, the dark wood stool with the circular seat', 'the polished wooden table with the white-painted base, the dark wood stool with the circular seat'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_45261129/Validation_45261129_45261129_3385.372.png", "Validation_45261129/Validation_45261129_45261129_3387.371.png", "Validation_45261129/Validation_45261129_45261129_3389.370.png", "Validation_45261129/Validation_45261129_45261129_3391.369.png" ], "turn_id": 6, "scan_id": "Validation_45261129", "origin_question": "Which two objects are on the same side of you now? \"the sleek black stainless steel kettle with the plastic handle\", \"the modern cream upholstered chair with the slender backrest\", \"the sleek white cylindrical bin with the dark lid\". ", "option": [ "the sleek black stainless steel kettle with the plastic handle, the modern cream upholstered chair with the slender backrest", "the sleek black stainless steel kettle with the plastic handle, the sleek white cylindrical bin with the dark lid", "the modern cream upholstered chair with the slender backrest, the sleek white cylindrical bin with the dark lid" ], "answer": "the sleek black stainless steel kettle with the plastic handle, the modern cream upholstered chair with the slender backrest", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the sleek black stainless steel kettle with the plastic handle\", \"the modern cream upholstered chair with the slender backrest\", \"the sleek white cylindrical bin with the dark lid\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the sleek black stainless steel kettle with the plastic handle, the modern cream upholstered chair with the slender backrest', 'the sleek black stainless steel kettle with the plastic handle, the sleek white cylindrical bin with the dark lid', 'the modern cream upholstered chair with the slender backrest, the sleek white cylindrical bin with the dark lid'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_45261129/Validation_45261129_45261129_3393.369.png", "Validation_45261129/Validation_45261129_45261129_3395.368.png", "Validation_45261129/Validation_45261129_45261129_3397.367.png", "Validation_45261129/Validation_45261129_45261129_3399.366.png" ], "turn_id": 7, "scan_id": "Validation_45261129", "origin_question": "Which direction is the modern black metal shelf design in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the modern black metal shelf design in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47331262/Validation_47331262_47331262_89914.507.png", "Validation_47331262/Validation_47331262_47331262_89916.489.png", "Validation_47331262/Validation_47331262_47331262_89918.505.png", "Validation_47331262/Validation_47331262_47331262_89920.504.png" ], "turn_id": 1, "scan_id": "Validation_47331262", "origin_question": "Remember, have you seen any table(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any table(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47331262/Validation_47331262_47331262_89922.504.png", "Validation_47331262/Validation_47331262_47331262_89924.503.png", "Validation_47331262/Validation_47331262_47331262_89926.502.png", "Validation_47331262/Validation_47331262_47331262_89928.501.png" ], "turn_id": 2, "scan_id": "Validation_47331262", "origin_question": "Remember, have you seen any chair(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any chair(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_47331262/Validation_47331262_47331262_89930.500.png", "Validation_47331262/Validation_47331262_47331262_89932.500.png", "Validation_47331262/Validation_47331262_47331262_89934.499.png", "Validation_47331262/Validation_47331262_47331262_89936.498.png" ], "turn_id": 3, "scan_id": "Validation_47331262", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the sleek black rectangular monitor with the glossy screen\" from your position?", "option": [], "answer": "100.84", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the sleek black rectangular monitor with the glossy screen\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_47331262/Validation_47331262_47331262_89938.497.png", "Validation_47331262/Validation_47331262_47331262_89940.496.png", "Validation_47331262/Validation_47331262_47331262_89942.496.png", "Validation_47331262/Validation_47331262_47331262_89944.495.png" ], "turn_id": 4, "scan_id": "Validation_47331262", "origin_question": "What will be the first-time appearance order of the following categories: monitor, table, window?", "option": [ "window, table, monitor", "window, monitor, table", "table, monitor, window", "table, window, monitor" ], "answer": "table, monitor, window", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: monitor, table, window? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['window, table, monitor', 'window, monitor, table', 'table, monitor, window', 'table, window, monitor'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_47331262/Validation_47331262_47331262_89946.494.png", "Validation_47331262/Validation_47331262_47331262_89948.493.png", "Validation_47331262/Validation_47331262_47331262_89950.492.png", "Validation_47331262/Validation_47331262_47331262_89952.492.png" ], "turn_id": 5, "scan_id": "Validation_47331262", "origin_question": "In which turn did you see both \"the sleek black rectangular monitor with the glossy screen\" and \"the round deep green ceramic flowerpot\" simultaneously? ", "option": [], "answer": "4", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the sleek black rectangular monitor with the glossy screen\" and \"the round deep green ceramic flowerpot\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_47331262/Validation_47331262_47331262_89954.491.png", "Validation_47331262/Validation_47331262_47331262_89956.490.png", "Validation_47331262/Validation_47331262_47331262_89958.489.png", "Validation_47331262/Validation_47331262_47331262_89960.505.png" ], "turn_id": 6, "scan_id": "Validation_47331262", "origin_question": "Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "145.84", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_47331262/Validation_47331262_47331262_89962.504.png", "Validation_47331262/Validation_47331262_47331262_89964.503.png", "Validation_47331262/Validation_47331262_47331262_89966.503.png", "Validation_47331262/Validation_47331262_47331262_89968.502.png" ], "turn_id": 7, "scan_id": "Validation_47331262", "origin_question": "By the end of the 5 turn, were you getting closer to or farther away from these two objects: X: \"the sleek black rectangular monitor with the glossy screen\", Y: \"the compact, white, rectangular metal radiator\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "farther away from both X and Y", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 5 turn, were you getting closer to or farther away from these two objects: X: \"the sleek black rectangular monitor with the glossy screen\", Y: \"the compact, white, rectangular metal radiator\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47430026/Validation_47430026_47430026_18812.043.png", "Validation_47430026/Validation_47430026_47430026_18813.542.png", "Validation_47430026/Validation_47430026_47430026_18815.042.png", "Validation_47430026/Validation_47430026_47430026_18816.541.png" ], "turn_id": 1, "scan_id": "Validation_47430026", "origin_question": "Remember, have you seen any vase(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any vase(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47430026/Validation_47430026_47430026_18818.040.png", "Validation_47430026/Validation_47430026_47430026_18819.540.png", "Validation_47430026/Validation_47430026_47430026_18821.039.png", "Validation_47430026/Validation_47430026_47430026_18822.538.png" ], "turn_id": 2, "scan_id": "Validation_47430026", "origin_question": "In which turn were you closest to \"the pastel blue, velvety couch\"? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the pastel blue, velvety couch\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47430026/Validation_47430026_47430026_18824.038.png", "Validation_47430026/Validation_47430026_47430026_18825.537.png", "Validation_47430026/Validation_47430026_47430026_18827.037.png", "Validation_47430026/Validation_47430026_47430026_18828.536.png" ], "turn_id": 3, "scan_id": "Validation_47430026", "origin_question": "", "option": [], "answer": "", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47430026/Validation_47430026_47430026_18830.036.png", "Validation_47430026/Validation_47430026_47430026_18831.535.png", "Validation_47430026/Validation_47430026_47430026_18833.034.png", "Validation_47430026/Validation_47430026_47430026_18834.534.png" ], "turn_id": 4, "scan_id": "Validation_47430026", "origin_question": "In which turn were you closest to \"the rectangular two-pane window with the light-colored frame\"? ", "option": [], "answer": "4", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the rectangular two-pane window with the light-colored frame\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47430026/Validation_47430026_47430026_18836.033.png", "Validation_47430026/Validation_47430026_47430026_18837.532.png", "Validation_47430026/Validation_47430026_47430026_18839.048.png", "Validation_47430026/Validation_47430026_47430026_18840.548.png" ], "turn_id": 5, "scan_id": "Validation_47430026", "origin_question": "", "option": [], "answer": "", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_47430026/Validation_47430026_47430026_18842.047.png", "Validation_47430026/Validation_47430026_47430026_18843.547.png", "Validation_47430026/Validation_47430026_47430026_18845.046.png", "Validation_47430026/Validation_47430026_47430026_18846.545.png" ], "turn_id": 6, "scan_id": "Validation_47430026", "origin_question": "Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42897651/Validation_42897651_42897651_478245.294.png", "Validation_42897651/Validation_42897651_42897651_478247.276.png", "Validation_42897651/Validation_42897651_42897651_478249.275.png", "Validation_42897651/Validation_42897651_42897651_478251.275.png", "Validation_42897651/Validation_42897651_42897651_478253.274.png" ], "turn_id": 1, "scan_id": "Validation_42897651", "origin_question": "Remember, have you seen any cup(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any cup(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_42897651/Validation_42897651_42897651_478255.273.png", "Validation_42897651/Validation_42897651_42897651_478257.272.png", "Validation_42897651/Validation_42897651_42897651_478259.288.png", "Validation_42897651/Validation_42897651_42897651_478261.287.png", "Validation_42897651/Validation_42897651_42897651_478263.286.png" ], "turn_id": 2, "scan_id": "Validation_42897651", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the vibrant red couch with soft fabric upholstery\", Y: \"the rectangular woven white basket made of synthetic material\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the vibrant red couch with soft fabric upholstery\", Y: \"the rectangular woven white basket made of synthetic material\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_42897651/Validation_42897651_42897651_478265.286.png", "Validation_42897651/Validation_42897651_42897651_478267.285.png", "Validation_42897651/Validation_42897651_42897651_478269.284.png", "Validation_42897651/Validation_42897651_42897651_478271.283.png", "Validation_42897651/Validation_42897651_42897651_478273.282.png" ], "turn_id": 3, "scan_id": "Validation_42897651", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the vibrant red couch with soft fabric upholstery\" from your position?", "option": [], "answer": "62.04", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the vibrant red couch with soft fabric upholstery\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_42897651/Validation_42897651_42897651_478275.281.png", "Validation_42897651/Validation_42897651_42897651_478277.281.png", "Validation_42897651/Validation_42897651_42897651_478279.280.png", "Validation_42897651/Validation_42897651_42897651_478281.279.png", "Validation_42897651/Validation_42897651_42897651_478283.278.png" ], "turn_id": 4, "scan_id": "Validation_42897651", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "51.07", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_42897651/Validation_42897651_42897651_478285.277.png", "Validation_42897651/Validation_42897651_42897651_478287.277.png", "Validation_42897651/Validation_42897651_42897651_478289.276.png", "Validation_42897651/Validation_42897651_42897651_478291.275.png", "Validation_42897651/Validation_42897651_42897651_478293.274.png" ], "turn_id": 5, "scan_id": "Validation_42897651", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_42897651/Validation_42897651_42897651_478295.273.png", "Validation_42897651/Validation_42897651_42897651_478297.273.png", "Validation_42897651/Validation_42897651_42897651_478299.288.png", "Validation_42897651/Validation_42897651_42897651_478301.287.png", "Validation_42897651/Validation_42897651_42897651_478303.287.png" ], "turn_id": 6, "scan_id": "Validation_42897651", "origin_question": "How far is your current position from where you were at the end of the turn 5? (in meters)", "option": [], "answer": "1.23", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 5? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_42897651/Validation_42897651_42897651_478305.286.png", "Validation_42897651/Validation_42897651_42897651_478307.285.png", "Validation_42897651/Validation_42897651_42897651_478309.284.png", "Validation_42897651/Validation_42897651_42897651_478311.283.png", "Validation_42897651/Validation_42897651_42897651_478313.283.png" ], "turn_id": 7, "scan_id": "Validation_42897651", "origin_question": "Which direction is the vintage lamp with the curved, warm-hued lampshade in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the vintage lamp with the curved, warm-hued lampshade in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_42897651/Validation_42897651_42897651_478315.282.png", "Validation_42897651/Validation_42897651_42897651_478317.281.png", "Validation_42897651/Validation_42897651_42897651_478319.280.png", "Validation_42897651/Validation_42897651_42897651_478321.279.png", "Validation_42897651/Validation_42897651_42897651_478323.279.png" ], "turn_id": 8, "scan_id": "Validation_42897651", "origin_question": "At the end of which turn were both of \"the vintage lamp with the curved, warm-hued lampshade\" and \"the plush circular stool with light-colored wooden legs\" on your left side?", "option": [], "answer": "7", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the vintage lamp with the curved, warm-hued lampshade\" and \"the plush circular stool with light-colored wooden legs\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_42897651/Validation_42897651_42897651_478325.278.png", "Validation_42897651/Validation_42897651_42897651_478327.277.png", "Validation_42897651/Validation_42897651_42897651_478329.276.png", "Validation_42897651/Validation_42897651_42897651_478331.275.png", "Validation_42897651/Validation_42897651_42897651_478333.274.png" ], "turn_id": 9, "scan_id": "Validation_42897651", "origin_question": "Which direction is the rectangular woven white basket made of synthetic material in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the rectangular woven white basket made of synthetic material in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_42897651/Validation_42897651_42897651_478335.274.png", "Validation_42897651/Validation_42897651_42897651_478337.273.png", "Validation_42897651/Validation_42897651_42897651_478339.289.png", "Validation_42897651/Validation_42897651_42897651_478341.288.png", "Validation_42897651/Validation_42897651_42897651_478343.287.png" ], "turn_id": 10, "scan_id": "Validation_42897651", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the rectangular woven white basket made of synthetic material\" from your position?", "option": [], "answer": "163.39", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the rectangular woven white basket made of synthetic material\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_45663165/Validation_45663165_45663165_65805.035.png", "Validation_45663165/Validation_45663165_45663165_65806.518.png", "Validation_45663165/Validation_45663165_45663165_65808.017.png", "Validation_45663165/Validation_45663165_45663165_65809.516.png" ], "turn_id": 1, "scan_id": "Validation_45663165", "origin_question": "Remember, have you seen any lamp(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any lamp(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_45663165/Validation_45663165_45663165_65811.016.png", "Validation_45663165/Validation_45663165_45663165_65812.515.png", "Validation_45663165/Validation_45663165_45663165_65814.015.png", "Validation_45663165/Validation_45663165_45663165_65815.514.png" ], "turn_id": 2, "scan_id": "Validation_45663165", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the sturdy wooden chair with the light brown surface\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the sturdy wooden chair with the light brown surface\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_45663165/Validation_45663165_45663165_65817.013.png", "Validation_45663165/Validation_45663165_45663165_65818.513.png", "Validation_45663165/Validation_45663165_45663165_65820.012.png", "Validation_45663165/Validation_45663165_45663165_65821.528.png" ], "turn_id": 3, "scan_id": "Validation_45663165", "origin_question": "What will be the first-time appearance order of the following categories: chandelier, door, refrigerator?", "option": [ "door, refrigerator, chandelier", "chandelier, refrigerator, door", "door, chandelier, refrigerator", "chandelier, door, refrigerator" ], "answer": "door, chandelier, refrigerator", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: chandelier, door, refrigerator? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['door, refrigerator, chandelier', 'chandelier, refrigerator, door', 'door, chandelier, refrigerator', 'chandelier, door, refrigerator'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_45663165/Validation_45663165_45663165_65823.028.png", "Validation_45663165/Validation_45663165_45663165_65824.527.png", "Validation_45663165/Validation_45663165_45663165_65826.026.png", "Validation_45663165/Validation_45663165_45663165_65827.526.png" ], "turn_id": 4, "scan_id": "Validation_45663165", "origin_question": "", "option": [], "answer": "", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_45663165/Validation_45663165_45663165_65829.025.png", "Validation_45663165/Validation_45663165_45663165_65830.525.png", "Validation_45663165/Validation_45663165_45663165_65832.024.png", "Validation_45663165/Validation_45663165_45663165_65833.523.png" ], "turn_id": 5, "scan_id": "Validation_45663165", "origin_question": "Is the window to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the window to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_45663165/Validation_45663165_45663165_65835.023.png", "Validation_45663165/Validation_45663165_45663165_65836.522.png", "Validation_45663165/Validation_45663165_45663165_65838.022.png", "Validation_45663165/Validation_45663165_45663165_65839.521.png" ], "turn_id": 6, "scan_id": "Validation_45663165", "origin_question": "Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Right", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_45663165/Validation_45663165_45663165_65841.020.png", "Validation_45663165/Validation_45663165_45663165_65842.520.png", "Validation_45663165/Validation_45663165_45663165_65844.019.png", "Validation_45663165/Validation_45663165_45663165_65845.519.png" ], "turn_id": 7, "scan_id": "Validation_45663165", "origin_question": "By the end of the 3 turn, were you getting closer to or farther away from these two objects: X: \"the sturdy wooden chair with the light brown surface\", Y: \"the medium-sized blue plastic bottle with the cylindrical shape\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to Y and farther away from X", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 3 turn, were you getting closer to or farther away from these two objects: X: \"the sturdy wooden chair with the light brown surface\", Y: \"the medium-sized blue plastic bottle with the cylindrical shape\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47430048/Validation_47430048_47430048_21045.757.png", "Validation_47430048/Validation_47430048_47430048_21047.240.png", "Validation_47430048/Validation_47430048_47430048_21048.739.png", "Validation_47430048/Validation_47430048_47430048_21050.239.png" ], "turn_id": 1, "scan_id": "Validation_47430048", "origin_question": "Remember, have you seen any adhesive tape(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any adhesive tape(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_47430048/Validation_47430048_47430048_21051.738.png", "Validation_47430048/Validation_47430048_47430048_21053.238.png", "Validation_47430048/Validation_47430048_47430048_21054.737.png", "Validation_47430048/Validation_47430048_47430048_21056.236.png" ], "turn_id": 2, "scan_id": "Validation_47430048", "origin_question": "Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Left", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_47430048/Validation_47430048_47430048_21057.736.png", "Validation_47430048/Validation_47430048_47430048_21059.235.png", "Validation_47430048/Validation_47430048_47430048_21060.734.png", "Validation_47430048/Validation_47430048_47430048_21062.234.png" ], "turn_id": 3, "scan_id": "Validation_47430048", "origin_question": "Please recall \"the classic round white porcelain toilet\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.08", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the classic round white porcelain toilet\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_47430048/Validation_47430048_47430048_21063.733.png", "Validation_47430048/Validation_47430048_47430048_21065.233.png", "Validation_47430048/Validation_47430048_47430048_21066.732.png", "Validation_47430048/Validation_47430048_47430048_21068.232.png" ], "turn_id": 4, "scan_id": "Validation_47430048", "origin_question": "Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "Validation_47430048/Validation_47430048_47430048_21069.731.png", "Validation_47430048/Validation_47430048_47430048_21071.230.png", "Validation_47430048/Validation_47430048_47430048_21072.730.png", "Validation_47430048/Validation_47430048_47430048_21074.229.png" ], "turn_id": 5, "scan_id": "Validation_47430048", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the grey rectangular plastic bucket with rounded edges\"; \"the compact white front-loading washing machine\"; \"the sleek white cylindrical bin\" ", "option": [ "the grey rectangular plastic bucket with rounded edges", "the compact white front-loading washing machine", "the sleek white cylindrical bin" ], "answer": "the sleek white cylindrical bin", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the grey rectangular plastic bucket with rounded edges\"; \"the compact white front-loading washing machine\"; \"the sleek white cylindrical bin\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the grey rectangular plastic bucket with rounded edges', 'the compact white front-loading washing machine', 'the sleek white cylindrical bin'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_47430048/Validation_47430048_47430048_21075.728.png", "Validation_47430048/Validation_47430048_47430048_21077.228.png", "Validation_47430048/Validation_47430048_47430048_21078.727.png", "Validation_47430048/Validation_47430048_47430048_21080.227.png" ], "turn_id": 6, "scan_id": "Validation_47430048", "origin_question": "Please recall \"an oval white glossy sink\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.45", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"an oval white glossy sink\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_42445966/Validation_42445966_42445966_347040.233.png", "Validation_42445966/Validation_42445966_42445966_347042.232.png", "Validation_42445966/Validation_42445966_42445966_347044.232.png", "Validation_42445966/Validation_42445966_42445966_347046.231.png", "Validation_42445966/Validation_42445966_42445966_347048.230.png" ], "turn_id": 1, "scan_id": "Validation_42445966", "origin_question": "", "option": [], "answer": "", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_42445966/Validation_42445966_42445966_347050.229.png", "Validation_42445966/Validation_42445966_42445966_347052.228.png", "Validation_42445966/Validation_42445966_42445966_347054.227.png", "Validation_42445966/Validation_42445966_42445966_347056.227.png", "Validation_42445966/Validation_42445966_42445966_347058.226.png" ], "turn_id": 2, "scan_id": "Validation_42445966", "origin_question": "When did you first discover \"the sturdy medium brown wooden chair with slatted backrest\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the sturdy medium brown wooden chair with slatted backrest\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_42445966/Validation_42445966_42445966_347060.242.png", "Validation_42445966/Validation_42445966_42445966_347062.241.png", "Validation_42445966/Validation_42445966_42445966_347064.240.png", "Validation_42445966/Validation_42445966_42445966_347066.239.png", "Validation_42445966/Validation_42445966_42445966_347068.238.png" ], "turn_id": 3, "scan_id": "Validation_42445966", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_42445966/Validation_42445966_42445966_347070.238.png", "Validation_42445966/Validation_42445966_42445966_347072.237.png", "Validation_42445966/Validation_42445966_42445966_347074.236.png", "Validation_42445966/Validation_42445966_42445966_347076.235.png", "Validation_42445966/Validation_42445966_42445966_347078.234.png" ], "turn_id": 4, "scan_id": "Validation_42445966", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the warm light-brown wooden cabinet with glass doors\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the warm light-brown wooden cabinet with glass doors\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_42445966/Validation_42445966_42445966_347080.233.png", "Validation_42445966/Validation_42445966_42445966_347082.233.png", "Validation_42445966/Validation_42445966_42445966_347084.232.png", "Validation_42445966/Validation_42445966_42445966_347086.231.png", "Validation_42445966/Validation_42445966_42445966_347088.230.png" ], "turn_id": 5, "scan_id": "Validation_42445966", "origin_question": "What will be the first-time appearance order of the following categories: chair, plant, window?", "option": [ "window, plant, chair", "plant, chair, window", "window, chair, plant", "chair, window, plant" ], "answer": "chair, window, plant", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: chair, plant, window? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['window, plant, chair', 'plant, chair, window', 'window, chair, plant', 'chair, window, plant'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_42445966/Validation_42445966_42445966_347090.229.png", "Validation_42445966/Validation_42445966_42445966_347092.229.png", "Validation_42445966/Validation_42445966_42445966_347094.228.png", "Validation_42445966/Validation_42445966_42445966_347096.227.png", "Validation_42445966/Validation_42445966_42445966_347098.226.png" ], "turn_id": 6, "scan_id": "Validation_42445966", "origin_question": "Compared to your position at the end of the 5 turn, are you now closer or farther away from \"the warm light-brown wooden cabinet with glass doors\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 5 turn, are you now closer or farther away from \"the warm light-brown wooden cabinet with glass doors\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_42445966/Validation_42445966_42445966_347100.242.png", "Validation_42445966/Validation_42445966_42445966_347102.241.png", "Validation_42445966/Validation_42445966_42445966_347104.240.png", "Validation_42445966/Validation_42445966_42445966_347106.240.png", "Validation_42445966/Validation_42445966_42445966_347108.239.png" ], "turn_id": 7, "scan_id": "Validation_42445966", "origin_question": "Assuming the direction you are facing at the end of the turn 4 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Left", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 4 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_42445966/Validation_42445966_42445966_347110.238.png", "Validation_42445966/Validation_42445966_42445966_347112.237.png", "Validation_42445966/Validation_42445966_42445966_347114.236.png", "Validation_42445966/Validation_42445966_42445966_347116.235.png", "Validation_42445966/Validation_42445966_42445966_347118.235.png" ], "turn_id": 8, "scan_id": "Validation_42445966", "origin_question": "When did you first discover \"an upside-down artificial plant with decorative colors\"? (index of the turn)", "option": [], "answer": "5", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"an upside-down artificial plant with decorative colors\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_42445966/Validation_42445966_42445966_347120.234.png", "Validation_42445966/Validation_42445966_42445966_347122.233.png", "Validation_42445966/Validation_42445966_42445966_347124.232.png", "Validation_42445966/Validation_42445966_42445966_347126.231.png", "Validation_42445966/Validation_42445966_42445966_347128.231.png" ], "turn_id": 9, "scan_id": "Validation_42445966", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "2.58", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_42445966/Validation_42445966_42445966_347130.230.png", "Validation_42445966/Validation_42445966_42445966_347132.229.png", "Validation_42445966/Validation_42445966_42445966_347134.228.png", "Validation_42445966/Validation_42445966_42445966_347136.227.png", "Validation_42445966/Validation_42445966_42445966_347138.226.png" ], "turn_id": 10, "scan_id": "Validation_42445966", "origin_question": "How far is your current position from where you were at the end of the turn 7? (in meters)", "option": [], "answer": "2.93", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 7? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "Validation_41159557/Validation_41159557_41159557_33191.279.png", "Validation_41159557/Validation_41159557_41159557_33193.262.png", "Validation_41159557/Validation_41159557_41159557_33195.261.png", "Validation_41159557/Validation_41159557_41159557_33197.260.png", "Validation_41159557/Validation_41159557_41159557_33199.260.png" ], "turn_id": 1, "scan_id": "Validation_41159557", "origin_question": "Remember, how many pillow(s) have you seen so far? ", "option": [], "answer": "3", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many pillow(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_41159557/Validation_41159557_41159557_33201.259.png", "Validation_41159557/Validation_41159557_41159557_33203.258.png", "Validation_41159557/Validation_41159557_41159557_33205.257.png", "Validation_41159557/Validation_41159557_41159557_33207.256.png", "Validation_41159557/Validation_41159557_41159557_33209.256.png" ], "turn_id": 2, "scan_id": "Validation_41159557", "origin_question": "Which direction is the calming green rectangular blanket with blue accents in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the calming green rectangular blanket with blue accents in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "Validation_41159557/Validation_41159557_41159557_33211.255.png", "Validation_41159557/Validation_41159557_41159557_33213.254.png", "Validation_41159557/Validation_41159557_41159557_33215.253.png", "Validation_41159557/Validation_41159557_41159557_33217.253.png", "Validation_41159557/Validation_41159557_41159557_33219.252.png" ], "turn_id": 3, "scan_id": "Validation_41159557", "origin_question": "Among these three objects(\"the sleek, medium brown wooden shelf\",\"the calming green rectangular blanket with blue accents\",\"the rustic brown wooden table\" ), which one is the closest to you?", "option": [ "the sleek, medium brown wooden shelf", "the calming green rectangular blanket with blue accents", "the rustic brown wooden table" ], "answer": "the sleek, medium brown wooden shelf", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the sleek, medium brown wooden shelf\",\"the calming green rectangular blanket with blue accents\",\"the rustic brown wooden table\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the sleek, medium brown wooden shelf', 'the calming green rectangular blanket with blue accents', 'the rustic brown wooden table'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_41159557/Validation_41159557_41159557_33221.251.png", "Validation_41159557/Validation_41159557_41159557_33223.250.png", "Validation_41159557/Validation_41159557_41159557_33225.249.png", "Validation_41159557/Validation_41159557_41159557_33227.249.png", "Validation_41159557/Validation_41159557_41159557_33229.264.png" ], "turn_id": 4, "scan_id": "Validation_41159557", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "1.69", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_41159557/Validation_41159557_41159557_33231.264.png", "Validation_41159557/Validation_41159557_41159557_33233.263.png", "Validation_41159557/Validation_41159557_41159557_33235.262.png", "Validation_41159557/Validation_41159557_41159557_33237.261.png", "Validation_41159557/Validation_41159557_41159557_33239.260.png" ], "turn_id": 5, "scan_id": "Validation_41159557", "origin_question": "When did you first discover \"the burgundy curtain made of durable patterned fabric\"? (index of the turn)", "option": [], "answer": "4", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the burgundy curtain made of durable patterned fabric\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_41159557/Validation_41159557_41159557_33241.260.png", "Validation_41159557/Validation_41159557_41159557_33243.259.png", "Validation_41159557/Validation_41159557_41159557_33245.258.png", "Validation_41159557/Validation_41159557_41159557_33247.257.png", "Validation_41159557/Validation_41159557_41159557_33249.257.png" ], "turn_id": 6, "scan_id": "Validation_41159557", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "1.53", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_41159557/Validation_41159557_41159557_33251.256.png", "Validation_41159557/Validation_41159557_41159557_33253.255.png", "Validation_41159557/Validation_41159557_41159557_33255.254.png", "Validation_41159557/Validation_41159557_41159557_33257.253.png", "Validation_41159557/Validation_41159557_41159557_33259.253.png" ], "turn_id": 7, "scan_id": "Validation_41159557", "origin_question": "Please recall \"the cylindrical light fixture with an off-white lampshade\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.74", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the cylindrical light fixture with an off-white lampshade\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_41159557/Validation_41159557_41159557_33261.252.png", "Validation_41159557/Validation_41159557_41159557_33263.251.png", "Validation_41159557/Validation_41159557_41159557_33265.250.png", "Validation_41159557/Validation_41159557_41159557_33267.249.png", "Validation_41159557/Validation_41159557_41159557_33269.249.png" ], "turn_id": 8, "scan_id": "Validation_41159557", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "40.37", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_41159557/Validation_41159557_41159557_33271.264.png", "Validation_41159557/Validation_41159557_41159557_33273.264.png", "Validation_41159557/Validation_41159557_41159557_33275.263.png", "Validation_41159557/Validation_41159557_41159557_33277.262.png", "Validation_41159557/Validation_41159557_41159557_33279.261.png" ], "turn_id": 9, "scan_id": "Validation_41159557", "origin_question": "Which direction is an elegant light brown wooden door in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is an elegant light brown wooden door in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_41159557/Validation_41159557_41159557_33281.260.png", "Validation_41159557/Validation_41159557_41159557_33283.260.png", "Validation_41159557/Validation_41159557_41159557_33285.259.png", "Validation_41159557/Validation_41159557_41159557_33287.258.png", "Validation_41159557/Validation_41159557_41159557_33289.257.png" ], "turn_id": 10, "scan_id": "Validation_41159557", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the vibrant geometric patterned carpet\" from your position?", "option": [], "answer": "157.57", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the vibrant geometric patterned carpet\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47334252/Validation_47334252_47334252_112080.992.png", "Validation_47334252/Validation_47334252_47334252_112082.992.png", "Validation_47334252/Validation_47334252_47334252_112084.991.png", "Validation_47334252/Validation_47334252_47334252_112086.990.png", "Validation_47334252/Validation_47334252_47334252_112088.989.png" ], "turn_id": 1, "scan_id": "Validation_47334252", "origin_question": "Remember, have you seen any table(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any table(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_47334252/Validation_47334252_47334252_112090.989.png", "Validation_47334252/Validation_47334252_47334252_112092.988.png", "Validation_47334252/Validation_47334252_47334252_112094.987.png", "Validation_47334252/Validation_47334252_47334252_112096.986.png", "Validation_47334252/Validation_47334252_47334252_112098.985.png" ], "turn_id": 2, "scan_id": "Validation_47334252", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the medium-sized rectangular window with the white frame\", Y: \"the sleek circular pan with the silver interior\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the medium-sized rectangular window with the white frame\", Y: \"the sleek circular pan with the silver interior\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_47334252/Validation_47334252_47334252_112100.985.png", "Validation_47334252/Validation_47334252_47334252_112102.984.png", "Validation_47334252/Validation_47334252_47334252_112105.000.png", "Validation_47334252/Validation_47334252_47334252_112106.999.png", "Validation_47334252/Validation_47334252_47334252_112108.998.png" ], "turn_id": 3, "scan_id": "Validation_47334252", "origin_question": "Please recall \"the small, round table with the warm brown finish\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.16", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the small, round table with the warm brown finish\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_47334252/Validation_47334252_47334252_112110.997.png", "Validation_47334252/Validation_47334252_47334252_112112.996.png", "Validation_47334252/Validation_47334252_47334252_112114.996.png", "Validation_47334252/Validation_47334252_47334252_112116.995.png", "Validation_47334252/Validation_47334252_47334252_112118.994.png" ], "turn_id": 4, "scan_id": "Validation_47334252", "origin_question": "When did you first discover \"the sleek circular pan with the silver interior\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the sleek circular pan with the silver interior\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47334252/Validation_47334252_47334252_112120.993.png", "Validation_47334252/Validation_47334252_47334252_112122.992.png", "Validation_47334252/Validation_47334252_47334252_112124.991.png", "Validation_47334252/Validation_47334252_47334252_112126.991.png", "Validation_47334252/Validation_47334252_47334252_112128.990.png" ], "turn_id": 5, "scan_id": "Validation_47334252", "origin_question": "In which turn were you farthest from \"the small, round table with the warm brown finish\"? ", "option": [], "answer": "1", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the small, round table with the warm brown finish\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_47334252/Validation_47334252_47334252_112130.989.png", "Validation_47334252/Validation_47334252_47334252_112132.988.png", "Validation_47334252/Validation_47334252_47334252_112134.987.png", "Validation_47334252/Validation_47334252_47334252_112136.987.png", "Validation_47334252/Validation_47334252_47334252_112138.986.png" ], "turn_id": 6, "scan_id": "Validation_47334252", "origin_question": "Please recall \"the small, round table with the warm brown finish\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.89", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the small, round table with the warm brown finish\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "Validation_47334252/Validation_47334252_47334252_112140.985.png", "Validation_47334252/Validation_47334252_47334252_112142.984.png", "Validation_47334252/Validation_47334252_47334252_112144.983.png", "Validation_47334252/Validation_47334252_47334252_112146.999.png", "Validation_47334252/Validation_47334252_47334252_112148.998.png" ], "turn_id": 7, "scan_id": "Validation_47334252", "origin_question": "Among these three objects(\"the small, round table with the warm brown finish\",\"the medium-sized rectangular window with the white frame\",\"the sleek circular pan with the silver interior\" ), which one is the closest to you?", "option": [ "the small, round table with the warm brown finish", "the medium-sized rectangular window with the white frame", "the sleek circular pan with the silver interior" ], "answer": "the sleek circular pan with the silver interior", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the small, round table with the warm brown finish\",\"the medium-sized rectangular window with the white frame\",\"the sleek circular pan with the silver interior\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the small, round table with the warm brown finish', 'the medium-sized rectangular window with the white frame', 'the sleek circular pan with the silver interior'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_47334252/Validation_47334252_47334252_112150.998.png", "Validation_47334252/Validation_47334252_47334252_112152.997.png", "Validation_47334252/Validation_47334252_47334252_112154.996.png", "Validation_47334252/Validation_47334252_47334252_112156.995.png", "Validation_47334252/Validation_47334252_47334252_112158.994.png" ], "turn_id": 8, "scan_id": "Validation_47334252", "origin_question": "Please recall \"the medium-sized rectangular window with the white frame\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.7", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the medium-sized rectangular window with the white frame\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_47334252/Validation_47334252_47334252_112160.993.png", "Validation_47334252/Validation_47334252_47334252_112162.993.png", "Validation_47334252/Validation_47334252_47334252_112164.992.png", "Validation_47334252/Validation_47334252_47334252_112166.991.png", "Validation_47334252/Validation_47334252_47334252_112168.990.png" ], "turn_id": 9, "scan_id": "Validation_47334252", "origin_question": "Which two objects are on the same side of you now? \"the small, round table with the warm brown finish\", \"the medium-sized rectangular window with the white frame\", \"the sink\". ", "option": [ "the small, round table with the warm brown finish, the medium-sized rectangular window with the white frame", "the small, round table with the warm brown finish, the sink", "the medium-sized rectangular window with the white frame, the sink" ], "answer": "the small, round table with the warm brown finish, the medium-sized rectangular window with the white frame", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the small, round table with the warm brown finish\", \"the medium-sized rectangular window with the white frame\", \"the sink\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the small, round table with the warm brown finish, the medium-sized rectangular window with the white frame', 'the small, round table with the warm brown finish, the sink', 'the medium-sized rectangular window with the white frame, the sink'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47334252/Validation_47334252_47334252_112170.989.png", "Validation_47334252/Validation_47334252_47334252_112172.989.png", "Validation_47334252/Validation_47334252_47334252_112174.988.png", "Validation_47334252/Validation_47334252_47334252_112176.987.png", "Validation_47334252/Validation_47334252_47334252_112178.986.png" ], "turn_id": 10, "scan_id": "Validation_47334252", "origin_question": "", "option": [], "answer": "", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_48018732/Validation_48018732_48018732_1913.575.png", "Validation_48018732/Validation_48018732_48018732_1915.574.png", "Validation_48018732/Validation_48018732_48018732_1917.573.png", "Validation_48018732/Validation_48018732_48018732_1919.573.png", "Validation_48018732/Validation_48018732_48018732_1921.572.png" ], "turn_id": 1, "scan_id": "Validation_48018732", "origin_question": "Remember, have you seen any candle(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any candle(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_48018732/Validation_48018732_48018732_1923.571.png", "Validation_48018732/Validation_48018732_48018732_1925.570.png", "Validation_48018732/Validation_48018732_48018732_1927.569.png", "Validation_48018732/Validation_48018732_48018732_1929.569.png", "Validation_48018732/Validation_48018732_48018732_1931.568.png" ], "turn_id": 2, "scan_id": "Validation_48018732", "origin_question": "In which turn did you see both \"the medium-brown wooden bed with the patchwork cover\" and \"the compact speaker in dark hues with plastic and metal\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the medium-brown wooden bed with the patchwork cover\" and \"the compact speaker in dark hues with plastic and metal\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_48018732/Validation_48018732_48018732_1933.567.png", "Validation_48018732/Validation_48018732_48018732_1935.566.png", "Validation_48018732/Validation_48018732_48018732_1937.565.png", "Validation_48018732/Validation_48018732_48018732_1939.564.png", "Validation_48018732/Validation_48018732_48018732_1941.564.png" ], "turn_id": 3, "scan_id": "Validation_48018732", "origin_question": "Please recall \"the spacious brown couch with sturdy armrests\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.18", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the spacious brown couch with sturdy armrests\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_48018732/Validation_48018732_48018732_1943.563.png", "Validation_48018732/Validation_48018732_48018732_1945.579.png", "Validation_48018732/Validation_48018732_48018732_1947.578.png", "Validation_48018732/Validation_48018732_48018732_1949.577.png", "Validation_48018732/Validation_48018732_48018732_1951.576.png" ], "turn_id": 4, "scan_id": "Validation_48018732", "origin_question": "Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "127.98", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_48018732/Validation_48018732_48018732_1953.575.png", "Validation_48018732/Validation_48018732_48018732_1955.575.png", "Validation_48018732/Validation_48018732_48018732_1957.574.png", "Validation_48018732/Validation_48018732_48018732_1959.573.png", "Validation_48018732/Validation_48018732_48018732_1961.572.png" ], "turn_id": 5, "scan_id": "Validation_48018732", "origin_question": "Is the spacious brown couch with sturdy armrests to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the spacious brown couch with sturdy armrests to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_48018732/Validation_48018732_48018732_1963.571.png", "Validation_48018732/Validation_48018732_48018732_1965.571.png", "Validation_48018732/Validation_48018732_48018732_1967.570.png", "Validation_48018732/Validation_48018732_48018732_1969.569.png", "Validation_48018732/Validation_48018732_48018732_1971.568.png" ], "turn_id": 6, "scan_id": "Validation_48018732", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the medium-brown wooden bed with the patchwork cover\" from your position?", "option": [], "answer": "178.46", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the medium-brown wooden bed with the patchwork cover\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_48018732/Validation_48018732_48018732_1973.567.png", "Validation_48018732/Validation_48018732_48018732_1975.566.png", "Validation_48018732/Validation_48018732_48018732_1977.566.png", "Validation_48018732/Validation_48018732_48018732_1979.565.png", "Validation_48018732/Validation_48018732_48018732_1981.564.png" ], "turn_id": 7, "scan_id": "Validation_48018732", "origin_question": "How far is your current position from where you were at the end of the turn 3? (in meters)", "option": [], "answer": "2.83", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 3? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_48018732/Validation_48018732_48018732_1983.563.png", "Validation_48018732/Validation_48018732_48018732_1985.579.png", "Validation_48018732/Validation_48018732_48018732_1987.578.png", "Validation_48018732/Validation_48018732_48018732_1989.578.png", "Validation_48018732/Validation_48018732_48018732_1991.577.png" ], "turn_id": 8, "scan_id": "Validation_48018732", "origin_question": "When did you first discover \"the large rectangular window framed in light material\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the large rectangular window framed in light material\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47204609/Validation_47204609_47204609_2453.013.png", "Validation_47204609/Validation_47204609_47204609_2454.995.png", "Validation_47204609/Validation_47204609_47204609_2456.994.png", "Validation_47204609/Validation_47204609_47204609_2458.994.png" ], "turn_id": 1, "scan_id": "Validation_47204609", "origin_question": "Remember, have you seen any dvd(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any dvd(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "Validation_47204609/Validation_47204609_47204609_2460.993.png", "Validation_47204609/Validation_47204609_47204609_2462.992.png", "Validation_47204609/Validation_47204609_47204609_2464.991.png", "Validation_47204609/Validation_47204609_47204609_2466.990.png" ], "turn_id": 2, "scan_id": "Validation_47204609", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the stylish black rectangular monitor with thin bezels\"; \"the sleek, modern dark glass table\"; \"the honey-toned wooden door with the smooth finish\" ", "option": [ "the stylish black rectangular monitor with thin bezels", "the sleek, modern dark glass table", "the honey-toned wooden door with the smooth finish" ], "answer": "the stylish black rectangular monitor with thin bezels", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the stylish black rectangular monitor with thin bezels\"; \"the sleek, modern dark glass table\"; \"the honey-toned wooden door with the smooth finish\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the stylish black rectangular monitor with thin bezels', 'the sleek, modern dark glass table', 'the honey-toned wooden door with the smooth finish'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_47204609/Validation_47204609_47204609_2468.990.png", "Validation_47204609/Validation_47204609_47204609_2470.989.png", "Validation_47204609/Validation_47204609_47204609_2473.005.png", "Validation_47204609/Validation_47204609_47204609_2475.004.png" ], "turn_id": 3, "scan_id": "Validation_47204609", "origin_question": "How far is your current position from where you were at the end of the turn 1? (in meters)", "option": [], "answer": "3.52", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 1? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_47204609/Validation_47204609_47204609_2477.003.png", "Validation_47204609/Validation_47204609_47204609_2479.002.png", "Validation_47204609/Validation_47204609_47204609_2481.002.png", "Validation_47204609/Validation_47204609_47204609_2483.001.png" ], "turn_id": 4, "scan_id": "Validation_47204609", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the honey-toned wooden door with the smooth finish\", Y: \"the stylish black rectangular monitor with thin bezels\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to Y and farther away from X", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the honey-toned wooden door with the smooth finish\", Y: \"the stylish black rectangular monitor with thin bezels\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_47204609/Validation_47204609_47204609_2485.000.png", "Validation_47204609/Validation_47204609_47204609_2486.999.png", "Validation_47204609/Validation_47204609_47204609_2488.998.png", "Validation_47204609/Validation_47204609_47204609_2490.998.png" ], "turn_id": 5, "scan_id": "Validation_47204609", "origin_question": "How far is your current position from where you were at the end of the turn 4? (in meters)", "option": [], "answer": "2.06", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 4? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_47204609/Validation_47204609_47204609_2492.997.png", "Validation_47204609/Validation_47204609_47204609_2494.996.png", "Validation_47204609/Validation_47204609_47204609_2496.995.png", "Validation_47204609/Validation_47204609_47204609_2498.994.png" ], "turn_id": 6, "scan_id": "Validation_47204609", "origin_question": "What will be the first-time appearance order of the following categories: carpet, monitor, window?", "option": [ "carpet, window, monitor", "window, monitor, carpet", "window, carpet, monitor", "carpet, monitor, window" ], "answer": "carpet, monitor, window", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: carpet, monitor, window? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['carpet, window, monitor', 'window, monitor, carpet', 'window, carpet, monitor', 'carpet, monitor, window'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47204609/Validation_47204609_47204609_2500.994.png", "Validation_47204609/Validation_47204609_47204609_2502.993.png", "Validation_47204609/Validation_47204609_47204609_2504.992.png", "Validation_47204609/Validation_47204609_47204609_2506.991.png" ], "turn_id": 7, "scan_id": "Validation_47204609", "origin_question": "In which turn were you closest to \"the rectangular natural wood tone frame\"? ", "option": [], "answer": "5", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the rectangular natural wood tone frame\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47895355/Validation_47895355_47895355_98754.653.png", "Validation_47895355/Validation_47895355_47895355_98756.636.png", "Validation_47895355/Validation_47895355_47895355_98758.635.png", "Validation_47895355/Validation_47895355_47895355_98760.634.png", "Validation_47895355/Validation_47895355_47895355_98762.633.png" ], "turn_id": 1, "scan_id": "Validation_47895355", "origin_question": "", "option": [], "answer": "", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47895355/Validation_47895355_47895355_98764.632.png", "Validation_47895355/Validation_47895355_47895355_98766.632.png", "Validation_47895355/Validation_47895355_47895355_98768.631.png", "Validation_47895355/Validation_47895355_47895355_98770.630.png", "Validation_47895355/Validation_47895355_47895355_98772.629.png" ], "turn_id": 2, "scan_id": "Validation_47895355", "origin_question": "Remember, have you seen any blackboard(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any blackboard(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47895355/Validation_47895355_47895355_98774.628.png", "Validation_47895355/Validation_47895355_47895355_98776.627.png", "Validation_47895355/Validation_47895355_47895355_98778.627.png", "Validation_47895355/Validation_47895355_47895355_98780.626.png", "Validation_47895355/Validation_47895355_47895355_98782.625.png" ], "turn_id": 3, "scan_id": "Validation_47895355", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the modern compact cabinet with the dark glossy finish\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the modern compact cabinet with the dark glossy finish\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47895355/Validation_47895355_47895355_98784.641.png", "Validation_47895355/Validation_47895355_47895355_98786.640.png", "Validation_47895355/Validation_47895355_47895355_98788.639.png", "Validation_47895355/Validation_47895355_47895355_98790.638.png", "Validation_47895355/Validation_47895355_47895355_98792.638.png" ], "turn_id": 4, "scan_id": "Validation_47895355", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the dark fabric stool with light-colored stars\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the dark fabric stool with light-colored stars\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_47895355/Validation_47895355_47895355_98794.637.png", "Validation_47895355/Validation_47895355_47895355_98796.636.png", "Validation_47895355/Validation_47895355_47895355_98798.635.png", "Validation_47895355/Validation_47895355_47895355_98800.634.png", "Validation_47895355/Validation_47895355_47895355_98802.633.png" ], "turn_id": 5, "scan_id": "Validation_47895355", "origin_question": "What will be the first-time appearance order of the following categories: cabinet, kettle, stool?", "option": [ "kettle, cabinet, stool", "cabinet, stool, kettle", "cabinet, kettle, stool", "stool, cabinet, kettle" ], "answer": "stool, cabinet, kettle", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: cabinet, kettle, stool? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['kettle, cabinet, stool', 'cabinet, stool, kettle', 'cabinet, kettle, stool', 'stool, cabinet, kettle'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_47895355/Validation_47895355_47895355_98804.633.png", "Validation_47895355/Validation_47895355_47895355_98806.632.png", "Validation_47895355/Validation_47895355_47895355_98808.631.png", "Validation_47895355/Validation_47895355_47895355_98810.630.png", "Validation_47895355/Validation_47895355_47895355_98812.629.png" ], "turn_id": 6, "scan_id": "Validation_47895355", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47895355/Validation_47895355_47895355_98814.629.png", "Validation_47895355/Validation_47895355_47895355_98816.628.png", "Validation_47895355/Validation_47895355_47895355_98818.627.png", "Validation_47895355/Validation_47895355_47895355_98820.626.png", "Validation_47895355/Validation_47895355_47895355_98822.625.png" ], "turn_id": 7, "scan_id": "Validation_47895355", "origin_question": "Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the sleek stainless steel kitchenware with the black handle\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the sleek stainless steel kitchenware with the black handle\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47895355/Validation_47895355_47895355_98824.641.png", "Validation_47895355/Validation_47895355_47895355_98826.640.png", "Validation_47895355/Validation_47895355_47895355_98828.640.png", "Validation_47895355/Validation_47895355_47895355_98830.639.png", "Validation_47895355/Validation_47895355_47895355_98832.638.png" ], "turn_id": 8, "scan_id": "Validation_47895355", "origin_question": "In which turn were you closest to \"the sleek, deep green glass bottle\"? ", "option": [], "answer": "8", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the sleek, deep green glass bottle\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_47895355/Validation_47895355_47895355_98834.637.png", "Validation_47895355/Validation_47895355_47895355_98836.636.png", "Validation_47895355/Validation_47895355_47895355_98838.636.png", "Validation_47895355/Validation_47895355_47895355_98840.635.png", "Validation_47895355/Validation_47895355_47895355_98842.634.png" ], "turn_id": 9, "scan_id": "Validation_47895355", "origin_question": "Please recall \"the dark fabric stool with light-colored stars\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.21", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the dark fabric stool with light-colored stars\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_47895355/Validation_47895355_47895355_98844.633.png", "Validation_47895355/Validation_47895355_47895355_98846.632.png", "Validation_47895355/Validation_47895355_47895355_98848.631.png", "Validation_47895355/Validation_47895355_47895355_98850.631.png", "Validation_47895355/Validation_47895355_47895355_98852.630.png" ], "turn_id": 10, "scan_id": "Validation_47895355", "origin_question": "Which two objects are on the same side of you now? \"the sleek oval table with the glossy white surface\", \"the sleek, deep green glass bottle\", \"the contemporary rectangular light beige sink\". ", "option": [ "the sleek oval table with the glossy white surface, the sleek, deep green glass bottle", "the sleek oval table with the glossy white surface, the contemporary rectangular light beige sink", "the sleek, deep green glass bottle, the contemporary rectangular light beige sink" ], "answer": "the sleek oval table with the glossy white surface, the sleek, deep green glass bottle", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the sleek oval table with the glossy white surface\", \"the sleek, deep green glass bottle\", \"the contemporary rectangular light beige sink\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the sleek oval table with the glossy white surface, the sleek, deep green glass bottle', 'the sleek oval table with the glossy white surface, the contemporary rectangular light beige sink', 'the sleek, deep green glass bottle, the contemporary rectangular light beige sink'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "Validation_44358440/Validation_44358440_44358440_42253.519.png", "Validation_44358440/Validation_44358440_44358440_42255.502.png", "Validation_44358440/Validation_44358440_44358440_42257.501.png", "Validation_44358440/Validation_44358440_44358440_42259.500.png" ], "turn_id": 1, "scan_id": "Validation_44358440", "origin_question": "Remember, how many chair(s) have you seen so far? ", "option": [], "answer": "2", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many chair(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_44358440/Validation_44358440_44358440_42261.499.png", "Validation_44358440/Validation_44358440_44358440_42263.498.png", "Validation_44358440/Validation_44358440_44358440_42265.497.png", "Validation_44358440/Validation_44358440_44358440_42267.497.png" ], "turn_id": 2, "scan_id": "Validation_44358440", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_44358440/Validation_44358440_44358440_42269.496.png", "Validation_44358440/Validation_44358440_44358440_42271.495.png", "Validation_44358440/Validation_44358440_44358440_42273.494.png", "Validation_44358440/Validation_44358440_44358440_42275.493.png" ], "turn_id": 3, "scan_id": "Validation_44358440", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the modern silver dome-shaped lamp with the stable base\" from your position?", "option": [], "answer": "63.67", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the modern silver dome-shaped lamp with the stable base\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_44358440/Validation_44358440_44358440_42277.493.png", "Validation_44358440/Validation_44358440_44358440_42279.492.png", "Validation_44358440/Validation_44358440_44358440_42281.491.png", "Validation_44358440/Validation_44358440_44358440_42283.490.png" ], "turn_id": 4, "scan_id": "Validation_44358440", "origin_question": "At the end of which turn were both of \"the modern silver dome-shaped lamp with the stable base\" and \"soft, white cylindrical toilet paper rolls\" on your right side?", "option": [], "answer": "3", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the modern silver dome-shaped lamp with the stable base\" and \"soft, white cylindrical toilet paper rolls\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "Validation_44358440/Validation_44358440_44358440_42285.489.png", "Validation_44358440/Validation_44358440_44358440_42287.488.png", "Validation_44358440/Validation_44358440_44358440_42289.504.png", "Validation_44358440/Validation_44358440_44358440_42291.503.png" ], "turn_id": 5, "scan_id": "Validation_44358440", "origin_question": "Remember, how many pool(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many pool(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_44358440/Validation_44358440_44358440_42293.503.png", "Validation_44358440/Validation_44358440_44358440_42295.502.png", "Validation_44358440/Validation_44358440_44358440_42297.501.png", "Validation_44358440/Validation_44358440_44358440_42299.500.png" ], "turn_id": 6, "scan_id": "Validation_44358440", "origin_question": "Which two objects are on the same side of you now? \"the cozy gray two-seater couch with plush armrests\", \"the modern silver dome-shaped lamp with the stable base\", \"the sleek black rectangular monitor with the matte finish\". ", "option": [ "the cozy gray two-seater couch with plush armrests, the modern silver dome-shaped lamp with the stable base", "the cozy gray two-seater couch with plush armrests, the sleek black rectangular monitor with the matte finish", "the modern silver dome-shaped lamp with the stable base, the sleek black rectangular monitor with the matte finish" ], "answer": "the modern silver dome-shaped lamp with the stable base, the sleek black rectangular monitor with the matte finish", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the cozy gray two-seater couch with plush armrests\", \"the modern silver dome-shaped lamp with the stable base\", \"the sleek black rectangular monitor with the matte finish\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the cozy gray two-seater couch with plush armrests, the modern silver dome-shaped lamp with the stable base', 'the cozy gray two-seater couch with plush armrests, the sleek black rectangular monitor with the matte finish', 'the modern silver dome-shaped lamp with the stable base, the sleek black rectangular monitor with the matte finish'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_44358440/Validation_44358440_44358440_42301.499.png", "Validation_44358440/Validation_44358440_44358440_42303.499.png", "Validation_44358440/Validation_44358440_44358440_42305.498.png", "Validation_44358440/Validation_44358440_44358440_42307.497.png" ], "turn_id": 7, "scan_id": "Validation_44358440", "origin_question": "Please recall \"the sturdy white wooden shelf with open compartments\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.53", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the sturdy white wooden shelf with open compartments\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42897667/Validation_42897667_42897667_477997.095.png", "Validation_42897667/Validation_42897667_42897667_477999.078.png", "Validation_42897667/Validation_42897667_42897667_478001.077.png", "Validation_42897667/Validation_42897667_42897667_478003.076.png", "Validation_42897667/Validation_42897667_42897667_478005.076.png" ], "turn_id": 1, "scan_id": "Validation_42897667", "origin_question": "Remember, have you seen any table(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any table(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_42897667/Validation_42897667_42897667_478007.075.png", "Validation_42897667/Validation_42897667_42897667_478009.091.png", "Validation_42897667/Validation_42897667_42897667_478011.090.png", "Validation_42897667/Validation_42897667_42897667_478013.089.png", "Validation_42897667/Validation_42897667_42897667_478015.088.png" ], "turn_id": 2, "scan_id": "Validation_42897667", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_42897667/Validation_42897667_42897667_478017.087.png", "Validation_42897667/Validation_42897667_42897667_478019.087.png", "Validation_42897667/Validation_42897667_42897667_478021.086.png", "Validation_42897667/Validation_42897667_42897667_478023.085.png", "Validation_42897667/Validation_42897667_42897667_478025.084.png" ], "turn_id": 3, "scan_id": "Validation_42897667", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular framed picture in warm hues\", Y: \"the large, vibrant green plant with bushy leaves\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular framed picture in warm hues\", Y: \"the large, vibrant green plant with bushy leaves\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_42897667/Validation_42897667_42897667_478027.083.png", "Validation_42897667/Validation_42897667_42897667_478029.083.png", "Validation_42897667/Validation_42897667_42897667_478031.082.png", "Validation_42897667/Validation_42897667_42897667_478033.081.png", "Validation_42897667/Validation_42897667_42897667_478035.080.png" ], "turn_id": 4, "scan_id": "Validation_42897667", "origin_question": "What will be the first-time appearance order of the following categories: cabinet, couch, stool?", "option": [ "couch, stool, cabinet", "stool, cabinet, couch", "stool, couch, cabinet", "couch, cabinet, stool" ], "answer": "couch, cabinet, stool", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: cabinet, couch, stool? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['couch, stool, cabinet', 'stool, cabinet, couch', 'stool, couch, cabinet', 'couch, cabinet, stool'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_42897667/Validation_42897667_42897667_478037.079.png", "Validation_42897667/Validation_42897667_42897667_478039.078.png", "Validation_42897667/Validation_42897667_42897667_478041.078.png", "Validation_42897667/Validation_42897667_42897667_478043.077.png", "Validation_42897667/Validation_42897667_42897667_478045.076.png" ], "turn_id": 5, "scan_id": "Validation_42897667", "origin_question": "Please recall \"the vibrant red rectangular couch with soft fabric\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.05", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the vibrant red rectangular couch with soft fabric\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_42897667/Validation_42897667_42897667_478047.075.png", "Validation_42897667/Validation_42897667_42897667_478049.091.png", "Validation_42897667/Validation_42897667_42897667_478051.090.png", "Validation_42897667/Validation_42897667_42897667_478053.089.png", "Validation_42897667/Validation_42897667_42897667_478055.089.png" ], "turn_id": 6, "scan_id": "Validation_42897667", "origin_question": "Please recall \"the rectangular framed picture in warm hues\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.49", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the rectangular framed picture in warm hues\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_42897667/Validation_42897667_42897667_478057.088.png", "Validation_42897667/Validation_42897667_42897667_478059.087.png", "Validation_42897667/Validation_42897667_42897667_478061.086.png", "Validation_42897667/Validation_42897667_42897667_478063.085.png", "Validation_42897667/Validation_42897667_42897667_478065.084.png" ], "turn_id": 7, "scan_id": "Validation_42897667", "origin_question": "", "option": [], "answer": "", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_42897667/Validation_42897667_42897667_478067.084.png", "Validation_42897667/Validation_42897667_42897667_478069.083.png", "Validation_42897667/Validation_42897667_42897667_478071.082.png", "Validation_42897667/Validation_42897667_42897667_478073.081.png", "Validation_42897667/Validation_42897667_42897667_478075.080.png" ], "turn_id": 8, "scan_id": "Validation_42897667", "origin_question": "In which turn were you closest to \"the large rectangular wooden door with glass inserts\"? ", "option": [], "answer": "8", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the large rectangular wooden door with glass inserts\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_42897667/Validation_42897667_42897667_478077.079.png", "Validation_42897667/Validation_42897667_42897667_478079.079.png", "Validation_42897667/Validation_42897667_42897667_478081.078.png", "Validation_42897667/Validation_42897667_42897667_478083.077.png", "Validation_42897667/Validation_42897667_42897667_478085.076.png" ], "turn_id": 9, "scan_id": "Validation_42897667", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the large, sleek black monitor with the thin bezel\" from your position?", "option": [], "answer": "131.88", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the large, sleek black monitor with the thin bezel\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_42897667/Validation_42897667_42897667_478087.075.png", "Validation_42897667/Validation_42897667_42897667_478089.075.png", "Validation_42897667/Validation_42897667_42897667_478091.090.png", "Validation_42897667/Validation_42897667_42897667_478093.090.png", "Validation_42897667/Validation_42897667_42897667_478095.089.png" ], "turn_id": 10, "scan_id": "Validation_42897667", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the light rectangular wooden chair without armrests\", Y: \"the round, rich brown faux fur cushion\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to Y and farther away from X", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the light rectangular wooden chair without armrests\", Y: \"the round, rich brown faux fur cushion\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_45261193/Validation_45261193_45261193_3619.208.png", "Validation_45261193/Validation_45261193_45261193_3621.190.png", "Validation_45261193/Validation_45261193_45261193_3623.189.png", "Validation_45261193/Validation_45261193_45261193_3625.189.png" ], "turn_id": 1, "scan_id": "Validation_45261193", "origin_question": "Remember, have you seen any bed(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any bed(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_45261193/Validation_45261193_45261193_3627.188.png", "Validation_45261193/Validation_45261193_45261193_3629.187.png", "Validation_45261193/Validation_45261193_45261193_3631.186.png", "Validation_45261193/Validation_45261193_45261193_3633.185.png" ], "turn_id": 2, "scan_id": "Validation_45261193", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_45261193/Validation_45261193_45261193_3635.185.png", "Validation_45261193/Validation_45261193_45261193_3637.184.png", "Validation_45261193/Validation_45261193_45261193_3639.183.png", "Validation_45261193/Validation_45261193_45261193_3641.182.png" ], "turn_id": 3, "scan_id": "Validation_45261193", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "33.66", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_45261193/Validation_45261193_45261193_3643.182.png", "Validation_45261193/Validation_45261193_45261193_3645.181.png", "Validation_45261193/Validation_45261193_45261193_3647.180.png", "Validation_45261193/Validation_45261193_45261193_3649.179.png" ], "turn_id": 4, "scan_id": "Validation_45261193", "origin_question": "How far is your current position from where you were at the end of the turn 1? (in meters)", "option": [], "answer": "2.0", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 1? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_45261193/Validation_45261193_45261193_3651.178.png", "Validation_45261193/Validation_45261193_45261193_3653.194.png", "Validation_45261193/Validation_45261193_45261193_3655.193.png", "Validation_45261193/Validation_45261193_45261193_3657.193.png" ], "turn_id": 5, "scan_id": "Validation_45261193", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the rectangular bed with the pink bedspread\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the rectangular bed with the pink bedspread\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_45261193/Validation_45261193_45261193_3659.192.png", "Validation_45261193/Validation_45261193_45261193_3661.191.png", "Validation_45261193/Validation_45261193_45261193_3663.190.png", "Validation_45261193/Validation_45261193_45261193_3665.189.png" ], "turn_id": 6, "scan_id": "Validation_45261193", "origin_question": "Using your orientation at the end of turn 5 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "98.06", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 5 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_45261193/Validation_45261193_45261193_3667.189.png", "Validation_45261193/Validation_45261193_45261193_3669.188.png", "Validation_45261193/Validation_45261193_45261193_3671.187.png", "Validation_45261193/Validation_45261193_45261193_3673.186.png" ], "turn_id": 7, "scan_id": "Validation_45261193", "origin_question": "In which turn did you see both \"the rectangular bed with the pink bedspread\" and \"the classic white ceramic toilet with the rounded shape\" simultaneously? ", "option": [], "answer": "7", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the rectangular bed with the pink bedspread\" and \"the classic white ceramic toilet with the rounded shape\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42445448/Validation_42445448_42445448_50409.603.png", "Validation_42445448/Validation_42445448_42445448_50411.585.png", "Validation_42445448/Validation_42445448_42445448_50413.584.png", "Validation_42445448/Validation_42445448_42445448_50415.584.png", "Validation_42445448/Validation_42445448_42445448_50417.583.png" ], "turn_id": 1, "scan_id": "Validation_42445448", "origin_question": "Remember, have you seen any lamp(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any lamp(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_42445448/Validation_42445448_42445448_50419.582.png", "Validation_42445448/Validation_42445448_42445448_50421.581.png", "Validation_42445448/Validation_42445448_42445448_50423.580.png", "Validation_42445448/Validation_42445448_42445448_50425.596.png", "Validation_42445448/Validation_42445448_42445448_50427.596.png" ], "turn_id": 2, "scan_id": "Validation_42445448", "origin_question": "Which direction is the sleek rectangular window with the gray metallic frame in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the sleek rectangular window with the gray metallic frame in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_42445448/Validation_42445448_42445448_50429.595.png", "Validation_42445448/Validation_42445448_42445448_50431.594.png", "Validation_42445448/Validation_42445448_42445448_50433.593.png", "Validation_42445448/Validation_42445448_42445448_50435.592.png", "Validation_42445448/Validation_42445448_42445448_50437.592.png" ], "turn_id": 3, "scan_id": "Validation_42445448", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the white door with raised panel detailing\" from your position?", "option": [], "answer": "171.38", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the white door with raised panel detailing\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "Validation_42445448/Validation_42445448_42445448_50439.591.png", "Validation_42445448/Validation_42445448_42445448_50441.590.png", "Validation_42445448/Validation_42445448_42445448_50443.589.png", "Validation_42445448/Validation_42445448_42445448_50445.588.png", "Validation_42445448/Validation_42445448_42445448_50447.588.png" ], "turn_id": 4, "scan_id": "Validation_42445448", "origin_question": "Among these three objects(\"the sleek dark rectangular polished table\",\"the modern floor lamp with the chrome stand and white shade\",\"the white door with raised panel detailing\" ), which one is the farthest from you?", "option": [ "the sleek dark rectangular polished table", "the modern floor lamp with the chrome stand and white shade", "the white door with raised panel detailing" ], "answer": "the modern floor lamp with the chrome stand and white shade", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the sleek dark rectangular polished table\",\"the modern floor lamp with the chrome stand and white shade\",\"the white door with raised panel detailing\" ), which one is the farthest from you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the sleek dark rectangular polished table', 'the modern floor lamp with the chrome stand and white shade', 'the white door with raised panel detailing'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42445448/Validation_42445448_42445448_50449.587.png", "Validation_42445448/Validation_42445448_42445448_50451.586.png", "Validation_42445448/Validation_42445448_42445448_50453.585.png", "Validation_42445448/Validation_42445448_42445448_50455.584.png", "Validation_42445448/Validation_42445448_42445448_50457.584.png" ], "turn_id": 5, "scan_id": "Validation_42445448", "origin_question": "Remember, have you seen any table(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any table(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_42445448/Validation_42445448_42445448_50459.583.png", "Validation_42445448/Validation_42445448_42445448_50461.582.png", "Validation_42445448/Validation_42445448_42445448_50463.581.png", "Validation_42445448/Validation_42445448_42445448_50465.580.png", "Validation_42445448/Validation_42445448_42445448_50467.580.png" ], "turn_id": 6, "scan_id": "Validation_42445448", "origin_question": "Compared to your position at the end of the 5 turn, are you now closer or farther away from \"the modern floor lamp with the chrome stand and white shade\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 5 turn, are you now closer or farther away from \"the modern floor lamp with the chrome stand and white shade\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_42445448/Validation_42445448_42445448_50469.596.png", "Validation_42445448/Validation_42445448_42445448_50471.595.png", "Validation_42445448/Validation_42445448_42445448_50473.594.png", "Validation_42445448/Validation_42445448_42445448_50475.593.png", "Validation_42445448/Validation_42445448_42445448_50477.592.png" ], "turn_id": 7, "scan_id": "Validation_42445448", "origin_question": "Compared to your position at the end of the 6 turn, are you now closer or farther away from \"the sleek dark rectangular polished table\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 6 turn, are you now closer or farther away from \"the sleek dark rectangular polished table\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_42445448/Validation_42445448_42445448_50479.592.png", "Validation_42445448/Validation_42445448_42445448_50481.591.png", "Validation_42445448/Validation_42445448_42445448_50483.590.png", "Validation_42445448/Validation_42445448_42445448_50485.589.png", "Validation_42445448/Validation_42445448_42445448_50487.588.png" ], "turn_id": 8, "scan_id": "Validation_42445448", "origin_question": "At the end of which turn were both of \"the sleek rectangular window with the gray metallic frame\" and \"the white door with raised panel detailing\" on your right side?", "option": [], "answer": "1", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the sleek rectangular window with the gray metallic frame\" and \"the white door with raised panel detailing\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_42445448/Validation_42445448_42445448_50489.588.png", "Validation_42445448/Validation_42445448_42445448_50491.587.png", "Validation_42445448/Validation_42445448_42445448_50493.586.png", "Validation_42445448/Validation_42445448_42445448_50495.585.png", "Validation_42445448/Validation_42445448_42445448_50497.584.png" ], "turn_id": 9, "scan_id": "Validation_42445448", "origin_question": "Which direction is the sleek rectangular window with the gray metallic frame in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the sleek rectangular window with the gray metallic frame in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_42445448/Validation_42445448_42445448_50499.584.png", "Validation_42445448/Validation_42445448_42445448_50501.583.png", "Validation_42445448/Validation_42445448_42445448_50503.582.png", "Validation_42445448/Validation_42445448_42445448_50505.581.png", "Validation_42445448/Validation_42445448_42445448_50507.597.png" ], "turn_id": 10, "scan_id": "Validation_42445448", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the sleek dark rectangular polished table\" from your position?", "option": [], "answer": "174.38", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the sleek dark rectangular polished table\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47332885/Validation_47332885_47332885_7484.882.png", "Validation_47332885/Validation_47332885_47332885_7486.865.png", "Validation_47332885/Validation_47332885_47332885_7488.864.png", "Validation_47332885/Validation_47332885_47332885_7490.863.png" ], "turn_id": 1, "scan_id": "Validation_47332885", "origin_question": "Remember, have you seen any excercise equipment(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any excercise equipment(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_47332885/Validation_47332885_47332885_7492.862.png", "Validation_47332885/Validation_47332885_47332885_7494.861.png", "Validation_47332885/Validation_47332885_47332885_7496.860.png", "Validation_47332885/Validation_47332885_47332885_7498.860.png" ], "turn_id": 2, "scan_id": "Validation_47332885", "origin_question": "When did you first discover \"the medium-sized, dark rectangular backpack with compartments\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the medium-sized, dark rectangular backpack with compartments\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47332885/Validation_47332885_47332885_7500.859.png", "Validation_47332885/Validation_47332885_47332885_7502.858.png", "Validation_47332885/Validation_47332885_47332885_7504.857.png", "Validation_47332885/Validation_47332885_47332885_7506.856.png" ], "turn_id": 3, "scan_id": "Validation_47332885", "origin_question": "", "option": [], "answer": "", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_47332885/Validation_47332885_47332885_7508.872.png", "Validation_47332885/Validation_47332885_47332885_7510.872.png", "Validation_47332885/Validation_47332885_47332885_7512.871.png", "Validation_47332885/Validation_47332885_47332885_7514.870.png" ], "turn_id": 4, "scan_id": "Validation_47332885", "origin_question": "In which turn did you see both \"the door\" and \"the medium-sized, dark rectangular backpack with compartments\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the door\" and \"the medium-sized, dark rectangular backpack with compartments\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "Validation_47332885/Validation_47332885_47332885_7516.869.png", "Validation_47332885/Validation_47332885_47332885_7518.868.png", "Validation_47332885/Validation_47332885_47332885_7520.868.png", "Validation_47332885/Validation_47332885_47332885_7522.867.png" ], "turn_id": 5, "scan_id": "Validation_47332885", "origin_question": "Among these three objects(\"the slender black monitor with the medium-sized screen\",\"the medium-sized, dark rectangular backpack with compartments\",\"the light\" ), which one is the closest to you?", "option": [ "the slender black monitor with the medium-sized screen", "the medium-sized, dark rectangular backpack with compartments", "the light" ], "answer": "the light", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects(\"the slender black monitor with the medium-sized screen\",\"the medium-sized, dark rectangular backpack with compartments\",\"the light\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the slender black monitor with the medium-sized screen', 'the medium-sized, dark rectangular backpack with compartments', 'the light'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47332885/Validation_47332885_47332885_7524.866.png", "Validation_47332885/Validation_47332885_47332885_7526.865.png", "Validation_47332885/Validation_47332885_47332885_7528.864.png", "Validation_47332885/Validation_47332885_47332885_7530.863.png" ], "turn_id": 6, "scan_id": "Validation_47332885", "origin_question": "Which direction is the large, vibrant red plastic bucket in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the large, vibrant red plastic bucket in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47115452/Validation_47115452_47115452_13056.585.png", "Validation_47115452/Validation_47115452_47115452_13058.567.png", "Validation_47115452/Validation_47115452_47115452_13060.583.png", "Validation_47115452/Validation_47115452_47115452_13062.582.png" ], "turn_id": 1, "scan_id": "Validation_47115452", "origin_question": "Remember, have you seen any treadmill(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any treadmill(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "Validation_47115452/Validation_47115452_47115452_13064.581.png", "Validation_47115452/Validation_47115452_47115452_13066.581.png", "Validation_47115452/Validation_47115452_47115452_13068.580.png", "Validation_47115452/Validation_47115452_47115452_13070.579.png" ], "turn_id": 2, "scan_id": "Validation_47115452", "origin_question": "Remember, how many sink(s) have you seen so far? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many sink(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47115452/Validation_47115452_47115452_13072.578.png", "Validation_47115452/Validation_47115452_47115452_13074.577.png", "Validation_47115452/Validation_47115452_47115452_13076.576.png", "Validation_47115452/Validation_47115452_47115452_13078.576.png" ], "turn_id": 3, "scan_id": "Validation_47115452", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the stainless steel rectangular oven with the glass door\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the stainless steel rectangular oven with the glass door\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_47115452/Validation_47115452_47115452_13080.575.png", "Validation_47115452/Validation_47115452_47115452_13082.574.png", "Validation_47115452/Validation_47115452_47115452_13084.573.png", "Validation_47115452/Validation_47115452_47115452_13086.572.png" ], "turn_id": 4, "scan_id": "Validation_47115452", "origin_question": "Which two objects are on the same side of you now? \"the modern rectangular stainless steel stove with control knobs\", \"the range hood\", \"the cylindrical stainless steel bin with the silver hue\". ", "option": [ "the modern rectangular stainless steel stove with control knobs, the range hood", "the modern rectangular stainless steel stove with control knobs, the cylindrical stainless steel bin with the silver hue", "the range hood, the cylindrical stainless steel bin with the silver hue" ], "answer": "the modern rectangular stainless steel stove with control knobs, the range hood", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the modern rectangular stainless steel stove with control knobs\", \"the range hood\", \"the cylindrical stainless steel bin with the silver hue\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the modern rectangular stainless steel stove with control knobs, the range hood', 'the modern rectangular stainless steel stove with control knobs, the cylindrical stainless steel bin with the silver hue', 'the range hood, the cylindrical stainless steel bin with the silver hue'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47115452/Validation_47115452_47115452_13088.571.png", "Validation_47115452/Validation_47115452_47115452_13090.570.png", "Validation_47115452/Validation_47115452_47115452_13092.570.png", "Validation_47115452/Validation_47115452_47115452_13094.569.png" ], "turn_id": 5, "scan_id": "Validation_47115452", "origin_question": "In which turn were you farthest from \"the range hood\"? ", "option": [], "answer": "1", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the range hood\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_47115452/Validation_47115452_47115452_13096.568.png", "Validation_47115452/Validation_47115452_47115452_13098.567.png", "Validation_47115452/Validation_47115452_47115452_13100.583.png", "Validation_47115452/Validation_47115452_47115452_13102.582.png" ], "turn_id": 6, "scan_id": "Validation_47115452", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the range hood\" from your position?", "option": [], "answer": "99.73", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the range hood\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47115452/Validation_47115452_47115452_13104.581.png", "Validation_47115452/Validation_47115452_47115452_13106.581.png", "Validation_47115452/Validation_47115452_47115452_13108.580.png", "Validation_47115452/Validation_47115452_47115452_13110.579.png" ], "turn_id": 7, "scan_id": "Validation_47115452", "origin_question": "Which direction is the modern rectangular stainless steel stove with control knobs in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the modern rectangular stainless steel stove with control knobs in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47115452/Validation_47115452_47115452_13112.578.png", "Validation_47115452/Validation_47115452_47115452_13114.577.png", "Validation_47115452/Validation_47115452_47115452_13116.576.png", "Validation_47115452/Validation_47115452_47115452_13118.576.png" ], "turn_id": 8, "scan_id": "Validation_47115452", "origin_question": "Which direction is the rectangular window with the white wooden frame in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the rectangular window with the white wooden frame in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47115452/Validation_47115452_47115452_13120.575.png", "Validation_47115452/Validation_47115452_47115452_13122.574.png", "Validation_47115452/Validation_47115452_47115452_13124.573.png", "Validation_47115452/Validation_47115452_47115452_13126.572.png" ], "turn_id": 9, "scan_id": "Validation_47115452", "origin_question": "Which direction is the range hood in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 9 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the range hood in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47331339/Validation_47331339_47331339_183397.851.png", "Validation_47331339/Validation_47331339_47331339_183399.317.png", "Validation_47331339/Validation_47331339_47331339_183400.833.png", "Validation_47331339/Validation_47331339_47331339_183402.333.png" ], "turn_id": 1, "scan_id": "Validation_47331339", "origin_question": "Remember, have you seen any door(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any door(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_47331339/Validation_47331339_47331339_183403.832.png", "Validation_47331339/Validation_47331339_47331339_183405.332.png", "Validation_47331339/Validation_47331339_47331339_183406.831.png", "Validation_47331339/Validation_47331339_47331339_183408.330.png" ], "turn_id": 2, "scan_id": "Validation_47331339", "origin_question": "In which turn did you see both \"the door\" and \"the white oval-shaped ceramic sink with the chrome faucet\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the door\" and \"the white oval-shaped ceramic sink with the chrome faucet\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_47331339/Validation_47331339_47331339_183409.830.png", "Validation_47331339/Validation_47331339_47331339_183411.329.png", "Validation_47331339/Validation_47331339_47331339_183412.828.png", "Validation_47331339/Validation_47331339_47331339_183414.328.png" ], "turn_id": 3, "scan_id": "Validation_47331339", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "87.58", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_47331339/Validation_47331339_47331339_183415.827.png", "Validation_47331339/Validation_47331339_47331339_183417.327.png", "Validation_47331339/Validation_47331339_47331339_183418.826.png", "Validation_47331339/Validation_47331339_47331339_183420.325.png" ], "turn_id": 4, "scan_id": "Validation_47331339", "origin_question": "When did you first discover \"the door\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the door\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47331339/Validation_47331339_47331339_183421.825.png", "Validation_47331339/Validation_47331339_47331339_183423.324.png", "Validation_47331339/Validation_47331339_47331339_183424.824.png", "Validation_47331339/Validation_47331339_47331339_183426.323.png" ], "turn_id": 5, "scan_id": "Validation_47331339", "origin_question": "Which direction is the door in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the door in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47331339/Validation_47331339_47331339_183427.822.png", "Validation_47331339/Validation_47331339_47331339_183429.322.png", "Validation_47331339/Validation_47331339_47331339_183430.821.png", "Validation_47331339/Validation_47331339_47331339_183432.321.png" ], "turn_id": 6, "scan_id": "Validation_47331339", "origin_question": "", "option": [], "answer": "", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_41254269/Validation_41254269_41254269_101655.220.png", "Validation_41254269/Validation_41254269_41254269_101657.202.png", "Validation_41254269/Validation_41254269_41254269_101659.202.png", "Validation_41254269/Validation_41254269_41254269_101661.201.png", "Validation_41254269/Validation_41254269_41254269_101663.200.png" ], "turn_id": 1, "scan_id": "Validation_41254269", "origin_question": "Remember, have you seen any cabinet(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any cabinet(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_41254269/Validation_41254269_41254269_101665.199.png", "Validation_41254269/Validation_41254269_41254269_101667.198.png", "Validation_41254269/Validation_41254269_41254269_101669.197.png", "Validation_41254269/Validation_41254269_41254269_101671.197.png", "Validation_41254269/Validation_41254269_41254269_101673.196.png" ], "turn_id": 2, "scan_id": "Validation_41254269", "origin_question": "Remember, have you seen any toaster(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any toaster(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_41254269/Validation_41254269_41254269_101675.195.png", "Validation_41254269/Validation_41254269_41254269_101677.194.png", "Validation_41254269/Validation_41254269_41254269_101679.193.png", "Validation_41254269/Validation_41254269_41254269_101681.209.png", "Validation_41254269/Validation_41254269_41254269_101683.208.png" ], "turn_id": 3, "scan_id": "Validation_41254269", "origin_question": "Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "89.81", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_41254269/Validation_41254269_41254269_101685.208.png", "Validation_41254269/Validation_41254269_41254269_101687.207.png", "Validation_41254269/Validation_41254269_41254269_101689.206.png", "Validation_41254269/Validation_41254269_41254269_101691.205.png", "Validation_41254269/Validation_41254269_41254269_101693.204.png" ], "turn_id": 4, "scan_id": "Validation_41254269", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "139.59", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_41254269/Validation_41254269_41254269_101695.203.png", "Validation_41254269/Validation_41254269_41254269_101697.203.png", "Validation_41254269/Validation_41254269_41254269_101699.202.png", "Validation_41254269/Validation_41254269_41254269_101701.201.png", "Validation_41254269/Validation_41254269_41254269_101703.200.png" ], "turn_id": 5, "scan_id": "Validation_41254269", "origin_question": "In which turn did you see both \"the bed\" and \"the window\" simultaneously? ", "option": [], "answer": "5", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the bed\" and \"the window\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_41254269/Validation_41254269_41254269_101705.199.png", "Validation_41254269/Validation_41254269_41254269_101707.198.png", "Validation_41254269/Validation_41254269_41254269_101709.198.png", "Validation_41254269/Validation_41254269_41254269_101711.197.png", "Validation_41254269/Validation_41254269_41254269_101713.196.png" ], "turn_id": 6, "scan_id": "Validation_41254269", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the window\" from your position?", "option": [], "answer": "89.65", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the window\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_41254269/Validation_41254269_41254269_101715.195.png", "Validation_41254269/Validation_41254269_41254269_101717.194.png", "Validation_41254269/Validation_41254269_41254269_101719.193.png", "Validation_41254269/Validation_41254269_41254269_101721.193.png", "Validation_41254269/Validation_41254269_41254269_101723.208.png" ], "turn_id": 7, "scan_id": "Validation_41254269", "origin_question": "Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the bed\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the bed\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_41254269/Validation_41254269_41254269_101725.208.png", "Validation_41254269/Validation_41254269_41254269_101727.207.png", "Validation_41254269/Validation_41254269_41254269_101729.206.png", "Validation_41254269/Validation_41254269_41254269_101731.205.png", "Validation_41254269/Validation_41254269_41254269_101733.204.png" ], "turn_id": 8, "scan_id": "Validation_41254269", "origin_question": "Using your orientation at the end of turn 5 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "51.58", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 5 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_41254269/Validation_41254269_41254269_101735.204.png", "Validation_41254269/Validation_41254269_41254269_101737.203.png", "Validation_41254269/Validation_41254269_41254269_101739.202.png", "Validation_41254269/Validation_41254269_41254269_101741.201.png", "Validation_41254269/Validation_41254269_41254269_101743.200.png" ], "turn_id": 9, "scan_id": "Validation_41254269", "origin_question": "Using your orientation at the end of turn 6 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "63.46", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 6 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42446517/Validation_42446517_42446517_197418.298.png", "Validation_42446517/Validation_42446517_42446517_197420.281.png", "Validation_42446517/Validation_42446517_42446517_197422.280.png", "Validation_42446517/Validation_42446517_42446517_197424.279.png" ], "turn_id": 1, "scan_id": "Validation_42446517", "origin_question": "Remember, have you seen any vase(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any vase(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_42446517/Validation_42446517_42446517_197426.278.png", "Validation_42446517/Validation_42446517_42446517_197428.277.png", "Validation_42446517/Validation_42446517_42446517_197430.277.png", "Validation_42446517/Validation_42446517_42446517_197432.276.png" ], "turn_id": 2, "scan_id": "Validation_42446517", "origin_question": "When did you first discover \"the black ergonomic chair with the high backrest\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the black ergonomic chair with the high backrest\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_42446517/Validation_42446517_42446517_197434.275.png", "Validation_42446517/Validation_42446517_42446517_197436.274.png", "Validation_42446517/Validation_42446517_42446517_197438.273.png", "Validation_42446517/Validation_42446517_42446517_197440.272.png" ], "turn_id": 3, "scan_id": "Validation_42446517", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_42446517/Validation_42446517_42446517_197442.272.png", "Validation_42446517/Validation_42446517_42446517_197444.271.png", "Validation_42446517/Validation_42446517_42446517_197446.270.png", "Validation_42446517/Validation_42446517_42446517_197448.269.png" ], "turn_id": 4, "scan_id": "Validation_42446517", "origin_question": "Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "139.44", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_42446517/Validation_42446517_42446517_197450.268.png", "Validation_42446517/Validation_42446517_42446517_197452.284.png", "Validation_42446517/Validation_42446517_42446517_197454.283.png", "Validation_42446517/Validation_42446517_42446517_197456.283.png" ], "turn_id": 5, "scan_id": "Validation_42446517", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the black ergonomic chair with the high backrest\", Y: \"the shelf\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the black ergonomic chair with the high backrest\", Y: \"the shelf\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "Validation_42446517/Validation_42446517_42446517_197458.282.png", "Validation_42446517/Validation_42446517_42446517_197460.281.png", "Validation_42446517/Validation_42446517_42446517_197462.280.png", "Validation_42446517/Validation_42446517_42446517_197464.279.png" ], "turn_id": 6, "scan_id": "Validation_42446517", "origin_question": "Among these three objects(\"the black ergonomic chair with the high backrest\",\"the shelf\",\"the sleek gray cylindrical bin made of plastic\" ), which one is the farthest from you?", "option": [ "the black ergonomic chair with the high backrest", "the shelf", "the sleek gray cylindrical bin made of plastic" ], "answer": "the black ergonomic chair with the high backrest", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects(\"the black ergonomic chair with the high backrest\",\"the shelf\",\"the sleek gray cylindrical bin made of plastic\" ), which one is the farthest from you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the black ergonomic chair with the high backrest', 'the shelf', 'the sleek gray cylindrical bin made of plastic'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_42446517/Validation_42446517_42446517_197466.278.png", "Validation_42446517/Validation_42446517_42446517_197468.278.png", "Validation_42446517/Validation_42446517_42446517_197470.277.png", "Validation_42446517/Validation_42446517_42446517_197472.276.png" ], "turn_id": 7, "scan_id": "Validation_42446517", "origin_question": "Please recall \"the sleek black flat-screen monitor with the thin bezel\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.49", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the sleek black flat-screen monitor with the thin bezel\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47430489/Validation_47430489_47430489_313766.980.png", "Validation_47430489/Validation_47430489_47430489_313768.979.png", "Validation_47430489/Validation_47430489_47430489_313770.978.png", "Validation_47430489/Validation_47430489_47430489_313772.977.png" ], "turn_id": 1, "scan_id": "Validation_47430489", "origin_question": "Remember, have you seen any couch(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any couch(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_47430489/Validation_47430489_47430489_313774.977.png", "Validation_47430489/Validation_47430489_47430489_313776.976.png", "Validation_47430489/Validation_47430489_47430489_313778.975.png", "Validation_47430489/Validation_47430489_47430489_313780.974.png" ], "turn_id": 2, "scan_id": "Validation_47430489", "origin_question": "In which turn did you see both \"the sleek, compact couch with the light wood frame\" and \"the medium, transparent plastic box\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the sleek, compact couch with the light wood frame\" and \"the medium, transparent plastic box\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_47430489/Validation_47430489_47430489_313782.973.png", "Validation_47430489/Validation_47430489_47430489_313784.973.png", "Validation_47430489/Validation_47430489_47430489_313786.972.png", "Validation_47430489/Validation_47430489_47430489_313788.971.png" ], "turn_id": 3, "scan_id": "Validation_47430489", "origin_question": "What will be the first-time appearance order of the following categories: box, clothes, couch?", "option": [ "couch, box, clothes", "clothes, box, couch", "box, couch, clothes", "couch, clothes, box" ], "answer": "couch, box, clothes", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: box, clothes, couch? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['couch, box, clothes', 'clothes, box, couch', 'box, couch, clothes', 'couch, clothes, box'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47430489/Validation_47430489_47430489_313790.970.png", "Validation_47430489/Validation_47430489_47430489_313792.969.png", "Validation_47430489/Validation_47430489_47430489_313794.969.png", "Validation_47430489/Validation_47430489_47430489_313796.968.png" ], "turn_id": 4, "scan_id": "Validation_47430489", "origin_question": "In which turn were you closest to \"the blue, comfortable chair\"? ", "option": [], "answer": "2", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the blue, comfortable chair\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_47430489/Validation_47430489_47430489_313798.984.png", "Validation_47430489/Validation_47430489_47430489_313800.983.png", "Validation_47430489/Validation_47430489_47430489_313802.982.png", "Validation_47430489/Validation_47430489_47430489_313804.981.png" ], "turn_id": 5, "scan_id": "Validation_47430489", "origin_question": "Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance forward or backward from that position?", "option": [ "Forward", "Backward" ], "answer": "Backward", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance forward or backward from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Forward', 'Backward'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_47430489/Validation_47430489_47430489_313806.980.png", "Validation_47430489/Validation_47430489_47430489_313808.980.png", "Validation_47430489/Validation_47430489_47430489_313810.979.png", "Validation_47430489/Validation_47430489_47430489_313812.978.png" ], "turn_id": 6, "scan_id": "Validation_47430489", "origin_question": "Which two objects are on the same side of you now? \"the light gray clothes made of soft fabric\", \"the colorful, paper book with the readable spine\", \"the large red and yellow plush toy\". ", "option": [ "the light gray clothes made of soft fabric, the colorful, paper book with the readable spine", "the light gray clothes made of soft fabric, the large red and yellow plush toy", "the colorful, paper book with the readable spine, the large red and yellow plush toy" ], "answer": "the light gray clothes made of soft fabric, the colorful, paper book with the readable spine", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the light gray clothes made of soft fabric\", \"the colorful, paper book with the readable spine\", \"the large red and yellow plush toy\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the light gray clothes made of soft fabric, the colorful, paper book with the readable spine', 'the light gray clothes made of soft fabric, the large red and yellow plush toy', 'the colorful, paper book with the readable spine, the large red and yellow plush toy'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47430489/Validation_47430489_47430489_313814.977.png", "Validation_47430489/Validation_47430489_47430489_313816.976.png", "Validation_47430489/Validation_47430489_47430489_313818.975.png", "Validation_47430489/Validation_47430489_47430489_313820.975.png" ], "turn_id": 7, "scan_id": "Validation_47430489", "origin_question": "Which direction is the colorful, paper book with the readable spine in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the colorful, paper book with the readable spine in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47204554/Validation_47204554_47204554_3211.203.png", "Validation_47204554/Validation_47204554_47204554_3213.202.png", "Validation_47204554/Validation_47204554_47204554_3215.201.png", "Validation_47204554/Validation_47204554_47204554_3217.200.png" ], "turn_id": 1, "scan_id": "Validation_47204554", "origin_question": "Remember, have you seen any light(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any light(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "Validation_47204554/Validation_47204554_47204554_3219.199.png", "Validation_47204554/Validation_47204554_47204554_3221.198.png", "Validation_47204554/Validation_47204554_47204554_3223.198.png", "Validation_47204554/Validation_47204554_47204554_3225.197.png" ], "turn_id": 2, "scan_id": "Validation_47204554", "origin_question": "Remember, how many light(s) have you seen so far? ", "option": [], "answer": "3", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many light(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47204554/Validation_47204554_47204554_3227.196.png", "Validation_47204554/Validation_47204554_47204554_3229.195.png", "Validation_47204554/Validation_47204554_47204554_3231.194.png", "Validation_47204554/Validation_47204554_47204554_3233.193.png" ], "turn_id": 3, "scan_id": "Validation_47204554", "origin_question": "Which direction is the rectangular, black ceramic glass stove in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the rectangular, black ceramic glass stove in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_47204554/Validation_47204554_47204554_3235.193.png", "Validation_47204554/Validation_47204554_47204554_3237.192.png", "Validation_47204554/Validation_47204554_47204554_3239.191.png", "Validation_47204554/Validation_47204554_47204554_3241.190.png" ], "turn_id": 4, "scan_id": "Validation_47204554", "origin_question": "Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance forward or backward from that position?", "option": [ "Forward", "Backward" ], "answer": "Backward", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance forward or backward from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Forward', 'Backward'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47204554/Validation_47204554_47204554_3243.189.png", "Validation_47204554/Validation_47204554_47204554_3245.189.png", "Validation_47204554/Validation_47204554_47204554_3247.188.png", "Validation_47204554/Validation_47204554_47204554_3249.204.png" ], "turn_id": 5, "scan_id": "Validation_47204554", "origin_question": "", "option": [], "answer": "", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_47204554/Validation_47204554_47204554_3251.203.png", "Validation_47204554/Validation_47204554_47204554_3253.202.png", "Validation_47204554/Validation_47204554_47204554_3255.201.png", "Validation_47204554/Validation_47204554_47204554_3257.200.png" ], "turn_id": 6, "scan_id": "Validation_47204554", "origin_question": "What will be the first-time appearance order of the following categories: flowerpot, monitor, oven?", "option": [ "monitor, oven, flowerpot", "flowerpot, monitor, oven", "oven, monitor, flowerpot", "oven, flowerpot, monitor" ], "answer": "oven, monitor, flowerpot", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: flowerpot, monitor, oven? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['monitor, oven, flowerpot', 'flowerpot, monitor, oven', 'oven, monitor, flowerpot', 'oven, flowerpot, monitor'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_47204554/Validation_47204554_47204554_3259.199.png", "Validation_47204554/Validation_47204554_47204554_3261.199.png", "Validation_47204554/Validation_47204554_47204554_3263.198.png", "Validation_47204554/Validation_47204554_47204554_3265.197.png" ], "turn_id": 7, "scan_id": "Validation_47204554", "origin_question": "What will be the first-time appearance order of the following categories: cabinet, monitor, pillow?", "option": [ "pillow, cabinet, monitor", "cabinet, pillow, monitor", "pillow, monitor, cabinet", "cabinet, monitor, pillow" ], "answer": "cabinet, monitor, pillow", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: cabinet, monitor, pillow? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['pillow, cabinet, monitor', 'cabinet, pillow, monitor', 'pillow, monitor, cabinet', 'cabinet, monitor, pillow'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47204554/Validation_47204554_47204554_3267.196.png", "Validation_47204554/Validation_47204554_47204554_3269.195.png", "Validation_47204554/Validation_47204554_47204554_3271.195.png", "Validation_47204554/Validation_47204554_47204554_3273.194.png" ], "turn_id": 8, "scan_id": "Validation_47204554", "origin_question": "", "option": [], "answer": "", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42899739/Validation_42899739_42899739_801457.144.png", "Validation_42899739/Validation_42899739_42899739_801459.143.png", "Validation_42899739/Validation_42899739_42899739_801461.142.png", "Validation_42899739/Validation_42899739_42899739_801463.142.png" ], "turn_id": 1, "scan_id": "Validation_42899739", "origin_question": "Remember, have you seen any bed(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any bed(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_42899739/Validation_42899739_42899739_801465.141.png", "Validation_42899739/Validation_42899739_42899739_801467.140.png", "Validation_42899739/Validation_42899739_42899739_801469.139.png", "Validation_42899739/Validation_42899739_42899739_801471.138.png" ], "turn_id": 2, "scan_id": "Validation_42899739", "origin_question": "When did you first discover \"the medium-sized cylindrical lamp with the metal base\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the medium-sized cylindrical lamp with the metal base\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_42899739/Validation_42899739_42899739_801473.137.png", "Validation_42899739/Validation_42899739_42899739_801475.137.png", "Validation_42899739/Validation_42899739_42899739_801477.136.png", "Validation_42899739/Validation_42899739_42899739_801479.135.png" ], "turn_id": 3, "scan_id": "Validation_42899739", "origin_question": "Is the medium-sized cylindrical lamp with the metal base to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the medium-sized cylindrical lamp with the metal base to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_42899739/Validation_42899739_42899739_801481.134.png", "Validation_42899739/Validation_42899739_42899739_801483.133.png", "Validation_42899739/Validation_42899739_42899739_801485.132.png", "Validation_42899739/Validation_42899739_42899739_801487.132.png" ], "turn_id": 4, "scan_id": "Validation_42899739", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the soft, white towel in medium to large size\" from your position?", "option": [], "answer": "101.19", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the soft, white towel in medium to large size\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_42899739/Validation_42899739_42899739_801489.131.png", "Validation_42899739/Validation_42899739_42899739_801491.130.png", "Validation_42899739/Validation_42899739_42899739_801493.129.png", "Validation_42899739/Validation_42899739_42899739_801495.128.png" ], "turn_id": 5, "scan_id": "Validation_42899739", "origin_question": "At the end of which turn were both of \"the rectangular bed with the dark brown duvet and light pillows\" and \"the soft, white towel in medium to large size\" on your left side?", "option": [], "answer": "1", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the rectangular bed with the dark brown duvet and light pillows\" and \"the soft, white towel in medium to large size\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_42899739/Validation_42899739_42899739_801497.127.png", "Validation_42899739/Validation_42899739_42899739_801499.143.png", "Validation_42899739/Validation_42899739_42899739_801501.142.png", "Validation_42899739/Validation_42899739_42899739_801503.142.png" ], "turn_id": 6, "scan_id": "Validation_42899739", "origin_question": "Compared to your position at the end of the 5 turn, are you now closer or farther away from \"the medium-sized cylindrical lamp with the metal base\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 5 turn, are you now closer or farther away from \"the medium-sized cylindrical lamp with the metal base\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_42899739/Validation_42899739_42899739_801505.141.png", "Validation_42899739/Validation_42899739_42899739_801507.140.png", "Validation_42899739/Validation_42899739_42899739_801509.139.png", "Validation_42899739/Validation_42899739_42899739_801511.138.png" ], "turn_id": 7, "scan_id": "Validation_42899739", "origin_question": "Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Right", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_45260854/Validation_45260854_45260854_557.093.png", "Validation_45260854/Validation_45260854_45260854_559.092.png", "Validation_45260854/Validation_45260854_45260854_561.091.png", "Validation_45260854/Validation_45260854_45260854_563.090.png" ], "turn_id": 1, "scan_id": "Validation_45260854", "origin_question": "Remember, have you seen any helmets(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any helmets(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_45260854/Validation_45260854_45260854_565.090.png", "Validation_45260854/Validation_45260854_45260854_567.089.png", "Validation_45260854/Validation_45260854_45260854_569.088.png", "Validation_45260854/Validation_45260854_45260854_571.087.png" ], "turn_id": 2, "scan_id": "Validation_45260854", "origin_question": "In which turn were you closest to \"the tall, elegant four-poster bed with the neutral shade\"? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the tall, elegant four-poster bed with the neutral shade\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_45260854/Validation_45260854_45260854_573.086.png", "Validation_45260854/Validation_45260854_45260854_575.085.png", "Validation_45260854/Validation_45260854_45260854_577.085.png", "Validation_45260854/Validation_45260854_45260854_579.084.png" ], "turn_id": 3, "scan_id": "Validation_45260854", "origin_question": "What will be the first-time appearance order of the following categories: bed, chair, fireplace?", "option": [ "fireplace, chair, bed", "fireplace, bed, chair", "chair, bed, fireplace", "bed, fireplace, chair" ], "answer": "chair, bed, fireplace", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bed, chair, fireplace? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['fireplace, chair, bed', 'fireplace, bed, chair', 'chair, bed, fireplace', 'bed, fireplace, chair'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_45260854/Validation_45260854_45260854_581.083.png", "Validation_45260854/Validation_45260854_45260854_583.082.png", "Validation_45260854/Validation_45260854_45260854_585.081.png", "Validation_45260854/Validation_45260854_45260854_587.080.png" ], "turn_id": 4, "scan_id": "Validation_45260854", "origin_question": "Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "130.67", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_45260854/Validation_45260854_45260854_589.080.png", "Validation_45260854/Validation_45260854_45260854_591.079.png", "Validation_45260854/Validation_45260854_45260854_593.078.png", "Validation_45260854/Validation_45260854_45260854_595.077.png" ], "turn_id": 5, "scan_id": "Validation_45260854", "origin_question": "What will be the first-time appearance order of the following categories: cabinet, chair, fireplace?", "option": [ "chair, cabinet, fireplace", "fireplace, cabinet, chair", "fireplace, chair, cabinet", "cabinet, chair, fireplace" ], "answer": "chair, cabinet, fireplace", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: cabinet, chair, fireplace? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['chair, cabinet, fireplace', 'fireplace, cabinet, chair', 'fireplace, chair, cabinet', 'cabinet, chair, fireplace'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_45260854/Validation_45260854_45260854_597.093.png", "Validation_45260854/Validation_45260854_45260854_599.092.png", "Validation_45260854/Validation_45260854_45260854_601.091.png", "Validation_45260854/Validation_45260854_45260854_603.090.png" ], "turn_id": 6, "scan_id": "Validation_45260854", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the tall, elegant four-poster bed with the neutral shade\", Y: \"the fireplace\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the tall, elegant four-poster bed with the neutral shade\", Y: \"the fireplace\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_45260854/Validation_45260854_45260854_605.090.png", "Validation_45260854/Validation_45260854_45260854_607.089.png", "Validation_45260854/Validation_45260854_45260854_609.088.png", "Validation_45260854/Validation_45260854_45260854_611.087.png" ], "turn_id": 7, "scan_id": "Validation_45260854", "origin_question": "At the end of which turn were both of \"the green vintage couch with the curved backrest\" and \"the fireplace\" on your right side?", "option": [], "answer": "7", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the green vintage couch with the curved backrest\" and \"the fireplace\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_45260854/Validation_45260854_45260854_613.086.png", "Validation_45260854/Validation_45260854_45260854_615.086.png", "Validation_45260854/Validation_45260854_45260854_617.085.png", "Validation_45260854/Validation_45260854_45260854_619.084.png" ], "turn_id": 8, "scan_id": "Validation_45260854", "origin_question": "Using your orientation at the end of turn 5 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "29.65", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 5 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_48018387/Validation_48018387_48018387_390379.656.png", "Validation_48018387/Validation_48018387_48018387_390381.139.png", "Validation_48018387/Validation_48018387_48018387_390382.639.png", "Validation_48018387/Validation_48018387_48018387_390384.138.png" ], "turn_id": 1, "scan_id": "Validation_48018387", "origin_question": "Remember, have you seen any carpet(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any carpet(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_48018387/Validation_48018387_48018387_390385.637.png", "Validation_48018387/Validation_48018387_48018387_390387.137.png", "Validation_48018387/Validation_48018387_48018387_390388.636.png", "Validation_48018387/Validation_48018387_48018387_390390.135.png" ], "turn_id": 2, "scan_id": "Validation_48018387", "origin_question": "In which turn did you see both \"the modern white round ceramic sink\" and \"the compact green plastic stool with the rectangular shape\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the modern white round ceramic sink\" and \"the compact green plastic stool with the rectangular shape\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_48018387/Validation_48018387_48018387_390391.635.png", "Validation_48018387/Validation_48018387_48018387_390393.134.png", "Validation_48018387/Validation_48018387_48018387_390394.634.png", "Validation_48018387/Validation_48018387_48018387_390396.150.png" ], "turn_id": 3, "scan_id": "Validation_48018387", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the modern white round ceramic sink\" from your position?", "option": [], "answer": "145.73", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the modern white round ceramic sink\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_48018387/Validation_48018387_48018387_390397.649.png", "Validation_48018387/Validation_48018387_48018387_390399.148.png", "Validation_48018387/Validation_48018387_48018387_390400.648.png", "Validation_48018387/Validation_48018387_48018387_390402.147.png" ], "turn_id": 4, "scan_id": "Validation_48018387", "origin_question": "Which two objects are on the same side of you now? \"the classic oval white ceramic toilet with the glossy finish\", \"the classic rectangular white bathtub made of porcelain-coated steel\", \"the sleek rectangular deep brown door with glass panels\". ", "option": [ "the classic oval white ceramic toilet with the glossy finish, the classic rectangular white bathtub made of porcelain-coated steel", "the classic oval white ceramic toilet with the glossy finish, the sleek rectangular deep brown door with glass panels", "the classic rectangular white bathtub made of porcelain-coated steel, the sleek rectangular deep brown door with glass panels" ], "answer": "the classic oval white ceramic toilet with the glossy finish, the classic rectangular white bathtub made of porcelain-coated steel", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the classic oval white ceramic toilet with the glossy finish\", \"the classic rectangular white bathtub made of porcelain-coated steel\", \"the sleek rectangular deep brown door with glass panels\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the classic oval white ceramic toilet with the glossy finish, the classic rectangular white bathtub made of porcelain-coated steel', 'the classic oval white ceramic toilet with the glossy finish, the sleek rectangular deep brown door with glass panels', 'the classic rectangular white bathtub made of porcelain-coated steel, the sleek rectangular deep brown door with glass panels'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_48018387/Validation_48018387_48018387_390403.646.png", "Validation_48018387/Validation_48018387_48018387_390405.146.png", "Validation_48018387/Validation_48018387_48018387_390406.645.png", "Validation_48018387/Validation_48018387_48018387_390408.145.png" ], "turn_id": 5, "scan_id": "Validation_48018387", "origin_question": "Which direction is the sleek rectangular deep brown door with glass panels in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the sleek rectangular deep brown door with glass panels in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_48018387/Validation_48018387_48018387_390409.644.png", "Validation_48018387/Validation_48018387_48018387_390411.143.png", "Validation_48018387/Validation_48018387_48018387_390412.643.png", "Validation_48018387/Validation_48018387_48018387_390414.142.png" ], "turn_id": 6, "scan_id": "Validation_48018387", "origin_question": "Which direction is the classic rectangular white bathtub made of porcelain-coated steel in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the classic rectangular white bathtub made of porcelain-coated steel in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_48458647/Validation_48458647_48458647_7011.925.png", "Validation_48458647/Validation_48458647_48458647_7013.408.png", "Validation_48458647/Validation_48458647_48458647_7014.907.png", "Validation_48458647/Validation_48458647_48458647_7016.406.png" ], "turn_id": 1, "scan_id": "Validation_48458647", "origin_question": "Remember, have you seen any wardrobe(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any wardrobe(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_48458647/Validation_48458647_48458647_7017.906.png", "Validation_48458647/Validation_48458647_48458647_7019.405.png", "Validation_48458647/Validation_48458647_48458647_7020.904.png", "Validation_48458647/Validation_48458647_48458647_7022.404.png" ], "turn_id": 2, "scan_id": "Validation_48458647", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the large rectangular mirror\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the large rectangular mirror\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_48458647/Validation_48458647_48458647_7023.903.png", "Validation_48458647/Validation_48458647_48458647_7025.403.png", "Validation_48458647/Validation_48458647_48458647_7026.902.png", "Validation_48458647/Validation_48458647_48458647_7028.418.png" ], "turn_id": 3, "scan_id": "Validation_48458647", "origin_question": "", "option": [], "answer": "", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "Validation_48458647/Validation_48458647_48458647_7029.917.png", "Validation_48458647/Validation_48458647_48458647_7031.417.png", "Validation_48458647/Validation_48458647_48458647_7032.916.png", "Validation_48458647/Validation_48458647_48458647_7034.416.png" ], "turn_id": 4, "scan_id": "Validation_48458647", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the contemporary white door with the silver handle\"; \"the medium to large rectangular window with the white frame\"; \"the light wood rectangular cabinet with closed doors\" ", "option": [ "the contemporary white door with the silver handle", "the medium to large rectangular window with the white frame", "the light wood rectangular cabinet with closed doors" ], "answer": "the light wood rectangular cabinet with closed doors", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the contemporary white door with the silver handle\"; \"the medium to large rectangular window with the white frame\"; \"the light wood rectangular cabinet with closed doors\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the contemporary white door with the silver handle', 'the medium to large rectangular window with the white frame', 'the light wood rectangular cabinet with closed doors'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_48458647/Validation_48458647_48458647_7035.915.png", "Validation_48458647/Validation_48458647_48458647_7037.414.png", "Validation_48458647/Validation_48458647_48458647_7038.914.png", "Validation_48458647/Validation_48458647_48458647_7040.413.png" ], "turn_id": 5, "scan_id": "Validation_48458647", "origin_question": "Please recall \"the large rectangular mirror\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.54", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the large rectangular mirror\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "Validation_48458647/Validation_48458647_48458647_7041.913.png", "Validation_48458647/Validation_48458647_48458647_7043.412.png", "Validation_48458647/Validation_48458647_48458647_7044.911.png", "Validation_48458647/Validation_48458647_48458647_7046.411.png" ], "turn_id": 6, "scan_id": "Validation_48458647", "origin_question": "Remember, how many globe(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many globe(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_48458647/Validation_48458647_48458647_7047.910.png", "Validation_48458647/Validation_48458647_48458647_7049.409.png", "Validation_48458647/Validation_48458647_48458647_7050.909.png", "Validation_48458647/Validation_48458647_48458647_7052.408.png" ], "turn_id": 7, "scan_id": "Validation_48458647", "origin_question": "Which two objects are on the same side of you now? \"the deep blue upholstered wooden chair with the sleek design\", \"the white cylindrical plastic bucket with the handle\", \"the contemporary white door with the silver handle\". ", "option": [ "the deep blue upholstered wooden chair with the sleek design, the white cylindrical plastic bucket with the handle", "the deep blue upholstered wooden chair with the sleek design, the contemporary white door with the silver handle", "the white cylindrical plastic bucket with the handle, the contemporary white door with the silver handle" ], "answer": "the deep blue upholstered wooden chair with the sleek design, the white cylindrical plastic bucket with the handle", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the deep blue upholstered wooden chair with the sleek design\", \"the white cylindrical plastic bucket with the handle\", \"the contemporary white door with the silver handle\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the deep blue upholstered wooden chair with the sleek design, the white cylindrical plastic bucket with the handle', 'the deep blue upholstered wooden chair with the sleek design, the contemporary white door with the silver handle', 'the white cylindrical plastic bucket with the handle, the contemporary white door with the silver handle'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "Validation_45663149/Validation_45663149_45663149_63296.729.png", "Validation_45663149/Validation_45663149_45663149_63298.711.png", "Validation_45663149/Validation_45663149_45663149_63300.711.png", "Validation_45663149/Validation_45663149_45663149_63302.710.png" ], "turn_id": 1, "scan_id": "Validation_45663149", "origin_question": "Remember, how many couch(s) have you seen so far? ", "option": [], "answer": "2", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many couch(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_45663149/Validation_45663149_45663149_63304.709.png", "Validation_45663149/Validation_45663149_45663149_63306.708.png", "Validation_45663149/Validation_45663149_45663149_63308.707.png", "Validation_45663149/Validation_45663149_45663149_63310.707.png" ], "turn_id": 2, "scan_id": "Validation_45663149", "origin_question": "Which direction is the sleek rectangular monitor with the glossy black bezel in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the sleek rectangular monitor with the glossy black bezel in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_45663149/Validation_45663149_45663149_63312.706.png", "Validation_45663149/Validation_45663149_45663149_63314.705.png", "Validation_45663149/Validation_45663149_45663149_63316.704.png", "Validation_45663149/Validation_45663149_45663149_63318.703.png" ], "turn_id": 3, "scan_id": "Validation_45663149", "origin_question": "At the end of which turn were both of \"the modern cylindrical lamp with the transparent base\" and \"the sleek black clock with the white face\" on your left side?", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the modern cylindrical lamp with the transparent base\" and \"the sleek black clock with the white face\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_45663149/Validation_45663149_45663149_63320.703.png", "Validation_45663149/Validation_45663149_45663149_63322.702.png", "Validation_45663149/Validation_45663149_45663149_63324.701.png", "Validation_45663149/Validation_45663149_45663149_63326.700.png" ], "turn_id": 4, "scan_id": "Validation_45663149", "origin_question": "Please recall \"the sleek black clock with the white face\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.76", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the sleek black clock with the white face\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_45663149/Validation_45663149_45663149_63328.699.png", "Validation_45663149/Validation_45663149_45663149_63330.715.png", "Validation_45663149/Validation_45663149_45663149_63332.715.png", "Validation_45663149/Validation_45663149_45663149_63334.714.png" ], "turn_id": 5, "scan_id": "Validation_45663149", "origin_question": "Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_45663149/Validation_45663149_45663149_63336.713.png", "Validation_45663149/Validation_45663149_45663149_63338.712.png", "Validation_45663149/Validation_45663149_45663149_63340.711.png", "Validation_45663149/Validation_45663149_45663149_63342.711.png" ], "turn_id": 6, "scan_id": "Validation_45663149", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the sleek black clock with the white face\", Y: \"the window\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the sleek black clock with the white face\", Y: \"the window\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_45663149/Validation_45663149_45663149_63344.710.png", "Validation_45663149/Validation_45663149_45663149_63346.709.png", "Validation_45663149/Validation_45663149_45663149_63348.708.png", "Validation_45663149/Validation_45663149_45663149_63350.707.png" ], "turn_id": 7, "scan_id": "Validation_45663149", "origin_question": "How far is your current position from where you were at the end of the turn 3? (in meters)", "option": [], "answer": "3.37", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 3? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42898521/Validation_42898521_42898521_750363.174.png", "Validation_42898521/Validation_42898521_42898521_750365.174.png", "Validation_42898521/Validation_42898521_42898521_750367.173.png", "Validation_42898521/Validation_42898521_42898521_750369.172.png", "Validation_42898521/Validation_42898521_42898521_750371.171.png" ], "turn_id": 1, "scan_id": "Validation_42898521", "origin_question": "Remember, have you seen any pillow(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any pillow(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_42898521/Validation_42898521_42898521_750373.170.png", "Validation_42898521/Validation_42898521_42898521_750375.169.png", "Validation_42898521/Validation_42898521_42898521_750377.169.png", "Validation_42898521/Validation_42898521_42898521_750379.168.png", "Validation_42898521/Validation_42898521_42898521_750381.167.png" ], "turn_id": 2, "scan_id": "Validation_42898521", "origin_question": "When did you first discover \"the modern rectangular neutral-colored bin\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the modern rectangular neutral-colored bin\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_42898521/Validation_42898521_42898521_750383.166.png", "Validation_42898521/Validation_42898521_42898521_750385.165.png", "Validation_42898521/Validation_42898521_42898521_750387.165.png", "Validation_42898521/Validation_42898521_42898521_750389.164.png", "Validation_42898521/Validation_42898521_42898521_750391.180.png" ], "turn_id": 3, "scan_id": "Validation_42898521", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the modern rectangular neutral-colored bin\", Y: \"the medium-sized rectangular soft pastel towel\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the modern rectangular neutral-colored bin\", Y: \"the medium-sized rectangular soft pastel towel\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_42898521/Validation_42898521_42898521_750393.179.png", "Validation_42898521/Validation_42898521_42898521_750395.178.png", "Validation_42898521/Validation_42898521_42898521_750397.177.png", "Validation_42898521/Validation_42898521_42898521_750399.176.png", "Validation_42898521/Validation_42898521_42898521_750401.176.png" ], "turn_id": 4, "scan_id": "Validation_42898521", "origin_question": "In which turn were you closest to \"the stylish white door with glass panels\"? ", "option": [], "answer": "2", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the stylish white door with glass panels\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "Validation_42898521/Validation_42898521_42898521_750403.175.png", "Validation_42898521/Validation_42898521_42898521_750405.174.png", "Validation_42898521/Validation_42898521_42898521_750407.173.png", "Validation_42898521/Validation_42898521_42898521_750409.172.png", "Validation_42898521/Validation_42898521_42898521_750411.171.png" ], "turn_id": 5, "scan_id": "Validation_42898521", "origin_question": "Among these three objects(\"the medium-sized rectangular soft pastel towel\",\"the modern stainless steel rectangular oven with black accents\",\"the stove\" ), which one is the farthest from you?", "option": [ "the medium-sized rectangular soft pastel towel", "the modern stainless steel rectangular oven with black accents", "the stove" ], "answer": "the medium-sized rectangular soft pastel towel", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the medium-sized rectangular soft pastel towel\",\"the modern stainless steel rectangular oven with black accents\",\"the stove\" ), which one is the farthest from you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the medium-sized rectangular soft pastel towel', 'the modern stainless steel rectangular oven with black accents', 'the stove'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_42898521/Validation_42898521_42898521_750413.171.png", "Validation_42898521/Validation_42898521_42898521_750415.170.png", "Validation_42898521/Validation_42898521_42898521_750417.169.png", "Validation_42898521/Validation_42898521_42898521_750419.168.png", "Validation_42898521/Validation_42898521_42898521_750421.167.png" ], "turn_id": 6, "scan_id": "Validation_42898521", "origin_question": "Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Right", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_42898521/Validation_42898521_42898521_750423.167.png", "Validation_42898521/Validation_42898521_42898521_750425.166.png", "Validation_42898521/Validation_42898521_42898521_750427.165.png", "Validation_42898521/Validation_42898521_42898521_750429.164.png", "Validation_42898521/Validation_42898521_42898521_750431.163.png" ], "turn_id": 7, "scan_id": "Validation_42898521", "origin_question": "Is the medium-sized rectangular soft pastel towel to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the medium-sized rectangular soft pastel towel to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_42898521/Validation_42898521_42898521_750433.179.png", "Validation_42898521/Validation_42898521_42898521_750435.178.png", "Validation_42898521/Validation_42898521_42898521_750437.177.png", "Validation_42898521/Validation_42898521_42898521_750439.177.png", "Validation_42898521/Validation_42898521_42898521_750441.176.png" ], "turn_id": 8, "scan_id": "Validation_42898521", "origin_question": "How far is your current position from where you were at the end of the turn 4? (in meters)", "option": [], "answer": "1.64", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 4? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_42898521/Validation_42898521_42898521_750443.175.png", "Validation_42898521/Validation_42898521_42898521_750445.174.png", "Validation_42898521/Validation_42898521_42898521_750447.173.png", "Validation_42898521/Validation_42898521_42898521_750449.173.png", "Validation_42898521/Validation_42898521_42898521_750451.172.png" ], "turn_id": 9, "scan_id": "Validation_42898521", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the range hood\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the range hood\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "Validation_42898521/Validation_42898521_42898521_750453.171.png", "Validation_42898521/Validation_42898521_42898521_750455.170.png", "Validation_42898521/Validation_42898521_42898521_750457.169.png", "Validation_42898521/Validation_42898521_42898521_750459.169.png", "Validation_42898521/Validation_42898521_42898521_750461.168.png" ], "turn_id": 10, "scan_id": "Validation_42898521", "origin_question": "Among these three objects(\"the modern stainless steel rectangular oven with black accents\",\"the range hood\",\"the glossy black cylindrical kettle\" ), which one is the farthest from you?", "option": [ "the modern stainless steel rectangular oven with black accents", "the range hood", "the glossy black cylindrical kettle" ], "answer": "the glossy black cylindrical kettle", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the modern stainless steel rectangular oven with black accents\",\"the range hood\",\"the glossy black cylindrical kettle\" ), which one is the farthest from you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the modern stainless steel rectangular oven with black accents', 'the range hood', 'the glossy black cylindrical kettle'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "Validation_42445441/Validation_42445441_42445441_50214.796.png", "Validation_42445441/Validation_42445441_42445441_50216.796.png", "Validation_42445441/Validation_42445441_42445441_50218.795.png", "Validation_42445441/Validation_42445441_42445441_50220.794.png", "Validation_42445441/Validation_42445441_42445441_50222.793.png" ], "turn_id": 1, "scan_id": "Validation_42445441", "origin_question": "Remember, how many table(s) have you seen so far? ", "option": [], "answer": "2", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many table(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_42445441/Validation_42445441_42445441_50224.793.png", "Validation_42445441/Validation_42445441_42445441_50226.792.png", "Validation_42445441/Validation_42445441_42445441_50228.791.png", "Validation_42445441/Validation_42445441_42445441_50230.790.png", "Validation_42445441/Validation_42445441_42445441_50232.789.png" ], "turn_id": 2, "scan_id": "Validation_42445441", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"an off-white paneled door\" from your position?", "option": [], "answer": "159.38", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"an off-white paneled door\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_42445441/Validation_42445441_42445441_50234.789.png", "Validation_42445441/Validation_42445441_42445441_50236.788.png", "Validation_42445441/Validation_42445441_42445441_50238.787.png", "Validation_42445441/Validation_42445441_42445441_50240.786.png", "Validation_42445441/Validation_42445441_42445441_50242.785.png" ], "turn_id": 3, "scan_id": "Validation_42445441", "origin_question": "Please recall \"the cylindrical, minimalist chandelier in the neutral color\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.64", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the cylindrical, minimalist chandelier in the neutral color\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_42445441/Validation_42445441_42445441_50244.785.png", "Validation_42445441/Validation_42445441_42445441_50246.784.png", "Validation_42445441/Validation_42445441_42445441_50248.783.png", "Validation_42445441/Validation_42445441_42445441_50250.782.png", "Validation_42445441/Validation_42445441_42445441_50252.781.png" ], "turn_id": 4, "scan_id": "Validation_42445441", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_42445441/Validation_42445441_42445441_50254.781.png", "Validation_42445441/Validation_42445441_42445441_50256.780.png", "Validation_42445441/Validation_42445441_42445441_50258.796.png", "Validation_42445441/Validation_42445441_42445441_50260.795.png", "Validation_42445441/Validation_42445441_42445441_50262.794.png" ], "turn_id": 5, "scan_id": "Validation_42445441", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular window with the clear glass pane\", Y: \"the cylindrical, minimalist chandelier in the neutral color\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular window with the clear glass pane\", Y: \"the cylindrical, minimalist chandelier in the neutral color\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_42445441/Validation_42445441_42445441_50264.793.png", "Validation_42445441/Validation_42445441_42445441_50266.793.png", "Validation_42445441/Validation_42445441_42445441_50268.792.png", "Validation_42445441/Validation_42445441_42445441_50270.791.png", "Validation_42445441/Validation_42445441_42445441_50272.790.png" ], "turn_id": 6, "scan_id": "Validation_42445441", "origin_question": "Please recall \"the rectangular window with the clear glass pane\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.45", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the rectangular window with the clear glass pane\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_42445441/Validation_42445441_42445441_50274.789.png", "Validation_42445441/Validation_42445441_42445441_50276.789.png", "Validation_42445441/Validation_42445441_42445441_50278.788.png", "Validation_42445441/Validation_42445441_42445441_50280.787.png", "Validation_42445441/Validation_42445441_42445441_50282.786.png" ], "turn_id": 7, "scan_id": "Validation_42445441", "origin_question": "In which turn did you see both \"the modern three-seater brown couch with rounded armrests\" and \"the lush green decorative plant\" simultaneously? ", "option": [], "answer": "4", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the modern three-seater brown couch with rounded armrests\" and \"the lush green decorative plant\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_42445441/Validation_42445441_42445441_50284.785.png", "Validation_42445441/Validation_42445441_42445441_50286.785.png", "Validation_42445441/Validation_42445441_42445441_50288.784.png", "Validation_42445441/Validation_42445441_42445441_50290.783.png", "Validation_42445441/Validation_42445441_42445441_50292.782.png" ], "turn_id": 8, "scan_id": "Validation_42445441", "origin_question": "Please recall \"the cylindrical, minimalist chandelier in the neutral color\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.7", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the cylindrical, minimalist chandelier in the neutral color\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47430471/Validation_47430471_47430471_311527.667.png", "Validation_47430471/Validation_47430471_47430471_311529.633.png", "Validation_47430471/Validation_47430471_47430471_311531.649.png", "Validation_47430471/Validation_47430471_47430471_311533.648.png", "Validation_47430471/Validation_47430471_47430471_311535.647.png" ], "turn_id": 1, "scan_id": "Validation_47430471", "origin_question": "Remember, have you seen any folder(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any folder(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_47430471/Validation_47430471_47430471_311537.646.png", "Validation_47430471/Validation_47430471_47430471_311539.646.png", "Validation_47430471/Validation_47430471_47430471_311541.645.png", "Validation_47430471/Validation_47430471_47430471_311543.644.png", "Validation_47430471/Validation_47430471_47430471_311545.643.png" ], "turn_id": 2, "scan_id": "Validation_47430471", "origin_question": "When did you first discover \"the sleek, glossy black monitor with the high-definition screen\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the sleek, glossy black monitor with the high-definition screen\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "Validation_47430471/Validation_47430471_47430471_311547.642.png", "Validation_47430471/Validation_47430471_47430471_311549.642.png", "Validation_47430471/Validation_47430471_47430471_311551.641.png", "Validation_47430471/Validation_47430471_47430471_311553.640.png", "Validation_47430471/Validation_47430471_47430471_311555.639.png" ], "turn_id": 3, "scan_id": "Validation_47430471", "origin_question": "Remember, how many stool(s) have you seen so far? ", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many stool(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_47430471/Validation_47430471_47430471_311557.638.png", "Validation_47430471/Validation_47430471_47430471_311559.637.png", "Validation_47430471/Validation_47430471_47430471_311561.637.png", "Validation_47430471/Validation_47430471_47430471_311563.636.png", "Validation_47430471/Validation_47430471_47430471_311565.635.png" ], "turn_id": 4, "scan_id": "Validation_47430471", "origin_question": "At the end of which turn were both of \"the medium-sized, rectangular plate\" and \"the lightweight, white rectangular plastic stool\" on your left side?", "option": [], "answer": "3", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the medium-sized, rectangular plate\" and \"the lightweight, white rectangular plastic stool\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_47430471/Validation_47430471_47430471_311567.634.png", "Validation_47430471/Validation_47430471_47430471_311569.633.png", "Validation_47430471/Validation_47430471_47430471_311571.633.png", "Validation_47430471/Validation_47430471_47430471_311573.648.png", "Validation_47430471/Validation_47430471_47430471_311575.648.png" ], "turn_id": 5, "scan_id": "Validation_47430471", "origin_question": "Which two objects are on the same side of you now? \"the medium-sized, rectangular plate\", \"the lightweight, white rectangular plastic stool\", \"the modern white wooden shelf with multiple levels\". ", "option": [ "the medium-sized, rectangular plate, the lightweight, white rectangular plastic stool", "the medium-sized, rectangular plate, the modern white wooden shelf with multiple levels", "the lightweight, white rectangular plastic stool, the modern white wooden shelf with multiple levels" ], "answer": "the medium-sized, rectangular plate, the lightweight, white rectangular plastic stool", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the medium-sized, rectangular plate\", \"the lightweight, white rectangular plastic stool\", \"the modern white wooden shelf with multiple levels\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the medium-sized, rectangular plate, the lightweight, white rectangular plastic stool', 'the medium-sized, rectangular plate, the modern white wooden shelf with multiple levels', 'the lightweight, white rectangular plastic stool, the modern white wooden shelf with multiple levels'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47430471/Validation_47430471_47430471_311577.647.png", "Validation_47430471/Validation_47430471_47430471_311579.646.png", "Validation_47430471/Validation_47430471_47430471_311581.645.png", "Validation_47430471/Validation_47430471_47430471_311583.644.png", "Validation_47430471/Validation_47430471_47430471_311585.644.png" ], "turn_id": 6, "scan_id": "Validation_47430471", "origin_question": "In which turn were you closest to \"the medium-sized, rectangular plate\"? ", "option": [], "answer": "1", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the medium-sized, rectangular plate\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "Validation_47430471/Validation_47430471_47430471_311587.643.png", "Validation_47430471/Validation_47430471_47430471_311589.642.png", "Validation_47430471/Validation_47430471_47430471_311591.641.png", "Validation_47430471/Validation_47430471_47430471_311593.640.png", "Validation_47430471/Validation_47430471_47430471_311595.639.png" ], "turn_id": 7, "scan_id": "Validation_47430471", "origin_question": "Among these three objects(\"the sleek, glossy black monitor with the high-definition screen\",\"the lightweight, white rectangular plastic stool\",\"the modern kitchenware organizer with the circular design\" ), which one is the closest to you?", "option": [ "the sleek, glossy black monitor with the high-definition screen", "the lightweight, white rectangular plastic stool", "the modern kitchenware organizer with the circular design" ], "answer": "the lightweight, white rectangular plastic stool", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the sleek, glossy black monitor with the high-definition screen\",\"the lightweight, white rectangular plastic stool\",\"the modern kitchenware organizer with the circular design\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the sleek, glossy black monitor with the high-definition screen', 'the lightweight, white rectangular plastic stool', 'the modern kitchenware organizer with the circular design'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "Validation_47430471/Validation_47430471_47430471_311597.639.png", "Validation_47430471/Validation_47430471_47430471_311599.638.png", "Validation_47430471/Validation_47430471_47430471_311601.637.png", "Validation_47430471/Validation_47430471_47430471_311603.636.png", "Validation_47430471/Validation_47430471_47430471_311605.635.png" ], "turn_id": 8, "scan_id": "Validation_47430471", "origin_question": "Among these three objects(\"the contemporary dark brown door with frosted glass panels\",\"the lightweight, white rectangular plastic stool\",\"the sleek, modern black glass cooktop stove\" ), which one is the closest to you?", "option": [ "the contemporary dark brown door with frosted glass panels", "the lightweight, white rectangular plastic stool", "the sleek, modern black glass cooktop stove" ], "answer": "the lightweight, white rectangular plastic stool", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the contemporary dark brown door with frosted glass panels\",\"the lightweight, white rectangular plastic stool\",\"the sleek, modern black glass cooktop stove\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the contemporary dark brown door with frosted glass panels', 'the lightweight, white rectangular plastic stool', 'the sleek, modern black glass cooktop stove'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47430471/Validation_47430471_47430471_311607.634.png", "Validation_47430471/Validation_47430471_47430471_311609.634.png", "Validation_47430471/Validation_47430471_47430471_311611.633.png", "Validation_47430471/Validation_47430471_47430471_311613.649.png", "Validation_47430471/Validation_47430471_47430471_311615.648.png" ], "turn_id": 9, "scan_id": "Validation_47430471", "origin_question": "Which direction is the contemporary dark brown door with frosted glass panels in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the contemporary dark brown door with frosted glass panels in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_47430471/Validation_47430471_47430471_311617.647.png", "Validation_47430471/Validation_47430471_47430471_311619.646.png", "Validation_47430471/Validation_47430471_47430471_311621.645.png", "Validation_47430471/Validation_47430471_47430471_311623.645.png", "Validation_47430471/Validation_47430471_47430471_311625.644.png" ], "turn_id": 10, "scan_id": "Validation_47430471", "origin_question": "When did you first discover \"the tall, stainless steel rectangular refrigerator\"? (index of the turn)", "option": [], "answer": "3", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the tall, stainless steel rectangular refrigerator\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47334367/Validation_47334367_47334367_63175.494.png", "Validation_47334367/Validation_47334367_47334367_63177.493.png", "Validation_47334367/Validation_47334367_47334367_63179.492.png", "Validation_47334367/Validation_47334367_47334367_63181.491.png", "Validation_47334367/Validation_47334367_47334367_63183.490.png" ], "turn_id": 1, "scan_id": "Validation_47334367", "origin_question": "Remember, have you seen any washing machine(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any washing machine(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_47334367/Validation_47334367_47334367_63185.490.png", "Validation_47334367/Validation_47334367_47334367_63187.489.png", "Validation_47334367/Validation_47334367_47334367_63189.488.png", "Validation_47334367/Validation_47334367_47334367_63191.487.png", "Validation_47334367/Validation_47334367_47334367_63193.486.png" ], "turn_id": 2, "scan_id": "Validation_47334367", "origin_question": "When did you first discover \"the sleek, white and gray washing machine\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the sleek, white and gray washing machine\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_47334367/Validation_47334367_47334367_63195.486.png", "Validation_47334367/Validation_47334367_47334367_63197.485.png", "Validation_47334367/Validation_47334367_47334367_63199.484.png", "Validation_47334367/Validation_47334367_47334367_63201.483.png", "Validation_47334367/Validation_47334367_47334367_63203.482.png" ], "turn_id": 3, "scan_id": "Validation_47334367", "origin_question": "Is the sleek white rectangular sink to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the sleek white rectangular sink to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47334367/Validation_47334367_47334367_63205.482.png", "Validation_47334367/Validation_47334367_47334367_63207.497.png", "Validation_47334367/Validation_47334367_47334367_63209.497.png", "Validation_47334367/Validation_47334367_47334367_63211.496.png", "Validation_47334367/Validation_47334367_47334367_63213.495.png" ], "turn_id": 4, "scan_id": "Validation_47334367", "origin_question": "Which direction is the sleek white rectangular sink in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the sleek white rectangular sink in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_47334367/Validation_47334367_47334367_63215.494.png", "Validation_47334367/Validation_47334367_47334367_63217.493.png", "Validation_47334367/Validation_47334367_47334367_63219.493.png", "Validation_47334367/Validation_47334367_47334367_63221.492.png", "Validation_47334367/Validation_47334367_47334367_63223.491.png" ], "turn_id": 5, "scan_id": "Validation_47334367", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_47334367/Validation_47334367_47334367_63225.490.png", "Validation_47334367/Validation_47334367_47334367_63227.489.png", "Validation_47334367/Validation_47334367_47334367_63229.488.png", "Validation_47334367/Validation_47334367_47334367_63231.488.png", "Validation_47334367/Validation_47334367_47334367_63233.487.png" ], "turn_id": 6, "scan_id": "Validation_47334367", "origin_question": "When did you first discover \"the sleek white rectangular sink\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the sleek white rectangular sink\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47334367/Validation_47334367_47334367_63235.486.png", "Validation_47334367/Validation_47334367_47334367_63237.485.png", "Validation_47334367/Validation_47334367_47334367_63239.484.png", "Validation_47334367/Validation_47334367_47334367_63241.484.png", "Validation_47334367/Validation_47334367_47334367_63243.483.png" ], "turn_id": 7, "scan_id": "Validation_47334367", "origin_question": "In which turn were you closest to \"the bathtub\"? ", "option": [], "answer": "5", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the bathtub\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_47334367/Validation_47334367_47334367_63245.482.png", "Validation_47334367/Validation_47334367_47334367_63247.498.png", "Validation_47334367/Validation_47334367_47334367_63249.497.png", "Validation_47334367/Validation_47334367_47334367_63251.496.png", "Validation_47334367/Validation_47334367_47334367_63253.495.png" ], "turn_id": 8, "scan_id": "Validation_47334367", "origin_question": "Please recall \"the collection of toiletry items in neutral and vibrant colors\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.88", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the collection of toiletry items in neutral and vibrant colors\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47334367/Validation_47334367_47334367_63255.495.png", "Validation_47334367/Validation_47334367_47334367_63257.494.png", "Validation_47334367/Validation_47334367_47334367_63259.493.png", "Validation_47334367/Validation_47334367_47334367_63261.492.png", "Validation_47334367/Validation_47334367_47334367_63263.491.png" ], "turn_id": 9, "scan_id": "Validation_47334367", "origin_question": "Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the sleek white rectangular sink\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the sleek white rectangular sink\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_48018353/Validation_48018353_48018353_389122.669.png", "Validation_48018353/Validation_48018353_48018353_389124.151.png", "Validation_48018353/Validation_48018353_48018353_389125.651.png", "Validation_48018353/Validation_48018353_48018353_389127.150.png" ], "turn_id": 1, "scan_id": "Validation_48018353", "origin_question": "Remember, have you seen any treadmill(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any treadmill(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_48018353/Validation_48018353_48018353_389128.649.png", "Validation_48018353/Validation_48018353_48018353_389130.149.png", "Validation_48018353/Validation_48018353_48018353_389131.648.png", "Validation_48018353/Validation_48018353_48018353_389133.148.png" ], "turn_id": 2, "scan_id": "Validation_48018353", "origin_question": "Is the rectangular light wood table with particle board to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the rectangular light wood table with particle board to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_48018353/Validation_48018353_48018353_389134.647.png", "Validation_48018353/Validation_48018353_48018353_389136.146.png", "Validation_48018353/Validation_48018353_48018353_389137.646.png", "Validation_48018353/Validation_48018353_48018353_389139.145.png" ], "turn_id": 3, "scan_id": "Validation_48018353", "origin_question": "", "option": [], "answer": "", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_48018353/Validation_48018353_48018353_389140.644.png", "Validation_48018353/Validation_48018353_48018353_389142.144.png", "Validation_48018353/Validation_48018353_48018353_389143.643.png", "Validation_48018353/Validation_48018353_48018353_389145.143.png" ], "turn_id": 4, "scan_id": "Validation_48018353", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular light wood table with particle board\", Y: \"the window\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular light wood table with particle board\", Y: \"the window\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_48018353/Validation_48018353_48018353_389146.642.png", "Validation_48018353/Validation_48018353_48018353_389148.141.png", "Validation_48018353/Validation_48018353_48018353_389149.641.png", "Validation_48018353/Validation_48018353_48018353_389151.157.png" ], "turn_id": 5, "scan_id": "Validation_48018353", "origin_question": "When did you first discover \"the compact-sized bed with sleek red upholstery\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the compact-sized bed with sleek red upholstery\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47334241/Validation_47334241_47334241_111233.938.png", "Validation_47334241/Validation_47334241_47334241_111235.920.png", "Validation_47334241/Validation_47334241_47334241_111237.919.png", "Validation_47334241/Validation_47334241_47334241_111239.918.png", "Validation_47334241/Validation_47334241_47334241_111241.918.png" ], "turn_id": 1, "scan_id": "Validation_47334241", "origin_question": "Remember, have you seen any towel(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any towel(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47334241/Validation_47334241_47334241_111243.917.png", "Validation_47334241/Validation_47334241_47334241_111245.916.png", "Validation_47334241/Validation_47334241_47334241_111247.915.png", "Validation_47334241/Validation_47334241_47334241_111249.914.png", "Validation_47334241/Validation_47334241_47334241_111251.914.png" ], "turn_id": 2, "scan_id": "Validation_47334241", "origin_question": "In which turn were you farthest from \"the vibrant striped towel made of soft, absorbent fabric\"? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the vibrant striped towel made of soft, absorbent fabric\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_47334241/Validation_47334241_47334241_111253.929.png", "Validation_47334241/Validation_47334241_47334241_111255.929.png", "Validation_47334241/Validation_47334241_47334241_111257.928.png", "Validation_47334241/Validation_47334241_47334241_111259.927.png", "Validation_47334241/Validation_47334241_47334241_111261.926.png" ], "turn_id": 3, "scan_id": "Validation_47334241", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_47334241/Validation_47334241_47334241_111263.925.png", "Validation_47334241/Validation_47334241_47334241_111265.925.png", "Validation_47334241/Validation_47334241_47334241_111267.924.png", "Validation_47334241/Validation_47334241_47334241_111269.923.png", "Validation_47334241/Validation_47334241_47334241_111271.922.png" ], "turn_id": 4, "scan_id": "Validation_47334241", "origin_question": "When did you first discover \"the vibrant striped towel made of soft, absorbent fabric\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the vibrant striped towel made of soft, absorbent fabric\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "Validation_47334241/Validation_47334241_47334241_111273.921.png", "Validation_47334241/Validation_47334241_47334241_111275.921.png", "Validation_47334241/Validation_47334241_47334241_111277.920.png", "Validation_47334241/Validation_47334241_47334241_111279.919.png", "Validation_47334241/Validation_47334241_47334241_111281.918.png" ], "turn_id": 5, "scan_id": "Validation_47334241", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the sleek, light-colored modern cabinet with closed doors\"; \"the modern glossy white rectangular sink\"; \"the medium-sized glossy white toilet with an oval shape\" ", "option": [ "the sleek, light-colored modern cabinet with closed doors", "the modern glossy white rectangular sink", "the medium-sized glossy white toilet with an oval shape" ], "answer": "the medium-sized glossy white toilet with an oval shape", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the sleek, light-colored modern cabinet with closed doors\"; \"the modern glossy white rectangular sink\"; \"the medium-sized glossy white toilet with an oval shape\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the sleek, light-colored modern cabinet with closed doors', 'the modern glossy white rectangular sink', 'the medium-sized glossy white toilet with an oval shape'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_47334241/Validation_47334241_47334241_111283.917.png", "Validation_47334241/Validation_47334241_47334241_111285.917.png", "Validation_47334241/Validation_47334241_47334241_111287.916.png", "Validation_47334241/Validation_47334241_47334241_111289.915.png", "Validation_47334241/Validation_47334241_47334241_111291.914.png" ], "turn_id": 6, "scan_id": "Validation_47334241", "origin_question": "At the end of which turn were both of \"the modern glossy white rectangular sink\" and \"the white bottle with the red cap\" on your right side?", "option": [], "answer": "6", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the modern glossy white rectangular sink\" and \"the white bottle with the red cap\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47334241/Validation_47334241_47334241_111293.930.png", "Validation_47334241/Validation_47334241_47334241_111295.929.png", "Validation_47334241/Validation_47334241_47334241_111297.928.png", "Validation_47334241/Validation_47334241_47334241_111299.928.png", "Validation_47334241/Validation_47334241_47334241_111301.927.png" ], "turn_id": 7, "scan_id": "Validation_47334241", "origin_question": "", "option": [], "answer": "", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_47334241/Validation_47334241_47334241_111303.926.png", "Validation_47334241/Validation_47334241_47334241_111305.925.png", "Validation_47334241/Validation_47334241_47334241_111307.924.png", "Validation_47334241/Validation_47334241_47334241_111309.923.png", "Validation_47334241/Validation_47334241_47334241_111311.923.png" ], "turn_id": 8, "scan_id": "Validation_47334241", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "1.17", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47334241/Validation_47334241_47334241_111313.922.png", "Validation_47334241/Validation_47334241_47334241_111315.921.png", "Validation_47334241/Validation_47334241_47334241_111317.920.png", "Validation_47334241/Validation_47334241_47334241_111319.919.png", "Validation_47334241/Validation_47334241_47334241_111321.919.png" ], "turn_id": 9, "scan_id": "Validation_47334241", "origin_question": "Which direction is the modern glossy white rectangular sink in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the modern glossy white rectangular sink in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_47334241/Validation_47334241_47334241_111323.918.png", "Validation_47334241/Validation_47334241_47334241_111325.917.png", "Validation_47334241/Validation_47334241_47334241_111327.916.png", "Validation_47334241/Validation_47334241_47334241_111329.915.png", "Validation_47334241/Validation_47334241_47334241_111331.914.png" ], "turn_id": 10, "scan_id": "Validation_47334241", "origin_question": "Is the modern glossy white rectangular sink to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the modern glossy white rectangular sink to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0647_00/00000.jpg", "scene0647_00/00040.jpg", "scene0647_00/00080.jpg", "scene0647_00/00120.jpg" ], "turn_id": 1, "scan_id": "scene0647_00", "origin_question": "Remember, how many stool(s) have you seen so far? ", "option": [], "answer": "2", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many stool(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0647_00/00160.jpg", "scene0647_00/00200.jpg", "scene0647_00/00240.jpg", "scene0647_00/00280.jpg" ], "turn_id": 2, "scan_id": "scene0647_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0647_00/00320.jpg", "scene0647_00/00360.jpg", "scene0647_00/00400.jpg", "scene0647_00/00440.jpg" ], "turn_id": 3, "scan_id": "scene0647_00", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "3.21", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0647_00/00480.jpg", "scene0647_00/00520.jpg", "scene0647_00/00560.jpg", "scene0647_00/00600.jpg" ], "turn_id": 4, "scan_id": "scene0647_00", "origin_question": "In which turn were you closest to \"the moderate-sized black flat-screen television\"? ", "option": [], "answer": "1", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the moderate-sized black flat-screen television\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0647_00/00640.jpg", "scene0647_00/00680.jpg", "scene0647_00/00720.jpg", "scene0647_00/00760.jpg" ], "turn_id": 5, "scan_id": "scene0647_00", "origin_question": "What will be the first-time appearance order of the following categories: backpack, couch, picture?", "option": [ "backpack, couch, picture", "picture, backpack, couch", "picture, couch, backpack", "backpack, picture, couch" ], "answer": "picture, couch, backpack", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: backpack, couch, picture? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['backpack, couch, picture', 'picture, backpack, couch', 'picture, couch, backpack', 'backpack, picture, couch'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0647_00/00800.jpg", "scene0647_00/00840.jpg", "scene0647_00/00880.jpg", "scene0647_00/00920.jpg" ], "turn_id": 6, "scan_id": "scene0647_00", "origin_question": "Is the light brown wooden chair with an elegant backrest to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the light brown wooden chair with an elegant backrest to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0193_01/00000.jpg", "scene0193_01/00030.jpg", "scene0193_01/00060.jpg", "scene0193_01/00090.jpg" ], "turn_id": 1, "scan_id": "scene0193_01", "origin_question": "Remember, have you seen any jar(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any jar(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 3 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0193_01/00120.jpg", "scene0193_01/00150.jpg", "scene0193_01/00180.jpg", "scene0193_01/00210.jpg" ], "turn_id": 2, "scan_id": "scene0193_01", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the medium rectangular white wooden door with a silver handle\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the medium rectangular white wooden door with a silver handle\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 3 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0193_01/00240.jpg", "scene0193_01/00270.jpg", "scene0193_01/00300.jpg", "scene0193_01/00330.jpg" ], "turn_id": 3, "scan_id": "scene0193_01", "origin_question": "In which turn were you farthest from \"the rectangular light-colored wooden bed with a burgundy mattress\"? ", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the rectangular light-colored wooden bed with a burgundy mattress\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 3 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0704_00/00000.jpg", "scene0704_00/00040.jpg", "scene0704_00/00080.jpg", "scene0704_00/00120.jpg", "scene0704_00/00160.jpg" ], "turn_id": 1, "scan_id": "scene0704_00", "origin_question": "Remember, have you seen any counter(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any counter(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0704_00/00200.jpg", "scene0704_00/00240.jpg", "scene0704_00/00280.jpg", "scene0704_00/00320.jpg", "scene0704_00/00360.jpg" ], "turn_id": 2, "scan_id": "scene0704_00", "origin_question": "At the end of which turn were both of \"the small rectangular silver-gray cart\" and \"the rectangular light brown wooden table\" on your right side?", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the small rectangular silver-gray cart\" and \"the rectangular light brown wooden table\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0704_00/00400.jpg", "scene0704_00/00440.jpg", "scene0704_00/00480.jpg", "scene0704_00/00520.jpg", "scene0704_00/00560.jpg" ], "turn_id": 3, "scan_id": "scene0704_00", "origin_question": "In which turn did you see both \"the broom\" and \"the medium rectangular brown cardboard box\" simultaneously? ", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the broom\" and \"the medium rectangular brown cardboard box\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0704_00/00600.jpg", "scene0704_00/00640.jpg", "scene0704_00/00680.jpg", "scene0704_00/00720.jpg", "scene0704_00/00760.jpg" ], "turn_id": 4, "scan_id": "scene0704_00", "origin_question": "Which two objects are on the same side of you now? \"the modern white multifunctional printer with a control panel\", \"the copier\", \"the small black metal shelf with compartments for bottles\". ", "option": [ "the modern white multifunctional printer with a control panel, the copier", "the modern white multifunctional printer with a control panel, the small black metal shelf with compartments for bottles", "the copier, the small black metal shelf with compartments for bottles" ], "answer": "the copier, the small black metal shelf with compartments for bottles", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the modern white multifunctional printer with a control panel\", \"the copier\", \"the small black metal shelf with compartments for bottles\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the modern white multifunctional printer with a control panel, the copier', 'the modern white multifunctional printer with a control panel, the small black metal shelf with compartments for bottles', 'the copier, the small black metal shelf with compartments for bottles'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0704_00/00800.jpg", "scene0704_00/00840.jpg", "scene0704_00/00880.jpg", "scene0704_00/00920.jpg", "scene0704_00/00960.jpg" ], "turn_id": 5, "scan_id": "scene0704_00", "origin_question": "What will be the first-time appearance order of the following categories: broom, printer, water cooler?", "option": [ "water cooler, broom, printer", "printer, broom, water cooler", "printer, water cooler, broom", "water cooler, printer, broom" ], "answer": "printer, broom, water cooler", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: broom, printer, water cooler? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['water cooler, broom, printer', 'printer, broom, water cooler', 'printer, water cooler, broom', 'water cooler, printer, broom'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0704_00/01000.jpg", "scene0704_00/01040.jpg", "scene0704_00/01080.jpg", "scene0704_00/01120.jpg", "scene0704_00/01160.jpg" ], "turn_id": 6, "scan_id": "scene0704_00", "origin_question": "Using your orientation at the end of turn 5 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "98.59", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 5 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0704_00/01200.jpg", "scene0704_00/01240.jpg", "scene0704_00/01280.jpg", "scene0704_00/01320.jpg", "scene0704_00/01360.jpg" ], "turn_id": 7, "scan_id": "scene0704_00", "origin_question": "Which direction is the modern white multifunctional printer with a control panel in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the modern white multifunctional printer with a control panel in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0704_00/01400.jpg", "scene0704_00/01440.jpg", "scene0704_00/01480.jpg", "scene0704_00/01520.jpg", "scene0704_00/01560.jpg" ], "turn_id": 8, "scan_id": "scene0704_00", "origin_question": "Using your orientation at the end of turn 6 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "111.79", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 6 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0704_00/01600.jpg", "scene0704_00/01640.jpg", "scene0704_00/01680.jpg", "scene0704_00/01720.jpg", "scene0704_00/01760.jpg" ], "turn_id": 9, "scan_id": "scene0704_00", "origin_question": "Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0704_00/01800.jpg", "scene0704_00/01840.jpg", "scene0704_00/01880.jpg", "scene0704_00/01920.jpg", "scene0704_00/01960.jpg" ], "turn_id": 10, "scan_id": "scene0704_00", "origin_question": "By the end of the 6 turn, were you getting closer to or farther away from these two objects: X: \"the modern white multifunctional printer with a control panel\", Y: \"the copier\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to X and farther away from Y", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 6 turn, were you getting closer to or farther away from these two objects: X: \"the modern white multifunctional printer with a control panel\", Y: \"the copier\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0187_01/00000.jpg", "scene0187_01/00040.jpg", "scene0187_01/00080.jpg", "scene0187_01/00120.jpg", "scene0187_01/00160.jpg" ], "turn_id": 1, "scan_id": "scene0187_01", "origin_question": "Remember, have you seen any player(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any player(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0187_01/00200.jpg", "scene0187_01/00240.jpg", "scene0187_01/00280.jpg", "scene0187_01/00320.jpg", "scene0187_01/00360.jpg" ], "turn_id": 2, "scan_id": "scene0187_01", "origin_question": "In which turn were you farthest from \"the round brown wooden plate with a smooth shape\"? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the round brown wooden plate with a smooth shape\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0187_01/00400.jpg", "scene0187_01/00440.jpg", "scene0187_01/00480.jpg", "scene0187_01/00520.jpg", "scene0187_01/00560.jpg" ], "turn_id": 3, "scan_id": "scene0187_01", "origin_question": "", "option": [], "answer": "", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0187_01/00600.jpg", "scene0187_01/00640.jpg", "scene0187_01/00680.jpg", "scene0187_01/00720.jpg", "scene0187_01/00760.jpg" ], "turn_id": 4, "scan_id": "scene0187_01", "origin_question": "Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0187_01/00800.jpg", "scene0187_01/00840.jpg", "scene0187_01/00880.jpg", "scene0187_01/00920.jpg", "scene0187_01/00960.jpg" ], "turn_id": 5, "scan_id": "scene0187_01", "origin_question": "In which turn were you closest to \"the round brown wooden plate with a smooth shape\"? ", "option": [], "answer": "3", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the round brown wooden plate with a smooth shape\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0187_01/01000.jpg", "scene0187_01/01040.jpg", "scene0187_01/01080.jpg", "scene0187_01/01120.jpg", "scene0187_01/01160.jpg" ], "turn_id": 6, "scan_id": "scene0187_01", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the medium-sized, dark-colored backpack with multiple compartments\" from your position?", "option": [], "answer": "177.26", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the medium-sized, dark-colored backpack with multiple compartments\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0187_01/01200.jpg", "scene0187_01/01240.jpg", "scene0187_01/01280.jpg", "scene0187_01/01320.jpg", "scene0187_01/01360.jpg" ], "turn_id": 7, "scan_id": "scene0187_01", "origin_question": "Which direction is the round brown wooden plate with a smooth shape in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the round brown wooden plate with a smooth shape in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0187_01/01400.jpg", "scene0187_01/01440.jpg", "scene0187_01/01480.jpg", "scene0187_01/01520.jpg", "scene0187_01/01560.jpg" ], "turn_id": 8, "scan_id": "scene0187_01", "origin_question": "Is the medium-sized, dark-colored backpack with multiple compartments to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the medium-sized, dark-colored backpack with multiple compartments to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0187_01/01600.jpg", "scene0187_01/01640.jpg", "scene0187_01/01680.jpg", "scene0187_01/01720.jpg", "scene0187_01/01760.jpg" ], "turn_id": 9, "scan_id": "scene0187_01", "origin_question": "Compared to your position at the end of the 6 turn, are you now closer or farther away from \"the medium-sized, dark-colored backpack with multiple compartments\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 6 turn, are you now closer or farther away from \"the medium-sized, dark-colored backpack with multiple compartments\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0359_01/00000.jpg", "scene0359_01/00030.jpg", "scene0359_01/00060.jpg", "scene0359_01/00090.jpg" ], "turn_id": 1, "scan_id": "scene0359_01", "origin_question": "Remember, have you seen any chair(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any chair(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0359_01/00120.jpg", "scene0359_01/00150.jpg", "scene0359_01/00180.jpg", "scene0359_01/00210.jpg" ], "turn_id": 2, "scan_id": "scene0359_01", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the cylindrical glossy plastic bottle with a green nozzle\" from your position?", "option": [], "answer": "115.83", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the cylindrical glossy plastic bottle with a green nozzle\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0359_01/00240.jpg", "scene0359_01/00270.jpg", "scene0359_01/00300.jpg", "scene0359_01/00400.jpg" ], "turn_id": 3, "scan_id": "scene0359_01", "origin_question": "What will be the first-time appearance order of the following categories: bed, tissue, window?", "option": [ "tissue, window, bed", "window, bed, tissue", "bed, tissue, window", "window, tissue, bed" ], "answer": "tissue, window, bed", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bed, tissue, window? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['tissue, window, bed', 'window, bed, tissue', 'bed, tissue, window', 'window, tissue, bed'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0359_01/00430.jpg", "scene0359_01/00460.jpg", "scene0359_01/00490.jpg", "scene0359_01/00520.jpg" ], "turn_id": 4, "scan_id": "scene0359_01", "origin_question": "Which two objects are on the same side of you now? \"the medium-sized rectangular wooden desk with a drawer unit\", \"the crumpled light beige paper\", \"the bed\". ", "option": [ "the medium-sized rectangular wooden desk with a drawer unit, the crumpled light beige paper", "the medium-sized rectangular wooden desk with a drawer unit, the bed", "the crumpled light beige paper, the bed" ], "answer": "the medium-sized rectangular wooden desk with a drawer unit, the crumpled light beige paper", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the medium-sized rectangular wooden desk with a drawer unit\", \"the crumpled light beige paper\", \"the bed\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the medium-sized rectangular wooden desk with a drawer unit, the crumpled light beige paper', 'the medium-sized rectangular wooden desk with a drawer unit, the bed', 'the crumpled light beige paper, the bed'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0017_02/00000.jpg", "scene0017_02/00030.jpg", "scene0017_02/00060.jpg", "scene0017_02/00090.jpg" ], "turn_id": 1, "scan_id": "scene0017_02", "origin_question": "Remember, how many monitor(s) have you seen so far? ", "option": [], "answer": "3", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many monitor(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0017_02/00120.jpg", "scene0017_02/00150.jpg", "scene0017_02/00180.jpg", "scene0017_02/00210.jpg" ], "turn_id": 2, "scan_id": "scene0017_02", "origin_question": "Which two objects are on the same side of you now? \"the stylish pair of black and silver headphones\", \"the white rectangular sheet of standard paper\", \"the tall, light-colored wooden cabinet with a sleek surface\". ", "option": [ "the stylish pair of black and silver headphones, the white rectangular sheet of standard paper", "the stylish pair of black and silver headphones, the tall, light-colored wooden cabinet with a sleek surface", "the white rectangular sheet of standard paper, the tall, light-colored wooden cabinet with a sleek surface" ], "answer": "the stylish pair of black and silver headphones, the white rectangular sheet of standard paper", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the stylish pair of black and silver headphones\", \"the white rectangular sheet of standard paper\", \"the tall, light-colored wooden cabinet with a sleek surface\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the stylish pair of black and silver headphones, the white rectangular sheet of standard paper', 'the stylish pair of black and silver headphones, the tall, light-colored wooden cabinet with a sleek surface', 'the white rectangular sheet of standard paper, the tall, light-colored wooden cabinet with a sleek surface'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0017_02/00240.jpg", "scene0017_02/00270.jpg", "scene0017_02/00300.jpg", "scene0017_02/00330.jpg" ], "turn_id": 3, "scan_id": "scene0017_02", "origin_question": "Which direction is the stylish pair of black and silver headphones in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the stylish pair of black and silver headphones in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0017_02/00360.jpg", "scene0017_02/00390.jpg", "scene0017_02/00420.jpg", "scene0017_02/00450.jpg" ], "turn_id": 4, "scan_id": "scene0017_02", "origin_question": "Which direction is the small red cylindrical container with a white lid in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the small red cylindrical container with a white lid in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0017_02/00480.jpg", "scene0017_02/00510.jpg", "scene0017_02/00540.jpg", "scene0017_02/00570.jpg" ], "turn_id": 5, "scan_id": "scene0017_02", "origin_question": "When did you first discover \"the small red cylindrical container with a white lid\"? (index of the turn)", "option": [], "answer": "3", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the small red cylindrical container with a white lid\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0017_02/00600.jpg", "scene0017_02/00630.jpg", "scene0017_02/00660.jpg", "scene0017_02/00690.jpg" ], "turn_id": 6, "scan_id": "scene0017_02", "origin_question": "What will be the first-time appearance order of the following categories: cabinet, computer, window?", "option": [ "computer, cabinet, window", "window, computer, cabinet", "cabinet, computer, window", "computer, window, cabinet" ], "answer": "cabinet, computer, window", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: cabinet, computer, window? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['computer, cabinet, window', 'window, computer, cabinet', 'cabinet, computer, window', 'computer, window, cabinet'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0017_02/00720.jpg", "scene0017_02/00750.jpg", "scene0017_02/00780.jpg", "scene0017_02/00810.jpg" ], "turn_id": 7, "scan_id": "scene0017_02", "origin_question": "Which direction is the medium-sized rectangular computer in light brown in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the medium-sized rectangular computer in light brown in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0017_02/00840.jpg", "scene0017_02/00870.jpg", "scene0017_02/00900.jpg", "scene0017_02/00930.jpg" ], "turn_id": 8, "scan_id": "scene0017_02", "origin_question": "In which turn did you see both \"the windowsill\" and \"the medium-sized rectangular computer in light brown\" simultaneously? ", "option": [], "answer": "3", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the windowsill\" and \"the medium-sized rectangular computer in light brown\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0610_01/00000.jpg", "scene0610_01/00040.jpg", "scene0610_01/00080.jpg", "scene0610_01/00120.jpg", "scene0610_01/00160.jpg" ], "turn_id": 1, "scan_id": "scene0610_01", "origin_question": "Remember, have you seen any tissue(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any tissue(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0610_01/00200.jpg", "scene0610_01/00240.jpg", "scene0610_01/00280.jpg", "scene0610_01/00320.jpg", "scene0610_01/00360.jpg" ], "turn_id": 2, "scan_id": "scene0610_01", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the fluffy white bath towel made of absorbent cotton\"; \"the sink\"; \"the vanity\" ", "option": [ "the fluffy white bath towel made of absorbent cotton", "the sink", "the vanity" ], "answer": "the fluffy white bath towel made of absorbent cotton", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the fluffy white bath towel made of absorbent cotton\"; \"the sink\"; \"the vanity\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the fluffy white bath towel made of absorbent cotton', 'the sink', 'the vanity'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0610_01/00660.jpg", "scene0610_01/00700.jpg", "scene0610_01/00740.jpg", "scene0610_01/00780.jpg", "scene0610_01/00820.jpg" ], "turn_id": 3, "scan_id": "scene0610_01", "origin_question": "Which direction is the bin in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the bin in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0610_01/00860.jpg", "scene0610_01/00900.jpg", "scene0610_01/00940.jpg", "scene0610_01/00980.jpg", "scene0610_01/01020.jpg" ], "turn_id": 4, "scan_id": "scene0610_01", "origin_question": "Which two objects are on the same side of you now? \"the vanity\", \"the fluffy white bath towel made of absorbent cotton\", \"the small wooden stool with a light brown square top\". ", "option": [ "the vanity, the fluffy white bath towel made of absorbent cotton", "the vanity, the small wooden stool with a light brown square top", "the fluffy white bath towel made of absorbent cotton, the small wooden stool with a light brown square top" ], "answer": "the vanity, the fluffy white bath towel made of absorbent cotton", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the vanity\", \"the fluffy white bath towel made of absorbent cotton\", \"the small wooden stool with a light brown square top\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the vanity, the fluffy white bath towel made of absorbent cotton', 'the vanity, the small wooden stool with a light brown square top', 'the fluffy white bath towel made of absorbent cotton, the small wooden stool with a light brown square top'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0610_01/01060.jpg", "scene0610_01/01100.jpg", "scene0610_01/01140.jpg", "scene0610_01/01180.jpg", "scene0610_01/01220.jpg" ], "turn_id": 5, "scan_id": "scene0610_01", "origin_question": "What will be the first-time appearance order of the following categories: curtain, faucet, switch?", "option": [ "faucet, curtain, switch", "curtain, switch, faucet", "curtain, faucet, switch", "switch, curtain, faucet" ], "answer": "curtain, faucet, switch", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: curtain, faucet, switch? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['faucet, curtain, switch', 'curtain, switch, faucet', 'curtain, faucet, switch', 'switch, curtain, faucet'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0610_01/01260.jpg", "scene0610_01/01300.jpg", "scene0610_01/01340.jpg", "scene0610_01/01380.jpg", "scene0610_01/01420.jpg" ], "turn_id": 6, "scan_id": "scene0610_01", "origin_question": "Please recall \"the vanity\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.08", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the vanity\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0610_01/01460.jpg", "scene0610_01/01500.jpg", "scene0610_01/01540.jpg", "scene0610_01/01580.jpg", "scene0610_01/01620.jpg" ], "turn_id": 7, "scan_id": "scene0610_01", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0610_01/01660.jpg", "scene0610_01/01700.jpg", "scene0610_01/01740.jpg", "scene0610_01/01780.jpg", "scene0610_01/01820.jpg" ], "turn_id": 8, "scan_id": "scene0610_01", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the white rectangular ceramic bathtub\"; \"the bin\"; \"the mirror\" ", "option": [ "the white rectangular ceramic bathtub", "the bin", "the mirror" ], "answer": "the mirror", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the white rectangular ceramic bathtub\"; \"the bin\"; \"the mirror\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the white rectangular ceramic bathtub', 'the bin', 'the mirror'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0610_01/01860.jpg", "scene0610_01/01910.jpg", "scene0610_01/01950.jpg", "scene0610_01/01990.jpg", "scene0610_01/02030.jpg" ], "turn_id": 9, "scan_id": "scene0610_01", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the faucet\" from your position?", "option": [], "answer": "118.79", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the faucet\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0610_01/02070.jpg", "scene0610_01/02110.jpg", "scene0610_01/02150.jpg", "scene0610_01/02190.jpg", "scene0610_01/02230.jpg" ], "turn_id": 10, "scan_id": "scene0610_01", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the small wooden stool with a light brown square top\" from your position?", "option": [], "answer": "85.6", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the small wooden stool with a light brown square top\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0044_region4/1dbb0638f04f486fa9226a986b243873_i2_4.jpg", "1mp3d_0044_region4/1dbb0638f04f486fa9226a986b243873_i1_5.jpg", "1mp3d_0044_region4/1dbb0638f04f486fa9226a986b243873_i2_5.jpg", "1mp3d_0044_region4/7372e123d5d94035a6d77c5f77bf0ff6_i1_5.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0044_region4", "origin_question": "Remember, have you seen any magazine(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any magazine(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 2 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0044_region4/7372e123d5d94035a6d77c5f77bf0ff6_i2_5.jpg", "1mp3d_0044_region4/1dbb0638f04f486fa9226a986b243873_i1_0.jpg", "1mp3d_0044_region4/1dbb0638f04f486fa9226a986b243873_i2_0.jpg", "1mp3d_0044_region4/7372e123d5d94035a6d77c5f77bf0ff6_i1_1.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0044_region4", "origin_question": "Remember, have you seen any door(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any door(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 2 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0441_00/00000.jpg", "scene0441_00/00030.jpg", "scene0441_00/00060.jpg", "scene0441_00/00090.jpg" ], "turn_id": 1, "scan_id": "scene0441_00", "origin_question": "Remember, have you seen any tissue box(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any tissue box(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0441_00/00120.jpg", "scene0441_00/00150.jpg", "scene0441_00/00180.jpg", "scene0441_00/00210.jpg" ], "turn_id": 2, "scan_id": "scene0441_00", "origin_question": "When did you first discover \"the standard white oval porcelain toilet\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the standard white oval porcelain toilet\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0441_00/00240.jpg", "scene0441_00/00270.jpg", "scene0441_00/00300.jpg", "scene0441_00/00330.jpg" ], "turn_id": 3, "scan_id": "scene0441_00", "origin_question": "Which direction is the tissue box in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the tissue box in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0441_00/00360.jpg", "scene0441_00/00390.jpg", "scene0441_00/00420.jpg", "scene0441_00/00450.jpg" ], "turn_id": 4, "scan_id": "scene0441_00", "origin_question": "Which two objects are on the same side of you now? \"the standard white oval porcelain toilet\", \"the shiny metallic stopcock with a classic shape\", \"the rectangular dark gray plastic bin\". ", "option": [ "the standard white oval porcelain toilet, the shiny metallic stopcock with a classic shape", "the standard white oval porcelain toilet, the rectangular dark gray plastic bin", "the shiny metallic stopcock with a classic shape, the rectangular dark gray plastic bin" ], "answer": "the standard white oval porcelain toilet, the shiny metallic stopcock with a classic shape", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the standard white oval porcelain toilet\", \"the shiny metallic stopcock with a classic shape\", \"the rectangular dark gray plastic bin\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the standard white oval porcelain toilet, the shiny metallic stopcock with a classic shape', 'the standard white oval porcelain toilet, the rectangular dark gray plastic bin', 'the shiny metallic stopcock with a classic shape, the rectangular dark gray plastic bin'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0441_00/00480.jpg", "scene0441_00/00510.jpg", "scene0441_00/00540.jpg", "scene0441_00/00570.jpg" ], "turn_id": 5, "scan_id": "scene0441_00", "origin_question": "What will be the first-time appearance order of the following categories: bin, sink, toilet?", "option": [ "toilet, sink, bin", "bin, toilet, sink", "toilet, bin, sink", "sink, bin, toilet" ], "answer": "toilet, bin, sink", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bin, sink, toilet? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['toilet, sink, bin', 'bin, toilet, sink', 'toilet, bin, sink', 'sink, bin, toilet'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0441_00/00600.jpg", "scene0441_00/00630.jpg", "scene0441_00/00660.jpg", "scene0441_00/00690.jpg" ], "turn_id": 6, "scan_id": "scene0441_00", "origin_question": "How far is your current position from where you were at the end of the turn 1? (in meters)", "option": [], "answer": "1.39", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 1? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0527_00/00000.jpg", "scene0527_00/00030.jpg", "scene0527_00/00060.jpg", "scene0527_00/00090.jpg" ], "turn_id": 1, "scan_id": "scene0527_00", "origin_question": "Remember, how many picture(s) have you seen so far? ", "option": [], "answer": "2", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many picture(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0527_00/00120.jpg", "scene0527_00/00150.jpg", "scene0527_00/00180.jpg", "scene0527_00/00210.jpg" ], "turn_id": 2, "scan_id": "scene0527_00", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the transparent cylindrical glass jar\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the transparent cylindrical glass jar\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0527_00/00240.jpg", "scene0527_00/00270.jpg", "scene0527_00/00300.jpg", "scene0527_00/00330.jpg" ], "turn_id": 3, "scan_id": "scene0527_00", "origin_question": "When did you first discover \"the white rectangular door with a smooth texture\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the white rectangular door with a smooth texture\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0527_00/00360.jpg", "scene0527_00/00390.jpg", "scene0527_00/00420.jpg", "scene0527_00/00450.jpg" ], "turn_id": 4, "scan_id": "scene0527_00", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0527_00/00480.jpg", "scene0527_00/00510.jpg", "scene0527_00/00540.jpg", "scene0527_00/00570.jpg" ], "turn_id": 5, "scan_id": "scene0527_00", "origin_question": "At the end of which turn were both of \"the coat hanger\" and \"the curtain\" on your left side?", "option": [], "answer": "5", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the coat hanger\" and \"the curtain\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0004_region0/255adf75751547fbb13b7f5d83c7166f_i1_5.jpg", "1mp3d_0004_region0/255adf75751547fbb13b7f5d83c7166f_i2_5.jpg", "1mp3d_0004_region0/7f6aa85262424e7da75f62063d397514_i1_5.jpg", "1mp3d_0004_region0/7f6aa85262424e7da75f62063d397514_i2_5.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0004_region0", "origin_question": "Remember, have you seen any lamp(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any lamp(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "1mp3d_0004_region0/9eedcaf6ec54441ba2454e247dd957db_i1_3.jpg", "1mp3d_0004_region0/9eedcaf6ec54441ba2454e247dd957db_i2_3.jpg", "1mp3d_0004_region0/7f6aa85262424e7da75f62063d397514_i1_0.jpg", "1mp3d_0004_region0/7f6aa85262424e7da75f62063d397514_i2_0.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0004_region0", "origin_question": "At the end of which turn were both of \"the soft gray plank roof with geometric patterns\" and \"the cylindrical lamp with a white textured lampshade\" on your right side?", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the soft gray plank roof with geometric patterns\" and \"the cylindrical lamp with a white textured lampshade\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "1mp3d_0004_region0/255adf75751547fbb13b7f5d83c7166f_i1_2.jpg", "1mp3d_0004_region0/255adf75751547fbb13b7f5d83c7166f_i2_2.jpg", "1mp3d_0004_region0/7aaefefd70ef448eb38651869d819d0e_i1_4.jpg", "1mp3d_0004_region0/7aaefefd70ef448eb38651869d819d0e_i2_4.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0004_region0", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0004_region0/fcc68e924e754661a3fa714421b9810c_i1_5.jpg", "1mp3d_0004_region0/fcc68e924e754661a3fa714421b9810c_i2_5.jpg", "1mp3d_0004_region0/fcc68e924e754661a3fa714421b9810c_i1_4.jpg", "1mp3d_0004_region0/fcc68e924e754661a3fa714421b9810c_i2_4.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0004_region0", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the white rectangular wooden shelf with a minimalist design\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the white rectangular wooden shelf with a minimalist design\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "1mp3d_0004_region0/fcc68e924e754661a3fa714421b9810c_i1_3.jpg", "1mp3d_0004_region0/fcc68e924e754661a3fa714421b9810c_i2_3.jpg", "1mp3d_0004_region0/fcc68e924e754661a3fa714421b9810c_i1_2.jpg", "1mp3d_0004_region0/fcc68e924e754661a3fa714421b9810c_i2_2.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0004_region0", "origin_question": "How far is your current position from where you were at the end of the turn 1? (in meters)", "option": [], "answer": "3.51", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 1? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "1mp3d_0004_region0/95d793d922e54671a4debca5c926c860_i1_2.jpg", "1mp3d_0004_region0/95d793d922e54671a4debca5c926c860_i2_2.jpg", "1mp3d_0004_region0/a89388278b38480a8e2a9cdedbbcb316_i1_0.jpg", "1mp3d_0004_region0/a89388278b38480a8e2a9cdedbbcb316_i2_0.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0004_region0", "origin_question": "Among these three objects(\"the white rectangular wooden shelf with a minimalist design\",\"the rectangular white wooden door with a minimalist design\",\"the thick, beige knitted blanket made of soft material\" ), which one is the closest to you?", "option": [ "the white rectangular wooden shelf with a minimalist design", "the rectangular white wooden door with a minimalist design", "the thick, beige knitted blanket made of soft material" ], "answer": "the thick, beige knitted blanket made of soft material", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects(\"the white rectangular wooden shelf with a minimalist design\",\"the rectangular white wooden door with a minimalist design\",\"the thick, beige knitted blanket made of soft material\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the white rectangular wooden shelf with a minimalist design', 'the rectangular white wooden door with a minimalist design', 'the thick, beige knitted blanket made of soft material'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "1mp3d_0004_region0/55b0e784538f4a80a9750ff8b1f3b4c6_i1_0.jpg", "1mp3d_0004_region0/55b0e784538f4a80a9750ff8b1f3b4c6_i2_0.jpg", "1mp3d_0004_region0/55b0e784538f4a80a9750ff8b1f3b4c6_i1_5.jpg", "1mp3d_0004_region0/55b0e784538f4a80a9750ff8b1f3b4c6_i2_5.jpg" ], "turn_id": 7, "scan_id": "1mp3d_0004_region0", "origin_question": "Assuming the direction you are facing at the end of the turn 5 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Right", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 5 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "1mp3d_0004_region0/a89388278b38480a8e2a9cdedbbcb316_i1_4.jpg", "1mp3d_0004_region0/a89388278b38480a8e2a9cdedbbcb316_i2_4.jpg", "1mp3d_0004_region0/a89388278b38480a8e2a9cdedbbcb316_i1_5.jpg", "1mp3d_0004_region0/a89388278b38480a8e2a9cdedbbcb316_i2_5.jpg" ], "turn_id": 8, "scan_id": "1mp3d_0004_region0", "origin_question": "In which turn were you closest to \"the rectangular white wooden door with a minimalist design\"? ", "option": [], "answer": "3", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the rectangular white wooden door with a minimalist design\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "1mp3d_0004_region0/95d793d922e54671a4debca5c926c860_i1_5.jpg", "1mp3d_0004_region0/95d793d922e54671a4debca5c926c860_i2_5.jpg", "1mp3d_0004_region0/95d793d922e54671a4debca5c926c860_i1_3.jpg", "1mp3d_0004_region0/95d793d922e54671a4debca5c926c860_i2_3.jpg" ], "turn_id": 9, "scan_id": "1mp3d_0004_region0", "origin_question": "Which two objects are on the same side of you now? \"the large rectangular wooden bed with a white bedsheet\", \"the rectangular white wooden door with a minimalist design\", \"the creamy yellow, one-person couch with a curved backrest\". ", "option": [ "the large rectangular wooden bed with a white bedsheet, the rectangular white wooden door with a minimalist design", "the large rectangular wooden bed with a white bedsheet, the creamy yellow, one-person couch with a curved backrest", "the rectangular white wooden door with a minimalist design, the creamy yellow, one-person couch with a curved backrest" ], "answer": "the large rectangular wooden bed with a white bedsheet, the rectangular white wooden door with a minimalist design", "user_message": "For the 9 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the large rectangular wooden bed with a white bedsheet\", \"the rectangular white wooden door with a minimalist design\", \"the creamy yellow, one-person couch with a curved backrest\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the large rectangular wooden bed with a white bedsheet, the rectangular white wooden door with a minimalist design', 'the large rectangular wooden bed with a white bedsheet, the creamy yellow, one-person couch with a curved backrest', 'the rectangular white wooden door with a minimalist design, the creamy yellow, one-person couch with a curved backrest'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0044_region13/19040a53d9c6459da092ab8702a6f421_i1_2.jpg", "1mp3d_0044_region13/19040a53d9c6459da092ab8702a6f421_i2_2.jpg", "1mp3d_0044_region13/19040a53d9c6459da092ab8702a6f421_i1_1.jpg", "1mp3d_0044_region13/19040a53d9c6459da092ab8702a6f421_i2_1.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0044_region13", "origin_question": "Remember, have you seen any arch(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any arch(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 2 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "1mp3d_0044_region13/72e6cca2e95c43ee988d422ff7247844_i1_1.jpg", "1mp3d_0044_region13/19040a53d9c6459da092ab8702a6f421_i1_3.jpg", "1mp3d_0044_region13/19040a53d9c6459da092ab8702a6f421_i2_3.jpg", "1mp3d_0044_region13/72e6cca2e95c43ee988d422ff7247844_i1_2.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0044_region13", "origin_question": "When did you first discover \"the step\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the step\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 2 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0229_01/00000.jpg", "scene0229_01/00040.jpg", "scene0229_01/00080.jpg", "scene0229_01/00120.jpg" ], "turn_id": 1, "scan_id": "scene0229_01", "origin_question": "Remember, how many helmets(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many helmets(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0229_01/00160.jpg", "scene0229_01/00200.jpg", "scene0229_01/00240.jpg", "scene0229_01/00280.jpg" ], "turn_id": 2, "scan_id": "scene0229_01", "origin_question": "When did you first discover \"the stack of varied-sized papers in white and yellow\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the stack of varied-sized papers in white and yellow\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0229_01/00320.jpg", "scene0229_01/00360.jpg", "scene0229_01/00400.jpg", "scene0229_01/00440.jpg" ], "turn_id": 3, "scan_id": "scene0229_01", "origin_question": "Which direction is the large yellow sun-shaped light fixture with rays in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the large yellow sun-shaped light fixture with rays in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0229_01/00480.jpg", "scene0229_01/00520.jpg", "scene0229_01/00560.jpg", "scene0229_01/00600.jpg" ], "turn_id": 4, "scan_id": "scene0229_01", "origin_question": "By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the large rectangular gray board\", Y: \"the mirror\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "farther away from both X and Y", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the large rectangular gray board\", Y: \"the mirror\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0229_01/00640.jpg", "scene0229_01/00680.jpg", "scene0229_01/00720.jpg", "scene0229_01/00760.jpg" ], "turn_id": 5, "scan_id": "scene0229_01", "origin_question": "In which turn were you closest to \"the large yellow sun-shaped light fixture with rays\"? ", "option": [], "answer": "4", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the large yellow sun-shaped light fixture with rays\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0229_01/00800.jpg", "scene0229_01/00840.jpg", "scene0229_01/00880.jpg", "scene0229_01/00920.jpg" ], "turn_id": 6, "scan_id": "scene0229_01", "origin_question": "Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the mirror\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the mirror\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0628_02/00000.jpg", "scene0628_02/00040.jpg", "scene0628_02/00080.jpg", "scene0628_02/00120.jpg" ], "turn_id": 1, "scan_id": "scene0628_02", "origin_question": "Remember, have you seen any paper(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any paper(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0628_02/00160.jpg", "scene0628_02/00200.jpg", "scene0628_02/00240.jpg", "scene0628_02/00280.jpg" ], "turn_id": 2, "scan_id": "scene0628_02", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the medium-sized, black nylon backpack\"; \"the soft, long-sleeved deep blue jacket\"; \"the window\" ", "option": [ "the medium-sized, black nylon backpack", "the soft, long-sleeved deep blue jacket", "the window" ], "answer": "the soft, long-sleeved deep blue jacket", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the medium-sized, black nylon backpack\"; \"the soft, long-sleeved deep blue jacket\"; \"the window\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the medium-sized, black nylon backpack', 'the soft, long-sleeved deep blue jacket', 'the window'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0628_02/00320.jpg", "scene0628_02/00360.jpg", "scene0628_02/00400.jpg", "scene0628_02/00440.jpg" ], "turn_id": 3, "scan_id": "scene0628_02", "origin_question": "What will be the first-time appearance order of the following categories: blackboard, desk, laptop?", "option": [ "desk, laptop, blackboard", "blackboard, desk, laptop", "laptop, desk, blackboard", "laptop, blackboard, desk" ], "answer": "desk, laptop, blackboard", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: blackboard, desk, laptop? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['desk, laptop, blackboard', 'blackboard, desk, laptop', 'laptop, desk, blackboard', 'laptop, blackboard, desk'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0628_02/00480.jpg", "scene0628_02/00520.jpg", "scene0628_02/00560.jpg", "scene0628_02/00600.jpg" ], "turn_id": 4, "scan_id": "scene0628_02", "origin_question": "What will be the first-time appearance order of the following categories: blackboard, laptop, paper?", "option": [ "paper, laptop, blackboard", "laptop, blackboard, paper", "laptop, paper, blackboard", "blackboard, laptop, paper" ], "answer": "paper, laptop, blackboard", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: blackboard, laptop, paper? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['paper, laptop, blackboard', 'laptop, blackboard, paper', 'laptop, paper, blackboard', 'blackboard, laptop, paper'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0628_02/00640.jpg", "scene0628_02/00680.jpg", "scene0628_02/00720.jpg", "scene0628_02/00760.jpg" ], "turn_id": 5, "scan_id": "scene0628_02", "origin_question": "In which turn did you see both \"the medium-sized, black nylon backpack\" and \"the soft, long-sleeved deep blue jacket\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the medium-sized, black nylon backpack\" and \"the soft, long-sleeved deep blue jacket\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0628_02/00800.jpg", "scene0628_02/00840.jpg", "scene0628_02/00880.jpg", "scene0628_02/00920.jpg" ], "turn_id": 6, "scan_id": "scene0628_02", "origin_question": "Which direction is the black laptop with a stylish modern appearance in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the black laptop with a stylish modern appearance in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0361_02/00000.jpg", "scene0361_02/00030.jpg", "scene0361_02/00060.jpg", "scene0361_02/00090.jpg" ], "turn_id": 1, "scan_id": "scene0361_02", "origin_question": "Remember, have you seen any fire extinguisher(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any fire extinguisher(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0361_02/00120.jpg", "scene0361_02/00150.jpg", "scene0361_02/00180.jpg", "scene0361_02/00210.jpg" ], "turn_id": 2, "scan_id": "scene0361_02", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "scene0361_02/00240.jpg", "scene0361_02/00270.jpg", "scene0361_02/00300.jpg", "scene0361_02/00330.jpg" ], "turn_id": 3, "scan_id": "scene0361_02", "origin_question": "Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance forward or backward from that position?", "option": [ "Forward", "Backward" ], "answer": "Forward", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance forward or backward from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Forward', 'Backward'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0361_02/00360.jpg", "scene0361_02/00390.jpg", "scene0361_02/00420.jpg", "scene0361_02/00450.jpg" ], "turn_id": 4, "scan_id": "scene0361_02", "origin_question": "", "option": [], "answer": "", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0361_02/00480.jpg", "scene0361_02/00510.jpg", "scene0361_02/00540.jpg", "scene0361_02/00570.jpg" ], "turn_id": 5, "scan_id": "scene0361_02", "origin_question": "In which turn were you closest to \"the small rectangular light brown wooden table\"? ", "option": [], "answer": "1", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the small rectangular light brown wooden table\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0361_02/00600.jpg", "scene0361_02/00630.jpg", "scene0361_02/00660.jpg", "scene0361_02/00690.jpg" ], "turn_id": 6, "scan_id": "scene0361_02", "origin_question": "How far is your current position from where you were at the end of the turn 4? (in meters)", "option": [], "answer": "3.17", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 4? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "scene0361_02/00720.jpg", "scene0361_02/00750.jpg", "scene0361_02/00780.jpg", "scene0361_02/00810.jpg" ], "turn_id": 7, "scan_id": "scene0361_02", "origin_question": "Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance forward or backward from that position?", "option": [ "Forward", "Backward" ], "answer": "Backward", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance forward or backward from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Forward', 'Backward'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0040_region21/f7c22572c3c9494db9473fe978992177_i1_2.jpg", "1mp3d_0040_region21/f7c22572c3c9494db9473fe978992177_i2_2.jpg", "1mp3d_0040_region21/6309f0ce3b9b4488bf9aa2dde1da9f3f_i1_2.jpg", "1mp3d_0040_region21/6309f0ce3b9b4488bf9aa2dde1da9f3f_i2_2.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0040_region21", "origin_question": "", "option": [], "answer": "", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0040_region21/4779786357a34dfe938caf804044cac7_i1_2.jpg", "1mp3d_0040_region21/4779786357a34dfe938caf804044cac7_i2_2.jpg", "1mp3d_0040_region21/b2c816b0e9a84252a6411c6cfd4996e5_i1_0.jpg", "1mp3d_0040_region21/b2c816b0e9a84252a6411c6cfd4996e5_i2_0.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0040_region21", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "1mp3d_0040_region21/4779786357a34dfe938caf804044cac7_i1_5.jpg", "1mp3d_0040_region21/4779786357a34dfe938caf804044cac7_i2_5.jpg", "1mp3d_0040_region21/b2c816b0e9a84252a6411c6cfd4996e5_i2_5.jpg", "1mp3d_0040_region21/4779786357a34dfe938caf804044cac7_i1_0.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0040_region21", "origin_question": "How far is your current position from where you were at the end of the turn 1? (in meters)", "option": [], "answer": "1.94", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 1? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0040_region21/4779786357a34dfe938caf804044cac7_i2_0.jpg", "1mp3d_0040_region21/b2c816b0e9a84252a6411c6cfd4996e5_i1_4.jpg", "1mp3d_0040_region21/b2c816b0e9a84252a6411c6cfd4996e5_i2_4.jpg", "1mp3d_0040_region21/b2c816b0e9a84252a6411c6cfd4996e5_i1_3.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0040_region21", "origin_question": "Remember, have you seen any treadmill(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any treadmill(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0050_01/00000.jpg", "scene0050_01/00040.jpg", "scene0050_01/00080.jpg", "scene0050_01/00120.jpg", "scene0050_01/00160.jpg" ], "turn_id": 1, "scan_id": "scene0050_01", "origin_question": "Remember, have you seen any bed(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any bed(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0050_01/00200.jpg", "scene0050_01/00240.jpg", "scene0050_01/00280.jpg", "scene0050_01/00320.jpg", "scene0050_01/00360.jpg" ], "turn_id": 2, "scan_id": "scene0050_01", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0050_01/00400.jpg", "scene0050_01/00440.jpg", "scene0050_01/00480.jpg", "scene0050_01/00520.jpg", "scene0050_01/00560.jpg" ], "turn_id": 3, "scan_id": "scene0050_01", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the rectangular, deep brown wooden chair\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the rectangular, deep brown wooden chair\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "scene0050_01/00600.jpg", "scene0050_01/00640.jpg", "scene0050_01/00680.jpg", "scene0050_01/00720.jpg", "scene0050_01/00760.jpg" ], "turn_id": 4, "scan_id": "scene0050_01", "origin_question": "Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Right", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0050_01/00800.jpg", "scene0050_01/00840.jpg", "scene0050_01/00880.jpg", "scene0050_01/00920.jpg", "scene0050_01/00960.jpg" ], "turn_id": 5, "scan_id": "scene0050_01", "origin_question": "Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0050_01/01000.jpg", "scene0050_01/01040.jpg", "scene0050_01/01080.jpg", "scene0050_01/01120.jpg", "scene0050_01/01160.jpg" ], "turn_id": 6, "scan_id": "scene0050_01", "origin_question": "Which two objects are on the same side of you now? \"the rectangular light blue tissue box with patterns\", \"the medium-sized hardcover book with a brightly colored cover\", \"the round analog clock with a dark-colored face\". ", "option": [ "the rectangular light blue tissue box with patterns, the medium-sized hardcover book with a brightly colored cover", "the rectangular light blue tissue box with patterns, the round analog clock with a dark-colored face", "the medium-sized hardcover book with a brightly colored cover, the round analog clock with a dark-colored face" ], "answer": "the medium-sized hardcover book with a brightly colored cover, the round analog clock with a dark-colored face", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the rectangular light blue tissue box with patterns\", \"the medium-sized hardcover book with a brightly colored cover\", \"the round analog clock with a dark-colored face\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular light blue tissue box with patterns, the medium-sized hardcover book with a brightly colored cover', 'the rectangular light blue tissue box with patterns, the round analog clock with a dark-colored face', 'the medium-sized hardcover book with a brightly colored cover, the round analog clock with a dark-colored face'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0050_01/01200.jpg", "scene0050_01/01240.jpg", "scene0050_01/01280.jpg", "scene0050_01/01320.jpg", "scene0050_01/01360.jpg" ], "turn_id": 7, "scan_id": "scene0050_01", "origin_question": "By the end of the 6 turn, were you getting closer to or farther away from these two objects: X: \"the medium-sized rectangular brown leather ottoman\", Y: \"the rectangular light blue tissue box with patterns\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 6 turn, were you getting closer to or farther away from these two objects: X: \"the medium-sized rectangular brown leather ottoman\", Y: \"the rectangular light blue tissue box with patterns\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0050_01/01400.jpg", "scene0050_01/01440.jpg", "scene0050_01/01480.jpg", "scene0050_01/01520.jpg", "scene0050_01/01560.jpg" ], "turn_id": 8, "scan_id": "scene0050_01", "origin_question": "Please recall \"the rectangular white wall-mounted wooden shelf\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.9", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the rectangular white wall-mounted wooden shelf\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0050_01/01600.jpg", "scene0050_01/01640.jpg", "scene0050_01/01680.jpg", "scene0050_01/01720.jpg", "scene0050_01/01760.jpg" ], "turn_id": 9, "scan_id": "scene0050_01", "origin_question": "", "option": [], "answer": "", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0050_01/01800.jpg", "scene0050_01/01840.jpg", "scene0050_01/01880.jpg", "scene0050_01/01920.jpg", "scene0050_01/01960.jpg" ], "turn_id": 10, "scan_id": "scene0050_01", "origin_question": "In which turn were you closest to \"the medium-sized rectangular wooden desk with a light brown surface\"? ", "option": [], "answer": "9", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the medium-sized rectangular wooden desk with a light brown surface\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0029_region7/450b5325b5bf48f0a8076bb973b42fdf_i1_5.jpg", "1mp3d_0029_region7/450b5325b5bf48f0a8076bb973b42fdf_i2_5.jpg", "1mp3d_0029_region7/152b7ad4b471410a8557ce5fc2c9147a_i1_3.jpg", "1mp3d_0029_region7/152b7ad4b471410a8557ce5fc2c9147a_i2_3.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0029_region7", "origin_question": "Remember, have you seen any light(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any light(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 3 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0029_region7/c50fc96abd854fd2a6766d52dfb4f355_i1_1.jpg", "1mp3d_0029_region7/c50fc96abd854fd2a6766d52dfb4f355_i2_1.jpg", "1mp3d_0029_region7/c50fc96abd854fd2a6766d52dfb4f355_i1_0.jpg", "1mp3d_0029_region7/c50fc96abd854fd2a6766d52dfb4f355_i2_0.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0029_region7", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 3 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0029_region7/152b7ad4b471410a8557ce5fc2c9147a_i1_1.jpg", "1mp3d_0029_region7/152b7ad4b471410a8557ce5fc2c9147a_i2_1.jpg", "1mp3d_0029_region7/450b5325b5bf48f0a8076bb973b42fdf_i2_2.jpg", "1mp3d_0029_region7/450b5325b5bf48f0a8076bb973b42fdf_i1_1.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0029_region7", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the door\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the door\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 3 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0651_00/00000.jpg", "scene0651_00/00040.jpg", "scene0651_00/00080.jpg", "scene0651_00/00120.jpg" ], "turn_id": 1, "scan_id": "scene0651_00", "origin_question": "Remember, have you seen any counter(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any counter(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0651_00/00160.jpg", "scene0651_00/00200.jpg", "scene0651_00/00240.jpg", "scene0651_00/00280.jpg" ], "turn_id": 2, "scan_id": "scene0651_00", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the simple white rectangular range hood made of smooth material\"; \"the rectangular stainless steel sink with a single faucet\"; \"the rectangular white electric stove with six burners\" ", "option": [ "the simple white rectangular range hood made of smooth material", "the rectangular stainless steel sink with a single faucet", "the rectangular white electric stove with six burners" ], "answer": "the rectangular stainless steel sink with a single faucet", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the simple white rectangular range hood made of smooth material\"; \"the rectangular stainless steel sink with a single faucet\"; \"the rectangular white electric stove with six burners\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the simple white rectangular range hood made of smooth material', 'the rectangular stainless steel sink with a single faucet', 'the rectangular white electric stove with six burners'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0651_00/00320.jpg", "scene0651_00/00360.jpg", "scene0651_00/00400.jpg", "scene0651_00/00440.jpg" ], "turn_id": 3, "scan_id": "scene0651_00", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0651_00/00480.jpg", "scene0651_00/00520.jpg", "scene0651_00/00560.jpg", "scene0651_00/00600.jpg" ], "turn_id": 4, "scan_id": "scene0651_00", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the rectangular white electric stove with six burners\" from your position?", "option": [], "answer": "173.78", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the rectangular white electric stove with six burners\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0651_00/00640.jpg", "scene0651_00/00680.jpg", "scene0651_00/00720.jpg", "scene0651_00/00760.jpg" ], "turn_id": 5, "scan_id": "scene0651_00", "origin_question": "By the end of the 3 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular stainless steel sink with a single faucet\", Y: \"the rectangular light-colored wooden table\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to Y and farther away from X", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 3 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular stainless steel sink with a single faucet\", Y: \"the rectangular light-colored wooden table\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0651_00/00800.jpg", "scene0651_00/00840.jpg", "scene0651_00/00880.jpg", "scene0651_00/00920.jpg" ], "turn_id": 6, "scan_id": "scene0651_00", "origin_question": "Which two objects are on the same side of you now? \"the counter\", \"the bin\", \"the rectangular light-colored wooden table\". ", "option": [ "the counter, the bin", "the counter, the rectangular light-colored wooden table", "the bin, the rectangular light-colored wooden table" ], "answer": "the bin, the rectangular light-colored wooden table", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the counter\", \"the bin\", \"the rectangular light-colored wooden table\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the counter, the bin', 'the counter, the rectangular light-colored wooden table', 'the bin, the rectangular light-colored wooden table'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0118_02/00000.jpg", "scene0118_02/00030.jpg", "scene0118_02/00060.jpg", "scene0118_02/00090.jpg" ], "turn_id": 1, "scan_id": "scene0118_02", "origin_question": "Remember, have you seen any table(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any table(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0118_02/00120.jpg", "scene0118_02/00150.jpg", "scene0118_02/00180.jpg", "scene0118_02/00210.jpg" ], "turn_id": 2, "scan_id": "scene0118_02", "origin_question": "In which turn did you see both \"the black medium-sized wool-polyester jacket\" and \"the small red hardcover book with a black spine\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the black medium-sized wool-polyester jacket\" and \"the small red hardcover book with a black spine\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0118_02/00240.jpg", "scene0118_02/00270.jpg", "scene0118_02/00300.jpg", "scene0118_02/00330.jpg" ], "turn_id": 3, "scan_id": "scene0118_02", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the black medium-sized wool-polyester jacket\" from your position?", "option": [], "answer": "100.5", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the black medium-sized wool-polyester jacket\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0118_02/00360.jpg", "scene0118_02/00390.jpg", "scene0118_02/00420.jpg", "scene0118_02/00450.jpg" ], "turn_id": 4, "scan_id": "scene0118_02", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the small red hardcover book with a black spine\", Y: \"the white metallic cabinet with a simple design\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to X and farther away from Y", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the small red hardcover book with a black spine\", Y: \"the white metallic cabinet with a simple design\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0645_00/00000.jpg", "scene0645_00/00040.jpg", "scene0645_00/00080.jpg", "scene0645_00/00120.jpg", "scene0645_00/00160.jpg" ], "turn_id": 1, "scan_id": "scene0645_00", "origin_question": "Remember, have you seen any cabinet(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any cabinet(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0645_00/00200.jpg", "scene0645_00/00240.jpg", "scene0645_00/00280.jpg", "scene0645_00/00320.jpg", "scene0645_00/00360.jpg" ], "turn_id": 2, "scan_id": "scene0645_00", "origin_question": "Please recall \"the white triangular semi-transparent plastic bag\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.71", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the white triangular semi-transparent plastic bag\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0645_00/00400.jpg", "scene0645_00/00440.jpg", "scene0645_00/00480.jpg", "scene0645_00/00520.jpg", "scene0645_00/00560.jpg" ], "turn_id": 3, "scan_id": "scene0645_00", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the sleek black rectangular remote control with buttons\", Y: \"the rectangular black glossy album with colorful design\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the sleek black rectangular remote control with buttons\", Y: \"the rectangular black glossy album with colorful design\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "scene0645_00/00600.jpg", "scene0645_00/00640.jpg", "scene0645_00/00680.jpg", "scene0645_00/00720.jpg", "scene0645_00/00760.jpg" ], "turn_id": 4, "scan_id": "scene0645_00", "origin_question": "Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Right", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0645_00/00800.jpg", "scene0645_00/00840.jpg", "scene0645_00/00880.jpg", "scene0645_00/00920.jpg", "scene0645_00/00960.jpg" ], "turn_id": 5, "scan_id": "scene0645_00", "origin_question": "Which direction is the white triangular semi-transparent plastic bag in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the white triangular semi-transparent plastic bag in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0645_00/01000.jpg", "scene0645_00/01040.jpg", "scene0645_00/01080.jpg", "scene0645_00/01120.jpg", "scene0645_00/01160.jpg" ], "turn_id": 6, "scan_id": "scene0645_00", "origin_question": "At the end of which turn were both of \"the small dark brown wooden stand with a glass top\" and \"the light beige rectangular velvet ottoman\" on your right side?", "option": [], "answer": "1", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the small dark brown wooden stand with a glass top\" and \"the light beige rectangular velvet ottoman\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0645_00/01200.jpg", "scene0645_00/01240.jpg", "scene0645_00/01280.jpg", "scene0645_00/01320.jpg", "scene0645_00/01360.jpg" ], "turn_id": 7, "scan_id": "scene0645_00", "origin_question": "At the end of which turn were both of \"the rectangular black glossy album with colorful design\" and \"the light beige rectangular velvet ottoman\" on your right side?", "option": [], "answer": "1", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the rectangular black glossy album with colorful design\" and \"the light beige rectangular velvet ottoman\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0645_00/01400.jpg", "scene0645_00/01440.jpg", "scene0645_00/01480.jpg", "scene0645_00/01520.jpg", "scene0645_00/01560.jpg" ], "turn_id": 8, "scan_id": "scene0645_00", "origin_question": "What will be the first-time appearance order of the following categories: couch, remote control, table?", "option": [ "remote control, couch, table", "couch, table, remote control", "remote control, table, couch", "couch, remote control, table" ], "answer": "remote control, couch, table", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: couch, remote control, table? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['remote control, couch, table', 'couch, table, remote control', 'remote control, table, couch', 'couch, remote control, table'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "scene0645_00/01600.jpg", "scene0645_00/01640.jpg", "scene0645_00/01680.jpg", "scene0645_00/01720.jpg", "scene0645_00/01760.jpg" ], "turn_id": 9, "scan_id": "scene0645_00", "origin_question": "Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance forward or backward from that position?", "option": [ "Forward", "Backward" ], "answer": "Backward", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance forward or backward from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Forward', 'Backward'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "scene0645_00/01800.jpg", "scene0645_00/01840.jpg", "scene0645_00/01880.jpg", "scene0645_00/01920.jpg", "scene0645_00/01960.jpg" ], "turn_id": 10, "scan_id": "scene0645_00", "origin_question": "Assuming the direction you are facing at the end of the turn 7 is forward, did you move a certain distance forward or backward from that position?", "option": [ "Forward", "Backward" ], "answer": "Forward", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 7 is forward, did you move a certain distance forward or backward from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Forward', 'Backward'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0146_00/00000.jpg", "scene0146_00/00040.jpg", "scene0146_00/00080.jpg", "scene0146_00/00120.jpg" ], "turn_id": 1, "scan_id": "scene0146_00", "origin_question": "Remember, have you seen any toilet(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any toilet(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0146_00/00160.jpg", "scene0146_00/00200.jpg", "scene0146_00/00240.jpg", "scene0146_00/00280.jpg" ], "turn_id": 2, "scan_id": "scene0146_00", "origin_question": "When did you first discover \"the toilet paper\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the toilet paper\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0146_00/00320.jpg", "scene0146_00/00360.jpg", "scene0146_00/00400.jpg", "scene0146_00/00440.jpg" ], "turn_id": 3, "scan_id": "scene0146_00", "origin_question": "Is the dispenser to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the dispenser to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0146_00/00480.jpg", "scene0146_00/00520.jpg", "scene0146_00/00560.jpg", "scene0146_00/00600.jpg" ], "turn_id": 4, "scan_id": "scene0146_00", "origin_question": "Please recall \"the white ceramic toilet of standard size\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.73", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the white ceramic toilet of standard size\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0146_00/00640.jpg", "scene0146_00/00680.jpg", "scene0146_00/00720.jpg", "scene0146_00/00760.jpg" ], "turn_id": 5, "scan_id": "scene0146_00", "origin_question": "Which direction is the rectangular plastic soap dispenser in black and grey in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the rectangular plastic soap dispenser in black and grey in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0146_00/00800.jpg", "scene0146_00/00840.jpg", "scene0146_00/00880.jpg", "scene0146_00/00920.jpg" ], "turn_id": 6, "scan_id": "scene0146_00", "origin_question": "In which turn did you see both \"the white ceramic toilet of standard size\" and \"the small silver metal stopcock with two handles\" simultaneously? ", "option": [], "answer": "3", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the white ceramic toilet of standard size\" and \"the small silver metal stopcock with two handles\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0146_00/00960.jpg", "scene0146_00/01000.jpg", "scene0146_00/01040.jpg", "scene0146_00/01080.jpg" ], "turn_id": 7, "scan_id": "scene0146_00", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0146_00/01120.jpg", "scene0146_00/01160.jpg", "scene0146_00/01200.jpg", "scene0146_00/01240.jpg" ], "turn_id": 8, "scan_id": "scene0146_00", "origin_question": "How far is your current position from where you were at the end of the turn 6? (in meters)", "option": [], "answer": "1.09", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 6? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0040_region32/fc349ffd7240462fb54becf070eb9cd9_i2_4.jpg", "1mp3d_0040_region32/ec9e10633513410da7c5542504e971ee_i2_4.jpg", "1mp3d_0040_region32/fc349ffd7240462fb54becf070eb9cd9_i1_5.jpg", "1mp3d_0040_region32/fc349ffd7240462fb54becf070eb9cd9_i2_5.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0040_region32", "origin_question": "Remember, have you seen any stairs(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any stairs(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "1mp3d_0040_region32/987071a0c61b4b9a8d0a4fcd8aadeb23_i1_4.jpg", "1mp3d_0040_region32/987071a0c61b4b9a8d0a4fcd8aadeb23_i2_4.jpg", "1mp3d_0040_region32/e2fad68fa1a04b8095bee4d1d378597d_i2_3.jpg", "1mp3d_0040_region32/e2fad68fa1a04b8095bee4d1d378597d_i1_4.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0040_region32", "origin_question": "When did you first discover \"the large black metal banister for stairs\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the large black metal banister for stairs\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "1mp3d_0040_region32/e2fad68fa1a04b8095bee4d1d378597d_i2_4.jpg", "1mp3d_0040_region32/987071a0c61b4b9a8d0a4fcd8aadeb23_i1_5.jpg", "1mp3d_0040_region32/987071a0c61b4b9a8d0a4fcd8aadeb23_i2_5.jpg", "1mp3d_0040_region32/987071a0c61b4b9a8d0a4fcd8aadeb23_i1_0.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0040_region32", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "1mp3d_0040_region32/987071a0c61b4b9a8d0a4fcd8aadeb23_i2_0.jpg", "1mp3d_0040_region32/e2fad68fa1a04b8095bee4d1d378597d_i2_1.jpg", "1mp3d_0040_region32/e2fad68fa1a04b8095bee4d1d378597d_i1_0.jpg", "1mp3d_0040_region32/e2fad68fa1a04b8095bee4d1d378597d_i2_0.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0040_region32", "origin_question": "When did you first discover \"the large black metal banister for stairs\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the large black metal banister for stairs\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0040_region2/153536175bae4c5abbd1a20de40be77c_i1_3.jpg", "1mp3d_0040_region2/153536175bae4c5abbd1a20de40be77c_i2_3.jpg", "1mp3d_0040_region2/c0e26df4e85f42d4bb21a9f9ee9e628b_i2_5.jpg", "1mp3d_0040_region2/c0e26df4e85f42d4bb21a9f9ee9e628b_i1_0.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0040_region2", "origin_question": "Remember, have you seen any window(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any window(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "1mp3d_0040_region2/c0e26df4e85f42d4bb21a9f9ee9e628b_i2_0.jpg", "1mp3d_0040_region2/c0e26df4e85f42d4bb21a9f9ee9e628b_i1_1.jpg", "1mp3d_0040_region2/c0e26df4e85f42d4bb21a9f9ee9e628b_i2_1.jpg", "1mp3d_0040_region2/153536175bae4c5abbd1a20de40be77c_i1_5.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0040_region2", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the moderate-sized plant with colorful flowers\"; \"the hexagonal wooden table with a shiny surface\"; \"the medium, dark-colored lamp with a beige lampshade\" ", "option": [ "the moderate-sized plant with colorful flowers", "the hexagonal wooden table with a shiny surface", "the medium, dark-colored lamp with a beige lampshade" ], "answer": "the medium, dark-colored lamp with a beige lampshade", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the moderate-sized plant with colorful flowers\"; \"the hexagonal wooden table with a shiny surface\"; \"the medium, dark-colored lamp with a beige lampshade\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the moderate-sized plant with colorful flowers', 'the hexagonal wooden table with a shiny surface', 'the medium, dark-colored lamp with a beige lampshade'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "1mp3d_0040_region2/153536175bae4c5abbd1a20de40be77c_i2_5.jpg", "1mp3d_0040_region2/d7fe8f872086493d81b333ca1a867d0d_i1_3.jpg", "1mp3d_0040_region2/d7fe8f872086493d81b333ca1a867d0d_i2_3.jpg", "1mp3d_0040_region2/c782e1f5ad79494882c9b6715060236c_i1_3.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0040_region2", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "1mp3d_0040_region2/c782e1f5ad79494882c9b6715060236c_i2_3.jpg", "1mp3d_0040_region2/c782e1f5ad79494882c9b6715060236c_i1_4.jpg", "1mp3d_0040_region2/c782e1f5ad79494882c9b6715060236c_i2_4.jpg", "1mp3d_0040_region2/ea8a6c77ddc24b0bab54602bd79fa277_i1_1.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0040_region2", "origin_question": "Assuming the direction you are facing at the end of the turn 3 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Left", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 3 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "1mp3d_0040_region2/ea8a6c77ddc24b0bab54602bd79fa277_i2_1.jpg", "1mp3d_0040_region2/0ec77a25298c4935b585db95e91ed83f_i1_2.jpg", "1mp3d_0040_region2/0ec77a25298c4935b585db95e91ed83f_i2_2.jpg", "1mp3d_0040_region2/ea8a6c77ddc24b0bab54602bd79fa277_i1_2.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0040_region2", "origin_question": "At the end of which turn were both of \"the hexagonal wooden table with a shiny surface\" and \"the rectangular light beige fabric ottoman\" on your right side?", "option": [], "answer": "1", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the hexagonal wooden table with a shiny surface\" and \"the rectangular light beige fabric ottoman\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "1mp3d_0040_region2/ea8a6c77ddc24b0bab54602bd79fa277_i2_2.jpg", "1mp3d_0040_region2/ea8a6c77ddc24b0bab54602bd79fa277_i1_3.jpg", "1mp3d_0040_region2/ea8a6c77ddc24b0bab54602bd79fa277_i2_3.jpg", "1mp3d_0040_region2/0ec77a25298c4935b585db95e91ed83f_i1_3.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0040_region2", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the rectangular light beige fabric ottoman\"; \"the rectangular canvas picture with a red background\"; \"the medium-sized fan with a brownish body and wooden blades\" ", "option": [ "the rectangular light beige fabric ottoman", "the rectangular canvas picture with a red background", "the medium-sized fan with a brownish body and wooden blades" ], "answer": "the medium-sized fan with a brownish body and wooden blades", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the rectangular light beige fabric ottoman\"; \"the rectangular canvas picture with a red background\"; \"the medium-sized fan with a brownish body and wooden blades\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular light beige fabric ottoman', 'the rectangular canvas picture with a red background', 'the medium-sized fan with a brownish body and wooden blades'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "1mp3d_0040_region2/0ec77a25298c4935b585db95e91ed83f_i2_3.jpg", "1mp3d_0040_region2/61891832db1a494cb1e163fc4f23a693_i1_3.jpg", "1mp3d_0040_region2/61891832db1a494cb1e163fc4f23a693_i2_3.jpg", "1mp3d_0040_region2/f1c328c6178144edbf4c4c85758f56a9_i1_2.jpg" ], "turn_id": 7, "scan_id": "1mp3d_0040_region2", "origin_question": "Assuming the direction you are facing at the end of the turn 6 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Right", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 6 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0040_region2/f1c328c6178144edbf4c4c85758f56a9_i2_2.jpg", "1mp3d_0040_region2/4fcd68dd08c4492aa47ddb5c31d7e1bf_i1_3.jpg", "1mp3d_0040_region2/4fcd68dd08c4492aa47ddb5c31d7e1bf_i2_3.jpg", "1mp3d_0040_region2/4fcd68dd08c4492aa47ddb5c31d7e1bf_i1_2.jpg" ], "turn_id": 8, "scan_id": "1mp3d_0040_region2", "origin_question": "Compared to your position at the end of the 7 turn, are you now closer or farther away from \"the large, deep brown wooden headboard with carvings\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 7 turn, are you now closer or farther away from \"the large, deep brown wooden headboard with carvings\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0067_region0/529ed2a6b09e4f52b3e861e1b6a46d3a_i1_0.jpg", "1mp3d_0067_region0/529ed2a6b09e4f52b3e861e1b6a46d3a_i2_0.jpg", "1mp3d_0067_region0/9136737ab6024d03b151110c83a00610_i1_4.jpg", "1mp3d_0067_region0/9136737ab6024d03b151110c83a00610_i2_4.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0067_region0", "origin_question": "Remember, have you seen any door(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any door(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "1mp3d_0067_region0/14dadef1c9c74e2392ac8b1d04407fc7_i1_4.jpg", "1mp3d_0067_region0/14dadef1c9c74e2392ac8b1d04407fc7_i2_4.jpg", "1mp3d_0067_region0/14dadef1c9c74e2392ac8b1d04407fc7_i1_5.jpg", "1mp3d_0067_region0/14dadef1c9c74e2392ac8b1d04407fc7_i2_5.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0067_region0", "origin_question": "In which turn were you closest to \"the white rectangular metal radiator\"? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the white rectangular metal radiator\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0067_region0/9136737ab6024d03b151110c83a00610_i1_0.jpg", "1mp3d_0067_region0/9136737ab6024d03b151110c83a00610_i2_0.jpg", "1mp3d_0067_region0/529ed2a6b09e4f52b3e861e1b6a46d3a_i1_1.jpg", "1mp3d_0067_region0/529ed2a6b09e4f52b3e861e1b6a46d3a_i2_1.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0067_region0", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the chair\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the chair\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "1mp3d_0067_region0/529ed2a6b09e4f52b3e861e1b6a46d3a_i1_2.jpg", "1mp3d_0067_region0/529ed2a6b09e4f52b3e861e1b6a46d3a_i2_2.jpg", "1mp3d_0067_region0/828b8f02ffd145ac8b351cbb70682dc2_i1_4.jpg", "1mp3d_0067_region0/828b8f02ffd145ac8b351cbb70682dc2_i2_4.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0067_region0", "origin_question": "In which turn did you see both \"the chair\" and \"the large, round, white ball\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the chair\" and \"the large, round, white ball\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "1mp3d_0067_region0/d710e361173943d189ad82b18c11572d_i1_0.jpg", "1mp3d_0067_region0/d710e361173943d189ad82b18c11572d_i2_0.jpg", "1mp3d_0067_region0/828b8f02ffd145ac8b351cbb70682dc2_i1_5.jpg", "1mp3d_0067_region0/828b8f02ffd145ac8b351cbb70682dc2_i2_5.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0067_region0", "origin_question": "Is the rectangular window with a deep brown frame to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the rectangular window with a deep brown frame to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "1mp3d_0067_region0/828b8f02ffd145ac8b351cbb70682dc2_i1_0.jpg", "1mp3d_0067_region0/828b8f02ffd145ac8b351cbb70682dc2_i2_0.jpg", "1mp3d_0067_region0/828b8f02ffd145ac8b351cbb70682dc2_i1_1.jpg", "1mp3d_0067_region0/828b8f02ffd145ac8b351cbb70682dc2_i2_1.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0067_region0", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the large, round, white ball\" from your position?", "option": [], "answer": "140.4", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the large, round, white ball\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0660_00/00000.jpg", "scene0660_00/00040.jpg", "scene0660_00/00080.jpg", "scene0660_00/00120.jpg" ], "turn_id": 1, "scan_id": "scene0660_00", "origin_question": "Remember, have you seen any table(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any table(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0660_00/00160.jpg", "scene0660_00/00200.jpg", "scene0660_00/00240.jpg", "scene0660_00/00280.jpg" ], "turn_id": 2, "scan_id": "scene0660_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0660_00/00320.jpg", "scene0660_00/00360.jpg", "scene0660_00/00400.jpg", "scene0660_00/00440.jpg" ], "turn_id": 3, "scan_id": "scene0660_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0660_00/00480.jpg", "scene0660_00/00520.jpg", "scene0660_00/00560.jpg", "scene0660_00/00600.jpg" ], "turn_id": 4, "scan_id": "scene0660_00", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0660_00/00640.jpg", "scene0660_00/00680.jpg", "scene0660_00/00720.jpg", "scene0660_00/00760.jpg" ], "turn_id": 5, "scan_id": "scene0660_00", "origin_question": "Remember, have you seen any chair(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any chair(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0660_00/00800.jpg", "scene0660_00/00840.jpg", "scene0660_00/00880.jpg", "scene0660_00/00920.jpg" ], "turn_id": 6, "scan_id": "scene0660_00", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the large rectangular wooden table with a glossy surface\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the large rectangular wooden table with a glossy surface\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0466_00/00000.jpg", "scene0466_00/00030.jpg", "scene0466_00/00060.jpg", "scene0466_00/00090.jpg" ], "turn_id": 1, "scan_id": "scene0466_00", "origin_question": "Remember, have you seen any table(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any table(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0466_00/00120.jpg", "scene0466_00/00150.jpg", "scene0466_00/00180.jpg", "scene0466_00/00210.jpg" ], "turn_id": 2, "scan_id": "scene0466_00", "origin_question": "When did you first discover \"the small rectangular blue plastic recycle bin\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the small rectangular blue plastic recycle bin\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0466_00/00240.jpg", "scene0466_00/00270.jpg", "scene0466_00/00300.jpg", "scene0466_00/00330.jpg" ], "turn_id": 3, "scan_id": "scene0466_00", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "1.04", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0466_00/00360.jpg", "scene0466_00/00390.jpg", "scene0466_00/00420.jpg", "scene0466_00/00450.jpg" ], "turn_id": 4, "scan_id": "scene0466_00", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular light brown wooden table\", Y: \"the small rectangular blue plastic recycle bin\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "farther away from both X and Y", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular light brown wooden table\", Y: \"the small rectangular blue plastic recycle bin\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0466_00/00480.jpg", "scene0466_00/00510.jpg", "scene0466_00/00540.jpg", "scene0466_00/00570.jpg" ], "turn_id": 5, "scan_id": "scene0466_00", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "68.76", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0466_00/00600.jpg", "scene0466_00/00630.jpg", "scene0466_00/00660.jpg", "scene0466_00/00690.jpg" ], "turn_id": 6, "scan_id": "scene0466_00", "origin_question": "Using your orientation at the end of turn 4 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "138.07", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 4 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0516_00/00000.jpg", "scene0516_00/00040.jpg", "scene0516_00/00080.jpg", "scene0516_00/00120.jpg", "scene0516_00/00160.jpg" ], "turn_id": 1, "scan_id": "scene0516_00", "origin_question": "Remember, how many bag(s) have you seen so far? ", "option": [], "answer": "2", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many bag(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0516_00/00200.jpg", "scene0516_00/00240.jpg", "scene0516_00/00280.jpg", "scene0516_00/00320.jpg", "scene0516_00/00360.jpg" ], "turn_id": 2, "scan_id": "scene0516_00", "origin_question": "At the end of which turn were both of \"Wooden stairs with a brown hue and traditional design\" and \"the slender brown wooden rail\" on your right side?", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"Wooden stairs with a brown hue and traditional design\" and \"the slender brown wooden rail\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0516_00/00400.jpg", "scene0516_00/00440.jpg", "scene0516_00/00480.jpg", "scene0516_00/00520.jpg", "scene0516_00/00560.jpg" ], "turn_id": 3, "scan_id": "scene0516_00", "origin_question": "Which direction is the slender brown wooden rail in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the slender brown wooden rail in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0516_00/00600.jpg", "scene0516_00/00640.jpg", "scene0516_00/00680.jpg", "scene0516_00/00720.jpg", "scene0516_00/00760.jpg" ], "turn_id": 4, "scan_id": "scene0516_00", "origin_question": "In which turn did you see both \"Wooden stairs with a brown hue and traditional design\" and \"the hamper\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"Wooden stairs with a brown hue and traditional design\" and \"the hamper\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0516_00/00800.jpg", "scene0516_00/00840.jpg", "scene0516_00/00880.jpg", "scene0516_00/00920.jpg", "scene0516_00/00960.jpg" ], "turn_id": 5, "scan_id": "scene0516_00", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the red and black rectangular fabric wrap\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the red and black rectangular fabric wrap\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0516_00/01000.jpg", "scene0516_00/01040.jpg", "scene0516_00/01080.jpg", "scene0516_00/01120.jpg", "scene0516_00/01160.jpg" ], "turn_id": 6, "scan_id": "scene0516_00", "origin_question": "Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the hamper\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the hamper\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0516_00/01200.jpg", "scene0516_00/01240.jpg", "scene0516_00/01280.jpg", "scene0516_00/01320.jpg", "scene0516_00/01370.jpg" ], "turn_id": 7, "scan_id": "scene0516_00", "origin_question": "What will be the first-time appearance order of the following categories: picture, stairs, toilet?", "option": [ "picture, toilet, stairs", "stairs, picture, toilet", "toilet, stairs, picture", "stairs, toilet, picture" ], "answer": "stairs, toilet, picture", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: picture, stairs, toilet? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['picture, toilet, stairs', 'stairs, picture, toilet', 'toilet, stairs, picture', 'stairs, toilet, picture'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0516_00/01410.jpg", "scene0516_00/01500.jpg", "scene0516_00/01540.jpg", "scene0516_00/01580.jpg", "scene0516_00/01620.jpg" ], "turn_id": 8, "scan_id": "scene0516_00", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the slender brown wooden rail\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the slender brown wooden rail\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0516_00/01660.jpg", "scene0516_00/01700.jpg", "scene0516_00/01740.jpg", "scene0516_00/01780.jpg", "scene0516_00/01820.jpg" ], "turn_id": 9, "scan_id": "scene0516_00", "origin_question": "What will be the first-time appearance order of the following categories: mirror, picture, stairs?", "option": [ "picture, mirror, stairs", "stairs, picture, mirror", "stairs, mirror, picture", "picture, stairs, mirror" ], "answer": "stairs, picture, mirror", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: mirror, picture, stairs? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['picture, mirror, stairs', 'stairs, picture, mirror', 'stairs, mirror, picture', 'picture, stairs, mirror'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0516_00/01860.jpg", "scene0516_00/01900.jpg", "scene0516_00/01940.jpg", "scene0516_00/01980.jpg", "scene0516_00/02020.jpg" ], "turn_id": 10, "scan_id": "scene0516_00", "origin_question": "When did you first discover \"Wooden stairs with a brown hue and traditional design\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"Wooden stairs with a brown hue and traditional design\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0067_region13/f1c9fe4b49c443e891b6ef6b41ec25bc_i1_1.jpg", "1mp3d_0067_region13/f1c9fe4b49c443e891b6ef6b41ec25bc_i2_1.jpg", "1mp3d_0067_region13/52fbea82ae7248eaaf97b1c6fbecb024_i1_2.jpg", "1mp3d_0067_region13/52fbea82ae7248eaaf97b1c6fbecb024_i2_2.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0067_region13", "origin_question": "Remember, have you seen any seasoning(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any seasoning(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "1mp3d_0067_region13/a9f1f6d6519e423a99303e3c7fdb836a_i1_5.jpg", "1mp3d_0067_region13/a9f1f6d6519e423a99303e3c7fdb836a_i2_5.jpg", "1mp3d_0067_region13/91c1b800925b4c2fa0a4fe57705b8a83_i1_0.jpg", "1mp3d_0067_region13/91c1b800925b4c2fa0a4fe57705b8a83_i2_0.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0067_region13", "origin_question": "In which turn did you see both \"the rectangular, white box with a beige border\" and \"the white hat with a wide brim and black band\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the rectangular, white box with a beige border\" and \"the white hat with a wide brim and black band\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "1mp3d_0067_region13/02d5d7f4541048d8b530bb190c13b6a2_i1_4.jpg", "1mp3d_0067_region13/02d5d7f4541048d8b530bb190c13b6a2_i2_4.jpg", "1mp3d_0067_region13/02d5d7f4541048d8b530bb190c13b6a2_i1_3.jpg", "1mp3d_0067_region13/02d5d7f4541048d8b530bb190c13b6a2_i2_3.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0067_region13", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the curtain\"; \"the fireplace\"; \"the large rectangular dark brown wooden table\" ", "option": [ "the curtain", "the fireplace", "the large rectangular dark brown wooden table" ], "answer": "the fireplace", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the curtain\"; \"the fireplace\"; \"the large rectangular dark brown wooden table\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the curtain', 'the fireplace', 'the large rectangular dark brown wooden table'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "1mp3d_0067_region13/91c1b800925b4c2fa0a4fe57705b8a83_i1_4.jpg", "1mp3d_0067_region13/91c1b800925b4c2fa0a4fe57705b8a83_i2_4.jpg", "1mp3d_0067_region13/91c1b800925b4c2fa0a4fe57705b8a83_i1_5.jpg", "1mp3d_0067_region13/91c1b800925b4c2fa0a4fe57705b8a83_i2_5.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0067_region13", "origin_question": "Which two objects are on the same side of you now? \"the curtain\", \"the large rectangular dark brown wooden table\", \"the light brown rectangular wooden door\". ", "option": [ "the curtain, the large rectangular dark brown wooden table", "the curtain, the light brown rectangular wooden door", "the large rectangular dark brown wooden table, the light brown rectangular wooden door" ], "answer": "the large rectangular dark brown wooden table, the light brown rectangular wooden door", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the curtain\", \"the large rectangular dark brown wooden table\", \"the light brown rectangular wooden door\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the curtain, the large rectangular dark brown wooden table', 'the curtain, the light brown rectangular wooden door', 'the large rectangular dark brown wooden table, the light brown rectangular wooden door'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "1mp3d_0067_region13/a9f1f6d6519e423a99303e3c7fdb836a_i1_3.jpg", "1mp3d_0067_region13/a9f1f6d6519e423a99303e3c7fdb836a_i2_3.jpg", "1mp3d_0067_region13/a9f1f6d6519e423a99303e3c7fdb836a_i1_1.jpg", "1mp3d_0067_region13/a9f1f6d6519e423a99303e3c7fdb836a_i2_1.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0067_region13", "origin_question": "Among these three objects(\"the small, square, soft brown woven bin\",\"the silver, standard-sized laptop\",\"the white hat with a wide brim and black band\" ), which one is the farthest from you?", "option": [ "the small, square, soft brown woven bin", "the silver, standard-sized laptop", "the white hat with a wide brim and black band" ], "answer": "the white hat with a wide brim and black band", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects(\"the small, square, soft brown woven bin\",\"the silver, standard-sized laptop\",\"the white hat with a wide brim and black band\" ), which one is the farthest from you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the small, square, soft brown woven bin', 'the silver, standard-sized laptop', 'the white hat with a wide brim and black band'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0067_region13/a9f1f6d6519e423a99303e3c7fdb836a_i1_0.jpg", "1mp3d_0067_region13/a9f1f6d6519e423a99303e3c7fdb836a_i2_0.jpg", "1mp3d_0067_region13/52fbea82ae7248eaaf97b1c6fbecb024_i1_0.jpg", "1mp3d_0067_region13/52fbea82ae7248eaaf97b1c6fbecb024_i2_0.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0067_region13", "origin_question": "Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the silver, standard-sized laptop\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the silver, standard-sized laptop\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "1mp3d_0067_region13/f1c9fe4b49c443e891b6ef6b41ec25bc_i1_0.jpg", "1mp3d_0067_region13/f1c9fe4b49c443e891b6ef6b41ec25bc_i2_0.jpg", "1mp3d_0067_region13/f1c9fe4b49c443e891b6ef6b41ec25bc_i1_5.jpg", "1mp3d_0067_region13/f1c9fe4b49c443e891b6ef6b41ec25bc_i2_5.jpg" ], "turn_id": 7, "scan_id": "1mp3d_0067_region13", "origin_question": "By the end of the 3 turn, were you getting closer to or farther away from these two objects: X: \"the small, square, soft brown woven bin\", Y: \"the silver, standard-sized laptop\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to Y and farther away from X", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 3 turn, were you getting closer to or farther away from these two objects: X: \"the small, square, soft brown woven bin\", Y: \"the silver, standard-sized laptop\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0030_region5/0fa0a4dc5aea4285b225cda527388ecb_i1_0.jpg", "1mp3d_0030_region5/0fa0a4dc5aea4285b225cda527388ecb_i2_0.jpg", "1mp3d_0030_region5/949a6c58781d4e80bff19ca32493efcd_i1_5.jpg", "1mp3d_0030_region5/949a6c58781d4e80bff19ca32493efcd_i2_5.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0030_region5", "origin_question": "", "option": [], "answer": "", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 2 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0030_region5/949a6c58781d4e80bff19ca32493efcd_i1_0.jpg", "1mp3d_0030_region5/949a6c58781d4e80bff19ca32493efcd_i2_0.jpg", "1mp3d_0030_region5/f5cf1375de4d46b38328dda8ebcd6ce0_i1_1.jpg", "1mp3d_0030_region5/f5cf1375de4d46b38328dda8ebcd6ce0_i2_1.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0030_region5", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 2 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0039_01/00000.jpg", "scene0039_01/00030.jpg", "scene0039_01/00060.jpg", "scene0039_01/00090.jpg" ], "turn_id": 1, "scan_id": "scene0039_01", "origin_question": "Remember, have you seen any lamp(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any lamp(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0039_01/00120.jpg", "scene0039_01/00150.jpg", "scene0039_01/00180.jpg", "scene0039_01/00210.jpg" ], "turn_id": 2, "scan_id": "scene0039_01", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the minimalist black vertical shelf with multiple compartments\"; \"the rectangular desk with a dark glass top\"; \"the chair\" ", "option": [ "the minimalist black vertical shelf with multiple compartments", "the rectangular desk with a dark glass top", "the chair" ], "answer": "the minimalist black vertical shelf with multiple compartments", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the minimalist black vertical shelf with multiple compartments\"; \"the rectangular desk with a dark glass top\"; \"the chair\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the minimalist black vertical shelf with multiple compartments', 'the rectangular desk with a dark glass top', 'the chair'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0039_01/00240.jpg", "scene0039_01/00270.jpg", "scene0039_01/00300.jpg", "scene0039_01/00330.jpg" ], "turn_id": 3, "scan_id": "scene0039_01", "origin_question": "Which direction is the modern rectangular stand with a shiny black surface in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the modern rectangular stand with a shiny black surface in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0039_01/00360.jpg", "scene0039_01/00390.jpg", "scene0039_01/00420.jpg", "scene0039_01/00450.jpg" ], "turn_id": 4, "scan_id": "scene0039_01", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the minimalist black vertical shelf with multiple compartments\" from your position?", "option": [], "answer": "141.56", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the minimalist black vertical shelf with multiple compartments\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0039_01/00480.jpg", "scene0039_01/00510.jpg", "scene0039_01/00540.jpg", "scene0039_01/00570.jpg" ], "turn_id": 5, "scan_id": "scene0039_01", "origin_question": "Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0686_01/00000.jpg", "scene0686_01/00040.jpg", "scene0686_01/00080.jpg", "scene0686_01/00120.jpg" ], "turn_id": 1, "scan_id": "scene0686_01", "origin_question": "Remember, have you seen any sink(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any sink(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0686_01/00160.jpg", "scene0686_01/00200.jpg", "scene0686_01/00240.jpg", "scene0686_01/00280.jpg" ], "turn_id": 2, "scan_id": "scene0686_01", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0686_01/00320.jpg", "scene0686_01/00360.jpg", "scene0686_01/00400.jpg", "scene0686_01/00440.jpg" ], "turn_id": 3, "scan_id": "scene0686_01", "origin_question": "In which turn did you see both \"the white soap dispenser with a grey top\" and \"the white, wall-mounted oval toilet\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the white soap dispenser with a grey top\" and \"the white, wall-mounted oval toilet\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0686_01/00480.jpg", "scene0686_01/00520.jpg", "scene0686_01/00560.jpg", "scene0686_01/00600.jpg" ], "turn_id": 4, "scan_id": "scene0686_01", "origin_question": "", "option": [], "answer": "", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0686_01/00640.jpg", "scene0686_01/00680.jpg", "scene0686_01/00720.jpg", "scene0686_01/00760.jpg" ], "turn_id": 5, "scan_id": "scene0686_01", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0686_01/00800.jpg", "scene0686_01/00840.jpg", "scene0686_01/00880.jpg", "scene0686_01/00920.jpg" ], "turn_id": 6, "scan_id": "scene0686_01", "origin_question": "Which direction is the white rectangular wall-mounted sink with two faucets in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the white rectangular wall-mounted sink with two faucets in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0686_01/00960.jpg", "scene0686_01/01000.jpg", "scene0686_01/01040.jpg", "scene0686_01/01080.jpg" ], "turn_id": 7, "scan_id": "scene0686_01", "origin_question": "When did you first discover \"the rectangular stainless steel bin\"? (index of the turn)", "option": [], "answer": "7", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the rectangular stainless steel bin\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0686_01/01120.jpg", "scene0686_01/01160.jpg", "scene0686_01/01200.jpg", "scene0686_01/01240.jpg" ], "turn_id": 8, "scan_id": "scene0686_01", "origin_question": "Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the rectangular stainless steel bin\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the rectangular stainless steel bin\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0218_01/00000.jpg", "scene0218_01/00030.jpg", "scene0218_01/00060.jpg", "scene0218_01/00090.jpg" ], "turn_id": 1, "scan_id": "scene0218_01", "origin_question": "Remember, have you seen any bed(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any bed(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0218_01/00120.jpg", "scene0218_01/00150.jpg", "scene0218_01/00180.jpg", "scene0218_01/00210.jpg" ], "turn_id": 2, "scan_id": "scene0218_01", "origin_question": "Remember, have you seen any desk(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any desk(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0218_01/00240.jpg", "scene0218_01/00270.jpg", "scene0218_01/00300.jpg", "scene0218_01/00330.jpg" ], "turn_id": 3, "scan_id": "scene0218_01", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0218_01/00360.jpg", "scene0218_01/00390.jpg", "scene0218_01/00420.jpg", "scene0218_01/00450.jpg" ], "turn_id": 4, "scan_id": "scene0218_01", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the desk\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the desk\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0218_01/00480.jpg", "scene0218_01/00510.jpg", "scene0218_01/00540.jpg", "scene0218_01/00570.jpg" ], "turn_id": 5, "scan_id": "scene0218_01", "origin_question": "When did you first discover \"the desk\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the desk\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0218_01/00600.jpg", "scene0218_01/00630.jpg", "scene0218_01/00660.jpg", "scene0218_01/00690.jpg" ], "turn_id": 6, "scan_id": "scene0218_01", "origin_question": "What will be the first-time appearance order of the following categories: book, mouse, socket?", "option": [ "socket, mouse, book", "socket, book, mouse", "book, mouse, socket", "mouse, socket, book" ], "answer": "book, mouse, socket", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: book, mouse, socket? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['socket, mouse, book', 'socket, book, mouse', 'book, mouse, socket', 'mouse, socket, book'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0377_02/00000.jpg", "scene0377_02/00030.jpg", "scene0377_02/00060.jpg", "scene0377_02/00090.jpg" ], "turn_id": 1, "scan_id": "scene0377_02", "origin_question": "Remember, have you seen any towel(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any towel(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0377_02/00120.jpg", "scene0377_02/00150.jpg", "scene0377_02/00180.jpg", "scene0377_02/00210.jpg" ], "turn_id": 2, "scan_id": "scene0377_02", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the dark rectangular soft cotton towel\"; \"the rectangular dark brown or beige sturdy crate\"; \"the wooden chair with a reddish-brown seat cushion\" ", "option": [ "the dark rectangular soft cotton towel", "the rectangular dark brown or beige sturdy crate", "the wooden chair with a reddish-brown seat cushion" ], "answer": "the rectangular dark brown or beige sturdy crate", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the dark rectangular soft cotton towel\"; \"the rectangular dark brown or beige sturdy crate\"; \"the wooden chair with a reddish-brown seat cushion\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the dark rectangular soft cotton towel', 'the rectangular dark brown or beige sturdy crate', 'the wooden chair with a reddish-brown seat cushion'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "scene0377_02/00240.jpg", "scene0377_02/00270.jpg", "scene0377_02/00300.jpg", "scene0377_02/00330.jpg" ], "turn_id": 3, "scan_id": "scene0377_02", "origin_question": "Among these three objects(\"the bowl\",\"the wooden chair with a reddish-brown seat cushion\",\"the rectangular dark brown or beige sturdy crate\" ), which one is the closest to you?", "option": [ "the bowl", "the wooden chair with a reddish-brown seat cushion", "the rectangular dark brown or beige sturdy crate" ], "answer": "the rectangular dark brown or beige sturdy crate", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects(\"the bowl\",\"the wooden chair with a reddish-brown seat cushion\",\"the rectangular dark brown or beige sturdy crate\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the bowl', 'the wooden chair with a reddish-brown seat cushion', 'the rectangular dark brown or beige sturdy crate'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0377_02/00360.jpg", "scene0377_02/00390.jpg", "scene0377_02/00420.jpg", "scene0377_02/00450.jpg" ], "turn_id": 4, "scan_id": "scene0377_02", "origin_question": "Which direction is the board in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the board in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0377_02/00480.jpg", "scene0377_02/00510.jpg", "scene0377_02/00540.jpg", "scene0377_02/00570.jpg" ], "turn_id": 5, "scan_id": "scene0377_02", "origin_question": "Please recall \"the rectangular, slightly yellowish plastic tray\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.44", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the rectangular, slightly yellowish plastic tray\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0377_02/00600.jpg", "scene0377_02/00630.jpg", "scene0377_02/00660.jpg", "scene0377_02/00690.jpg" ], "turn_id": 6, "scan_id": "scene0377_02", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the wooden chair with a reddish-brown seat cushion\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the wooden chair with a reddish-brown seat cushion\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0469_01/00000.jpg", "scene0469_01/00030.jpg", "scene0469_01/00060.jpg", "scene0469_01/00090.jpg" ], "turn_id": 1, "scan_id": "scene0469_01", "origin_question": "Remember, how many bin(s) have you seen so far? ", "option": [], "answer": "2", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many bin(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0469_01/00120.jpg", "scene0469_01/00150.jpg", "scene0469_01/00180.jpg", "scene0469_01/00210.jpg" ], "turn_id": 2, "scan_id": "scene0469_01", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0469_01/00240.jpg", "scene0469_01/00270.jpg", "scene0469_01/00300.jpg", "scene0469_01/00330.jpg" ], "turn_id": 3, "scan_id": "scene0469_01", "origin_question": "In which turn did you see both \"the rectangular blackboard with a white background\" and \"the medium-sized rectangular black backpack\" simultaneously? ", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the rectangular blackboard with a white background\" and \"the medium-sized rectangular black backpack\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0469_01/00360.jpg", "scene0469_01/00390.jpg", "scene0469_01/00420.jpg", "scene0469_01/00450.jpg" ], "turn_id": 4, "scan_id": "scene0469_01", "origin_question": "In which turn were you closest to \"the window\"? ", "option": [], "answer": "3", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the window\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0469_01/00480.jpg", "scene0469_01/00510.jpg", "scene0469_01/00540.jpg", "scene0469_01/00570.jpg" ], "turn_id": 5, "scan_id": "scene0469_01", "origin_question": "Which two objects are on the same side of you now? \"the medium-sized rectangular black backpack\", \"the rectangular, white paper\", \"the vibrant blue wooden door with a smooth surface\". ", "option": [ "the medium-sized rectangular black backpack, the rectangular, white paper", "the medium-sized rectangular black backpack, the vibrant blue wooden door with a smooth surface", "the rectangular, white paper, the vibrant blue wooden door with a smooth surface" ], "answer": "the medium-sized rectangular black backpack, the rectangular, white paper", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the medium-sized rectangular black backpack\", \"the rectangular, white paper\", \"the vibrant blue wooden door with a smooth surface\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the medium-sized rectangular black backpack, the rectangular, white paper', 'the medium-sized rectangular black backpack, the vibrant blue wooden door with a smooth surface', 'the rectangular, white paper, the vibrant blue wooden door with a smooth surface'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0469_01/00600.jpg", "scene0469_01/00630.jpg", "scene0469_01/00660.jpg", "scene0469_01/00690.jpg" ], "turn_id": 6, "scan_id": "scene0469_01", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0469_01/00720.jpg", "scene0469_01/00750.jpg", "scene0469_01/00780.jpg", "scene0469_01/00810.jpg" ], "turn_id": 7, "scan_id": "scene0469_01", "origin_question": "How far is your current position from where you were at the end of the turn 3? (in meters)", "option": [], "answer": "1.05", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 3? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0074_region7/f1b191033043441987b8ebf1bb55002c_i1_5.jpg", "1mp3d_0074_region7/f1b191033043441987b8ebf1bb55002c_i2_5.jpg", "1mp3d_0074_region7/c20f944e9c434804944b0c1251247202_i1_1.jpg", "1mp3d_0074_region7/c20f944e9c434804944b0c1251247202_i2_1.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0074_region7", "origin_question": "Remember, have you seen any door(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any door(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0074_region7/1c2209fa44fc48c7b7ce7981bafc4ccf_i1_1.jpg", "1mp3d_0074_region7/1c2209fa44fc48c7b7ce7981bafc4ccf_i2_1.jpg", "1mp3d_0074_region7/c20f944e9c434804944b0c1251247202_i1_0.jpg", "1mp3d_0074_region7/c20f944e9c434804944b0c1251247202_i2_0.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0074_region7", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "1mp3d_0074_region7/4ea1cfa67ec145b39f90056a02a43d9f_i1_1.jpg", "1mp3d_0074_region7/4ea1cfa67ec145b39f90056a02a43d9f_i2_1.jpg", "1mp3d_0074_region7/c20f944e9c434804944b0c1251247202_i1_2.jpg", "1mp3d_0074_region7/c20f944e9c434804944b0c1251247202_i2_2.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0074_region7", "origin_question": "Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "61.06", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "1mp3d_0074_region7/1c2209fa44fc48c7b7ce7981bafc4ccf_i1_2.jpg", "1mp3d_0074_region7/1c2209fa44fc48c7b7ce7981bafc4ccf_i2_2.jpg", "1mp3d_0074_region7/595ef74c51bb4448ae709297121d4c10_i1_2.jpg", "1mp3d_0074_region7/595ef74c51bb4448ae709297121d4c10_i2_2.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0074_region7", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "89.61", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "1mp3d_0074_region7/e4a543521c44412e8f264d083ac5fc0c_i1_1.jpg", "1mp3d_0074_region7/e4a543521c44412e8f264d083ac5fc0c_i2_1.jpg", "1mp3d_0074_region7/e4a543521c44412e8f264d083ac5fc0c_i1_3.jpg", "1mp3d_0074_region7/e4a543521c44412e8f264d083ac5fc0c_i2_3.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0074_region7", "origin_question": "Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance forward or backward from that position?", "option": [ "Forward", "Backward" ], "answer": "Backward", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance forward or backward from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Forward', 'Backward'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0702_02/00000.jpg", "scene0702_02/00030.jpg", "scene0702_02/00060.jpg", "scene0702_02/00090.jpg" ], "turn_id": 1, "scan_id": "scene0702_02", "origin_question": "Remember, have you seen any toilet(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any toilet(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0702_02/00120.jpg", "scene0702_02/00150.jpg", "scene0702_02/00180.jpg", "scene0702_02/00210.jpg" ], "turn_id": 2, "scan_id": "scene0702_02", "origin_question": "In which turn were you closest to \"the door\"? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the door\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0702_02/00240.jpg", "scene0702_02/00270.jpg", "scene0702_02/00300.jpg", "scene0702_02/00330.jpg" ], "turn_id": 3, "scan_id": "scene0702_02", "origin_question": "When did you first discover \"the door\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the door\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0702_02/00360.jpg", "scene0702_02/00390.jpg", "scene0702_02/00420.jpg", "scene0702_02/00450.jpg" ], "turn_id": 4, "scan_id": "scene0702_02", "origin_question": "At the end of which turn were both of \"the white ceramic toilet with a round bowl\" and \"the small rectangular white telephone\" on your right side?", "option": [], "answer": "4", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the white ceramic toilet with a round bowl\" and \"the small rectangular white telephone\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "scene0702_02/00480.jpg", "scene0702_02/00510.jpg", "scene0702_02/00540.jpg", "scene0702_02/00570.jpg" ], "turn_id": 5, "scan_id": "scene0702_02", "origin_question": "Among these three objects(\"the white ceramic toilet with a round bowl\",\"the rectangular grey plastic bin, moderate in size\",\"the small rectangular white telephone\" ), which one is the closest to you?", "option": [ "the white ceramic toilet with a round bowl", "the rectangular grey plastic bin, moderate in size", "the small rectangular white telephone" ], "answer": "the small rectangular white telephone", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects(\"the white ceramic toilet with a round bowl\",\"the rectangular grey plastic bin, moderate in size\",\"the small rectangular white telephone\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the white ceramic toilet with a round bowl', 'the rectangular grey plastic bin, moderate in size', 'the small rectangular white telephone'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0044_region36/ca9d8289f61244288b696bf832fa2713_i1_1.jpg", "1mp3d_0044_region36/ca9d8289f61244288b696bf832fa2713_i2_1.jpg", "1mp3d_0044_region36/ca9d8289f61244288b696bf832fa2713_i1_0.jpg", "1mp3d_0044_region36/ca9d8289f61244288b696bf832fa2713_i2_0.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0044_region36", "origin_question": "Remember, have you seen any salt(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any salt(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0044_region36/ca9d8289f61244288b696bf832fa2713_i1_5.jpg", "1mp3d_0044_region36/ca9d8289f61244288b696bf832fa2713_i2_5.jpg", "1mp3d_0044_region36/ca9d8289f61244288b696bf832fa2713_i1_4.jpg", "1mp3d_0044_region36/ca9d8289f61244288b696bf832fa2713_i2_4.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0044_region36", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0044_region36/b01bbddd12ee462abc14ce13d62be35b_i1_0.jpg", "1mp3d_0044_region36/b01bbddd12ee462abc14ce13d62be35b_i2_0.jpg", "1mp3d_0044_region36/f4054160a6f445239913e0d2e20a8448_i1_5.jpg", "1mp3d_0044_region36/f4054160a6f445239913e0d2e20a8448_i2_5.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0044_region36", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the large oval white porcelain sink on granite\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the large oval white porcelain sink on granite\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "1mp3d_0044_region36/f4054160a6f445239913e0d2e20a8448_i1_4.jpg", "1mp3d_0044_region36/f4054160a6f445239913e0d2e20a8448_i2_4.jpg", "1mp3d_0044_region36/f4054160a6f445239913e0d2e20a8448_i1_3.jpg", "1mp3d_0044_region36/f4054160a6f445239913e0d2e20a8448_i2_3.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0044_region36", "origin_question": "What will be the first-time appearance order of the following categories: light, sink, towel?", "option": [ "sink, towel, light", "light, sink, towel", "towel, light, sink", "towel, sink, light" ], "answer": "towel, light, sink", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: light, sink, towel? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['sink, towel, light', 'light, sink, towel', 'towel, light, sink', 'towel, sink, light'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "1mp3d_0044_region36/b01bbddd12ee462abc14ce13d62be35b_i1_5.jpg", "1mp3d_0044_region36/b01bbddd12ee462abc14ce13d62be35b_i2_5.jpg", "1mp3d_0044_region36/b01bbddd12ee462abc14ce13d62be35b_i1_4.jpg", "1mp3d_0044_region36/b01bbddd12ee462abc14ce13d62be35b_i2_4.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0044_region36", "origin_question": "Which direction is the moderate-sized canvas picture in a black frame in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the moderate-sized canvas picture in a black frame in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "1mp3d_0044_region36/7998913bccf443c5812520c2e24b76bc_i1_2.jpg", "1mp3d_0044_region36/7998913bccf443c5812520c2e24b76bc_i2_2.jpg", "1mp3d_0044_region36/1d1d8cad4ab84fd99208019d77bda55e_i1_0.jpg", "1mp3d_0044_region36/1d1d8cad4ab84fd99208019d77bda55e_i2_0.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0044_region36", "origin_question": "Assuming the direction you are facing at the end of the turn 3 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Left", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 3 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0474_02/00000.jpg", "scene0474_02/00040.jpg", "scene0474_02/00080.jpg", "scene0474_02/00120.jpg", "scene0474_02/00160.jpg" ], "turn_id": 1, "scan_id": "scene0474_02", "origin_question": "Remember, how many bench(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many bench(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0474_02/00200.jpg", "scene0474_02/00240.jpg", "scene0474_02/00280.jpg", "scene0474_02/00320.jpg", "scene0474_02/00360.jpg" ], "turn_id": 2, "scan_id": "scene0474_02", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the rectangular wooden desk with signs of use\"; \"the smooth, white rectangular sheet of paper\"; \"the rectangular black plastic keyboard of standard design\" ", "option": [ "the rectangular wooden desk with signs of use", "the smooth, white rectangular sheet of paper", "the rectangular black plastic keyboard of standard design" ], "answer": "the smooth, white rectangular sheet of paper", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the rectangular wooden desk with signs of use\"; \"the smooth, white rectangular sheet of paper\"; \"the rectangular black plastic keyboard of standard design\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular wooden desk with signs of use', 'the smooth, white rectangular sheet of paper', 'the rectangular black plastic keyboard of standard design'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0474_02/00400.jpg", "scene0474_02/00440.jpg", "scene0474_02/00480.jpg", "scene0474_02/00520.jpg", "scene0474_02/00560.jpg" ], "turn_id": 3, "scan_id": "scene0474_02", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the rectangular black and silver telephone\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the rectangular black and silver telephone\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0474_02/00600.jpg", "scene0474_02/00640.jpg", "scene0474_02/00680.jpg", "scene0474_02/00720.jpg", "scene0474_02/00760.jpg" ], "turn_id": 4, "scan_id": "scene0474_02", "origin_question": "What will be the first-time appearance order of the following categories: keyboard, laptop, telephone?", "option": [ "keyboard, laptop, telephone", "keyboard, telephone, laptop", "telephone, laptop, keyboard", "laptop, keyboard, telephone" ], "answer": "keyboard, laptop, telephone", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: keyboard, laptop, telephone? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['keyboard, laptop, telephone', 'keyboard, telephone, laptop', 'telephone, laptop, keyboard', 'laptop, keyboard, telephone'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0474_02/00800.jpg", "scene0474_02/00840.jpg", "scene0474_02/00880.jpg", "scene0474_02/00920.jpg", "scene0474_02/00960.jpg" ], "turn_id": 5, "scan_id": "scene0474_02", "origin_question": "By the end of the 3 turn, were you getting closer to or farther away from these two objects: X: \"the cup\", Y: \"the rectangular black and silver telephone\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 3 turn, were you getting closer to or farther away from these two objects: X: \"the cup\", Y: \"the rectangular black and silver telephone\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0474_02/01000.jpg", "scene0474_02/01040.jpg", "scene0474_02/01080.jpg", "scene0474_02/01120.jpg", "scene0474_02/01160.jpg" ], "turn_id": 6, "scan_id": "scene0474_02", "origin_question": "At the end of which turn were both of \"the cup\" and \"the rectangular black and silver telephone\" on your right side?", "option": [], "answer": "6", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the cup\" and \"the rectangular black and silver telephone\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "scene0474_02/01200.jpg", "scene0474_02/01240.jpg", "scene0474_02/01280.jpg", "scene0474_02/01320.jpg", "scene0474_02/01360.jpg" ], "turn_id": 7, "scan_id": "scene0474_02", "origin_question": "Among these three objects(\"the silver modern laptop made of high-quality materials\",\"the large, colorful triangular decoration made of paper or fabric\",\"the cup\" ), which one is the farthest from you?", "option": [ "the silver modern laptop made of high-quality materials", "the large, colorful triangular decoration made of paper or fabric", "the cup" ], "answer": "the large, colorful triangular decoration made of paper or fabric", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the silver modern laptop made of high-quality materials\",\"the large, colorful triangular decoration made of paper or fabric\",\"the cup\" ), which one is the farthest from you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the silver modern laptop made of high-quality materials', 'the large, colorful triangular decoration made of paper or fabric', 'the cup'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0474_02/01400.jpg", "scene0474_02/01440.jpg", "scene0474_02/01480.jpg", "scene0474_02/01520.jpg", "scene0474_02/01560.jpg" ], "turn_id": 8, "scan_id": "scene0474_02", "origin_question": "By the end of the 5 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular wooden desk with signs of use\", Y: \"the rectangular black and silver telephone\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 5 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular wooden desk with signs of use\", Y: \"the rectangular black and silver telephone\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0474_02/01600.jpg", "scene0474_02/01640.jpg", "scene0474_02/01680.jpg", "scene0474_02/01720.jpg", "scene0474_02/01760.jpg" ], "turn_id": 9, "scan_id": "scene0474_02", "origin_question": "Is the couch to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the couch to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0474_02/01800.jpg", "scene0474_02/01840.jpg", "scene0474_02/01880.jpg", "scene0474_02/01920.jpg", "scene0474_02/01960.jpg" ], "turn_id": 10, "scan_id": "scene0474_02", "origin_question": "Which two objects are on the same side of you now? \"the silver modern laptop made of high-quality materials\", \"the couch\", \"the small rectangular blackboard with a white background\". ", "option": [ "the silver modern laptop made of high-quality materials, the couch", "the silver modern laptop made of high-quality materials, the small rectangular blackboard with a white background", "the couch, the small rectangular blackboard with a white background" ], "answer": "the couch, the small rectangular blackboard with a white background", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the silver modern laptop made of high-quality materials\", \"the couch\", \"the small rectangular blackboard with a white background\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the silver modern laptop made of high-quality materials, the couch', 'the silver modern laptop made of high-quality materials, the small rectangular blackboard with a white background', 'the couch, the small rectangular blackboard with a white background'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0256_01/00000.jpg", "scene0256_01/00040.jpg", "scene0256_01/00080.jpg", "scene0256_01/00120.jpg", "scene0256_01/00160.jpg" ], "turn_id": 1, "scan_id": "scene0256_01", "origin_question": "Remember, have you seen any table(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any table(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0256_01/00200.jpg", "scene0256_01/00240.jpg", "scene0256_01/00280.jpg", "scene0256_01/00320.jpg", "scene0256_01/00360.jpg" ], "turn_id": 2, "scan_id": "scene0256_01", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the three-tiered shelf with a white frame\"; \"the white rectangular table with a smooth surface\"; \"the light brown wooden chair with a gray seat\" ", "option": [ "the three-tiered shelf with a white frame", "the white rectangular table with a smooth surface", "the light brown wooden chair with a gray seat" ], "answer": "the three-tiered shelf with a white frame", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the three-tiered shelf with a white frame\"; \"the white rectangular table with a smooth surface\"; \"the light brown wooden chair with a gray seat\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the three-tiered shelf with a white frame', 'the white rectangular table with a smooth surface', 'the light brown wooden chair with a gray seat'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "scene0256_01/00400.jpg", "scene0256_01/00440.jpg", "scene0256_01/00480.jpg", "scene0256_01/00520.jpg", "scene0256_01/00560.jpg" ], "turn_id": 3, "scan_id": "scene0256_01", "origin_question": "Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Right", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0256_01/00600.jpg", "scene0256_01/00640.jpg", "scene0256_01/00680.jpg", "scene0256_01/00720.jpg", "scene0256_01/00760.jpg" ], "turn_id": 4, "scan_id": "scene0256_01", "origin_question": "What will be the first-time appearance order of the following categories: cup, hat, table?", "option": [ "hat, table, cup", "table, cup, hat", "cup, hat, table", "hat, cup, table" ], "answer": "table, cup, hat", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: cup, hat, table? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['hat, table, cup', 'table, cup, hat', 'cup, hat, table', 'hat, cup, table'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0256_01/00800.jpg", "scene0256_01/00840.jpg", "scene0256_01/00880.jpg", "scene0256_01/00920.jpg", "scene0256_01/00960.jpg" ], "turn_id": 5, "scan_id": "scene0256_01", "origin_question": "When did you first discover \"the clothes\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the clothes\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0256_01/01000.jpg", "scene0256_01/01040.jpg", "scene0256_01/01080.jpg", "scene0256_01/01120.jpg", "scene0256_01/01160.jpg" ], "turn_id": 6, "scan_id": "scene0256_01", "origin_question": "Which two objects are on the same side of you now? \"the white rectangular table with a smooth surface\", \"the small rectangular black tissue box\", \"the clothes\". ", "option": [ "the white rectangular table with a smooth surface, the small rectangular black tissue box", "the white rectangular table with a smooth surface, the clothes", "the small rectangular black tissue box, the clothes" ], "answer": "the white rectangular table with a smooth surface, the small rectangular black tissue box", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the white rectangular table with a smooth surface\", \"the small rectangular black tissue box\", \"the clothes\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the white rectangular table with a smooth surface, the small rectangular black tissue box', 'the white rectangular table with a smooth surface, the clothes', 'the small rectangular black tissue box, the clothes'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0256_01/01200.jpg", "scene0256_01/01240.jpg", "scene0256_01/01280.jpg", "scene0256_01/01320.jpg", "scene0256_01/01360.jpg" ], "turn_id": 7, "scan_id": "scene0256_01", "origin_question": "", "option": [], "answer": "", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0256_01/01400.jpg", "scene0256_01/01440.jpg", "scene0256_01/01480.jpg", "scene0256_01/01520.jpg", "scene0256_01/01560.jpg" ], "turn_id": 8, "scan_id": "scene0256_01", "origin_question": "At the end of which turn were both of \"the light brown wooden chair with a gray seat\" and \"the clothes\" on your right side?", "option": [], "answer": "8", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the light brown wooden chair with a gray seat\" and \"the clothes\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0256_01/01600.jpg", "scene0256_01/01640.jpg", "scene0256_01/01680.jpg", "scene0256_01/01720.jpg", "scene0256_01/01760.jpg" ], "turn_id": 9, "scan_id": "scene0256_01", "origin_question": "What will be the first-time appearance order of the following categories: chair, light, pillow?", "option": [ "chair, light, pillow", "pillow, light, chair", "chair, pillow, light", "light, pillow, chair" ], "answer": "chair, pillow, light", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: chair, light, pillow? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['chair, light, pillow', 'pillow, light, chair', 'chair, pillow, light', 'light, pillow, chair'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0136_01/00000.jpg", "scene0136_01/00030.jpg", "scene0136_01/00060.jpg", "scene0136_01/00090.jpg" ], "turn_id": 1, "scan_id": "scene0136_01", "origin_question": "Remember, have you seen any bin(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any bin(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0136_01/00120.jpg", "scene0136_01/00150.jpg", "scene0136_01/00180.jpg", "scene0136_01/00210.jpg" ], "turn_id": 2, "scan_id": "scene0136_01", "origin_question": "When did you first discover \"the white multifunction copier with a touchscreen interface\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the white multifunction copier with a touchscreen interface\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0136_01/00240.jpg", "scene0136_01/00270.jpg", "scene0136_01/00300.jpg", "scene0136_01/00330.jpg" ], "turn_id": 3, "scan_id": "scene0136_01", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the medium-sized green and white picture\" from your position?", "option": [], "answer": "133.19", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the medium-sized green and white picture\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0136_01/00360.jpg", "scene0136_01/00390.jpg", "scene0136_01/00420.jpg", "scene0136_01/00450.jpg" ], "turn_id": 4, "scan_id": "scene0136_01", "origin_question": "Which direction is the medium-sized green and white picture in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the medium-sized green and white picture in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0136_01/00480.jpg", "scene0136_01/00510.jpg", "scene0136_01/00540.jpg", "scene0136_01/00570.jpg" ], "turn_id": 5, "scan_id": "scene0136_01", "origin_question": "Which two objects are on the same side of you now? \"the medium-sized green and white picture\", \"the white multifunction copier with a touchscreen interface\", \"the medium rectangular white box with rounded corners\". ", "option": [ "the medium-sized green and white picture, the white multifunction copier with a touchscreen interface", "the medium-sized green and white picture, the medium rectangular white box with rounded corners", "the white multifunction copier with a touchscreen interface, the medium rectangular white box with rounded corners" ], "answer": "the medium-sized green and white picture, the white multifunction copier with a touchscreen interface", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the medium-sized green and white picture\", \"the white multifunction copier with a touchscreen interface\", \"the medium rectangular white box with rounded corners\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the medium-sized green and white picture, the white multifunction copier with a touchscreen interface', 'the medium-sized green and white picture, the medium rectangular white box with rounded corners', 'the white multifunction copier with a touchscreen interface, the medium rectangular white box with rounded corners'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0136_01/00600.jpg", "scene0136_01/00630.jpg", "scene0136_01/00660.jpg", "scene0136_01/00690.jpg" ], "turn_id": 6, "scan_id": "scene0136_01", "origin_question": "Which direction is the medium rectangular white box with rounded corners in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the medium rectangular white box with rounded corners in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0057_region19/dcb7130402734c3e8735b24d8afc2406_i1_1.jpg", "1mp3d_0057_region19/dcb7130402734c3e8735b24d8afc2406_i2_1.jpg", "1mp3d_0057_region19/dcb7130402734c3e8735b24d8afc2406_i1_0.jpg", "1mp3d_0057_region19/dcb7130402734c3e8735b24d8afc2406_i2_0.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0057_region19", "origin_question": "Remember, have you seen any tv(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any tv(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "1mp3d_0057_region19/dcb7130402734c3e8735b24d8afc2406_i1_5.jpg", "1mp3d_0057_region19/dcb7130402734c3e8735b24d8afc2406_i2_5.jpg", "1mp3d_0057_region19/dcb7130402734c3e8735b24d8afc2406_i1_4.jpg", "1mp3d_0057_region19/dcb7130402734c3e8735b24d8afc2406_i2_4.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0057_region19", "origin_question": "In which turn did you see both \"the large rectangular window with a wooden frame\" and \"the rectangular wooden bed with a light brown frame\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the large rectangular window with a wooden frame\" and \"the rectangular wooden bed with a light brown frame\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "1mp3d_0057_region19/dcb7130402734c3e8735b24d8afc2406_i1_3.jpg", "1mp3d_0057_region19/dcb7130402734c3e8735b24d8afc2406_i2_3.jpg", "1mp3d_0057_region19/dcb7130402734c3e8735b24d8afc2406_i1_2.jpg", "1mp3d_0057_region19/dcb7130402734c3e8735b24d8afc2406_i2_2.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0057_region19", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "122.02", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0057_region19/4868d91f4f544213acd2c266c98befaa_i1_0.jpg", "1mp3d_0057_region19/4868d91f4f544213acd2c266c98befaa_i2_0.jpg", "1mp3d_0057_region19/3825796e4d5b48558b36001b2f96b31c_i1_5.jpg", "1mp3d_0057_region19/3825796e4d5b48558b36001b2f96b31c_i2_5.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0057_region19", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the large rectangular window with a wooden frame\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the large rectangular window with a wooden frame\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "1mp3d_0057_region19/3825796e4d5b48558b36001b2f96b31c_i1_0.jpg", "1mp3d_0057_region19/3825796e4d5b48558b36001b2f96b31c_i2_0.jpg", "1mp3d_0057_region19/4868d91f4f544213acd2c266c98befaa_i1_1.jpg", "1mp3d_0057_region19/4868d91f4f544213acd2c266c98befaa_i2_1.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0057_region19", "origin_question": "Using your orientation at the end of turn 4 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "109.47", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 4 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "1mp3d_0057_region19/79103bf93d81464786b23dc26991d23e_i1_2.jpg", "1mp3d_0057_region19/79103bf93d81464786b23dc26991d23e_i2_2.jpg", "1mp3d_0057_region19/79103bf93d81464786b23dc26991d23e_i1_0.jpg", "1mp3d_0057_region19/79103bf93d81464786b23dc26991d23e_i2_0.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0057_region19", "origin_question": "What will be the first-time appearance order of the following categories: door, tv, window?", "option": [ "tv, door, window", "tv, window, door", "door, tv, window", "window, door, tv" ], "answer": "tv, window, door", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: door, tv, window? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['tv, door, window', 'tv, window, door', 'door, tv, window', 'window, door, tv'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0541_01/00000.jpg", "scene0541_01/00030.jpg", "scene0541_01/00060.jpg", "scene0541_01/00090.jpg" ], "turn_id": 1, "scan_id": "scene0541_01", "origin_question": "Remember, how many picture(s) have you seen so far? ", "option": [], "answer": "7", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many picture(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0541_01/00120.jpg", "scene0541_01/00150.jpg", "scene0541_01/00180.jpg", "scene0541_01/00210.jpg" ], "turn_id": 2, "scan_id": "scene0541_01", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the jar\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the jar\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0541_01/00240.jpg", "scene0541_01/00270.jpg", "scene0541_01/00300.jpg", "scene0541_01/00330.jpg" ], "turn_id": 3, "scan_id": "scene0541_01", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the tall, slender ceramic bottle with a reddish-brown hue\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the tall, slender ceramic bottle with a reddish-brown hue\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0541_01/00360.jpg", "scene0541_01/00390.jpg", "scene0541_01/00420.jpg", "scene0541_01/00450.jpg" ], "turn_id": 4, "scan_id": "scene0541_01", "origin_question": "Which two objects are on the same side of you now? \"the brown rectangular wooden bed with a curved headboard\", \"the window\", \"the three-step silver metal ladder with red accents\". ", "option": [ "the brown rectangular wooden bed with a curved headboard, the window", "the brown rectangular wooden bed with a curved headboard, the three-step silver metal ladder with red accents", "the window, the three-step silver metal ladder with red accents" ], "answer": "the brown rectangular wooden bed with a curved headboard, the window", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the brown rectangular wooden bed with a curved headboard\", \"the window\", \"the three-step silver metal ladder with red accents\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the brown rectangular wooden bed with a curved headboard, the window', 'the brown rectangular wooden bed with a curved headboard, the three-step silver metal ladder with red accents', 'the window, the three-step silver metal ladder with red accents'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0541_01/00480.jpg", "scene0541_01/00510.jpg", "scene0541_01/00540.jpg", "scene0541_01/00570.jpg" ], "turn_id": 5, "scan_id": "scene0541_01", "origin_question": "Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the three-step silver metal ladder with red accents\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the three-step silver metal ladder with red accents\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0541_01/00600.jpg", "scene0541_01/00630.jpg", "scene0541_01/00660.jpg", "scene0541_01/00690.jpg" ], "turn_id": 6, "scan_id": "scene0541_01", "origin_question": "Is the small, silver lamp with a round white shade to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the small, silver lamp with a round white shade to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0541_01/00720.jpg", "scene0541_01/00750.jpg", "scene0541_01/00780.jpg", "scene0541_01/00810.jpg" ], "turn_id": 7, "scan_id": "scene0541_01", "origin_question": "Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0425_01/00000.jpg", "scene0425_01/00040.jpg", "scene0425_01/00080.jpg", "scene0425_01/00120.jpg", "scene0425_01/00160.jpg" ], "turn_id": 1, "scan_id": "scene0425_01", "origin_question": "Remember, how many candlestick(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many candlestick(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0425_01/00200.jpg", "scene0425_01/00240.jpg", "scene0425_01/00280.jpg", "scene0425_01/00320.jpg", "scene0425_01/00360.jpg" ], "turn_id": 2, "scan_id": "scene0425_01", "origin_question": "Which direction is the transparent cylindrical pitcher with a white lid in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the transparent cylindrical pitcher with a white lid in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0425_01/00400.jpg", "scene0425_01/00440.jpg", "scene0425_01/00480.jpg", "scene0425_01/00520.jpg", "scene0425_01/00560.jpg" ], "turn_id": 3, "scan_id": "scene0425_01", "origin_question": "Please recall \"the square white foldable plastic stool with holes\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.0", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the square white foldable plastic stool with holes\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0425_01/00600.jpg", "scene0425_01/00640.jpg", "scene0425_01/00680.jpg", "scene0425_01/00720.jpg", "scene0425_01/00760.jpg" ], "turn_id": 4, "scan_id": "scene0425_01", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "scene0425_01/00800.jpg", "scene0425_01/00840.jpg", "scene0425_01/00880.jpg", "scene0425_01/00920.jpg", "scene0425_01/00960.jpg" ], "turn_id": 5, "scan_id": "scene0425_01", "origin_question": "Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance forward or backward from that position?", "option": [ "Forward", "Backward" ], "answer": "Backward", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance forward or backward from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Forward', 'Backward'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0425_01/01000.jpg", "scene0425_01/01040.jpg", "scene0425_01/01080.jpg", "scene0425_01/01120.jpg", "scene0425_01/01160.jpg" ], "turn_id": 6, "scan_id": "scene0425_01", "origin_question": "Which direction is the dark blue cylindrical hamper made of sturdy fabric in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the dark blue cylindrical hamper made of sturdy fabric in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0425_01/01200.jpg", "scene0425_01/01240.jpg", "scene0425_01/01280.jpg", "scene0425_01/01320.jpg", "scene0425_01/01360.jpg" ], "turn_id": 7, "scan_id": "scene0425_01", "origin_question": "Which direction is the square white foldable plastic stool with holes in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the square white foldable plastic stool with holes in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0425_01/01400.jpg", "scene0425_01/01440.jpg", "scene0425_01/01480.jpg", "scene0425_01/01520.jpg", "scene0425_01/01560.jpg" ], "turn_id": 8, "scan_id": "scene0425_01", "origin_question": "What will be the first-time appearance order of the following categories: bed, door, tissue box?", "option": [ "bed, tissue box, door", "tissue box, bed, door", "tissue box, door, bed", "bed, door, tissue box" ], "answer": "bed, tissue box, door", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bed, door, tissue box? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['bed, tissue box, door', 'tissue box, bed, door', 'tissue box, door, bed', 'bed, door, tissue box'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "1mp3d_0067_region25/13faa4c86406405ba802f5bdcba744ee_i1_2.jpg", "1mp3d_0067_region25/13faa4c86406405ba802f5bdcba744ee_i2_2.jpg", "1mp3d_0067_region25/500a6e1357ea4860b4971a3011195e05_i1_3.jpg", "1mp3d_0067_region25/500a6e1357ea4860b4971a3011195e05_i2_3.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0067_region25", "origin_question": "Remember, how many ball(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many ball(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "1mp3d_0067_region25/13faa4c86406405ba802f5bdcba744ee_i1_1.jpg", "1mp3d_0067_region25/13faa4c86406405ba802f5bdcba744ee_i2_1.jpg", "1mp3d_0067_region25/329e410f057b4056a2b61cab3322d815_i1_1.jpg", "1mp3d_0067_region25/329e410f057b4056a2b61cab3322d815_i2_1.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0067_region25", "origin_question": "When did you first discover \"the rectangular bed in dark colors with plush material\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the rectangular bed in dark colors with plush material\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0067_region25/1374647ec3f94d46a28310bdf04bb8d8_i1_5.jpg", "1mp3d_0067_region25/1374647ec3f94d46a28310bdf04bb8d8_i2_5.jpg", "1mp3d_0067_region25/5dee09fa6cca450bbffec66329368e89_i1_4.jpg", "1mp3d_0067_region25/5dee09fa6cca450bbffec66329368e89_i2_4.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0067_region25", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the rectangular bed in dark colors with plush material\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the rectangular bed in dark colors with plush material\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "1mp3d_0067_region25/8b6a76eb79104e6cbfe32db46e5fb58c_i1_4.jpg", "1mp3d_0067_region25/8b6a76eb79104e6cbfe32db46e5fb58c_i2_4.jpg", "1mp3d_0067_region25/8b6a76eb79104e6cbfe32db46e5fb58c_i1_5.jpg", "1mp3d_0067_region25/8b6a76eb79104e6cbfe32db46e5fb58c_i2_5.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0067_region25", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the fan\"; \"the rectangular, vibrant woven mat in red, yellow, and blue\"; \"the curved padded headboard in neutral tones\" ", "option": [ "the fan", "the rectangular, vibrant woven mat in red, yellow, and blue", "the curved padded headboard in neutral tones" ], "answer": "the fan", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the fan\"; \"the rectangular, vibrant woven mat in red, yellow, and blue\"; \"the curved padded headboard in neutral tones\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the fan', 'the rectangular, vibrant woven mat in red, yellow, and blue', 'the curved padded headboard in neutral tones'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "1mp3d_0067_region25/8b6a76eb79104e6cbfe32db46e5fb58c_i1_0.jpg", "1mp3d_0067_region25/8b6a76eb79104e6cbfe32db46e5fb58c_i2_0.jpg", "1mp3d_0067_region25/5dee09fa6cca450bbffec66329368e89_i1_0.jpg", "1mp3d_0067_region25/5dee09fa6cca450bbffec66329368e89_i2_0.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0067_region25", "origin_question": "In which turn were you closest to \"the rectangular bed in dark colors with plush material\"? ", "option": [], "answer": "2", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the rectangular bed in dark colors with plush material\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "1mp3d_0067_region25/5dee09fa6cca450bbffec66329368e89_i1_5.jpg", "1mp3d_0067_region25/5dee09fa6cca450bbffec66329368e89_i2_5.jpg", "1mp3d_0067_region25/1374647ec3f94d46a28310bdf04bb8d8_i1_0.jpg", "1mp3d_0067_region25/1374647ec3f94d46a28310bdf04bb8d8_i2_0.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0067_region25", "origin_question": "Is the long, light-colored curtain to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the long, light-colored curtain to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "1mp3d_0067_region25/1374647ec3f94d46a28310bdf04bb8d8_i1_1.jpg", "1mp3d_0067_region25/1374647ec3f94d46a28310bdf04bb8d8_i2_1.jpg", "1mp3d_0067_region25/329e410f057b4056a2b61cab3322d815_i1_0.jpg", "1mp3d_0067_region25/329e410f057b4056a2b61cab3322d815_i2_0.jpg" ], "turn_id": 7, "scan_id": "1mp3d_0067_region25", "origin_question": "Please recall \"the rectangular, vibrant woven mat in red, yellow, and blue\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.91", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the rectangular, vibrant woven mat in red, yellow, and blue\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "1mp3d_0067_region25/329e410f057b4056a2b61cab3322d815_i1_5.jpg", "1mp3d_0067_region25/329e410f057b4056a2b61cab3322d815_i2_5.jpg", "1mp3d_0067_region25/329e410f057b4056a2b61cab3322d815_i1_4.jpg", "1mp3d_0067_region25/329e410f057b4056a2b61cab3322d815_i2_4.jpg" ], "turn_id": 8, "scan_id": "1mp3d_0067_region25", "origin_question": "How far is your current position from where you were at the end of the turn 5? (in meters)", "option": [], "answer": "1.91", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 5? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0111_01/00000.jpg", "scene0111_01/00040.jpg", "scene0111_01/00080.jpg", "scene0111_01/00120.jpg", "scene0111_01/00160.jpg" ], "turn_id": 1, "scan_id": "scene0111_01", "origin_question": "Remember, how many player(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many player(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0111_01/00200.jpg", "scene0111_01/00240.jpg", "scene0111_01/00280.jpg", "scene0111_01/00320.jpg", "scene0111_01/00360.jpg" ], "turn_id": 2, "scan_id": "scene0111_01", "origin_question": "When did you first discover \"the rectangular wooden table with a natural brown color\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the rectangular wooden table with a natural brown color\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0111_01/00400.jpg", "scene0111_01/00440.jpg", "scene0111_01/00480.jpg", "scene0111_01/00520.jpg", "scene0111_01/00560.jpg" ], "turn_id": 3, "scan_id": "scene0111_01", "origin_question": "Remember, how many shoe(s) have you seen so far? ", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many shoe(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0111_01/00600.jpg", "scene0111_01/00640.jpg", "scene0111_01/00680.jpg", "scene0111_01/00720.jpg", "scene0111_01/00760.jpg" ], "turn_id": 4, "scan_id": "scene0111_01", "origin_question": "Which direction is the rectangular white range hood made of shiny metal in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the rectangular white range hood made of shiny metal in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0111_01/00800.jpg", "scene0111_01/00840.jpg", "scene0111_01/00880.jpg", "scene0111_01/00920.jpg", "scene0111_01/00960.jpg" ], "turn_id": 5, "scan_id": "scene0111_01", "origin_question": "Is the telephone to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the telephone to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0111_01/01000.jpg", "scene0111_01/01040.jpg", "scene0111_01/01080.jpg", "scene0111_01/01120.jpg", "scene0111_01/01160.jpg" ], "turn_id": 6, "scan_id": "scene0111_01", "origin_question": "When did you first discover \"the black vertical metal stand with a reflective surface\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the black vertical metal stand with a reflective surface\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0111_01/01200.jpg", "scene0111_01/01240.jpg", "scene0111_01/01280.jpg", "scene0111_01/01320.jpg", "scene0111_01/01360.jpg" ], "turn_id": 7, "scan_id": "scene0111_01", "origin_question": "Using your orientation at the end of turn 6 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "21.58", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 6 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0111_01/01410.jpg", "scene0111_01/01450.jpg", "scene0111_01/01490.jpg", "scene0111_01/01530.jpg", "scene0111_01/01570.jpg" ], "turn_id": 8, "scan_id": "scene0111_01", "origin_question": "Which direction is the small round white light fixture made of plastic in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the small round white light fixture made of plastic in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0111_01/01610.jpg", "scene0111_01/01650.jpg", "scene0111_01/01700.jpg", "scene0111_01/01740.jpg", "scene0111_01/01780.jpg" ], "turn_id": 9, "scan_id": "scene0111_01", "origin_question": "Which direction is the microwave in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the microwave in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0111_01/01820.jpg", "scene0111_01/01860.jpg", "scene0111_01/01900.jpg", "scene0111_01/01990.jpg", "scene0111_01/02030.jpg" ], "turn_id": 10, "scan_id": "scene0111_01", "origin_question": "Using your orientation at the end of turn 4 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "153.08", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 4 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0207_02/00000.jpg", "scene0207_02/00040.jpg", "scene0207_02/00080.jpg", "scene0207_02/00120.jpg", "scene0207_02/00160.jpg" ], "turn_id": 1, "scan_id": "scene0207_02", "origin_question": "Remember, have you seen any mop(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any mop(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0207_02/00200.jpg", "scene0207_02/00240.jpg", "scene0207_02/00280.jpg", "scene0207_02/00320.jpg", "scene0207_02/00360.jpg" ], "turn_id": 2, "scan_id": "scene0207_02", "origin_question": "Which direction is the medium light beige rectangular cotton pillow in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the medium light beige rectangular cotton pillow in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0207_02/00400.jpg", "scene0207_02/00440.jpg", "scene0207_02/00480.jpg", "scene0207_02/00520.jpg", "scene0207_02/00560.jpg" ], "turn_id": 3, "scan_id": "scene0207_02", "origin_question": "Is the rectangular wooden shelf with four layers to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the rectangular wooden shelf with four layers to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "scene0207_02/00600.jpg", "scene0207_02/00640.jpg", "scene0207_02/00680.jpg", "scene0207_02/00720.jpg", "scene0207_02/00760.jpg" ], "turn_id": 4, "scan_id": "scene0207_02", "origin_question": "Among these three objects(\"the black cushioned chair with a metal frame\",\"the dark rectangular couch designed for comfort\",\"the cylindrical light with a slight greenish hue\" ), which one is the farthest from you?", "option": [ "the black cushioned chair with a metal frame", "the dark rectangular couch designed for comfort", "the cylindrical light with a slight greenish hue" ], "answer": "the dark rectangular couch designed for comfort", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the black cushioned chair with a metal frame\",\"the dark rectangular couch designed for comfort\",\"the cylindrical light with a slight greenish hue\" ), which one is the farthest from you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the black cushioned chair with a metal frame', 'the dark rectangular couch designed for comfort', 'the cylindrical light with a slight greenish hue'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0207_02/00800.jpg", "scene0207_02/00840.jpg", "scene0207_02/00880.jpg", "scene0207_02/00920.jpg", "scene0207_02/00960.jpg" ], "turn_id": 5, "scan_id": "scene0207_02", "origin_question": "In which turn were you closest to \"the black cushioned chair with a metal frame\"? ", "option": [], "answer": "1", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the black cushioned chair with a metal frame\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "scene0207_02/01000.jpg", "scene0207_02/01040.jpg", "scene0207_02/01080.jpg", "scene0207_02/01120.jpg", "scene0207_02/01160.jpg" ], "turn_id": 6, "scan_id": "scene0207_02", "origin_question": "Among these three objects(\"the rectangular single bed with a wooden frame\",\"the large, cylindrical roll of white tissue\",\"the rectangular dark wooden stand for small items\" ), which one is the closest to you?", "option": [ "the rectangular single bed with a wooden frame", "the large, cylindrical roll of white tissue", "the rectangular dark wooden stand for small items" ], "answer": "the large, cylindrical roll of white tissue", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the rectangular single bed with a wooden frame\",\"the large, cylindrical roll of white tissue\",\"the rectangular dark wooden stand for small items\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular single bed with a wooden frame', 'the large, cylindrical roll of white tissue', 'the rectangular dark wooden stand for small items'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0207_02/01200.jpg", "scene0207_02/01240.jpg", "scene0207_02/01280.jpg", "scene0207_02/01320.jpg", "scene0207_02/01360.jpg" ], "turn_id": 7, "scan_id": "scene0207_02", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0207_02/01400.jpg", "scene0207_02/01440.jpg", "scene0207_02/01480.jpg", "scene0207_02/01520.jpg", "scene0207_02/01560.jpg" ], "turn_id": 8, "scan_id": "scene0207_02", "origin_question": "Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "scene0207_02/01600.jpg", "scene0207_02/01640.jpg", "scene0207_02/01680.jpg", "scene0207_02/01720.jpg", "scene0207_02/01760.jpg" ], "turn_id": 9, "scan_id": "scene0207_02", "origin_question": "Among these three objects(\"the standard vacuum cleaner with white and black stripes\",\"the white microwave with black stripes\",\"the rectangular white counter made of smooth material\" ), which one is the closest to you?", "option": [ "the standard vacuum cleaner with white and black stripes", "the white microwave with black stripes", "the rectangular white counter made of smooth material" ], "answer": "the standard vacuum cleaner with white and black stripes", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the standard vacuum cleaner with white and black stripes\",\"the white microwave with black stripes\",\"the rectangular white counter made of smooth material\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the standard vacuum cleaner with white and black stripes', 'the white microwave with black stripes', 'the rectangular white counter made of smooth material'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0207_02/01800.jpg", "scene0207_02/01840.jpg", "scene0207_02/01880.jpg", "scene0207_02/01920.jpg", "scene0207_02/01960.jpg" ], "turn_id": 10, "scan_id": "scene0207_02", "origin_question": "Which direction is the large, rectangular green soft blanket in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the large, rectangular green soft blanket in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0281_00/00000.jpg", "scene0281_00/00040.jpg", "scene0281_00/00080.jpg", "scene0281_00/00120.jpg", "scene0281_00/00160.jpg" ], "turn_id": 1, "scan_id": "scene0281_00", "origin_question": "Remember, how many garbage(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many garbage(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0281_00/00200.jpg", "scene0281_00/00240.jpg", "scene0281_00/00280.jpg", "scene0281_00/00320.jpg", "scene0281_00/00360.jpg" ], "turn_id": 2, "scan_id": "scene0281_00", "origin_question": "In which turn did you see both \"the horizontal, white panel stripe radiator\" and \"the rectangular green plush stool without armrests\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the horizontal, white panel stripe radiator\" and \"the rectangular green plush stool without armrests\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0281_00/00400.jpg", "scene0281_00/00440.jpg", "scene0281_00/00480.jpg", "scene0281_00/00520.jpg", "scene0281_00/00560.jpg" ], "turn_id": 3, "scan_id": "scene0281_00", "origin_question": "Remember, have you seen any picture(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any picture(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0281_00/00600.jpg", "scene0281_00/00640.jpg", "scene0281_00/00680.jpg", "scene0281_00/00720.jpg", "scene0281_00/00760.jpg" ], "turn_id": 4, "scan_id": "scene0281_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0281_00/00800.jpg", "scene0281_00/00840.jpg", "scene0281_00/00880.jpg", "scene0281_00/00920.jpg", "scene0281_00/00960.jpg" ], "turn_id": 5, "scan_id": "scene0281_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0281_00/01000.jpg", "scene0281_00/01040.jpg", "scene0281_00/01080.jpg", "scene0281_00/01120.jpg", "scene0281_00/01160.jpg" ], "turn_id": 6, "scan_id": "scene0281_00", "origin_question": "Using your orientation at the end of turn 5 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "63.85", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 5 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0281_00/01200.jpg", "scene0281_00/01240.jpg", "scene0281_00/01280.jpg", "scene0281_00/01320.jpg", "scene0281_00/01360.jpg" ], "turn_id": 7, "scan_id": "scene0281_00", "origin_question": "What will be the first-time appearance order of the following categories: blackboard, picture, window?", "option": [ "picture, window, blackboard", "window, picture, blackboard", "window, blackboard, picture", "picture, blackboard, window" ], "answer": "window, picture, blackboard", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: blackboard, picture, window? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['picture, window, blackboard', 'window, picture, blackboard', 'window, blackboard, picture', 'picture, blackboard, window'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0281_00/01400.jpg", "scene0281_00/01440.jpg", "scene0281_00/01480.jpg", "scene0281_00/01520.jpg", "scene0281_00/01560.jpg" ], "turn_id": 8, "scan_id": "scene0281_00", "origin_question": "Which two objects are on the same side of you now? \"the horizontal, white panel stripe radiator\", \"the window\", \"the standard rectangular white sheet of paper\". ", "option": [ "the horizontal, white panel stripe radiator, the window", "the horizontal, white panel stripe radiator, the standard rectangular white sheet of paper", "the window, the standard rectangular white sheet of paper" ], "answer": "the horizontal, white panel stripe radiator, the window", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the horizontal, white panel stripe radiator\", \"the window\", \"the standard rectangular white sheet of paper\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the horizontal, white panel stripe radiator, the window', 'the horizontal, white panel stripe radiator, the standard rectangular white sheet of paper', 'the window, the standard rectangular white sheet of paper'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0029_region25/6f6a10929810436f894094cbcd00e9e2_i1_2.jpg", "1mp3d_0029_region25/6f6a10929810436f894094cbcd00e9e2_i2_2.jpg", "1mp3d_0029_region25/6f6a10929810436f894094cbcd00e9e2_i1_0.jpg", "1mp3d_0029_region25/6f6a10929810436f894094cbcd00e9e2_i2_0.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0029_region25", "origin_question": "Remember, have you seen any mailbox(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any mailbox(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 2 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "1mp3d_0029_region25/60ce99b0264148c09db7ef836ad77e3f_i1_3.jpg", "1mp3d_0029_region25/60ce99b0264148c09db7ef836ad77e3f_i2_3.jpg", "1mp3d_0029_region25/60ce99b0264148c09db7ef836ad77e3f_i1_2.jpg", "1mp3d_0029_region25/60ce99b0264148c09db7ef836ad77e3f_i2_2.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0029_region25", "origin_question": "When did you first discover \"the rectangular wooden window frame painted white\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the rectangular wooden window frame painted white\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 2 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0451_02/00000.jpg", "scene0451_02/00040.jpg", "scene0451_02/00080.jpg", "scene0451_02/00120.jpg", "scene0451_02/00160.jpg" ], "turn_id": 1, "scan_id": "scene0451_02", "origin_question": "Remember, how many ottoman(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many ottoman(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0451_02/00200.jpg", "scene0451_02/00240.jpg", "scene0451_02/00280.jpg", "scene0451_02/00320.jpg", "scene0451_02/00360.jpg" ], "turn_id": 2, "scan_id": "scene0451_02", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the dark leather clothes made of thick material\" from your position?", "option": [], "answer": "144.6", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the dark leather clothes made of thick material\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0451_02/00400.jpg", "scene0451_02/00440.jpg", "scene0451_02/00480.jpg", "scene0451_02/00520.jpg", "scene0451_02/00560.jpg" ], "turn_id": 3, "scan_id": "scene0451_02", "origin_question": "Which direction is the white rectangular metal radiator in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the white rectangular metal radiator in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0451_02/00600.jpg", "scene0451_02/00640.jpg", "scene0451_02/00680.jpg", "scene0451_02/00720.jpg", "scene0451_02/00760.jpg" ], "turn_id": 4, "scan_id": "scene0451_02", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the white rectangular metal radiator\" from your position?", "option": [], "answer": "63.29", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the white rectangular metal radiator\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0451_02/00800.jpg", "scene0451_02/00840.jpg", "scene0451_02/00880.jpg", "scene0451_02/00920.jpg", "scene0451_02/00960.jpg" ], "turn_id": 5, "scan_id": "scene0451_02", "origin_question": "At the end of which turn were both of \"the comfortable brown single-seater couch made of plush material\" and \"the dark leather clothes made of thick material\" on your left side?", "option": [], "answer": "4", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the comfortable brown single-seater couch made of plush material\" and \"the dark leather clothes made of thick material\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0451_02/01000.jpg", "scene0451_02/01040.jpg", "scene0451_02/01080.jpg", "scene0451_02/01120.jpg", "scene0451_02/01160.jpg" ], "turn_id": 6, "scan_id": "scene0451_02", "origin_question": "Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0451_02/01200.jpg", "scene0451_02/01240.jpg", "scene0451_02/01280.jpg", "scene0451_02/01320.jpg", "scene0451_02/01360.jpg" ], "turn_id": 7, "scan_id": "scene0451_02", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0451_02/01400.jpg", "scene0451_02/01440.jpg", "scene0451_02/01480.jpg", "scene0451_02/01520.jpg", "scene0451_02/01560.jpg" ], "turn_id": 8, "scan_id": "scene0451_02", "origin_question": "Compared to your position at the end of the 5 turn, are you now closer or farther away from \"the bag\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 5 turn, are you now closer or farther away from \"the bag\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0451_02/01600.jpg", "scene0451_02/01640.jpg", "scene0451_02/01680.jpg", "scene0451_02/01720.jpg", "scene0451_02/01760.jpg" ], "turn_id": 9, "scan_id": "scene0451_02", "origin_question": "Which direction is the dark leather clothes made of thick material in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the dark leather clothes made of thick material in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0451_02/01800.jpg", "scene0451_02/01840.jpg", "scene0451_02/01880.jpg", "scene0451_02/01920.jpg", "scene0451_02/01960.jpg" ], "turn_id": 10, "scan_id": "scene0451_02", "origin_question": "By the end of the 7 turn, were you getting closer to or farther away from these two objects: X: \"the dark leather clothes made of thick material\", Y: \"the small rectangular dark brown wooden bench\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 7 turn, were you getting closer to or farther away from these two objects: X: \"the dark leather clothes made of thick material\", Y: \"the small rectangular dark brown wooden bench\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0057_region10/600eeada310e4c8d8e941cffd26421e8_i1_0.jpg", "1mp3d_0057_region10/600eeada310e4c8d8e941cffd26421e8_i2_0.jpg", "1mp3d_0057_region10/6ed853a792384773975309cad93b10f2_i1_0.jpg", "1mp3d_0057_region10/6ed853a792384773975309cad93b10f2_i2_0.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0057_region10", "origin_question": "Remember, have you seen any toilet(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any toilet(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "1mp3d_0057_region10/600eeada310e4c8d8e941cffd26421e8_i1_3.jpg", "1mp3d_0057_region10/600eeada310e4c8d8e941cffd26421e8_i2_3.jpg", "1mp3d_0057_region10/6df197e57ca54b64a24d622eeadf8a7a_i1_4.jpg", "1mp3d_0057_region10/6df197e57ca54b64a24d622eeadf8a7a_i2_4.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0057_region10", "origin_question": "Is the cylindrical roll of white toilet paper to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the cylindrical roll of white toilet paper to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "1mp3d_0057_region10/6df197e57ca54b64a24d622eeadf8a7a_i1_5.jpg", "1mp3d_0057_region10/6df197e57ca54b64a24d622eeadf8a7a_i2_5.jpg", "1mp3d_0057_region10/6df197e57ca54b64a24d622eeadf8a7a_i1_0.jpg", "1mp3d_0057_region10/6df197e57ca54b64a24d622eeadf8a7a_i2_0.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0057_region10", "origin_question": "What will be the first-time appearance order of the following categories: curtain, mirror, sink?", "option": [ "sink, mirror, curtain", "sink, curtain, mirror", "curtain, sink, mirror", "mirror, curtain, sink" ], "answer": "curtain, sink, mirror", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: curtain, mirror, sink? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['sink, mirror, curtain', 'sink, curtain, mirror', 'curtain, sink, mirror', 'mirror, curtain, sink'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0057_region10/6df197e57ca54b64a24d622eeadf8a7a_i1_3.jpg", "1mp3d_0057_region10/6df197e57ca54b64a24d622eeadf8a7a_i2_3.jpg", "1mp3d_0057_region10/600eeada310e4c8d8e941cffd26421e8_i1_2.jpg", "1mp3d_0057_region10/600eeada310e4c8d8e941cffd26421e8_i2_2.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0057_region10", "origin_question": "", "option": [], "answer": "", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "1mp3d_0057_region10/6ed853a792384773975309cad93b10f2_i1_1.jpg", "1mp3d_0057_region10/6ed853a792384773975309cad93b10f2_i2_1.jpg", "1mp3d_0057_region10/6ed853a792384773975309cad93b10f2_i1_2.jpg", "1mp3d_0057_region10/6ed853a792384773975309cad93b10f2_i2_2.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0057_region10", "origin_question": "Among these three objects(\"the white ceramic toilet with a matching tank\",\"the cylindrical roll of white toilet paper\",\"the table\" ), which one is the closest to you?", "option": [ "the white ceramic toilet with a matching tank", "the cylindrical roll of white toilet paper", "the table" ], "answer": "the table", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects(\"the white ceramic toilet with a matching tank\",\"the cylindrical roll of white toilet paper\",\"the table\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the white ceramic toilet with a matching tank', 'the cylindrical roll of white toilet paper', 'the table'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "1mp3d_0057_region10/6ed853a792384773975309cad93b10f2_i1_5.jpg", "1mp3d_0057_region10/6ed853a792384773975309cad93b10f2_i2_5.jpg", "1mp3d_0057_region10/6ed853a792384773975309cad93b10f2_i1_3.jpg", "1mp3d_0057_region10/6ed853a792384773975309cad93b10f2_i2_3.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0057_region10", "origin_question": "Is the white ceramic toilet with a matching tank to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the white ceramic toilet with a matching tank to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0689_00/00000.jpg", "scene0689_00/00030.jpg", "scene0689_00/00060.jpg", "scene0689_00/00090.jpg" ], "turn_id": 1, "scan_id": "scene0689_00", "origin_question": "Remember, have you seen any window(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any window(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0689_00/00120.jpg", "scene0689_00/00150.jpg", "scene0689_00/00180.jpg", "scene0689_00/00210.jpg" ], "turn_id": 2, "scan_id": "scene0689_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "scene0689_00/00240.jpg", "scene0689_00/00270.jpg", "scene0689_00/00300.jpg", "scene0689_00/00330.jpg" ], "turn_id": 3, "scan_id": "scene0689_00", "origin_question": "Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance forward or backward from that position?", "option": [ "Forward", "Backward" ], "answer": "Forward", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance forward or backward from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Forward', 'Backward'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "scene0689_00/00360.jpg", "scene0689_00/00390.jpg", "scene0689_00/00420.jpg", "scene0689_00/00450.jpg" ], "turn_id": 4, "scan_id": "scene0689_00", "origin_question": "Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Left", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0689_00/00480.jpg", "scene0689_00/00510.jpg", "scene0689_00/00540.jpg", "scene0689_00/00570.jpg" ], "turn_id": 5, "scan_id": "scene0689_00", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the dull cylindrical bin\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the dull cylindrical bin\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0689_00/00600.jpg", "scene0689_00/00630.jpg", "scene0689_00/00660.jpg", "scene0689_00/00690.jpg" ], "turn_id": 6, "scan_id": "scene0689_00", "origin_question": "In which turn were you closest to \"the door\"? ", "option": [], "answer": "3", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the door\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0365_00/00000.jpg", "scene0365_00/00040.jpg", "scene0365_00/00080.jpg", "scene0365_00/00120.jpg" ], "turn_id": 1, "scan_id": "scene0365_00", "origin_question": "Remember, have you seen any jacket(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any jacket(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0365_00/00160.jpg", "scene0365_00/00200.jpg", "scene0365_00/00240.jpg", "scene0365_00/00280.jpg" ], "turn_id": 2, "scan_id": "scene0365_00", "origin_question": "When did you first discover \"the cylindrical, matte black soap dispenser\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the cylindrical, matte black soap dispenser\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0365_00/00320.jpg", "scene0365_00/00360.jpg", "scene0365_00/00400.jpg", "scene0365_00/00440.jpg" ], "turn_id": 3, "scan_id": "scene0365_00", "origin_question": "Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "120.66", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0365_00/00480.jpg", "scene0365_00/00520.jpg", "scene0365_00/00560.jpg", "scene0365_00/00600.jpg" ], "turn_id": 4, "scan_id": "scene0365_00", "origin_question": "Is the classic rectangular white wall-mounted sink to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the classic rectangular white wall-mounted sink to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0365_00/00640.jpg", "scene0365_00/00680.jpg", "scene0365_00/00720.jpg", "scene0365_00/00760.jpg" ], "turn_id": 5, "scan_id": "scene0365_00", "origin_question": "Please recall \"the rectangular black plastic bin without a lid\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.4", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the rectangular black plastic bin without a lid\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0365_00/00800.jpg", "scene0365_00/00840.jpg", "scene0365_00/00880.jpg", "scene0365_00/00920.jpg" ], "turn_id": 6, "scan_id": "scene0365_00", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the cylindrical, matte black soap dispenser\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the cylindrical, matte black soap dispenser\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0365_00/00960.jpg", "scene0365_00/01000.jpg", "scene0365_00/01040.jpg", "scene0365_00/01080.jpg" ], "turn_id": 7, "scan_id": "scene0365_00", "origin_question": "Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "36.62", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0365_00/01120.jpg", "scene0365_00/01160.jpg", "scene0365_00/01200.jpg", "scene0365_00/01240.jpg" ], "turn_id": 8, "scan_id": "scene0365_00", "origin_question": "Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "57.4", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0365_00/01280.jpg", "scene0365_00/01320.jpg", "scene0365_00/01360.jpg", "scene0365_00/01400.jpg" ], "turn_id": 9, "scan_id": "scene0365_00", "origin_question": "Using your orientation at the end of turn 4 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "21.62", "user_message": "For the 9 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 4 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0044_region25/19728bafee484db7bc223039b9b85e5b_i1_2.jpg", "1mp3d_0044_region25/19728bafee484db7bc223039b9b85e5b_i2_2.jpg", "1mp3d_0044_region25/74dc39e3a33d49bca867e047364300e0_i1_2.jpg", "1mp3d_0044_region25/74dc39e3a33d49bca867e047364300e0_i2_2.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0044_region25", "origin_question": "Remember, have you seen any body loofah(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any body loofah(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 2 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "1mp3d_0044_region25/74dc39e3a33d49bca867e047364300e0_i1_1.jpg", "1mp3d_0044_region25/74dc39e3a33d49bca867e047364300e0_i2_1.jpg", "1mp3d_0044_region25/19728bafee484db7bc223039b9b85e5b_i1_0.jpg", "1mp3d_0044_region25/19728bafee484db7bc223039b9b85e5b_i2_0.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0044_region25", "origin_question": "At the end of which turn were both of \"the sink\" and \"the light-colored marble counter with a black sink\" on your left side?", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the sink\" and \"the light-colored marble counter with a black sink\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 2 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0078_region6/36ab9254a358420881b436d5fe34281b_i1_2.jpg", "1mp3d_0078_region6/36ab9254a358420881b436d5fe34281b_i2_2.jpg", "1mp3d_0078_region6/b6beaaff9e014de8b04eb79045739b49_i1_1.jpg", "1mp3d_0078_region6/b6beaaff9e014de8b04eb79045739b49_i2_1.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0078_region6", "origin_question": "Remember, have you seen any bench(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any bench(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0078_region6/dfb3e7ca0659455a93ee8493df3f8784_i1_3.jpg", "1mp3d_0078_region6/dfb3e7ca0659455a93ee8493df3f8784_i2_3.jpg", "1mp3d_0078_region6/dfb3e7ca0659455a93ee8493df3f8784_i1_2.jpg", "1mp3d_0078_region6/dfb3e7ca0659455a93ee8493df3f8784_i2_2.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0078_region6", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0078_region6/b6beaaff9e014de8b04eb79045739b49_i1_2.jpg", "1mp3d_0078_region6/b6beaaff9e014de8b04eb79045739b49_i2_2.jpg", "1mp3d_0078_region6/4d574a196c884caa93b2da69dd36f201_i1_4.jpg", "1mp3d_0078_region6/4d574a196c884caa93b2da69dd36f201_i2_4.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0078_region6", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the light rectangular wooden door with horizontal slats\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the light rectangular wooden door with horizontal slats\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0078_region6/4d574a196c884caa93b2da69dd36f201_i1_5.jpg", "1mp3d_0078_region6/4d574a196c884caa93b2da69dd36f201_i2_5.jpg", "1mp3d_0078_region6/4d574a196c884caa93b2da69dd36f201_i1_0.jpg", "1mp3d_0078_region6/4d574a196c884caa93b2da69dd36f201_i2_0.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0078_region6", "origin_question": "Remember, have you seen any light(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any light(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0078_region6/4d574a196c884caa93b2da69dd36f201_i1_1.jpg", "1mp3d_0078_region6/4d574a196c884caa93b2da69dd36f201_i2_1.jpg", "1mp3d_0078_region6/4d574a196c884caa93b2da69dd36f201_i1_3.jpg", "1mp3d_0078_region6/4d574a196c884caa93b2da69dd36f201_i2_3.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0078_region6", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the rectangular window with light-colored wooden decorations\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the rectangular window with light-colored wooden decorations\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0648_01/00000.jpg", "scene0648_01/00040.jpg", "scene0648_01/00080.jpg", "scene0648_01/00120.jpg", "scene0648_01/00160.jpg" ], "turn_id": 1, "scan_id": "scene0648_01", "origin_question": "Remember, have you seen any mouse(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any mouse(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0648_01/00200.jpg", "scene0648_01/00240.jpg", "scene0648_01/00280.jpg", "scene0648_01/00320.jpg", "scene0648_01/00360.jpg" ], "turn_id": 2, "scan_id": "scene0648_01", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0648_01/00400.jpg", "scene0648_01/00440.jpg", "scene0648_01/00480.jpg", "scene0648_01/00520.jpg", "scene0648_01/00560.jpg" ], "turn_id": 3, "scan_id": "scene0648_01", "origin_question": "Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "87.39", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0648_01/00600.jpg", "scene0648_01/00640.jpg", "scene0648_01/00680.jpg", "scene0648_01/00720.jpg", "scene0648_01/00760.jpg" ], "turn_id": 4, "scan_id": "scene0648_01", "origin_question": "Which two objects are on the same side of you now? \"the rectangular wooden shelf with two levels\", \"the rectangular wooden desk made of oak\", \"the black shoe with a simple design\". ", "option": [ "the rectangular wooden shelf with two levels, the rectangular wooden desk made of oak", "the rectangular wooden shelf with two levels, the black shoe with a simple design", "the rectangular wooden desk made of oak, the black shoe with a simple design" ], "answer": "the rectangular wooden shelf with two levels, the rectangular wooden desk made of oak", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the rectangular wooden shelf with two levels\", \"the rectangular wooden desk made of oak\", \"the black shoe with a simple design\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular wooden shelf with two levels, the rectangular wooden desk made of oak', 'the rectangular wooden shelf with two levels, the black shoe with a simple design', 'the rectangular wooden desk made of oak, the black shoe with a simple design'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0648_01/00800.jpg", "scene0648_01/00840.jpg", "scene0648_01/00880.jpg", "scene0648_01/00920.jpg", "scene0648_01/00960.jpg" ], "turn_id": 5, "scan_id": "scene0648_01", "origin_question": "In which turn were you closest to \"the rectangular wooden shelf with two levels\"? ", "option": [], "answer": "3", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the rectangular wooden shelf with two levels\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0648_01/01000.jpg", "scene0648_01/01040.jpg", "scene0648_01/01080.jpg", "scene0648_01/01120.jpg", "scene0648_01/01160.jpg" ], "turn_id": 6, "scan_id": "scene0648_01", "origin_question": "Which direction is the fan in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the fan in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0648_01/01200.jpg", "scene0648_01/01240.jpg", "scene0648_01/01280.jpg", "scene0648_01/01320.jpg", "scene0648_01/01360.jpg" ], "turn_id": 7, "scan_id": "scene0648_01", "origin_question": "Please recall \"the fan\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.7", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the fan\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0648_01/01400.jpg", "scene0648_01/01440.jpg", "scene0648_01/01480.jpg", "scene0648_01/01520.jpg", "scene0648_01/01560.jpg" ], "turn_id": 8, "scan_id": "scene0648_01", "origin_question": "", "option": [], "answer": "", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0648_01/01600.jpg", "scene0648_01/01640.jpg", "scene0648_01/01680.jpg", "scene0648_01/01720.jpg", "scene0648_01/01760.jpg" ], "turn_id": 9, "scan_id": "scene0648_01", "origin_question": "Using your orientation at the end of turn 5 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "114.06", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 5 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0648_01/01800.jpg", "scene0648_01/01840.jpg", "scene0648_01/01880.jpg", "scene0648_01/01920.jpg", "scene0648_01/01960.jpg" ], "turn_id": 10, "scan_id": "scene0648_01", "origin_question": "Please recall \"the rectangular decoration with artificial fabric flowers\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.58", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the rectangular decoration with artificial fabric flowers\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0354_00/00000.jpg", "scene0354_00/00040.jpg", "scene0354_00/00080.jpg", "scene0354_00/00120.jpg" ], "turn_id": 1, "scan_id": "scene0354_00", "origin_question": "Remember, have you seen any table(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any table(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0354_00/00160.jpg", "scene0354_00/00200.jpg", "scene0354_00/00240.jpg", "scene0354_00/00280.jpg" ], "turn_id": 2, "scan_id": "scene0354_00", "origin_question": "When did you first discover \"the standard-sized cylindrical dark bin\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the standard-sized cylindrical dark bin\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0354_00/00320.jpg", "scene0354_00/00360.jpg", "scene0354_00/00400.jpg", "scene0354_00/00440.jpg" ], "turn_id": 3, "scan_id": "scene0354_00", "origin_question": "In which turn did you see both \"the window\" and \"the door\" simultaneously? ", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the window\" and \"the door\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0354_00/00480.jpg", "scene0354_00/00520.jpg", "scene0354_00/00560.jpg", "scene0354_00/00600.jpg" ], "turn_id": 4, "scan_id": "scene0354_00", "origin_question": "What will be the first-time appearance order of the following categories: bin, table, tv?", "option": [ "bin, tv, table", "tv, bin, table", "table, bin, tv", "tv, table, bin" ], "answer": "table, bin, tv", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bin, table, tv? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['bin, tv, table', 'tv, bin, table', 'table, bin, tv', 'tv, table, bin'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0354_00/00640.jpg", "scene0354_00/00680.jpg", "scene0354_00/00720.jpg", "scene0354_00/00760.jpg" ], "turn_id": 5, "scan_id": "scene0354_00", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0354_00/00800.jpg", "scene0354_00/00840.jpg", "scene0354_00/00880.jpg", "scene0354_00/00920.jpg" ], "turn_id": 6, "scan_id": "scene0354_00", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the black medium-sized flat-screen TV\" from your position?", "option": [], "answer": "46.51", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the black medium-sized flat-screen TV\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0354_00/00960.jpg", "scene0354_00/01000.jpg", "scene0354_00/01040.jpg", "scene0354_00/01080.jpg" ], "turn_id": 7, "scan_id": "scene0354_00", "origin_question": "Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0500_00/00000.jpg", "scene0500_00/00040.jpg", "scene0500_00/00080.jpg", "scene0500_00/00120.jpg", "scene0500_00/00160.jpg" ], "turn_id": 1, "scan_id": "scene0500_00", "origin_question": "Remember, have you seen any candlestick(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any candlestick(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0500_00/00200.jpg", "scene0500_00/00240.jpg", "scene0500_00/00280.jpg", "scene0500_00/00320.jpg", "scene0500_00/00360.jpg" ], "turn_id": 2, "scan_id": "scene0500_00", "origin_question": "In which turn were you farthest from \"the rectangular brown wooden table with a smooth surface\"? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the rectangular brown wooden table with a smooth surface\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0500_00/00400.jpg", "scene0500_00/00440.jpg", "scene0500_00/00480.jpg", "scene0500_00/00520.jpg", "scene0500_00/00560.jpg" ], "turn_id": 3, "scan_id": "scene0500_00", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "1.21", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0500_00/00600.jpg", "scene0500_00/00640.jpg", "scene0500_00/00680.jpg", "scene0500_00/00720.jpg", "scene0500_00/00760.jpg" ], "turn_id": 4, "scan_id": "scene0500_00", "origin_question": "When did you first discover \"the rectangular dark slate blackboard in excellent condition\"? (index of the turn)", "option": [], "answer": "3", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the rectangular dark slate blackboard in excellent condition\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0500_00/00800.jpg", "scene0500_00/00840.jpg", "scene0500_00/00880.jpg", "scene0500_00/00920.jpg", "scene0500_00/00960.jpg" ], "turn_id": 5, "scan_id": "scene0500_00", "origin_question": "How far is your current position from where you were at the end of the turn 3? (in meters)", "option": [], "answer": "2.21", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 3? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0500_00/01000.jpg", "scene0500_00/01040.jpg", "scene0500_00/01080.jpg", "scene0500_00/01120.jpg", "scene0500_00/01220.jpg" ], "turn_id": 6, "scan_id": "scene0500_00", "origin_question": "What will be the first-time appearance order of the following categories: chair, cup, projector?", "option": [ "chair, cup, projector", "cup, chair, projector", "chair, projector, cup", "projector, chair, cup" ], "answer": "chair, cup, projector", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: chair, cup, projector? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['chair, cup, projector', 'cup, chair, projector', 'chair, projector, cup', 'projector, chair, cup'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0500_00/01260.jpg", "scene0500_00/01300.jpg", "scene0500_00/01340.jpg", "scene0500_00/01380.jpg", "scene0500_00/01420.jpg" ], "turn_id": 7, "scan_id": "scene0500_00", "origin_question": "Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the rectangular white projector with a sleek design\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the rectangular white projector with a sleek design\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0500_00/01460.jpg", "scene0500_00/01500.jpg", "scene0500_00/01540.jpg", "scene0500_00/01580.jpg", "scene0500_00/01620.jpg" ], "turn_id": 8, "scan_id": "scene0500_00", "origin_question": "In which turn were you farthest from \"the standard-sized white cup\"? ", "option": [], "answer": "8", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the standard-sized white cup\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0500_00/01660.jpg", "scene0500_00/01700.jpg", "scene0500_00/01740.jpg", "scene0500_00/01780.jpg", "scene0500_00/01820.jpg" ], "turn_id": 9, "scan_id": "scene0500_00", "origin_question": "In which turn did you see both \"the rectangular dark slate blackboard in excellent condition\" and \"the standard brown wooden door with a silver handle\" simultaneously? ", "option": [], "answer": "4", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the rectangular dark slate blackboard in excellent condition\" and \"the standard brown wooden door with a silver handle\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "scene0500_00/01860.jpg", "scene0500_00/01900.jpg", "scene0500_00/01940.jpg", "scene0500_00/01980.jpg", "scene0500_00/02020.jpg" ], "turn_id": 10, "scan_id": "scene0500_00", "origin_question": "Among these three objects(\"the standard-sized white cup\",\"the rectangular dark slate blackboard in excellent condition\",\"the rectangular white projector with a sleek design\" ), which one is the closest to you?", "option": [ "the standard-sized white cup", "the rectangular dark slate blackboard in excellent condition", "the rectangular white projector with a sleek design" ], "answer": "the rectangular dark slate blackboard in excellent condition", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the standard-sized white cup\",\"the rectangular dark slate blackboard in excellent condition\",\"the rectangular white projector with a sleek design\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the standard-sized white cup', 'the rectangular dark slate blackboard in excellent condition', 'the rectangular white projector with a sleek design'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0606_01/00000.jpg", "scene0606_01/00040.jpg", "scene0606_01/00080.jpg", "scene0606_01/00120.jpg", "scene0606_01/00160.jpg" ], "turn_id": 1, "scan_id": "scene0606_01", "origin_question": "Remember, have you seen any bin(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any bin(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0606_01/00200.jpg", "scene0606_01/00240.jpg", "scene0606_01/00280.jpg", "scene0606_01/00320.jpg", "scene0606_01/00360.jpg" ], "turn_id": 2, "scan_id": "scene0606_01", "origin_question": "In which turn were you closest to \"the medium brown wooden cabinet with raised paneling\"? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the medium brown wooden cabinet with raised paneling\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0606_01/00400.jpg", "scene0606_01/00440.jpg", "scene0606_01/00480.jpg", "scene0606_01/00520.jpg", "scene0606_01/00560.jpg" ], "turn_id": 3, "scan_id": "scene0606_01", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the tissue\"; \"the medium brown wooden cabinet with raised paneling\"; \"the rectangular white-framed glass window\" ", "option": [ "the tissue", "the medium brown wooden cabinet with raised paneling", "the rectangular white-framed glass window" ], "answer": "the tissue", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the tissue\"; \"the medium brown wooden cabinet with raised paneling\"; \"the rectangular white-framed glass window\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the tissue', 'the medium brown wooden cabinet with raised paneling', 'the rectangular white-framed glass window'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0606_01/00600.jpg", "scene0606_01/00640.jpg", "scene0606_01/00680.jpg", "scene0606_01/00720.jpg", "scene0606_01/00760.jpg" ], "turn_id": 4, "scan_id": "scene0606_01", "origin_question": "Please recall \"the small, healthy green potted plant with multiple branches\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "3.24", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the small, healthy green potted plant with multiple branches\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "scene0606_01/00800.jpg", "scene0606_01/00840.jpg", "scene0606_01/00880.jpg", "scene0606_01/00920.jpg", "scene0606_01/00960.jpg" ], "turn_id": 5, "scan_id": "scene0606_01", "origin_question": "Among these three objects(\"the blue rectangular plastic bin\",\"the large rectangular white bag with orange handles\",\"the case\" ), which one is the closest to you?", "option": [ "the blue rectangular plastic bin", "the large rectangular white bag with orange handles", "the case" ], "answer": "the case", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the blue rectangular plastic bin\",\"the large rectangular white bag with orange handles\",\"the case\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the blue rectangular plastic bin', 'the large rectangular white bag with orange handles', 'the case'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0606_01/01000.jpg", "scene0606_01/01040.jpg", "scene0606_01/01080.jpg", "scene0606_01/01120.jpg", "scene0606_01/01160.jpg" ], "turn_id": 6, "scan_id": "scene0606_01", "origin_question": "Which direction is the round clock with a black border and white surface in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the round clock with a black border and white surface in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0606_01/01200.jpg", "scene0606_01/01240.jpg", "scene0606_01/01280.jpg", "scene0606_01/01360.jpg", "scene0606_01/01400.jpg" ], "turn_id": 7, "scan_id": "scene0606_01", "origin_question": "Which direction is the dark rectangular picture with a wooden frame in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the dark rectangular picture with a wooden frame in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "scene0606_01/01440.jpg", "scene0606_01/01480.jpg", "scene0606_01/01520.jpg", "scene0606_01/01560.jpg", "scene0606_01/01600.jpg" ], "turn_id": 8, "scan_id": "scene0606_01", "origin_question": "Among these three objects(\"the small, healthy green potted plant with multiple branches\",\"the silver rectangular oven\",\"the round clock with a black border and white surface\" ), which one is the farthest from you?", "option": [ "the small, healthy green potted plant with multiple branches", "the silver rectangular oven", "the round clock with a black border and white surface" ], "answer": "the small, healthy green potted plant with multiple branches", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the small, healthy green potted plant with multiple branches\",\"the silver rectangular oven\",\"the round clock with a black border and white surface\" ), which one is the farthest from you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the small, healthy green potted plant with multiple branches', 'the silver rectangular oven', 'the round clock with a black border and white surface'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0606_01/01640.jpg", "scene0606_01/01680.jpg", "scene0606_01/01720.jpg", "scene0606_01/01760.jpg", "scene0606_01/01800.jpg" ], "turn_id": 9, "scan_id": "scene0606_01", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0606_01/01840.jpg", "scene0606_01/01880.jpg", "scene0606_01/01920.jpg", "scene0606_01/01960.jpg", "scene0606_01/02000.jpg" ], "turn_id": 10, "scan_id": "scene0606_01", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the large rectangular brown woven tray\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the large rectangular brown woven tray\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0040_region55/ceaee8b594de4d2d8c9f7625c0dbe09d_i1_0.jpg", "1mp3d_0040_region55/ceaee8b594de4d2d8c9f7625c0dbe09d_i2_0.jpg", "1mp3d_0040_region55/72db3245ca954e239e8a1ae3b78726e7_i1_4.jpg", "1mp3d_0040_region55/72db3245ca954e239e8a1ae3b78726e7_i2_4.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0040_region55", "origin_question": "Remember, have you seen any plant(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any plant(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0040_region55/ceaee8b594de4d2d8c9f7625c0dbe09d_i1_5.jpg", "1mp3d_0040_region55/ceaee8b594de4d2d8c9f7625c0dbe09d_i2_5.jpg", "1mp3d_0040_region55/ceaee8b594de4d2d8c9f7625c0dbe09d_i1_3.jpg", "1mp3d_0040_region55/ceaee8b594de4d2d8c9f7625c0dbe09d_i2_3.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0040_region55", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "1mp3d_0040_region55/72db3245ca954e239e8a1ae3b78726e7_i1_0.jpg", "1mp3d_0040_region55/72db3245ca954e239e8a1ae3b78726e7_i2_0.jpg", "1mp3d_0040_region55/72db3245ca954e239e8a1ae3b78726e7_i1_5.jpg", "1mp3d_0040_region55/72db3245ca954e239e8a1ae3b78726e7_i2_5.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0040_region55", "origin_question": "When did you first discover \"the vibrant ivy plant in a brown basket\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the vibrant ivy plant in a brown basket\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0040_region55/3d871416717845ab8642e969c803fbd9_i1_1.jpg", "1mp3d_0040_region55/3d871416717845ab8642e969c803fbd9_i2_1.jpg", "1mp3d_0040_region55/1f4dbc516bac46a1a3ea1ae2c0e30f60_i1_2.jpg", "1mp3d_0040_region55/1f4dbc516bac46a1a3ea1ae2c0e30f60_i2_2.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0040_region55", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the rich brown wooden door with an arched design\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the rich brown wooden door with an arched design\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0040_region55/3d871416717845ab8642e969c803fbd9_i1_5.jpg", "1mp3d_0040_region55/3d871416717845ab8642e969c803fbd9_i2_5.jpg", "1mp3d_0040_region55/1f4dbc516bac46a1a3ea1ae2c0e30f60_i1_5.jpg", "1mp3d_0040_region55/1f4dbc516bac46a1a3ea1ae2c0e30f60_i2_5.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0040_region55", "origin_question": "", "option": [], "answer": "", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "1mp3d_0040_region55/3d871416717845ab8642e969c803fbd9_i1_4.jpg", "1mp3d_0040_region55/3d871416717845ab8642e969c803fbd9_i2_4.jpg", "1mp3d_0040_region55/3d871416717845ab8642e969c803fbd9_i1_3.jpg", "1mp3d_0040_region55/3d871416717845ab8642e969c803fbd9_i2_3.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0040_region55", "origin_question": "Which direction is the vibrant ivy plant in a brown basket in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the vibrant ivy plant in a brown basket in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0580_00/00000.jpg", "scene0580_00/00040.jpg", "scene0580_00/00080.jpg", "scene0580_00/00120.jpg", "scene0580_00/00160.jpg" ], "turn_id": 1, "scan_id": "scene0580_00", "origin_question": "Remember, have you seen any bag(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any bag(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0580_00/00200.jpg", "scene0580_00/00240.jpg", "scene0580_00/00280.jpg", "scene0580_00/00320.jpg", "scene0580_00/00360.jpg" ], "turn_id": 2, "scan_id": "scene0580_00", "origin_question": "Which two objects are on the same side of you now? \"the red bottle-shaped plastic lamp with a gold base\", \"the rectangular wooden dresser with a natural shine\", \"the table\". ", "option": [ "the red bottle-shaped plastic lamp with a gold base, the rectangular wooden dresser with a natural shine", "the red bottle-shaped plastic lamp with a gold base, the table", "the rectangular wooden dresser with a natural shine, the table" ], "answer": "the rectangular wooden dresser with a natural shine, the table", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the red bottle-shaped plastic lamp with a gold base\", \"the rectangular wooden dresser with a natural shine\", \"the table\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the red bottle-shaped plastic lamp with a gold base, the rectangular wooden dresser with a natural shine', 'the red bottle-shaped plastic lamp with a gold base, the table', 'the rectangular wooden dresser with a natural shine, the table'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0580_00/00400.jpg", "scene0580_00/00440.jpg", "scene0580_00/00480.jpg", "scene0580_00/00520.jpg", "scene0580_00/00560.jpg" ], "turn_id": 3, "scan_id": "scene0580_00", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the table\" from your position?", "option": [], "answer": "77.04", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the table\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0580_00/00600.jpg", "scene0580_00/00640.jpg", "scene0580_00/00680.jpg", "scene0580_00/00720.jpg", "scene0580_00/00760.jpg" ], "turn_id": 4, "scan_id": "scene0580_00", "origin_question": "By the end of the 3 turn, were you getting closer to or farther away from these two objects: X: \"the table\", Y: \"the round metallic door knob with a shiny surface\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to Y and farther away from X", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 3 turn, were you getting closer to or farther away from these two objects: X: \"the table\", Y: \"the round metallic door knob with a shiny surface\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0580_00/00800.jpg", "scene0580_00/00840.jpg", "scene0580_00/00880.jpg", "scene0580_00/00920.jpg", "scene0580_00/00960.jpg" ], "turn_id": 5, "scan_id": "scene0580_00", "origin_question": "What will be the first-time appearance order of the following categories: door knob, mirror, pillow?", "option": [ "mirror, pillow, door knob", "door knob, pillow, mirror", "pillow, door knob, mirror", "pillow, mirror, door knob" ], "answer": "pillow, mirror, door knob", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: door knob, mirror, pillow? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['mirror, pillow, door knob', 'door knob, pillow, mirror', 'pillow, door knob, mirror', 'pillow, mirror, door knob'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0580_00/01000.jpg", "scene0580_00/01040.jpg", "scene0580_00/01080.jpg", "scene0580_00/01120.jpg", "scene0580_00/01160.jpg" ], "turn_id": 6, "scan_id": "scene0580_00", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "155.31", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "scene0580_00/01200.jpg", "scene0580_00/01240.jpg", "scene0580_00/01280.jpg", "scene0580_00/01320.jpg", "scene0580_00/01360.jpg" ], "turn_id": 7, "scan_id": "scene0580_00", "origin_question": "Among these three objects(\"the rectangular wooden dresser with a natural shine\",\"the hat\",\"the round metallic door knob with a shiny surface\" ), which one is the closest to you?", "option": [ "the rectangular wooden dresser with a natural shine", "the hat", "the round metallic door knob with a shiny surface" ], "answer": "the rectangular wooden dresser with a natural shine", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the rectangular wooden dresser with a natural shine\",\"the hat\",\"the round metallic door knob with a shiny surface\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular wooden dresser with a natural shine', 'the hat', 'the round metallic door knob with a shiny surface'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0580_00/01400.jpg", "scene0580_00/01440.jpg", "scene0580_00/01480.jpg", "scene0580_00/01520.jpg", "scene0580_00/01560.jpg" ], "turn_id": 8, "scan_id": "scene0580_00", "origin_question": "Which two objects are on the same side of you now? \"the chair\", \"the curtain\", \"the hat\". ", "option": [ "the chair, the curtain", "the chair, the hat", "the curtain, the hat" ], "answer": "the chair, the curtain", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the chair\", \"the curtain\", \"the hat\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the chair, the curtain', 'the chair, the hat', 'the curtain, the hat'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0580_00/01600.jpg", "scene0580_00/01640.jpg", "scene0580_00/01680.jpg", "scene0580_00/01720.jpg", "scene0580_00/01760.jpg" ], "turn_id": 9, "scan_id": "scene0580_00", "origin_question": "In which turn were you closest to \"the red bottle-shaped plastic lamp with a gold base\"? ", "option": [], "answer": "7", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the red bottle-shaped plastic lamp with a gold base\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0580_00/01800.jpg", "scene0580_00/01840.jpg", "scene0580_00/01880.jpg", "scene0580_00/01920.jpg", "scene0580_00/01960.jpg" ], "turn_id": 10, "scan_id": "scene0580_00", "origin_question": "Compared to your position at the end of the 8 turn, are you now closer or farther away from \"the chair\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 8 turn, are you now closer or farther away from \"the chair\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0108_00/00000.jpg", "scene0108_00/00030.jpg", "scene0108_00/00060.jpg", "scene0108_00/00090.jpg" ], "turn_id": 1, "scan_id": "scene0108_00", "origin_question": "Remember, have you seen any table(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any table(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0108_00/00120.jpg", "scene0108_00/00150.jpg", "scene0108_00/00180.jpg", "scene0108_00/00210.jpg" ], "turn_id": 2, "scan_id": "scene0108_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0108_00/00240.jpg", "scene0108_00/00270.jpg", "scene0108_00/00300.jpg", "scene0108_00/00330.jpg" ], "turn_id": 3, "scan_id": "scene0108_00", "origin_question": "Please recall \"the round, blue plush pillow\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.7", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the round, blue plush pillow\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0108_00/00360.jpg", "scene0108_00/00390.jpg", "scene0108_00/00420.jpg", "scene0108_00/00450.jpg" ], "turn_id": 4, "scan_id": "scene0108_00", "origin_question": "Please recall \"the white ceramic cup with a handle\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.99", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the white ceramic cup with a handle\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0074_region10/96490c4da78240058d1f76a07ed95c9d_i1_4.jpg", "1mp3d_0074_region10/96490c4da78240058d1f76a07ed95c9d_i2_4.jpg", "1mp3d_0074_region10/96490c4da78240058d1f76a07ed95c9d_i2_2.jpg", "1mp3d_0074_region10/68e4c4477aa442cbbb9a9f1e54b9b831_i1_3.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0074_region10", "origin_question": "Remember, have you seen any stairs(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any stairs(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 3 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "1mp3d_0074_region10/68e4c4477aa442cbbb9a9f1e54b9b831_i2_3.jpg", "1mp3d_0074_region10/f1b191033043441987b8ebf1bb55002c_i2_1.jpg", "1mp3d_0074_region10/68e4c4477aa442cbbb9a9f1e54b9b831_i2_1.jpg", "1mp3d_0074_region10/96490c4da78240058d1f76a07ed95c9d_i2_3.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0074_region10", "origin_question": "In which turn were you closest to \"the modern staircase with sleek metal steps\"? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the modern staircase with sleek metal steps\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 3 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "1mp3d_0074_region10/68e4c4477aa442cbbb9a9f1e54b9b831_i1_5.jpg", "1mp3d_0074_region10/68e4c4477aa442cbbb9a9f1e54b9b831_i2_5.jpg", "1mp3d_0074_region10/68e4c4477aa442cbbb9a9f1e54b9b831_i1_4.jpg", "1mp3d_0074_region10/68e4c4477aa442cbbb9a9f1e54b9b831_i2_4.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0074_region10", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 3 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0139_00/00000.jpg", "scene0139_00/00030.jpg", "scene0139_00/00060.jpg", "scene0139_00/00090.jpg" ], "turn_id": 1, "scan_id": "scene0139_00", "origin_question": "Remember, have you seen any water heater(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any water heater(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0139_00/00120.jpg", "scene0139_00/00150.jpg", "scene0139_00/00180.jpg", "scene0139_00/00210.jpg" ], "turn_id": 2, "scan_id": "scene0139_00", "origin_question": "Please recall \"the white rectangular plastic soap dispenser\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.03", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the white rectangular plastic soap dispenser\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0139_00/00240.jpg", "scene0139_00/00270.jpg", "scene0139_00/00300.jpg", "scene0139_00/00330.jpg" ], "turn_id": 3, "scan_id": "scene0139_00", "origin_question": "At the end of which turn were both of \"the small black metal pipe without extra decorations\" and \"then orange plastic bottle with a blue cap\" on your right side?", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the small black metal pipe without extra decorations\" and \"then orange plastic bottle with a blue cap\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0139_00/00360.jpg", "scene0139_00/00390.jpg", "scene0139_00/00420.jpg", "scene0139_00/00450.jpg" ], "turn_id": 4, "scan_id": "scene0139_00", "origin_question": "Which direction is the small black metal pipe without extra decorations in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the small black metal pipe without extra decorations in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0139_00/00480.jpg", "scene0139_00/00510.jpg", "scene0139_00/00540.jpg", "scene0139_00/00570.jpg" ], "turn_id": 5, "scan_id": "scene0139_00", "origin_question": "Which direction is the white rectangular porcelain sink with a chrome faucet in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the white rectangular porcelain sink with a chrome faucet in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0139_00/00600.jpg", "scene0139_00/00630.jpg", "scene0139_00/00660.jpg", "scene0139_00/00690.jpg" ], "turn_id": 6, "scan_id": "scene0139_00", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0139_00/00720.jpg", "scene0139_00/00750.jpg", "scene0139_00/01180.jpg", "scene0139_00/01210.jpg" ], "turn_id": 7, "scan_id": "scene0139_00", "origin_question": "Is the simple white rectangular shelf made of wood to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the simple white rectangular shelf made of wood to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0139_00/01240.jpg", "scene0139_00/01270.jpg", "scene0139_00/01300.jpg", "scene0139_00/01330.jpg" ], "turn_id": 8, "scan_id": "scene0139_00", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the rectangular white paper with black text\" from your position?", "option": [], "answer": "52.94", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the rectangular white paper with black text\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "1mp3d_0067_region9/54db65dba94741f0875f24c433a6f12d_i1_0.jpg", "1mp3d_0067_region9/54db65dba94741f0875f24c433a6f12d_i2_0.jpg", "1mp3d_0067_region9/d16d14bd9d194980b5339c4029bc4a54_i1_5.jpg", "1mp3d_0067_region9/d16d14bd9d194980b5339c4029bc4a54_i2_5.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0067_region9", "origin_question": "Remember, how many window(s) have you seen so far? ", "option": [], "answer": "3", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many window(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "1mp3d_0067_region9/54db65dba94741f0875f24c433a6f12d_i1_2.jpg", "1mp3d_0067_region9/54db65dba94741f0875f24c433a6f12d_i2_2.jpg", "1mp3d_0067_region9/d16d14bd9d194980b5339c4029bc4a54_i1_0.jpg", "1mp3d_0067_region9/d16d14bd9d194980b5339c4029bc4a54_i2_0.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0067_region9", "origin_question": "At the end of which turn were both of \"the three-seater curved couch in warm beige fabric\" and \"the rectangular, striped pillow with brown, orange, and beige\" on your left side?", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the three-seater curved couch in warm beige fabric\" and \"the rectangular, striped pillow with brown, orange, and beige\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "1mp3d_0067_region9/8c49b7ce35b8412f85553cdf572baa2c_i1_2.jpg", "1mp3d_0067_region9/8c49b7ce35b8412f85553cdf572baa2c_i2_2.jpg", "1mp3d_0067_region9/6967b846bd9c4c7a9c9c9dfab6d5eb34_i1_0.jpg", "1mp3d_0067_region9/6967b846bd9c4c7a9c9c9dfab6d5eb34_i2_0.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0067_region9", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the fireplace\" from your position?", "option": [], "answer": "63.97", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the fireplace\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "1mp3d_0067_region9/8c49b7ce35b8412f85553cdf572baa2c_i1_1.jpg", "1mp3d_0067_region9/8c49b7ce35b8412f85553cdf572baa2c_i2_1.jpg", "1mp3d_0067_region9/8c49b7ce35b8412f85553cdf572baa2c_i1_0.jpg", "1mp3d_0067_region9/8c49b7ce35b8412f85553cdf572baa2c_i2_0.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0067_region9", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "105.21", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0067_region9/6967b846bd9c4c7a9c9c9dfab6d5eb34_i1_5.jpg", "1mp3d_0067_region9/6967b846bd9c4c7a9c9c9dfab6d5eb34_i2_5.jpg", "1mp3d_0067_region9/10b5aaa0dd094b00a74dc52823b7aa66_i1_4.jpg", "1mp3d_0067_region9/10b5aaa0dd094b00a74dc52823b7aa66_i2_4.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0067_region9", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the rectangular, striped pillow with brown, orange, and beige\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the rectangular, striped pillow with brown, orange, and beige\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "1mp3d_0067_region9/10b5aaa0dd094b00a74dc52823b7aa66_i1_5.jpg", "1mp3d_0067_region9/10b5aaa0dd094b00a74dc52823b7aa66_i2_5.jpg", "1mp3d_0067_region9/73892c22eef14d169e24e007a8e544d9_i1_3.jpg", "1mp3d_0067_region9/73892c22eef14d169e24e007a8e544d9_i2_3.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0067_region9", "origin_question": "Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Right", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "1mp3d_0067_region9/3317348467784b19925a007204cb7f17_i1_3.jpg", "1mp3d_0067_region9/3317348467784b19925a007204cb7f17_i2_3.jpg", "1mp3d_0067_region9/4382fba86a6a4956b7007f079a990268_i1_2.jpg", "1mp3d_0067_region9/4382fba86a6a4956b7007f079a990268_i2_2.jpg" ], "turn_id": 7, "scan_id": "1mp3d_0067_region9", "origin_question": "Please recall \"the rectangular, striped pillow with brown, orange, and beige\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.55", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the rectangular, striped pillow with brown, orange, and beige\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "1mp3d_0067_region9/0260ad22c1f64b88b7c7ffa843d86aa8_i1_0.jpg", "1mp3d_0067_region9/0260ad22c1f64b88b7c7ffa843d86aa8_i2_0.jpg", "1mp3d_0067_region9/caa5b0dce8fc45a8982f917d041cf441_i1_5.jpg", "1mp3d_0067_region9/caa5b0dce8fc45a8982f917d041cf441_i2_5.jpg" ], "turn_id": 8, "scan_id": "1mp3d_0067_region9", "origin_question": "Assuming the direction you are facing at the end of the turn 5 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Right", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 5 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "1mp3d_0067_region9/e394a9eecb63432b804ebb2e96a563b5_i1_0.jpg", "1mp3d_0067_region9/e394a9eecb63432b804ebb2e96a563b5_i2_0.jpg", "1mp3d_0067_region9/24a5c4d4de2e47fba4cd35a2cf63bafe_i1_5.jpg", "1mp3d_0067_region9/24a5c4d4de2e47fba4cd35a2cf63bafe_i2_5.jpg" ], "turn_id": 9, "scan_id": "1mp3d_0067_region9", "origin_question": "By the end of the 7 turn, were you getting closer to or farther away from these two objects: X: \"the three-seater curved couch in warm beige fabric\", Y: \"the rectangular, striped pillow with brown, orange, and beige\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "farther away from both X and Y", "user_message": "For the 9 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 7 turn, were you getting closer to or farther away from these two objects: X: \"the three-seater curved couch in warm beige fabric\", Y: \"the rectangular, striped pillow with brown, orange, and beige\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0557_01/00000.jpg", "scene0557_01/00040.jpg", "scene0557_01/00080.jpg", "scene0557_01/00120.jpg" ], "turn_id": 1, "scan_id": "scene0557_01", "origin_question": "Remember, have you seen any blanket(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any blanket(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0557_01/00160.jpg", "scene0557_01/00200.jpg", "scene0557_01/00240.jpg", "scene0557_01/00280.jpg" ], "turn_id": 2, "scan_id": "scene0557_01", "origin_question": "In which turn did you see both \"the couch\" and \"the rectangular, black and white zebra print ottoman\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the couch\" and \"the rectangular, black and white zebra print ottoman\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0557_01/00320.jpg", "scene0557_01/00360.jpg", "scene0557_01/00400.jpg", "scene0557_01/00440.jpg" ], "turn_id": 3, "scan_id": "scene0557_01", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the couch\"; \"the bed\"; \"the black hard shell case with a simple design\" ", "option": [ "the couch", "the bed", "the black hard shell case with a simple design" ], "answer": "the black hard shell case with a simple design", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the couch\"; \"the bed\"; \"the black hard shell case with a simple design\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the couch', 'the bed', 'the black hard shell case with a simple design'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0557_01/00480.jpg", "scene0557_01/00520.jpg", "scene0557_01/00560.jpg", "scene0557_01/00600.jpg" ], "turn_id": 4, "scan_id": "scene0557_01", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the bed\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the bed\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0557_01/00640.jpg", "scene0557_01/00680.jpg", "scene0557_01/00720.jpg", "scene0557_01/00760.jpg" ], "turn_id": 5, "scan_id": "scene0557_01", "origin_question": "In which turn were you closest to \"the dark rectangular wooden stand with two drawers\"? ", "option": [], "answer": "1", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the dark rectangular wooden stand with two drawers\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0557_01/00800.jpg", "scene0557_01/00840.jpg", "scene0557_01/00880.jpg", "scene0557_01/00920.jpg" ], "turn_id": 6, "scan_id": "scene0557_01", "origin_question": "Is the black hard shell case with a simple design to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the black hard shell case with a simple design to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0557_01/00960.jpg", "scene0557_01/01000.jpg", "scene0557_01/01040.jpg", "scene0557_01/01080.jpg" ], "turn_id": 7, "scan_id": "scene0557_01", "origin_question": "Which direction is the rectangular, black and white zebra print ottoman in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the rectangular, black and white zebra print ottoman in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0081_02/00000.jpg", "scene0081_02/00040.jpg", "scene0081_02/00080.jpg", "scene0081_02/00120.jpg" ], "turn_id": 1, "scan_id": "scene0081_02", "origin_question": "Remember, how many chair(s) have you seen so far? ", "option": [], "answer": "9", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many chair(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0081_02/00160.jpg", "scene0081_02/00200.jpg", "scene0081_02/00240.jpg", "scene0081_02/00280.jpg" ], "turn_id": 2, "scan_id": "scene0081_02", "origin_question": "Remember, how many soap dish(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many soap dish(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0081_02/00320.jpg", "scene0081_02/00360.jpg", "scene0081_02/00400.jpg", "scene0081_02/00440.jpg" ], "turn_id": 3, "scan_id": "scene0081_02", "origin_question": "Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "31.83", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0081_02/00480.jpg", "scene0081_02/00520.jpg", "scene0081_02/00560.jpg", "scene0081_02/00600.jpg" ], "turn_id": 4, "scan_id": "scene0081_02", "origin_question": "How far is your current position from where you were at the end of the turn 1? (in meters)", "option": [], "answer": "3.56", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 1? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0081_02/00640.jpg", "scene0081_02/00680.jpg", "scene0081_02/00720.jpg", "scene0081_02/00760.jpg" ], "turn_id": 5, "scan_id": "scene0081_02", "origin_question": "", "option": [], "answer": "", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0081_02/00800.jpg", "scene0081_02/00840.jpg", "scene0081_02/00880.jpg", "scene0081_02/00920.jpg" ], "turn_id": 6, "scan_id": "scene0081_02", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "2.41", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0081_02/00960.jpg", "scene0081_02/01000.jpg", "scene0081_02/01040.jpg", "scene0081_02/01080.jpg" ], "turn_id": 7, "scan_id": "scene0081_02", "origin_question": "How far is your current position from where you were at the end of the turn 4? (in meters)", "option": [], "answer": "3.78", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 4? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0047_region19/848ce72594e44d42a022ec51edda9e7d_i1_3.jpg", "1mp3d_0047_region19/848ce72594e44d42a022ec51edda9e7d_i2_3.jpg", "1mp3d_0047_region19/79f827b430cd4a4eb707895d80009b26_i1_0.jpg", "1mp3d_0047_region19/79f827b430cd4a4eb707895d80009b26_i2_0.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0047_region19", "origin_question": "Remember, have you seen any cabinet(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any cabinet(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0047_region19/6cca94cc96d94a249c664435547e600d_i1_1.jpg", "1mp3d_0047_region19/6cca94cc96d94a249c664435547e600d_i2_1.jpg", "1mp3d_0047_region19/79f827b430cd4a4eb707895d80009b26_i1_1.jpg", "1mp3d_0047_region19/79f827b430cd4a4eb707895d80009b26_i2_1.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0047_region19", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0047_region19/848ce72594e44d42a022ec51edda9e7d_i1_5.jpg", "1mp3d_0047_region19/848ce72594e44d42a022ec51edda9e7d_i2_5.jpg", "1mp3d_0047_region19/848ce72594e44d42a022ec51edda9e7d_i1_4.jpg", "1mp3d_0047_region19/848ce72594e44d42a022ec51edda9e7d_i2_4.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0047_region19", "origin_question": "", "option": [], "answer": "", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "1mp3d_0047_region19/693940a64c4f4029a3082626f995c5ff_i1_0.jpg", "1mp3d_0047_region19/693940a64c4f4029a3082626f995c5ff_i2_0.jpg", "1mp3d_0047_region19/b3fb0bde1a234d4995f90a7c5aede983_i1_1.jpg", "1mp3d_0047_region19/b3fb0bde1a234d4995f90a7c5aede983_i2_1.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0047_region19", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0047_region19/0daf49767cb645798576ac58467a7356_i1_3.jpg", "1mp3d_0047_region19/0daf49767cb645798576ac58467a7356_i2_3.jpg", "1mp3d_0047_region19/5172278505b14dbfba2e900d89ec3d3d_i1_4.jpg", "1mp3d_0047_region19/5172278505b14dbfba2e900d89ec3d3d_i2_4.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0047_region19", "origin_question": "", "option": [], "answer": "", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0047_region19/b9f3fbcc0af34a888fc547751f2e2832_i1_3.jpg", "1mp3d_0047_region19/b9f3fbcc0af34a888fc547751f2e2832_i2_3.jpg", "1mp3d_0047_region19/5172278505b14dbfba2e900d89ec3d3d_i1_3.jpg", "1mp3d_0047_region19/5172278505b14dbfba2e900d89ec3d3d_i2_3.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0047_region19", "origin_question": "", "option": [], "answer": "", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0047_region19/0daf49767cb645798576ac58467a7356_i1_5.jpg", "1mp3d_0047_region19/0daf49767cb645798576ac58467a7356_i2_5.jpg", "1mp3d_0047_region19/b3fb0bde1a234d4995f90a7c5aede983_i1_4.jpg", "1mp3d_0047_region19/b3fb0bde1a234d4995f90a7c5aede983_i2_4.jpg" ], "turn_id": 7, "scan_id": "1mp3d_0047_region19", "origin_question": "", "option": [], "answer": "", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0047_region19/693940a64c4f4029a3082626f995c5ff_i1_4.jpg", "1mp3d_0047_region19/693940a64c4f4029a3082626f995c5ff_i2_4.jpg", "1mp3d_0047_region19/cea8b9f8251c40f5a9a2c99bd4ff2a56_i1_4.jpg", "1mp3d_0047_region19/cea8b9f8251c40f5a9a2c99bd4ff2a56_i2_4.jpg" ], "turn_id": 8, "scan_id": "1mp3d_0047_region19", "origin_question": "", "option": [], "answer": "", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "1mp3d_0047_region19/8e94a8cd62924a3086424fe9d6a23d64_i1_4.jpg", "1mp3d_0047_region19/8e94a8cd62924a3086424fe9d6a23d64_i2_4.jpg", "1mp3d_0047_region19/cea8b9f8251c40f5a9a2c99bd4ff2a56_i1_3.jpg", "1mp3d_0047_region19/cea8b9f8251c40f5a9a2c99bd4ff2a56_i2_3.jpg" ], "turn_id": 9, "scan_id": "1mp3d_0047_region19", "origin_question": "Using your orientation at the end of turn 6 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "151.34", "user_message": "For the 9 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 6 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0458_01/00000.jpg", "scene0458_01/00040.jpg", "scene0458_01/00080.jpg", "scene0458_01/00120.jpg", "scene0458_01/00160.jpg" ], "turn_id": 1, "scan_id": "scene0458_01", "origin_question": "Remember, have you seen any toothbrush(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any toothbrush(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0458_01/00200.jpg", "scene0458_01/00240.jpg", "scene0458_01/00280.jpg", "scene0458_01/00320.jpg", "scene0458_01/00360.jpg" ], "turn_id": 2, "scan_id": "scene0458_01", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0458_01/00400.jpg", "scene0458_01/00440.jpg", "scene0458_01/00480.jpg", "scene0458_01/00520.jpg", "scene0458_01/00560.jpg" ], "turn_id": 3, "scan_id": "scene0458_01", "origin_question": "", "option": [], "answer": "", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0458_01/00600.jpg", "scene0458_01/00640.jpg", "scene0458_01/00680.jpg", "scene0458_01/00720.jpg", "scene0458_01/00760.jpg" ], "turn_id": 4, "scan_id": "scene0458_01", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the blue and white standard size toothbrush\" from your position?", "option": [], "answer": "93.85", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the blue and white standard size toothbrush\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0458_01/00800.jpg", "scene0458_01/00840.jpg", "scene0458_01/00880.jpg", "scene0458_01/00920.jpg", "scene0458_01/00960.jpg" ], "turn_id": 5, "scan_id": "scene0458_01", "origin_question": "Is the vanity to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the vanity to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0458_01/01000.jpg", "scene0458_01/01040.jpg", "scene0458_01/01080.jpg", "scene0458_01/01120.jpg", "scene0458_01/01160.jpg" ], "turn_id": 6, "scan_id": "scene0458_01", "origin_question": "Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the toilet paper\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the toilet paper\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0458_01/01200.jpg", "scene0458_01/01240.jpg", "scene0458_01/01280.jpg", "scene0458_01/01320.jpg", "scene0458_01/01360.jpg" ], "turn_id": 7, "scan_id": "scene0458_01", "origin_question": "What will be the first-time appearance order of the following categories: bin, toilet paper, toothbrush?", "option": [ "toilet paper, bin, toothbrush", "bin, toothbrush, toilet paper", "toilet paper, toothbrush, bin", "toothbrush, bin, toilet paper" ], "answer": "toothbrush, bin, toilet paper", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bin, toilet paper, toothbrush? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['toilet paper, bin, toothbrush', 'bin, toothbrush, toilet paper', 'toilet paper, toothbrush, bin', 'toothbrush, bin, toilet paper'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0458_01/01400.jpg", "scene0458_01/01440.jpg", "scene0458_01/01480.jpg", "scene0458_01/01520.jpg", "scene0458_01/01560.jpg" ], "turn_id": 8, "scan_id": "scene0458_01", "origin_question": "Using your orientation at the end of turn 5 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 5 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0458_01/01600.jpg", "scene0458_01/01640.jpg", "scene0458_01/01680.jpg", "scene0458_01/01720.jpg", "scene0458_01/01770.jpg" ], "turn_id": 9, "scan_id": "scene0458_01", "origin_question": "At the end of which turn were both of \"the vanity\" and \"the blue and white standard size toothbrush\" on your left side?", "option": [], "answer": "9", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the vanity\" and \"the blue and white standard size toothbrush\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0458_01/01810.jpg", "scene0458_01/01850.jpg", "scene0458_01/01890.jpg", "scene0458_01/01930.jpg", "scene0458_01/01970.jpg" ], "turn_id": 10, "scan_id": "scene0458_01", "origin_question": "Please recall \"the small rectangular gray plastic dustpan\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.5", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the small rectangular gray plastic dustpan\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0510_02/00000.jpg", "scene0510_02/00040.jpg", "scene0510_02/00080.jpg", "scene0510_02/00120.jpg" ], "turn_id": 1, "scan_id": "scene0510_02", "origin_question": "Remember, have you seen any stopcock(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any stopcock(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0510_02/00160.jpg", "scene0510_02/00200.jpg", "scene0510_02/00240.jpg", "scene0510_02/00280.jpg" ], "turn_id": 2, "scan_id": "scene0510_02", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the dark black jacket made of synthetic blend material\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the dark black jacket made of synthetic blend material\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0510_02/00320.jpg", "scene0510_02/00360.jpg", "scene0510_02/00400.jpg", "scene0510_02/00440.jpg" ], "turn_id": 3, "scan_id": "scene0510_02", "origin_question": "What will be the first-time appearance order of the following categories: bin, broom, sign?", "option": [ "broom, sign, bin", "sign, broom, bin", "sign, bin, broom", "bin, sign, broom" ], "answer": "bin, sign, broom", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bin, broom, sign? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['broom, sign, bin', 'sign, broom, bin', 'sign, bin, broom', 'bin, sign, broom'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0510_02/00480.jpg", "scene0510_02/00520.jpg", "scene0510_02/00560.jpg", "scene0510_02/00600.jpg" ], "turn_id": 4, "scan_id": "scene0510_02", "origin_question": "", "option": [], "answer": "", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0510_02/00640.jpg", "scene0510_02/00680.jpg", "scene0510_02/00720.jpg", "scene0510_02/00760.jpg" ], "turn_id": 5, "scan_id": "scene0510_02", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the toilet\"; \"the circular metallic shower head with a stainless steel finish\"; \"the bucket\" ", "option": [ "the toilet", "the circular metallic shower head with a stainless steel finish", "the bucket" ], "answer": "the toilet", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the toilet\"; \"the circular metallic shower head with a stainless steel finish\"; \"the bucket\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the toilet', 'the circular metallic shower head with a stainless steel finish', 'the bucket'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0510_02/00800.jpg", "scene0510_02/00840.jpg", "scene0510_02/00880.jpg", "scene0510_02/00920.jpg" ], "turn_id": 6, "scan_id": "scene0510_02", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the toilet\"; \"the stall\"; \"the rectangular metallic holder made of stainless steel\" ", "option": [ "the toilet", "the stall", "the rectangular metallic holder made of stainless steel" ], "answer": "the stall", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the toilet\"; \"the stall\"; \"the rectangular metallic holder made of stainless steel\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the toilet', 'the stall', 'the rectangular metallic holder made of stainless steel'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0510_02/00960.jpg", "scene0510_02/01000.jpg", "scene0510_02/01040.jpg", "scene0510_02/01080.jpg" ], "turn_id": 7, "scan_id": "scene0510_02", "origin_question": "How far is your current position from where you were at the end of the turn 4? (in meters)", "option": [], "answer": "1.9", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 4? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0692_01/00000.jpg", "scene0692_01/00040.jpg", "scene0692_01/00080.jpg", "scene0692_01/00120.jpg" ], "turn_id": 1, "scan_id": "scene0692_01", "origin_question": "Remember, have you seen any book(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any book(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0692_01/00160.jpg", "scene0692_01/00200.jpg", "scene0692_01/00240.jpg", "scene0692_01/00280.jpg" ], "turn_id": 2, "scan_id": "scene0692_01", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0692_01/00320.jpg", "scene0692_01/00360.jpg", "scene0692_01/00400.jpg", "scene0692_01/00440.jpg" ], "turn_id": 3, "scan_id": "scene0692_01", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the standard lamp with a white lampshade\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the standard lamp with a white lampshade\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0692_01/00480.jpg", "scene0692_01/00520.jpg", "scene0692_01/00560.jpg", "scene0692_01/00600.jpg" ], "turn_id": 4, "scan_id": "scene0692_01", "origin_question": "Please recall \"the standard lamp with a white lampshade\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.69", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the standard lamp with a white lampshade\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0692_01/00640.jpg", "scene0692_01/00680.jpg", "scene0692_01/00720.jpg", "scene0692_01/00760.jpg" ], "turn_id": 5, "scan_id": "scene0692_01", "origin_question": "Using your orientation at the end of turn 4 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "60.06", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 4 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0692_01/00800.jpg", "scene0692_01/00840.jpg", "scene0692_01/00880.jpg", "scene0692_01/00920.jpg" ], "turn_id": 6, "scan_id": "scene0692_01", "origin_question": "Which two objects are on the same side of you now? \"the rectangular light brown wooden door\", \"the tall wooden cabinet with glass doors\", \"the standard lamp with a white lampshade\". ", "option": [ "the rectangular light brown wooden door, the tall wooden cabinet with glass doors", "the rectangular light brown wooden door, the standard lamp with a white lampshade", "the tall wooden cabinet with glass doors, the standard lamp with a white lampshade" ], "answer": "the rectangular light brown wooden door, the tall wooden cabinet with glass doors", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the rectangular light brown wooden door\", \"the tall wooden cabinet with glass doors\", \"the standard lamp with a white lampshade\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular light brown wooden door, the tall wooden cabinet with glass doors', 'the rectangular light brown wooden door, the standard lamp with a white lampshade', 'the tall wooden cabinet with glass doors, the standard lamp with a white lampshade'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0579_02/00000.jpg", "scene0579_02/00030.jpg", "scene0579_02/00060.jpg", "scene0579_02/00090.jpg" ], "turn_id": 1, "scan_id": "scene0579_02", "origin_question": "Remember, how many faucet(s) have you seen so far? ", "option": [], "answer": "2", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many faucet(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0579_02/00120.jpg", "scene0579_02/00150.jpg", "scene0579_02/00180.jpg", "scene0579_02/00210.jpg" ], "turn_id": 2, "scan_id": "scene0579_02", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0579_02/00240.jpg", "scene0579_02/00270.jpg", "scene0579_02/00300.jpg", "scene0579_02/00330.jpg" ], "turn_id": 3, "scan_id": "scene0579_02", "origin_question": "Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "122.52", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0579_02/00360.jpg", "scene0579_02/00390.jpg", "scene0579_02/00420.jpg", "scene0579_02/00450.jpg" ], "turn_id": 4, "scan_id": "scene0579_02", "origin_question": "Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "94.2", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0579_02/00480.jpg", "scene0579_02/00510.jpg", "scene0579_02/00540.jpg", "scene0579_02/00570.jpg" ], "turn_id": 5, "scan_id": "scene0579_02", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0077_01/00000.jpg", "scene0077_01/00030.jpg", "scene0077_01/00060.jpg", "scene0077_01/00090.jpg" ], "turn_id": 1, "scan_id": "scene0077_01", "origin_question": "Remember, have you seen any paper cutter(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any paper cutter(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0077_01/00120.jpg", "scene0077_01/00150.jpg", "scene0077_01/00180.jpg", "scene0077_01/00210.jpg" ], "turn_id": 2, "scan_id": "scene0077_01", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the counter\"; \"the sleek black rectangular plastic remote control\"; \"the sturdy rectangular deep brown wooden table\" ", "option": [ "the counter", "the sleek black rectangular plastic remote control", "the sturdy rectangular deep brown wooden table" ], "answer": "the sleek black rectangular plastic remote control", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the counter\"; \"the sleek black rectangular plastic remote control\"; \"the sturdy rectangular deep brown wooden table\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the counter', 'the sleek black rectangular plastic remote control', 'the sturdy rectangular deep brown wooden table'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0077_01/00240.jpg", "scene0077_01/00270.jpg", "scene0077_01/00300.jpg", "scene0077_01/00330.jpg" ], "turn_id": 3, "scan_id": "scene0077_01", "origin_question": "Is the computer to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the computer to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0077_01/00360.jpg", "scene0077_01/00390.jpg", "scene0077_01/00420.jpg", "scene0077_01/00450.jpg" ], "turn_id": 4, "scan_id": "scene0077_01", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the rectangular blue picture on paper or canvas\"; \"the counter\"; \"the shelf\" ", "option": [ "the rectangular blue picture on paper or canvas", "the counter", "the shelf" ], "answer": "the rectangular blue picture on paper or canvas", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the rectangular blue picture on paper or canvas\"; \"the counter\"; \"the shelf\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular blue picture on paper or canvas', 'the counter', 'the shelf'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0077_01/00480.jpg", "scene0077_01/00510.jpg", "scene0077_01/00540.jpg", "scene0077_01/00570.jpg" ], "turn_id": 5, "scan_id": "scene0077_01", "origin_question": "Please recall \"the shelf\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.08", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the shelf\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0077_01/00600.jpg", "scene0077_01/00630.jpg", "scene0077_01/00660.jpg", "scene0077_01/00690.jpg" ], "turn_id": 6, "scan_id": "scene0077_01", "origin_question": "Is the sleek black rectangular plastic remote control to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the sleek black rectangular plastic remote control to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0077_01/00720.jpg", "scene0077_01/00750.jpg", "scene0077_01/00780.jpg", "scene0077_01/00810.jpg" ], "turn_id": 7, "scan_id": "scene0077_01", "origin_question": "By the end of the 3 turn, were you getting closer to or farther away from these two objects: X: \"the sleek black rectangular plastic remote control\", Y: \"the large rectangular light brown cardboard box\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to Y and farther away from X", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 3 turn, were you getting closer to or farther away from these two objects: X: \"the sleek black rectangular plastic remote control\", Y: \"the large rectangular light brown cardboard box\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0047_region10/1d2e677ec3d54f62b50958b31a1c7864_i1_4.jpg", "1mp3d_0047_region10/1d2e677ec3d54f62b50958b31a1c7864_i2_4.jpg", "1mp3d_0047_region10/1d2e677ec3d54f62b50958b31a1c7864_i1_3.jpg", "1mp3d_0047_region10/1d2e677ec3d54f62b50958b31a1c7864_i2_3.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0047_region10", "origin_question": "Remember, have you seen any bed(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any bed(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "1mp3d_0047_region10/a12a4082b6c6428695e8cb00ec69a001_i1_4.jpg", "1mp3d_0047_region10/a12a4082b6c6428695e8cb00ec69a001_i2_4.jpg", "1mp3d_0047_region10/b9f3fbcc0af34a888fc547751f2e2832_i1_4.jpg", "1mp3d_0047_region10/b9f3fbcc0af34a888fc547751f2e2832_i2_4.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0047_region10", "origin_question": "In which turn did you see both \"the simple wooden bed with a dark blue blanket\" and \"the modern, shiny metal refrigerator\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the simple wooden bed with a dark blue blanket\" and \"the modern, shiny metal refrigerator\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "1mp3d_0047_region10/b9f3fbcc0af34a888fc547751f2e2832_i1_5.jpg", "1mp3d_0047_region10/b9f3fbcc0af34a888fc547751f2e2832_i2_5.jpg", "1mp3d_0047_region10/a12a4082b6c6428695e8cb00ec69a001_i1_5.jpg", "1mp3d_0047_region10/a12a4082b6c6428695e8cb00ec69a001_i2_5.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0047_region10", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the simple wooden bed with a dark blue blanket\"; \"the tall, dark-colored modern cabinet with multiple doors\"; \"the bright orange cylindrical pillow made of soft fabric\" ", "option": [ "the simple wooden bed with a dark blue blanket", "the tall, dark-colored modern cabinet with multiple doors", "the bright orange cylindrical pillow made of soft fabric" ], "answer": "the tall, dark-colored modern cabinet with multiple doors", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the simple wooden bed with a dark blue blanket\"; \"the tall, dark-colored modern cabinet with multiple doors\"; \"the bright orange cylindrical pillow made of soft fabric\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the simple wooden bed with a dark blue blanket', 'the tall, dark-colored modern cabinet with multiple doors', 'the bright orange cylindrical pillow made of soft fabric'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "1mp3d_0047_region10/1d2e677ec3d54f62b50958b31a1c7864_i1_0.jpg", "1mp3d_0047_region10/1d2e677ec3d54f62b50958b31a1c7864_i2_0.jpg", "1mp3d_0047_region10/1d2e677ec3d54f62b50958b31a1c7864_i1_1.jpg", "1mp3d_0047_region10/1d2e677ec3d54f62b50958b31a1c7864_i2_1.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0047_region10", "origin_question": "Remember, how many scale(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many scale(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0047_region10/1d2e677ec3d54f62b50958b31a1c7864_i1_2.jpg", "1mp3d_0047_region10/1d2e677ec3d54f62b50958b31a1c7864_i2_2.jpg", "1mp3d_0047_region10/a12a4082b6c6428695e8cb00ec69a001_i1_2.jpg", "1mp3d_0047_region10/a12a4082b6c6428695e8cb00ec69a001_i2_2.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0047_region10", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the modern, shiny metal refrigerator\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the modern, shiny metal refrigerator\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "1mp3d_0047_region10/a12a4082b6c6428695e8cb00ec69a001_i1_3.jpg", "1mp3d_0047_region10/a12a4082b6c6428695e8cb00ec69a001_i2_3.jpg", "1mp3d_0047_region10/b9f3fbcc0af34a888fc547751f2e2832_i1_0.jpg", "1mp3d_0047_region10/b9f3fbcc0af34a888fc547751f2e2832_i2_0.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0047_region10", "origin_question": "Among these three objects(\"the simple wooden bed with a dark blue blanket\",\"the bright orange cylindrical pillow made of soft fabric\",\"the tall, dark-colored modern cabinet with multiple doors\" ), which one is the closest to you?", "option": [ "the simple wooden bed with a dark blue blanket", "the bright orange cylindrical pillow made of soft fabric", "the tall, dark-colored modern cabinet with multiple doors" ], "answer": "the simple wooden bed with a dark blue blanket", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects(\"the simple wooden bed with a dark blue blanket\",\"the bright orange cylindrical pillow made of soft fabric\",\"the tall, dark-colored modern cabinet with multiple doors\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the simple wooden bed with a dark blue blanket', 'the bright orange cylindrical pillow made of soft fabric', 'the tall, dark-colored modern cabinet with multiple doors'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0047_region3/5eb5aa410e654d61900f3d22ad4efa40_i1_0.jpg", "1mp3d_0047_region3/5eb5aa410e654d61900f3d22ad4efa40_i2_0.jpg", "1mp3d_0047_region3/12a12badbdf44335b299d7656bac26b8_i1_5.jpg", "1mp3d_0047_region3/12a12badbdf44335b299d7656bac26b8_i2_5.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0047_region3", "origin_question": "Remember, have you seen any sink(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any sink(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "1mp3d_0047_region3/6ccae1ff9e0149bdb2e21cf2c734d03a_i1_5.jpg", "1mp3d_0047_region3/6ccae1ff9e0149bdb2e21cf2c734d03a_i2_5.jpg", "1mp3d_0047_region3/6ccae1ff9e0149bdb2e21cf2c734d03a_i1_0.jpg", "1mp3d_0047_region3/6ccae1ff9e0149bdb2e21cf2c734d03a_i2_0.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0047_region3", "origin_question": "When did you first discover \"the sleek arch-shaped table with a shiny black surface\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the sleek arch-shaped table with a shiny black surface\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0047_region3/6ccae1ff9e0149bdb2e21cf2c734d03a_i1_1.jpg", "1mp3d_0047_region3/6ccae1ff9e0149bdb2e21cf2c734d03a_i2_1.jpg", "1mp3d_0047_region3/12a12badbdf44335b299d7656bac26b8_i1_1.jpg", "1mp3d_0047_region3/12a12badbdf44335b299d7656bac26b8_i2_1.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0047_region3", "origin_question": "", "option": [], "answer": "", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "1mp3d_0047_region3/5eb5aa410e654d61900f3d22ad4efa40_i1_2.jpg", "1mp3d_0047_region3/5eb5aa410e654d61900f3d22ad4efa40_i2_2.jpg", "1mp3d_0047_region3/5eb5aa410e654d61900f3d22ad4efa40_i1_1.jpg", "1mp3d_0047_region3/5eb5aa410e654d61900f3d22ad4efa40_i2_1.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0047_region3", "origin_question": "In which turn did you see both \"the sleek arch-shaped table with a shiny black surface\" and \"the black metal shelf of standard size\" simultaneously? ", "option": [], "answer": "4", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the sleek arch-shaped table with a shiny black surface\" and \"the black metal shelf of standard size\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "1mp3d_0047_region3/6a5c5a1b3ae8449cae3a0f784a7a5a7a_i1_3.jpg", "1mp3d_0047_region3/6a5c5a1b3ae8449cae3a0f784a7a5a7a_i2_3.jpg", "1mp3d_0047_region3/6a5c5a1b3ae8449cae3a0f784a7a5a7a_i1_2.jpg", "1mp3d_0047_region3/6a5c5a1b3ae8449cae3a0f784a7a5a7a_i2_2.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0047_region3", "origin_question": "Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "73.75", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "1mp3d_0047_region3/6a5c5a1b3ae8449cae3a0f784a7a5a7a_i1_1.jpg", "1mp3d_0047_region3/6a5c5a1b3ae8449cae3a0f784a7a5a7a_i2_1.jpg", "1mp3d_0047_region3/6a5c5a1b3ae8449cae3a0f784a7a5a7a_i1_0.jpg", "1mp3d_0047_region3/6a5c5a1b3ae8449cae3a0f784a7a5a7a_i2_0.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0047_region3", "origin_question": "Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Right", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "1mp3d_0047_region3/6028b7e2036c448999b5bd48d16a4175_i1_0.jpg", "1mp3d_0047_region3/6028b7e2036c448999b5bd48d16a4175_i2_0.jpg", "1mp3d_0047_region3/3b24477fb2e84338b1c1b2d2ac085aa3_i1_2.jpg", "1mp3d_0047_region3/3b24477fb2e84338b1c1b2d2ac085aa3_i2_2.jpg" ], "turn_id": 7, "scan_id": "1mp3d_0047_region3", "origin_question": "Using your orientation at the end of turn 6 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "71.4", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 6 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0047_region3/6028b7e2036c448999b5bd48d16a4175_i1_5.jpg", "1mp3d_0047_region3/6028b7e2036c448999b5bd48d16a4175_i2_5.jpg", "1mp3d_0047_region3/6028b7e2036c448999b5bd48d16a4175_i1_4.jpg", "1mp3d_0047_region3/6028b7e2036c448999b5bd48d16a4175_i2_4.jpg" ], "turn_id": 8, "scan_id": "1mp3d_0047_region3", "origin_question": "Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the rectangular red counter with a glossy finish\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the rectangular red counter with a glossy finish\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "1mp3d_0047_region3/7520b36bc71b43ea80d10f921f8724b3_i1_3.jpg", "1mp3d_0047_region3/7520b36bc71b43ea80d10f921f8724b3_i2_3.jpg", "1mp3d_0047_region3/7520b36bc71b43ea80d10f921f8724b3_i1_2.jpg", "1mp3d_0047_region3/7520b36bc71b43ea80d10f921f8724b3_i2_2.jpg" ], "turn_id": 9, "scan_id": "1mp3d_0047_region3", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular red counter with a glossy finish\", Y: \"the black metal shelf of standard size\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "farther away from both X and Y", "user_message": "For the 9 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular red counter with a glossy finish\", Y: \"the black metal shelf of standard size\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0428_00/00000.jpg", "scene0428_00/00040.jpg", "scene0428_00/00080.jpg", "scene0428_00/00130.jpg" ], "turn_id": 1, "scan_id": "scene0428_00", "origin_question": "Remember, have you seen any wood(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any wood(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0428_00/00170.jpg", "scene0428_00/00210.jpg", "scene0428_00/00250.jpg", "scene0428_00/00290.jpg" ], "turn_id": 2, "scan_id": "scene0428_00", "origin_question": "When did you first discover \"the rectangular wooden-framed mirror made of reflective glass\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the rectangular wooden-framed mirror made of reflective glass\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0428_00/00330.jpg", "scene0428_00/00370.jpg", "scene0428_00/00410.jpg", "scene0428_00/00450.jpg" ], "turn_id": 3, "scan_id": "scene0428_00", "origin_question": "In which turn did you see both \"the rectangular wooden-framed mirror made of reflective glass\" and \"the recumbent stationary bike with a gray-silver frame\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the rectangular wooden-framed mirror made of reflective glass\" and \"the recumbent stationary bike with a gray-silver frame\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0428_00/00490.jpg", "scene0428_00/00530.jpg", "scene0428_00/00570.jpg", "scene0428_00/00610.jpg" ], "turn_id": 4, "scan_id": "scene0428_00", "origin_question": "Remember, have you seen any folder(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any folder(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0428_00/00650.jpg", "scene0428_00/00690.jpg", "scene0428_00/00730.jpg", "scene0428_00/00770.jpg" ], "turn_id": 5, "scan_id": "scene0428_00", "origin_question": "Remember, have you seen any radiator(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any radiator(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0428_00/00810.jpg", "scene0428_00/00850.jpg", "scene0428_00/00890.jpg", "scene0428_00/00930.jpg" ], "turn_id": 6, "scan_id": "scene0428_00", "origin_question": "Using your orientation at the end of turn 5 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 5 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0428_00/00970.jpg", "scene0428_00/01010.jpg", "scene0428_00/01050.jpg", "scene0428_00/01140.jpg" ], "turn_id": 7, "scan_id": "scene0428_00", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the recumbent stationary bike with a gray-silver frame\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the recumbent stationary bike with a gray-silver frame\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0428_00/01180.jpg", "scene0428_00/01220.jpg", "scene0428_00/01260.jpg", "scene0428_00/01300.jpg" ], "turn_id": 8, "scan_id": "scene0428_00", "origin_question": "By the end of the 5 turn, were you getting closer to or farther away from these two objects: X: \"the recumbent stationary bike with a gray-silver frame\", Y: \"the medium flat-panel TV with a sleek dark matte finish\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to X and farther away from Y", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 5 turn, were you getting closer to or farther away from these two objects: X: \"the recumbent stationary bike with a gray-silver frame\", Y: \"the medium flat-panel TV with a sleek dark matte finish\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0428_00/01340.jpg", "scene0428_00/01380.jpg", "scene0428_00/01420.jpg", "scene0428_00/01460.jpg" ], "turn_id": 9, "scan_id": "scene0428_00", "origin_question": "Remember, how many kettle(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 9 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many kettle(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0030_02/00000.jpg", "scene0030_02/00040.jpg", "scene0030_02/00080.jpg", "scene0030_02/00120.jpg", "scene0030_02/00160.jpg" ], "turn_id": 1, "scan_id": "scene0030_02", "origin_question": "Remember, have you seen any paper(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any paper(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0030_02/00200.jpg", "scene0030_02/00240.jpg", "scene0030_02/00280.jpg", "scene0030_02/00320.jpg", "scene0030_02/00360.jpg" ], "turn_id": 2, "scan_id": "scene0030_02", "origin_question": "In which turn did you see both \"the round silver clock with a white face\" and \"the window\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the round silver clock with a white face\" and \"the window\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0030_02/00400.jpg", "scene0030_02/00440.jpg", "scene0030_02/00480.jpg", "scene0030_02/00520.jpg", "scene0030_02/00560.jpg" ], "turn_id": 3, "scan_id": "scene0030_02", "origin_question": "In which turn did you see both \"the rectangular brown paper bag with red logos\" and \"the small rectangular wooden tray with greenish shine\" simultaneously? ", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the rectangular brown paper bag with red logos\" and \"the small rectangular wooden tray with greenish shine\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0030_02/00600.jpg", "scene0030_02/00640.jpg", "scene0030_02/00680.jpg", "scene0030_02/00720.jpg", "scene0030_02/00760.jpg" ], "turn_id": 4, "scan_id": "scene0030_02", "origin_question": "Which two objects are on the same side of you now? \"the round silver clock with a white face\", \"the windowsill\", \"the small rectangular wooden tray with greenish shine\". ", "option": [ "the round silver clock with a white face, the windowsill", "the round silver clock with a white face, the small rectangular wooden tray with greenish shine", "the windowsill, the small rectangular wooden tray with greenish shine" ], "answer": "the windowsill, the small rectangular wooden tray with greenish shine", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the round silver clock with a white face\", \"the windowsill\", \"the small rectangular wooden tray with greenish shine\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the round silver clock with a white face, the windowsill', 'the round silver clock with a white face, the small rectangular wooden tray with greenish shine', 'the windowsill, the small rectangular wooden tray with greenish shine'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0030_02/00800.jpg", "scene0030_02/00840.jpg", "scene0030_02/00880.jpg", "scene0030_02/00920.jpg", "scene0030_02/00960.jpg" ], "turn_id": 5, "scan_id": "scene0030_02", "origin_question": "By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular brown paper bag with red logos\", Y: \"the small rectangular wooden tray with greenish shine\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "farther away from both X and Y", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular brown paper bag with red logos\", Y: \"the small rectangular wooden tray with greenish shine\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0030_02/01000.jpg", "scene0030_02/01040.jpg", "scene0030_02/01080.jpg", "scene0030_02/01120.jpg", "scene0030_02/01160.jpg" ], "turn_id": 6, "scan_id": "scene0030_02", "origin_question": "Which direction is the small rectangular wooden tray with greenish shine in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the small rectangular wooden tray with greenish shine in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0030_02/01200.jpg", "scene0030_02/01240.jpg", "scene0030_02/01280.jpg", "scene0030_02/01320.jpg", "scene0030_02/01360.jpg" ], "turn_id": 7, "scan_id": "scene0030_02", "origin_question": "Please recall \"the small rectangular wooden tray with greenish shine\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "4.34", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the small rectangular wooden tray with greenish shine\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0030_02/01400.jpg", "scene0030_02/01440.jpg", "scene0030_02/01480.jpg", "scene0030_02/01520.jpg", "scene0030_02/01560.jpg" ], "turn_id": 8, "scan_id": "scene0030_02", "origin_question": "Which two objects are on the same side of you now? \"the windowsill\", \"the small rectangular wooden tray with greenish shine\", \"the red cylindrical jar\". ", "option": [ "the windowsill, the small rectangular wooden tray with greenish shine", "the windowsill, the red cylindrical jar", "the small rectangular wooden tray with greenish shine, the red cylindrical jar" ], "answer": "the windowsill, the small rectangular wooden tray with greenish shine", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the windowsill\", \"the small rectangular wooden tray with greenish shine\", \"the red cylindrical jar\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the windowsill, the small rectangular wooden tray with greenish shine', 'the windowsill, the red cylindrical jar', 'the small rectangular wooden tray with greenish shine, the red cylindrical jar'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0030_02/01600.jpg", "scene0030_02/01640.jpg", "scene0030_02/01680.jpg", "scene0030_02/01720.jpg", "scene0030_02/01760.jpg" ], "turn_id": 9, "scan_id": "scene0030_02", "origin_question": "By the end of the 4 turn, were you getting closer to or farther away from these two objects: X: \"the windowsill\", Y: \"the red cylindrical jar\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to Y and farther away from X", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 4 turn, were you getting closer to or farther away from these two objects: X: \"the windowsill\", Y: \"the red cylindrical jar\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0690_01/00000.jpg", "scene0690_01/00030.jpg", "scene0690_01/00060.jpg", "scene0690_01/00090.jpg" ], "turn_id": 1, "scan_id": "scene0690_01", "origin_question": "Remember, have you seen any lamp(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any lamp(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0690_01/00120.jpg", "scene0690_01/00150.jpg", "scene0690_01/00180.jpg", "scene0690_01/00210.jpg" ], "turn_id": 2, "scan_id": "scene0690_01", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0690_01/00240.jpg", "scene0690_01/00270.jpg", "scene0690_01/00300.jpg", "scene0690_01/00330.jpg" ], "turn_id": 3, "scan_id": "scene0690_01", "origin_question": "When did you first discover \"the rectangular light brown wooden cabinet\"? (index of the turn)", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the rectangular light brown wooden cabinet\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0690_01/00360.jpg", "scene0690_01/00390.jpg", "scene0690_01/00420.jpg", "scene0690_01/00450.jpg" ], "turn_id": 4, "scan_id": "scene0690_01", "origin_question": "What will be the first-time appearance order of the following categories: chair, paper, table?", "option": [ "chair, paper, table", "chair, table, paper", "paper, table, chair", "paper, chair, table" ], "answer": "chair, table, paper", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: chair, paper, table? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['chair, paper, table', 'chair, table, paper', 'paper, table, chair', 'paper, chair, table'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0690_01/00480.jpg", "scene0690_01/00510.jpg", "scene0690_01/00540.jpg", "scene0690_01/00570.jpg" ], "turn_id": 5, "scan_id": "scene0690_01", "origin_question": "Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0690_01/00600.jpg", "scene0690_01/00630.jpg", "scene0690_01/00660.jpg", "scene0690_01/00690.jpg" ], "turn_id": 6, "scan_id": "scene0690_01", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "64.62", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0134_01/00000.jpg", "scene0134_01/00040.jpg", "scene0134_01/00080.jpg", "scene0134_01/00120.jpg" ], "turn_id": 1, "scan_id": "scene0134_01", "origin_question": "Remember, how many table(s) have you seen so far? ", "option": [], "answer": "2", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many table(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "scene0134_01/00160.jpg", "scene0134_01/00200.jpg", "scene0134_01/00240.jpg", "scene0134_01/00280.jpg" ], "turn_id": 2, "scan_id": "scene0134_01", "origin_question": "Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance forward or backward from that position?", "option": [ "Forward", "Backward" ], "answer": "Backward", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance forward or backward from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Forward', 'Backward'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0134_01/00320.jpg", "scene0134_01/00360.jpg", "scene0134_01/00400.jpg", "scene0134_01/00440.jpg" ], "turn_id": 3, "scan_id": "scene0134_01", "origin_question": "How far is your current position from where you were at the end of the turn 1? (in meters)", "option": [], "answer": "1.63", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 1? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0134_01/00480.jpg", "scene0134_01/00520.jpg", "scene0134_01/00560.jpg", "scene0134_01/00600.jpg" ], "turn_id": 4, "scan_id": "scene0134_01", "origin_question": "Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "22.04", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0134_01/00640.jpg", "scene0134_01/00680.jpg", "scene0134_01/00720.jpg", "scene0134_01/00760.jpg" ], "turn_id": 5, "scan_id": "scene0134_01", "origin_question": "Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0134_01/00800.jpg", "scene0134_01/00840.jpg", "scene0134_01/00880.jpg", "scene0134_01/00920.jpg" ], "turn_id": 6, "scan_id": "scene0134_01", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "1.54", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0134_01/00960.jpg", "scene0134_01/01000.jpg", "scene0134_01/01040.jpg", "scene0134_01/01080.jpg" ], "turn_id": 7, "scan_id": "scene0134_01", "origin_question": "In which turn were you closest to \"the window\"? ", "option": [], "answer": "5", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the window\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0612_01/00000.jpg", "scene0612_01/00030.jpg", "scene0612_01/00060.jpg", "scene0612_01/00090.jpg" ], "turn_id": 1, "scan_id": "scene0612_01", "origin_question": "Remember, have you seen any bin(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any bin(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0612_01/00120.jpg", "scene0612_01/00150.jpg", "scene0612_01/00180.jpg", "scene0612_01/00210.jpg" ], "turn_id": 2, "scan_id": "scene0612_01", "origin_question": "When did you first discover \"the rectangular stainless steel dispenser\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the rectangular stainless steel dispenser\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0612_01/00240.jpg", "scene0612_01/00270.jpg", "scene0612_01/00300.jpg", "scene0612_01/00330.jpg" ], "turn_id": 3, "scan_id": "scene0612_01", "origin_question": "When did you first discover \"the circular white ceramic sink\"? (index of the turn)", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the circular white ceramic sink\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0612_01/00360.jpg", "scene0612_01/00390.jpg", "scene0612_01/00420.jpg", "scene0612_01/00450.jpg" ], "turn_id": 4, "scan_id": "scene0612_01", "origin_question": "What will be the first-time appearance order of the following categories: door, sink, toilet paper?", "option": [ "toilet paper, door, sink", "door, toilet paper, sink", "door, sink, toilet paper", "toilet paper, sink, door" ], "answer": "toilet paper, door, sink", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: door, sink, toilet paper? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['toilet paper, door, sink', 'door, toilet paper, sink', 'door, sink, toilet paper', 'toilet paper, sink, door'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0612_01/00480.jpg", "scene0612_01/00510.jpg", "scene0612_01/00540.jpg", "scene0612_01/00570.jpg" ], "turn_id": 5, "scan_id": "scene0612_01", "origin_question": "", "option": [], "answer": "", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0612_01/00600.jpg", "scene0612_01/00630.jpg", "scene0612_01/00660.jpg", "scene0612_01/00690.jpg" ], "turn_id": 6, "scan_id": "scene0612_01", "origin_question": "Which direction is the long silver metal bar securely fixed to the wall in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the long silver metal bar securely fixed to the wall in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0612_01/00720.jpg", "scene0612_01/00750.jpg", "scene0612_01/00780.jpg", "scene0612_01/00810.jpg" ], "turn_id": 7, "scan_id": "scene0612_01", "origin_question": "Remember, how many door(s) have you seen so far? ", "option": [], "answer": "1", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many door(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0011_00/00000.jpg", "scene0011_00/00040.jpg", "scene0011_00/00080.jpg", "scene0011_00/00120.jpg", "scene0011_00/00160.jpg" ], "turn_id": 1, "scan_id": "scene0011_00", "origin_question": "Remember, have you seen any blanket(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any blanket(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0011_00/00200.jpg", "scene0011_00/00240.jpg", "scene0011_00/00280.jpg", "scene0011_00/00320.jpg", "scene0011_00/00360.jpg" ], "turn_id": 2, "scan_id": "scene0011_00", "origin_question": "Remember, have you seen any counter(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any counter(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0011_00/00400.jpg", "scene0011_00/00440.jpg", "scene0011_00/00480.jpg", "scene0011_00/00520.jpg", "scene0011_00/00560.jpg" ], "turn_id": 3, "scan_id": "scene0011_00", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the large white granite counter\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the large white granite counter\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0011_00/00600.jpg", "scene0011_00/00640.jpg", "scene0011_00/00680.jpg", "scene0011_00/00720.jpg", "scene0011_00/00760.jpg" ], "turn_id": 4, "scan_id": "scene0011_00", "origin_question": "What will be the first-time appearance order of the following categories: oven, refrigerator, window?", "option": [ "window, oven, refrigerator", "oven, refrigerator, window", "oven, window, refrigerator", "window, refrigerator, oven" ], "answer": "window, refrigerator, oven", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: oven, refrigerator, window? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['window, oven, refrigerator', 'oven, refrigerator, window', 'oven, window, refrigerator', 'window, refrigerator, oven'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0011_00/00800.jpg", "scene0011_00/00840.jpg", "scene0011_00/00880.jpg", "scene0011_00/00920.jpg", "scene0011_00/00960.jpg" ], "turn_id": 5, "scan_id": "scene0011_00", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "57.29", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0011_00/01000.jpg", "scene0011_00/01040.jpg", "scene0011_00/01080.jpg", "scene0011_00/01120.jpg", "scene0011_00/01160.jpg" ], "turn_id": 6, "scan_id": "scene0011_00", "origin_question": "At the end of which turn were both of \"the rectangular silver stainless steel sink\" and \"the bin\" on your right side?", "option": [], "answer": "2", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the rectangular silver stainless steel sink\" and \"the bin\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0011_00/01200.jpg", "scene0011_00/01240.jpg", "scene0011_00/01280.jpg", "scene0011_00/01320.jpg", "scene0011_00/01360.jpg" ], "turn_id": 7, "scan_id": "scene0011_00", "origin_question": "Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the rectangular silver stainless steel sink\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the rectangular silver stainless steel sink\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0011_00/01400.jpg", "scene0011_00/01440.jpg", "scene0011_00/01480.jpg", "scene0011_00/01520.jpg", "scene0011_00/01560.jpg" ], "turn_id": 8, "scan_id": "scene0011_00", "origin_question": "Please recall \"the large flat-screen TV with a black frame\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.36", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the large flat-screen TV with a black frame\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0011_00/01600.jpg", "scene0011_00/01640.jpg", "scene0011_00/01680.jpg", "scene0011_00/01720.jpg", "scene0011_00/01760.jpg" ], "turn_id": 9, "scan_id": "scene0011_00", "origin_question": "Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the rectangular silver stainless steel sink\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the rectangular silver stainless steel sink\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0011_00/01800.jpg", "scene0011_00/01840.jpg", "scene0011_00/01880.jpg", "scene0011_00/01920.jpg", "scene0011_00/01960.jpg" ], "turn_id": 10, "scan_id": "scene0011_00", "origin_question": "How far is your current position from where you were at the end of the turn 6? (in meters)", "option": [], "answer": "3.65", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 6? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0664_02/00000.jpg", "scene0664_02/00040.jpg", "scene0664_02/00080.jpg", "scene0664_02/00120.jpg" ], "turn_id": 1, "scan_id": "scene0664_02", "origin_question": "Remember, have you seen any bin(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any bin(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0664_02/00160.jpg", "scene0664_02/00200.jpg", "scene0664_02/00240.jpg", "scene0664_02/00280.jpg" ], "turn_id": 2, "scan_id": "scene0664_02", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0664_02/00320.jpg", "scene0664_02/00360.jpg", "scene0664_02/00400.jpg", "scene0664_02/00440.jpg" ], "turn_id": 3, "scan_id": "scene0664_02", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0664_02/00480.jpg", "scene0664_02/00520.jpg", "scene0664_02/00560.jpg", "scene0664_02/00600.jpg" ], "turn_id": 4, "scan_id": "scene0664_02", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0664_02/00640.jpg", "scene0664_02/00680.jpg", "scene0664_02/00720.jpg", "scene0664_02/00760.jpg" ], "turn_id": 5, "scan_id": "scene0664_02", "origin_question": "Which direction is the standard white oval-shaped toilet fixture in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the standard white oval-shaped toilet fixture in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "scene0664_02/00800.jpg", "scene0664_02/00840.jpg", "scene0664_02/00880.jpg", "scene0664_02/00920.jpg" ], "turn_id": 6, "scan_id": "scene0664_02", "origin_question": "Among these three objects(\"the cylindrical stainless steel bin\",\"the modern, wall-mounted shower with a silver head\",\"the rectangular framed picture with colorful leaves\" ), which one is the closest to you?", "option": [ "the cylindrical stainless steel bin", "the modern, wall-mounted shower with a silver head", "the rectangular framed picture with colorful leaves" ], "answer": "the rectangular framed picture with colorful leaves", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects(\"the cylindrical stainless steel bin\",\"the modern, wall-mounted shower with a silver head\",\"the rectangular framed picture with colorful leaves\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the cylindrical stainless steel bin', 'the modern, wall-mounted shower with a silver head', 'the rectangular framed picture with colorful leaves'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0664_02/00960.jpg", "scene0664_02/01000.jpg", "scene0664_02/01040.jpg", "scene0664_02/01080.jpg" ], "turn_id": 7, "scan_id": "scene0664_02", "origin_question": "Which direction is the bathtub in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the bathtub in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0435_03/00000.jpg", "scene0435_03/00040.jpg", "scene0435_03/00080.jpg", "scene0435_03/00120.jpg", "scene0435_03/00160.jpg" ], "turn_id": 1, "scan_id": "scene0435_03", "origin_question": "Remember, have you seen any chair(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any chair(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0435_03/00200.jpg", "scene0435_03/00240.jpg", "scene0435_03/00280.jpg", "scene0435_03/00320.jpg", "scene0435_03/00360.jpg" ], "turn_id": 2, "scan_id": "scene0435_03", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the rectangular single bed with a white sheet\"; \"the black rectangular case made of fabric\"; \"the lamp\" ", "option": [ "the rectangular single bed with a white sheet", "the black rectangular case made of fabric", "the lamp" ], "answer": "the lamp", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the rectangular single bed with a white sheet\"; \"the black rectangular case made of fabric\"; \"the lamp\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular single bed with a white sheet', 'the black rectangular case made of fabric', 'the lamp'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0435_03/00400.jpg", "scene0435_03/00440.jpg", "scene0435_03/00480.jpg", "scene0435_03/00520.jpg", "scene0435_03/00560.jpg" ], "turn_id": 3, "scan_id": "scene0435_03", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "84.22", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0435_03/00600.jpg", "scene0435_03/00640.jpg", "scene0435_03/00680.jpg", "scene0435_03/00720.jpg", "scene0435_03/00760.jpg" ], "turn_id": 4, "scan_id": "scene0435_03", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0435_03/00800.jpg", "scene0435_03/00840.jpg", "scene0435_03/00880.jpg", "scene0435_03/00920.jpg", "scene0435_03/00960.jpg" ], "turn_id": 5, "scan_id": "scene0435_03", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the black casual lightweight clothes made of soft fabric\" from your position?", "option": [], "answer": "117.26", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the black casual lightweight clothes made of soft fabric\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0435_03/01000.jpg", "scene0435_03/01040.jpg", "scene0435_03/01080.jpg", "scene0435_03/01120.jpg", "scene0435_03/01160.jpg" ], "turn_id": 6, "scan_id": "scene0435_03", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the black flat-screen TV made of plastic\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the black flat-screen TV made of plastic\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "scene0435_03/01210.jpg", "scene0435_03/01250.jpg", "scene0435_03/01290.jpg", "scene0435_03/01330.jpg", "scene0435_03/01370.jpg" ], "turn_id": 7, "scan_id": "scene0435_03", "origin_question": "Among these three objects(\"the black and white picture in a black frame\",\"the small black square refrigerator\",\"the small black plastic bin\" ), which one is the farthest from you?", "option": [ "the black and white picture in a black frame", "the small black square refrigerator", "the small black plastic bin" ], "answer": "the small black square refrigerator", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the black and white picture in a black frame\",\"the small black square refrigerator\",\"the small black plastic bin\" ), which one is the farthest from you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the black and white picture in a black frame', 'the small black square refrigerator', 'the small black plastic bin'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0435_03/01410.jpg", "scene0435_03/01450.jpg", "scene0435_03/01490.jpg", "scene0435_03/01530.jpg", "scene0435_03/01570.jpg" ], "turn_id": 8, "scan_id": "scene0435_03", "origin_question": "Which two objects are on the same side of you now? \"the black flat-screen TV made of plastic\", \"the medium-sized beige rectangular couch with a plush finish\", \"the black corded rectangular telephone with a keypad\". ", "option": [ "the black flat-screen TV made of plastic, the medium-sized beige rectangular couch with a plush finish", "the black flat-screen TV made of plastic, the black corded rectangular telephone with a keypad", "the medium-sized beige rectangular couch with a plush finish, the black corded rectangular telephone with a keypad" ], "answer": "the medium-sized beige rectangular couch with a plush finish, the black corded rectangular telephone with a keypad", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the black flat-screen TV made of plastic\", \"the medium-sized beige rectangular couch with a plush finish\", \"the black corded rectangular telephone with a keypad\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the black flat-screen TV made of plastic, the medium-sized beige rectangular couch with a plush finish', 'the black flat-screen TV made of plastic, the black corded rectangular telephone with a keypad', 'the medium-sized beige rectangular couch with a plush finish, the black corded rectangular telephone with a keypad'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "scene0435_03/01610.jpg", "scene0435_03/01650.jpg", "scene0435_03/01690.jpg", "scene0435_03/01730.jpg", "scene0435_03/01770.jpg" ], "turn_id": 9, "scan_id": "scene0435_03", "origin_question": "Among these three objects(\"the small black square refrigerator\",\"the medium-sized beige rectangular couch with a plush finish\",\"the black corded rectangular telephone with a keypad\" ), which one is the closest to you?", "option": [ "the small black square refrigerator", "the medium-sized beige rectangular couch with a plush finish", "the black corded rectangular telephone with a keypad" ], "answer": "the black corded rectangular telephone with a keypad", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the small black square refrigerator\",\"the medium-sized beige rectangular couch with a plush finish\",\"the black corded rectangular telephone with a keypad\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the small black square refrigerator', 'the medium-sized beige rectangular couch with a plush finish', 'the black corded rectangular telephone with a keypad'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0046_01/00000.jpg", "scene0046_01/00040.jpg", "scene0046_01/00080.jpg", "scene0046_01/00120.jpg", "scene0046_01/00160.jpg" ], "turn_id": 1, "scan_id": "scene0046_01", "origin_question": "Remember, how many table(s) have you seen so far? ", "option": [], "answer": "2", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many table(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "scene0046_01/00200.jpg", "scene0046_01/00240.jpg", "scene0046_01/00280.jpg", "scene0046_01/00320.jpg", "scene0046_01/00360.jpg" ], "turn_id": 2, "scan_id": "scene0046_01", "origin_question": "Among these three objects(\"the sleek black rectangular plastic remote control\",\"the tall, light brown wooden wardrobe with silver handles\",\"the shelf\" ), which one is the closest to you?", "option": [ "the sleek black rectangular plastic remote control", "the tall, light brown wooden wardrobe with silver handles", "the shelf" ], "answer": "the tall, light brown wooden wardrobe with silver handles", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the sleek black rectangular plastic remote control\",\"the tall, light brown wooden wardrobe with silver handles\",\"the shelf\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the sleek black rectangular plastic remote control', 'the tall, light brown wooden wardrobe with silver handles', 'the shelf'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0046_01/00400.jpg", "scene0046_01/00440.jpg", "scene0046_01/00480.jpg", "scene0046_01/00520.jpg", "scene0046_01/00560.jpg" ], "turn_id": 3, "scan_id": "scene0046_01", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the medium rectangular black case made of plastic\" from your position?", "option": [], "answer": "145.06", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the medium rectangular black case made of plastic\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0046_01/00600.jpg", "scene0046_01/00640.jpg", "scene0046_01/00680.jpg", "scene0046_01/00720.jpg", "scene0046_01/00760.jpg" ], "turn_id": 4, "scan_id": "scene0046_01", "origin_question": "", "option": [], "answer": "", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0046_01/00800.jpg", "scene0046_01/00840.jpg", "scene0046_01/00880.jpg", "scene0046_01/00920.jpg", "scene0046_01/00960.jpg" ], "turn_id": 5, "scan_id": "scene0046_01", "origin_question": "Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0046_01/01000.jpg", "scene0046_01/01040.jpg", "scene0046_01/01080.jpg", "scene0046_01/01120.jpg", "scene0046_01/01160.jpg" ], "turn_id": 6, "scan_id": "scene0046_01", "origin_question": "Which direction is the medium rectangular black case made of plastic in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the medium rectangular black case made of plastic in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0046_01/01200.jpg", "scene0046_01/01240.jpg", "scene0046_01/01280.jpg", "scene0046_01/01320.jpg", "scene0046_01/01360.jpg" ], "turn_id": 7, "scan_id": "scene0046_01", "origin_question": "Which two objects are on the same side of you now? \"the blue bath towel made of soft material\", \"the vent\", \"the small rectangular light beige plastic switch\". ", "option": [ "the blue bath towel made of soft material, the vent", "the blue bath towel made of soft material, the small rectangular light beige plastic switch", "the vent, the small rectangular light beige plastic switch" ], "answer": "the blue bath towel made of soft material, the vent", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the blue bath towel made of soft material\", \"the vent\", \"the small rectangular light beige plastic switch\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the blue bath towel made of soft material, the vent', 'the blue bath towel made of soft material, the small rectangular light beige plastic switch', 'the vent, the small rectangular light beige plastic switch'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0046_01/01400.jpg", "scene0046_01/01440.jpg", "scene0046_01/01480.jpg", "scene0046_01/01520.jpg", "scene0046_01/01560.jpg" ], "turn_id": 8, "scan_id": "scene0046_01", "origin_question": "What will be the first-time appearance order of the following categories: curtain, light, sink?", "option": [ "light, curtain, sink", "curtain, light, sink", "sink, light, curtain", "light, sink, curtain" ], "answer": "curtain, light, sink", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: curtain, light, sink? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['light, curtain, sink', 'curtain, light, sink', 'sink, light, curtain', 'light, sink, curtain'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0046_01/01600.jpg", "scene0046_01/01640.jpg", "scene0046_01/01680.jpg", "scene0046_01/01720.jpg", "scene0046_01/01760.jpg" ], "turn_id": 9, "scan_id": "scene0046_01", "origin_question": "Which direction is the black oval optical mouse made of plastic in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the black oval optical mouse made of plastic in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0046_01/01800.jpg", "scene0046_01/01840.jpg", "scene0046_01/01880.jpg", "scene0046_01/01920.jpg", "scene0046_01/01960.jpg" ], "turn_id": 10, "scan_id": "scene0046_01", "origin_question": "By the end of the 3 turn, were you getting closer to or farther away from these two objects: X: \"the sleek black laptop with a black keyboard\", Y: \"the white vertical rod with a smooth surface\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to Y and farther away from X", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 3 turn, were you getting closer to or farther away from these two objects: X: \"the sleek black laptop with a black keyboard\", Y: \"the white vertical rod with a smooth surface\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0291_02/00000.jpg", "scene0291_02/00040.jpg", "scene0291_02/00080.jpg", "scene0291_02/00120.jpg" ], "turn_id": 1, "scan_id": "scene0291_02", "origin_question": "Remember, how many sponge(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many sponge(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0291_02/00160.jpg", "scene0291_02/00200.jpg", "scene0291_02/00240.jpg", "scene0291_02/00280.jpg" ], "turn_id": 2, "scan_id": "scene0291_02", "origin_question": "When did you first discover \"the large rectangular wooden table with a smooth surface\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the large rectangular wooden table with a smooth surface\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0291_02/00320.jpg", "scene0291_02/00360.jpg", "scene0291_02/00400.jpg", "scene0291_02/00440.jpg" ], "turn_id": 3, "scan_id": "scene0291_02", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0291_02/00480.jpg", "scene0291_02/00520.jpg", "scene0291_02/00560.jpg", "scene0291_02/00600.jpg" ], "turn_id": 4, "scan_id": "scene0291_02", "origin_question": "When did you first discover \"the medium-sized, dark rectangular box made of hard cardboard\"? (index of the turn)", "option": [], "answer": "3", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the medium-sized, dark rectangular box made of hard cardboard\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0291_02/00640.jpg", "scene0291_02/00680.jpg", "scene0291_02/00720.jpg", "scene0291_02/00760.jpg" ], "turn_id": 5, "scan_id": "scene0291_02", "origin_question": "When did you first discover \"the medium rectangular wooden door with a classic design\"? (index of the turn)", "option": [], "answer": "4", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the medium rectangular wooden door with a classic design\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0291_02/00800.jpg", "scene0291_02/00840.jpg", "scene0291_02/00880.jpg", "scene0291_02/00920.jpg" ], "turn_id": 6, "scan_id": "scene0291_02", "origin_question": "Is the rectangular black chalkboard made of plastic to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the rectangular black chalkboard made of plastic to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0291_02/00960.jpg", "scene0291_02/01000.jpg", "scene0291_02/01040.jpg", "scene0291_02/01080.jpg" ], "turn_id": 7, "scan_id": "scene0291_02", "origin_question": "In which turn did you see both \"the medium-sized, dark rectangular box made of hard cardboard\" and \"the medium rectangular wooden door with a classic design\" simultaneously? ", "option": [], "answer": "4", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the medium-sized, dark rectangular box made of hard cardboard\" and \"the medium rectangular wooden door with a classic design\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0291_02/01120.jpg", "scene0291_02/01160.jpg", "scene0291_02/01200.jpg", "scene0291_02/01240.jpg" ], "turn_id": 8, "scan_id": "scene0291_02", "origin_question": "What will be the first-time appearance order of the following categories: blackboard, door, window?", "option": [ "blackboard, door, window", "door, blackboard, window", "window, blackboard, door", "door, window, blackboard" ], "answer": "window, blackboard, door", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: blackboard, door, window? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['blackboard, door, window', 'door, blackboard, window', 'window, blackboard, door', 'door, window, blackboard'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0291_02/01280.jpg", "scene0291_02/01320.jpg", "scene0291_02/01360.jpg", "scene0291_02/01400.jpg" ], "turn_id": 9, "scan_id": "scene0291_02", "origin_question": "", "option": [], "answer": "", "user_message": "For the 9 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0030_region10/9b3beab962d040428acc300ff7a9326f_i2_0.jpg", "1mp3d_0030_region10/ce06069f651f49d19d05fde4a7e231b5_i1_4.jpg", "1mp3d_0030_region10/ce06069f651f49d19d05fde4a7e231b5_i2_4.jpg", "1mp3d_0030_region10/ce06069f651f49d19d05fde4a7e231b5_i1_3.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0030_region10", "origin_question": "Remember, have you seen any table(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any table(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "1mp3d_0030_region10/ce06069f651f49d19d05fde4a7e231b5_i2_3.jpg", "1mp3d_0030_region10/ce06069f651f49d19d05fde4a7e231b5_i1_2.jpg", "1mp3d_0030_region10/ce06069f651f49d19d05fde4a7e231b5_i2_2.jpg", "1mp3d_0030_region10/9b3beab962d040428acc300ff7a9326f_i1_3.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0030_region10", "origin_question": "Which two objects are on the same side of you now? \"the dark brown rectangular wooden table with a glossy finish\", \"the round, deep gray ceramic bowl\", \"the tall wooden door with visible wood grains\". ", "option": [ "the dark brown rectangular wooden table with a glossy finish, the round, deep gray ceramic bowl", "the dark brown rectangular wooden table with a glossy finish, the tall wooden door with visible wood grains", "the round, deep gray ceramic bowl, the tall wooden door with visible wood grains" ], "answer": "the dark brown rectangular wooden table with a glossy finish, the round, deep gray ceramic bowl", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the dark brown rectangular wooden table with a glossy finish\", \"the round, deep gray ceramic bowl\", \"the tall wooden door with visible wood grains\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the dark brown rectangular wooden table with a glossy finish, the round, deep gray ceramic bowl', 'the dark brown rectangular wooden table with a glossy finish, the tall wooden door with visible wood grains', 'the round, deep gray ceramic bowl, the tall wooden door with visible wood grains'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0030_region10/9b3beab962d040428acc300ff7a9326f_i2_3.jpg", "1mp3d_0030_region10/9b3beab962d040428acc300ff7a9326f_i1_2.jpg", "1mp3d_0030_region10/9b3beab962d040428acc300ff7a9326f_i2_2.jpg", "1mp3d_0030_region10/c69e398f10eb40938d19c2c736b0c084_i1_4.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0030_region10", "origin_question": "", "option": [], "answer": "", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "1mp3d_0030_region10/c69e398f10eb40938d19c2c736b0c084_i2_4.jpg", "1mp3d_0030_region10/91afeb3c3c214d2da1205dd505bf976b_i1_2.jpg", "1mp3d_0030_region10/91afeb3c3c214d2da1205dd505bf976b_i2_2.jpg", "1mp3d_0030_region10/c69e398f10eb40938d19c2c736b0c084_i1_3.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0030_region10", "origin_question": "Which two objects are on the same side of you now? \"the toy\", \"the petite white statue depicting a person\", \"the tall wooden door with visible wood grains\". ", "option": [ "the toy, the petite white statue depicting a person", "the toy, the tall wooden door with visible wood grains", "the petite white statue depicting a person, the tall wooden door with visible wood grains" ], "answer": "the toy, the petite white statue depicting a person", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the toy\", \"the petite white statue depicting a person\", \"the tall wooden door with visible wood grains\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the toy, the petite white statue depicting a person', 'the toy, the tall wooden door with visible wood grains', 'the petite white statue depicting a person, the tall wooden door with visible wood grains'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "1mp3d_0030_region10/c69e398f10eb40938d19c2c736b0c084_i2_3.jpg", "1mp3d_0030_region10/91afeb3c3c214d2da1205dd505bf976b_i1_3.jpg", "1mp3d_0030_region10/91afeb3c3c214d2da1205dd505bf976b_i2_3.jpg", "1mp3d_0030_region10/3d10b31a0dcc44148941f3906c54d30b_i1_5.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0030_region10", "origin_question": "Is the petite white statue depicting a person to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the petite white statue depicting a person to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0072_region3/9a896fadb819490089bf51005088f464_i1_2.jpg", "1mp3d_0072_region3/9a896fadb819490089bf51005088f464_i2_2.jpg", "1mp3d_0072_region3/6eddf5877b284e3184ddd6aec1e34101_i1_3.jpg", "1mp3d_0072_region3/6eddf5877b284e3184ddd6aec1e34101_i2_3.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0072_region3", "origin_question": "Remember, have you seen any picture(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any picture(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "1mp3d_0072_region3/35b41dcbfcf84f96878f6ca28c70e5af_i1_5.jpg", "1mp3d_0072_region3/35b41dcbfcf84f96878f6ca28c70e5af_i2_5.jpg", "1mp3d_0072_region3/6eddf5877b284e3184ddd6aec1e34101_i1_4.jpg", "1mp3d_0072_region3/6eddf5877b284e3184ddd6aec1e34101_i2_4.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0072_region3", "origin_question": "When did you first discover \"the vertical picture with vivid geometric patterns\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the vertical picture with vivid geometric patterns\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0072_region3/9a896fadb819490089bf51005088f464_i1_3.jpg", "1mp3d_0072_region3/9a896fadb819490089bf51005088f464_i2_3.jpg", "1mp3d_0072_region3/9a896fadb819490089bf51005088f464_i1_4.jpg", "1mp3d_0072_region3/9a896fadb819490089bf51005088f464_i2_4.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0072_region3", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the rectangular white plastic switch\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the rectangular white plastic switch\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "1mp3d_0072_region3/6eddf5877b284e3184ddd6aec1e34101_i1_5.jpg", "1mp3d_0072_region3/6eddf5877b284e3184ddd6aec1e34101_i2_5.jpg", "1mp3d_0072_region3/1a7fe152f6f34697bdcee712008f84e4_i1_4.jpg", "1mp3d_0072_region3/1a7fe152f6f34697bdcee712008f84e4_i2_4.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0072_region3", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "72.18", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "1mp3d_0072_region3/1a7fe152f6f34697bdcee712008f84e4_i1_3.jpg", "1mp3d_0072_region3/1a7fe152f6f34697bdcee712008f84e4_i2_3.jpg", "1mp3d_0072_region3/ca77dd4132f044b7b6b999d3795d09d3_i1_5.jpg", "1mp3d_0072_region3/ca77dd4132f044b7b6b999d3795d09d3_i2_5.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0072_region3", "origin_question": "Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0226_00/00000.jpg", "scene0226_00/00040.jpg", "scene0226_00/00080.jpg", "scene0226_00/00120.jpg", "scene0226_00/00160.jpg" ], "turn_id": 1, "scan_id": "scene0226_00", "origin_question": "Remember, how many pen(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many pen(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0226_00/00200.jpg", "scene0226_00/00240.jpg", "scene0226_00/00280.jpg", "scene0226_00/00320.jpg", "scene0226_00/00360.jpg" ], "turn_id": 2, "scan_id": "scene0226_00", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the round metal stool with three white legs\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the round metal stool with three white legs\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0226_00/00400.jpg", "scene0226_00/00440.jpg", "scene0226_00/00480.jpg", "scene0226_00/00520.jpg", "scene0226_00/00560.jpg" ], "turn_id": 3, "scan_id": "scene0226_00", "origin_question": "Which direction is the moderate-sized rectangular dark metal laptop in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the moderate-sized rectangular dark metal laptop in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0226_00/00600.jpg", "scene0226_00/00640.jpg", "scene0226_00/00680.jpg", "scene0226_00/00720.jpg", "scene0226_00/00760.jpg" ], "turn_id": 4, "scan_id": "scene0226_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0226_00/00800.jpg", "scene0226_00/00840.jpg", "scene0226_00/00880.jpg", "scene0226_00/00920.jpg", "scene0226_00/00960.jpg" ], "turn_id": 5, "scan_id": "scene0226_00", "origin_question": "Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0226_00/01000.jpg", "scene0226_00/01040.jpg", "scene0226_00/01080.jpg", "scene0226_00/01120.jpg", "scene0226_00/01160.jpg" ], "turn_id": 6, "scan_id": "scene0226_00", "origin_question": "By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the round metal stool with three white legs\", Y: \"the beige cylindrical lampshade with a copper-colored base\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "farther away from both X and Y", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the round metal stool with three white legs\", Y: \"the beige cylindrical lampshade with a copper-colored base\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0226_00/01200.jpg", "scene0226_00/01240.jpg", "scene0226_00/01280.jpg", "scene0226_00/01320.jpg", "scene0226_00/01360.jpg" ], "turn_id": 7, "scan_id": "scene0226_00", "origin_question": "Please recall \"the rectangular picture with a red background and white border\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.33", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the rectangular picture with a red background and white border\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0226_00/01400.jpg", "scene0226_00/01440.jpg", "scene0226_00/01480.jpg", "scene0226_00/01520.jpg", "scene0226_00/01560.jpg" ], "turn_id": 8, "scan_id": "scene0226_00", "origin_question": "When did you first discover \"the rectangular picture with a red background and white border\"? (index of the turn)", "option": [], "answer": "4", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the rectangular picture with a red background and white border\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0700_02/00000.jpg", "scene0700_02/00040.jpg", "scene0700_02/00080.jpg", "scene0700_02/00120.jpg", "scene0700_02/00160.jpg" ], "turn_id": 1, "scan_id": "scene0700_02", "origin_question": "Remember, how many telephone(s) have you seen so far? ", "option": [], "answer": "2", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many telephone(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0700_02/00200.jpg", "scene0700_02/00240.jpg", "scene0700_02/00280.jpg", "scene0700_02/00320.jpg", "scene0700_02/00360.jpg" ], "turn_id": 2, "scan_id": "scene0700_02", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the small rectangular dark-colored shelf\"; \"the windowsill\"; \"the white vertical metal radiator with a textured pattern\" ", "option": [ "the small rectangular dark-colored shelf", "the windowsill", "the white vertical metal radiator with a textured pattern" ], "answer": "the small rectangular dark-colored shelf", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the small rectangular dark-colored shelf\"; \"the windowsill\"; \"the white vertical metal radiator with a textured pattern\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the small rectangular dark-colored shelf', 'the windowsill', 'the white vertical metal radiator with a textured pattern'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0700_02/00400.jpg", "scene0700_02/00440.jpg", "scene0700_02/00480.jpg", "scene0700_02/00520.jpg", "scene0700_02/00560.jpg" ], "turn_id": 3, "scan_id": "scene0700_02", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the cylindrical yellow pen with black text\" from your position?", "option": [], "answer": "83.85", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the cylindrical yellow pen with black text\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "scene0700_02/00600.jpg", "scene0700_02/00640.jpg", "scene0700_02/00680.jpg", "scene0700_02/00720.jpg", "scene0700_02/00760.jpg" ], "turn_id": 4, "scan_id": "scene0700_02", "origin_question": "Among these three objects(\"the windowsill\",\"the rectangular plastic box with a green lid\",\"the rectangular white paper bag with a red logo\" ), which one is the closest to you?", "option": [ "the windowsill", "the rectangular plastic box with a green lid", "the rectangular white paper bag with a red logo" ], "answer": "the rectangular plastic box with a green lid", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the windowsill\",\"the rectangular plastic box with a green lid\",\"the rectangular white paper bag with a red logo\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the windowsill', 'the rectangular plastic box with a green lid', 'the rectangular white paper bag with a red logo'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0700_02/00800.jpg", "scene0700_02/00840.jpg", "scene0700_02/00880.jpg", "scene0700_02/00920.jpg", "scene0700_02/00960.jpg" ], "turn_id": 5, "scan_id": "scene0700_02", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the windowsill\" from your position?", "option": [], "answer": "40.3", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the windowsill\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0700_02/01000.jpg", "scene0700_02/01040.jpg", "scene0700_02/01080.jpg", "scene0700_02/01120.jpg", "scene0700_02/01160.jpg" ], "turn_id": 6, "scan_id": "scene0700_02", "origin_question": "Is the rectangular white paper bag with a red logo to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the rectangular white paper bag with a red logo to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0700_02/01200.jpg", "scene0700_02/01240.jpg", "scene0700_02/01280.jpg", "scene0700_02/01320.jpg", "scene0700_02/01360.jpg" ], "turn_id": 7, "scan_id": "scene0700_02", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the rectangular plastic box with a green lid\" from your position?", "option": [], "answer": "123.93", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the rectangular plastic box with a green lid\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0700_02/01400.jpg", "scene0700_02/01440.jpg", "scene0700_02/01480.jpg", "scene0700_02/01520.jpg", "scene0700_02/01560.jpg" ], "turn_id": 8, "scan_id": "scene0700_02", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the rectangular matte black keyboard made of plastic\" from your position?", "option": [], "answer": "116.5", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the rectangular matte black keyboard made of plastic\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0700_02/01600.jpg", "scene0700_02/01640.jpg", "scene0700_02/01680.jpg", "scene0700_02/01720.jpg", "scene0700_02/01760.jpg" ], "turn_id": 9, "scan_id": "scene0700_02", "origin_question": "In which turn did you see both \"the window\" and \"the small rectangular dark-colored shelf\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the window\" and \"the small rectangular dark-colored shelf\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0700_02/01800.jpg", "scene0700_02/01840.jpg", "scene0700_02/01880.jpg", "scene0700_02/01920.jpg", "scene0700_02/01960.jpg" ], "turn_id": 10, "scan_id": "scene0700_02", "origin_question": "Compared to your position at the end of the 5 turn, are you now closer or farther away from \"the small, black plastic mouse with a smooth shape\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 5 turn, are you now closer or farther away from \"the small, black plastic mouse with a smooth shape\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0317_01/00000.jpg", "scene0317_01/00040.jpg", "scene0317_01/00080.jpg", "scene0317_01/00120.jpg" ], "turn_id": 1, "scan_id": "scene0317_01", "origin_question": "Remember, have you seen any car(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any car(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0317_01/00160.jpg", "scene0317_01/00200.jpg", "scene0317_01/00240.jpg", "scene0317_01/00280.jpg" ], "turn_id": 2, "scan_id": "scene0317_01", "origin_question": "In which turn did you see both \"the medium-sized orange plush pillow\" and \"the white vertical stainless steel refrigerator\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the medium-sized orange plush pillow\" and \"the white vertical stainless steel refrigerator\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0317_01/00320.jpg", "scene0317_01/00360.jpg", "scene0317_01/00400.jpg", "scene0317_01/00440.jpg" ], "turn_id": 3, "scan_id": "scene0317_01", "origin_question": "Which direction is the bright red rectangular stool with a smooth surface in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the bright red rectangular stool with a smooth surface in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0317_01/00480.jpg", "scene0317_01/00520.jpg", "scene0317_01/00560.jpg", "scene0317_01/00600.jpg" ], "turn_id": 4, "scan_id": "scene0317_01", "origin_question": "Which direction is the white vertical stainless steel refrigerator in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the white vertical stainless steel refrigerator in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0317_01/00640.jpg", "scene0317_01/00680.jpg", "scene0317_01/00720.jpg", "scene0317_01/00760.jpg" ], "turn_id": 5, "scan_id": "scene0317_01", "origin_question": "Remember, how many backpack(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many backpack(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0317_01/00800.jpg", "scene0317_01/00840.jpg", "scene0317_01/00880.jpg", "scene0317_01/00920.jpg" ], "turn_id": 6, "scan_id": "scene0317_01", "origin_question": "Please recall \"the small paper cup with a black and white design\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.05", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the small paper cup with a black and white design\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0317_01/00960.jpg", "scene0317_01/01000.jpg", "scene0317_01/01040.jpg", "scene0317_01/01080.jpg" ], "turn_id": 7, "scan_id": "scene0317_01", "origin_question": "In which turn did you see both \"the medium-sized orange plush pillow\" and \"the white vertical stainless steel refrigerator\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the medium-sized orange plush pillow\" and \"the white vertical stainless steel refrigerator\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0317_01/01120.jpg", "scene0317_01/01160.jpg", "scene0317_01/01200.jpg", "scene0317_01/01240.jpg" ], "turn_id": 8, "scan_id": "scene0317_01", "origin_question": "Please recall \"the book\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.79", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the book\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0599_01/00000.jpg", "scene0599_01/00040.jpg", "scene0599_01/00080.jpg", "scene0599_01/00120.jpg", "scene0599_01/00160.jpg" ], "turn_id": 1, "scan_id": "scene0599_01", "origin_question": "Remember, have you seen any door(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any door(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0599_01/00200.jpg", "scene0599_01/00240.jpg", "scene0599_01/00280.jpg", "scene0599_01/00320.jpg", "scene0599_01/00360.jpg" ], "turn_id": 2, "scan_id": "scene0599_01", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the sleek gray rectangular plastic bin\" from your position?", "option": [], "answer": "80.1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the sleek gray rectangular plastic bin\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0599_01/00400.jpg", "scene0599_01/00440.jpg", "scene0599_01/00480.jpg", "scene0599_01/00520.jpg", "scene0599_01/00560.jpg" ], "turn_id": 3, "scan_id": "scene0599_01", "origin_question": "Is the rectangular, dark brown plastic dispenser to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the rectangular, dark brown plastic dispenser to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0599_01/00600.jpg", "scene0599_01/00640.jpg", "scene0599_01/00680.jpg", "scene0599_01/00720.jpg", "scene0599_01/00760.jpg" ], "turn_id": 4, "scan_id": "scene0599_01", "origin_question": "Please recall \"the sleek gray rectangular plastic bin\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.86", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the sleek gray rectangular plastic bin\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0599_01/00800.jpg", "scene0599_01/00840.jpg", "scene0599_01/00880.jpg", "scene0599_01/00920.jpg", "scene0599_01/00960.jpg" ], "turn_id": 5, "scan_id": "scene0599_01", "origin_question": "In which turn did you see both \"the sleek gray rectangular plastic bin\" and \"the blackboard\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the sleek gray rectangular plastic bin\" and \"the blackboard\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0599_01/01000.jpg", "scene0599_01/01040.jpg", "scene0599_01/01080.jpg", "scene0599_01/01120.jpg", "scene0599_01/01160.jpg" ], "turn_id": 6, "scan_id": "scene0599_01", "origin_question": "Which two objects are on the same side of you now? \"the long rectangular light brown wooden table\", \"the sleek gray rectangular plastic bin\", \"the rectangular, dark brown plastic dispenser\". ", "option": [ "the long rectangular light brown wooden table, the sleek gray rectangular plastic bin", "the long rectangular light brown wooden table, the rectangular, dark brown plastic dispenser", "the sleek gray rectangular plastic bin, the rectangular, dark brown plastic dispenser" ], "answer": "the sleek gray rectangular plastic bin, the rectangular, dark brown plastic dispenser", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the long rectangular light brown wooden table\", \"the sleek gray rectangular plastic bin\", \"the rectangular, dark brown plastic dispenser\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the long rectangular light brown wooden table, the sleek gray rectangular plastic bin', 'the long rectangular light brown wooden table, the rectangular, dark brown plastic dispenser', 'the sleek gray rectangular plastic bin, the rectangular, dark brown plastic dispenser'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0599_01/01200.jpg", "scene0599_01/01240.jpg", "scene0599_01/01280.jpg", "scene0599_01/01320.jpg", "scene0599_01/01360.jpg" ], "turn_id": 7, "scan_id": "scene0599_01", "origin_question": "At the end of which turn were both of \"the rectangular, dark brown plastic dispenser\" and \"the medium to large black and grey backpack\" on your right side?", "option": [], "answer": "5", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the rectangular, dark brown plastic dispenser\" and \"the medium to large black and grey backpack\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "scene0599_01/01400.jpg", "scene0599_01/01440.jpg", "scene0599_01/01480.jpg", "scene0599_01/01520.jpg", "scene0599_01/01560.jpg" ], "turn_id": 8, "scan_id": "scene0599_01", "origin_question": "Among these three objects(\"the stylish flat-panel TV with a dark matte surface\",\"the standard-sized white sheet of paper\",\"the medium to large black and grey backpack\" ), which one is the closest to you?", "option": [ "the stylish flat-panel TV with a dark matte surface", "the standard-sized white sheet of paper", "the medium to large black and grey backpack" ], "answer": "the medium to large black and grey backpack", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the stylish flat-panel TV with a dark matte surface\",\"the standard-sized white sheet of paper\",\"the medium to large black and grey backpack\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the stylish flat-panel TV with a dark matte surface', 'the standard-sized white sheet of paper', 'the medium to large black and grey backpack'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0599_01/01600.jpg", "scene0599_01/01640.jpg", "scene0599_01/01680.jpg", "scene0599_01/01720.jpg", "scene0599_01/01760.jpg" ], "turn_id": 9, "scan_id": "scene0599_01", "origin_question": "What will be the first-time appearance order of the following categories: backpack, blackboard, paper?", "option": [ "backpack, paper, blackboard", "backpack, blackboard, paper", "blackboard, paper, backpack", "blackboard, backpack, paper" ], "answer": "blackboard, paper, backpack", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: backpack, blackboard, paper? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['backpack, paper, blackboard', 'backpack, blackboard, paper', 'blackboard, paper, backpack', 'blackboard, backpack, paper'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0599_01/01800.jpg", "scene0599_01/01840.jpg", "scene0599_01/01880.jpg", "scene0599_01/01920.jpg", "scene0599_01/01960.jpg" ], "turn_id": 10, "scan_id": "scene0599_01", "origin_question": "In which turn did you see both \"the stylish flat-panel TV with a dark matte surface\" and \"the medium to large black and grey backpack\" simultaneously? ", "option": [], "answer": "10", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the stylish flat-panel TV with a dark matte surface\" and \"the medium to large black and grey backpack\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0609_00/00000.jpg", "scene0609_00/00040.jpg", "scene0609_00/00080.jpg", "scene0609_00/00120.jpg" ], "turn_id": 1, "scan_id": "scene0609_00", "origin_question": "Remember, how many chair(s) have you seen so far? ", "option": [], "answer": "3", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many chair(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0609_00/00160.jpg", "scene0609_00/00200.jpg", "scene0609_00/00240.jpg", "scene0609_00/00280.jpg" ], "turn_id": 2, "scan_id": "scene0609_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0609_00/00320.jpg", "scene0609_00/00360.jpg", "scene0609_00/00400.jpg", "scene0609_00/00440.jpg" ], "turn_id": 3, "scan_id": "scene0609_00", "origin_question": "How far is your current position from where you were at the end of the turn 1? (in meters)", "option": [], "answer": "1.19", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 1? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0609_00/00480.jpg", "scene0609_00/00520.jpg", "scene0609_00/00560.jpg", "scene0609_00/00600.jpg" ], "turn_id": 4, "scan_id": "scene0609_00", "origin_question": "Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0609_00/00640.jpg", "scene0609_00/00680.jpg", "scene0609_00/00720.jpg", "scene0609_00/00760.jpg" ], "turn_id": 5, "scan_id": "scene0609_00", "origin_question": "Remember, have you seen any chair(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any chair(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0609_00/00800.jpg", "scene0609_00/00840.jpg", "scene0609_00/00880.jpg", "scene0609_00/00920.jpg" ], "turn_id": 6, "scan_id": "scene0609_00", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "1.61", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0057_region3/186ca6b9eba74f038ac4ce7a6c0204fd_i1_4.jpg", "1mp3d_0057_region3/186ca6b9eba74f038ac4ce7a6c0204fd_i2_4.jpg", "1mp3d_0057_region3/186ca6b9eba74f038ac4ce7a6c0204fd_i1_3.jpg", "1mp3d_0057_region3/186ca6b9eba74f038ac4ce7a6c0204fd_i2_3.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0057_region3", "origin_question": "Remember, have you seen any food(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any food(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 2 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "1mp3d_0057_region3/186ca6b9eba74f038ac4ce7a6c0204fd_i1_2.jpg", "1mp3d_0057_region3/186ca6b9eba74f038ac4ce7a6c0204fd_i2_2.jpg", "1mp3d_0057_region3/1309491e21ea45f2ba31c18c929ccb78_i1_1.jpg", "1mp3d_0057_region3/1309491e21ea45f2ba31c18c929ccb78_i2_1.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0057_region3", "origin_question": "In which turn were you closest to \"the rectangular mirror with a black frame\"? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the rectangular mirror with a black frame\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 2 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0040_region46/54af55e2794d41cbae1f57d9e152809d_i1_4.jpg", "1mp3d_0040_region46/54af55e2794d41cbae1f57d9e152809d_i2_4.jpg", "1mp3d_0040_region46/1b27ae30f0ec43408a18f21de3655325_i1_3.jpg", "1mp3d_0040_region46/1b27ae30f0ec43408a18f21de3655325_i2_3.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0040_region46", "origin_question": "Remember, have you seen any window(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any window(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "1mp3d_0040_region46/f710d435547d4d21b537cb16d85f3b11_i1_3.jpg", "1mp3d_0040_region46/f710d435547d4d21b537cb16d85f3b11_i2_3.jpg", "1mp3d_0040_region46/e9aa0797514b40c2ae7da8193fe1afcb_i1_5.jpg", "1mp3d_0040_region46/e9aa0797514b40c2ae7da8193fe1afcb_i2_5.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0040_region46", "origin_question": "In which turn were you closest to \"the window\"? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the window\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "1mp3d_0040_region46/3aee6e0b4ca94608b3669b0660532664_i1_0.jpg", "1mp3d_0040_region46/3aee6e0b4ca94608b3669b0660532664_i2_0.jpg", "1mp3d_0040_region46/e9aa0797514b40c2ae7da8193fe1afcb_i1_2.jpg", "1mp3d_0040_region46/e9aa0797514b40c2ae7da8193fe1afcb_i2_2.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0040_region46", "origin_question": "In which turn were you farthest from \"the round, light-colored stone table with a polished surface\"? ", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the round, light-colored stone table with a polished surface\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "1mp3d_0040_region46/f710d435547d4d21b537cb16d85f3b11_i1_1.jpg", "1mp3d_0040_region46/f710d435547d4d21b537cb16d85f3b11_i2_1.jpg", "1mp3d_0040_region46/eed4c37211c6481eb0c30a11ff1f8f69_i1_3.jpg", "1mp3d_0040_region46/eed4c37211c6481eb0c30a11ff1f8f69_i2_3.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0040_region46", "origin_question": "At the end of which turn were both of \"the window\" and \"the round, light-colored stone table with a polished surface\" on your right side?", "option": [], "answer": "4", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the window\" and \"the round, light-colored stone table with a polished surface\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "1mp3d_0040_region46/f710d435547d4d21b537cb16d85f3b11_i1_2.jpg", "1mp3d_0040_region46/f710d435547d4d21b537cb16d85f3b11_i2_2.jpg", "1mp3d_0040_region46/d88c8468a85d47b7bf445cd3fae67262_i1_2.jpg", "1mp3d_0040_region46/d88c8468a85d47b7bf445cd3fae67262_i2_2.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0040_region46", "origin_question": "Assuming the direction you are facing at the end of the turn 4 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Left", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 4 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "1mp3d_0040_region46/466ef42243d641888f8044a19d263a56_i1_2.jpg", "1mp3d_0040_region46/466ef42243d641888f8044a19d263a56_i2_2.jpg", "1mp3d_0040_region46/ce025bf7dc1048f890d587c1f9435ee4_i1_3.jpg", "1mp3d_0040_region46/ce025bf7dc1048f890d587c1f9435ee4_i2_3.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0040_region46", "origin_question": "Which direction is the round, light-colored stone table with a polished surface in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the round, light-colored stone table with a polished surface in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0040_region28/6624bc0717af4b1a87d984537d569bf2_i1_4.jpg", "1mp3d_0040_region28/6624bc0717af4b1a87d984537d569bf2_i2_4.jpg", "1mp3d_0040_region28/ceca6966416b4a079aae199fe44b34ae_i1_5.jpg", "1mp3d_0040_region28/ceca6966416b4a079aae199fe44b34ae_i2_5.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0040_region28", "origin_question": "Remember, have you seen any door(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any door(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 2 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "1mp3d_0040_region28/6624bc0717af4b1a87d984537d569bf2_i1_3.jpg", "1mp3d_0040_region28/6624bc0717af4b1a87d984537d569bf2_i2_3.jpg", "1mp3d_0040_region28/ceca6966416b4a079aae199fe44b34ae_i1_4.jpg", "1mp3d_0040_region28/ceca6966416b4a079aae199fe44b34ae_i2_4.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0040_region28", "origin_question": "When did you first discover \"the door\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the door\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 2 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0144_00/00000.jpg", "scene0144_00/00040.jpg", "scene0144_00/00080.jpg", "scene0144_00/00120.jpg" ], "turn_id": 1, "scan_id": "scene0144_00", "origin_question": "Remember, have you seen any bed(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any bed(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0144_00/00160.jpg", "scene0144_00/00200.jpg", "scene0144_00/00240.jpg", "scene0144_00/00280.jpg" ], "turn_id": 2, "scan_id": "scene0144_00", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the white pillow with blue stripes, made of plush material\"; \"the stand\"; \"the rectangular brown wooden table with rounded edges\" ", "option": [ "the white pillow with blue stripes, made of plush material", "the stand", "the rectangular brown wooden table with rounded edges" ], "answer": "the stand", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the white pillow with blue stripes, made of plush material\"; \"the stand\"; \"the rectangular brown wooden table with rounded edges\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the white pillow with blue stripes, made of plush material', 'the stand', 'the rectangular brown wooden table with rounded edges'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0144_00/00320.jpg", "scene0144_00/00360.jpg", "scene0144_00/00400.jpg", "scene0144_00/00440.jpg" ], "turn_id": 3, "scan_id": "scene0144_00", "origin_question": "In which turn did you see both \"the white pillow with blue stripes, made of plush material\" and \"the stand\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the white pillow with blue stripes, made of plush material\" and \"the stand\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0144_00/00480.jpg", "scene0144_00/00520.jpg", "scene0144_00/00560.jpg", "scene0144_00/00600.jpg" ], "turn_id": 4, "scan_id": "scene0144_00", "origin_question": "Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the rectangular picture with a medium-sized brown frame\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the rectangular picture with a medium-sized brown frame\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0144_00/00640.jpg", "scene0144_00/00680.jpg", "scene0144_00/00720.jpg", "scene0144_00/00760.jpg" ], "turn_id": 5, "scan_id": "scene0144_00", "origin_question": "Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the rod\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the rod\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "scene0144_00/00800.jpg", "scene0144_00/00840.jpg", "scene0144_00/00880.jpg", "scene0144_00/00920.jpg" ], "turn_id": 6, "scan_id": "scene0144_00", "origin_question": "Among these three objects(\"the white pillow with blue stripes, made of plush material\",\"the stand\",\"the monitor\" ), which one is the farthest from you?", "option": [ "the white pillow with blue stripes, made of plush material", "the stand", "the monitor" ], "answer": "the monitor", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects(\"the white pillow with blue stripes, made of plush material\",\"the stand\",\"the monitor\" ), which one is the farthest from you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the white pillow with blue stripes, made of plush material', 'the stand', 'the monitor'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0144_00/00960.jpg", "scene0144_00/01000.jpg", "scene0144_00/01040.jpg", "scene0144_00/01080.jpg" ], "turn_id": 7, "scan_id": "scene0144_00", "origin_question": "By the end of the 4 turn, were you getting closer to or farther away from these two objects: X: \"the white pillow with blue stripes, made of plush material\", Y: \"the stand\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to X and farther away from Y", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 4 turn, were you getting closer to or farther away from these two objects: X: \"the white pillow with blue stripes, made of plush material\", Y: \"the stand\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0144_00/01120.jpg", "scene0144_00/01160.jpg", "scene0144_00/01200.jpg", "scene0144_00/01240.jpg" ], "turn_id": 8, "scan_id": "scene0144_00", "origin_question": "How far is your current position from where you were at the end of the turn 7? (in meters)", "option": [], "answer": "1.52", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 7? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "scene0144_00/01280.jpg", "scene0144_00/01320.jpg", "scene0144_00/01360.jpg", "scene0144_00/01400.jpg" ], "turn_id": 9, "scan_id": "scene0144_00", "origin_question": "Assuming the direction you are facing at the end of the turn 5 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Right", "user_message": "For the 9 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 5 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0217_00/00000.jpg", "scene0217_00/00040.jpg", "scene0217_00/00080.jpg", "scene0217_00/00120.jpg" ], "turn_id": 1, "scan_id": "scene0217_00", "origin_question": "Remember, have you seen any body loofah(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any body loofah(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0217_00/00160.jpg", "scene0217_00/00200.jpg", "scene0217_00/00240.jpg", "scene0217_00/00280.jpg" ], "turn_id": 2, "scan_id": "scene0217_00", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the light-colored rectangular wooden desk with drawers\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the light-colored rectangular wooden desk with drawers\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0217_00/00320.jpg", "scene0217_00/00360.jpg", "scene0217_00/00400.jpg", "scene0217_00/00440.jpg" ], "turn_id": 3, "scan_id": "scene0217_00", "origin_question": "What will be the first-time appearance order of the following categories: backpack, desk, window?", "option": [ "desk, window, backpack", "desk, backpack, window", "window, desk, backpack", "backpack, desk, window" ], "answer": "desk, window, backpack", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: backpack, desk, window? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['desk, window, backpack', 'desk, backpack, window', 'window, desk, backpack', 'backpack, desk, window'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0217_00/00480.jpg", "scene0217_00/00520.jpg", "scene0217_00/00560.jpg", "scene0217_00/00600.jpg" ], "turn_id": 4, "scan_id": "scene0217_00", "origin_question": "At the end of which turn were both of \"the vibrant red and black sturdy backpack\" and \"the standard-sized white paper with black text\" on your left side?", "option": [], "answer": "4", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the vibrant red and black sturdy backpack\" and \"the standard-sized white paper with black text\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0217_00/00640.jpg", "scene0217_00/00680.jpg", "scene0217_00/00720.jpg", "scene0217_00/00760.jpg" ], "turn_id": 5, "scan_id": "scene0217_00", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0217_00/00800.jpg", "scene0217_00/00840.jpg", "scene0217_00/00880.jpg", "scene0217_00/00920.jpg" ], "turn_id": 6, "scan_id": "scene0217_00", "origin_question": "In which turn did you see both \"the standard-sized white paper with black text\" and \"the rectangular white wooden door with a silver handle\" simultaneously? ", "option": [], "answer": "4", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the standard-sized white paper with black text\" and \"the rectangular white wooden door with a silver handle\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0217_00/00960.jpg", "scene0217_00/01000.jpg", "scene0217_00/01040.jpg", "scene0217_00/01080.jpg" ], "turn_id": 7, "scan_id": "scene0217_00", "origin_question": "Please recall \"the vibrant red and black sturdy backpack\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.87", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the vibrant red and black sturdy backpack\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0706_00/00000.jpg", "scene0706_00/00040.jpg", "scene0706_00/00080.jpg", "scene0706_00/00120.jpg", "scene0706_00/00160.jpg" ], "turn_id": 1, "scan_id": "scene0706_00", "origin_question": "Remember, have you seen any shoe(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any shoe(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0706_00/00200.jpg", "scene0706_00/00240.jpg", "scene0706_00/00280.jpg", "scene0706_00/00320.jpg", "scene0706_00/00360.jpg" ], "turn_id": 2, "scan_id": "scene0706_00", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the rectangular white wooden stand with a drawer\"; \"the white wooden single bed with a white headboard\"; \"the light brown rattan chair with a curved backrest\" ", "option": [ "the rectangular white wooden stand with a drawer", "the white wooden single bed with a white headboard", "the light brown rattan chair with a curved backrest" ], "answer": "the rectangular white wooden stand with a drawer", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the rectangular white wooden stand with a drawer\"; \"the white wooden single bed with a white headboard\"; \"the light brown rattan chair with a curved backrest\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular white wooden stand with a drawer', 'the white wooden single bed with a white headboard', 'the light brown rattan chair with a curved backrest'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0706_00/00400.jpg", "scene0706_00/00440.jpg", "scene0706_00/00480.jpg", "scene0706_00/00520.jpg", "scene0706_00/00560.jpg" ], "turn_id": 3, "scan_id": "scene0706_00", "origin_question": "In which turn did you see both \"the light brown rattan chair with a curved backrest\" and \"the flat-screen TV with a dark matte surface\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the light brown rattan chair with a curved backrest\" and \"the flat-screen TV with a dark matte surface\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0706_00/00600.jpg", "scene0706_00/00640.jpg", "scene0706_00/00680.jpg", "scene0706_00/00720.jpg", "scene0706_00/00760.jpg" ], "turn_id": 4, "scan_id": "scene0706_00", "origin_question": "Which direction is the rectangular white wooden stand with a drawer in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the rectangular white wooden stand with a drawer in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0706_00/00800.jpg", "scene0706_00/00840.jpg", "scene0706_00/00880.jpg", "scene0706_00/00920.jpg", "scene0706_00/00960.jpg" ], "turn_id": 5, "scan_id": "scene0706_00", "origin_question": "Which direction is the flat-screen TV with a dark matte surface in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the flat-screen TV with a dark matte surface in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0706_00/01000.jpg", "scene0706_00/01040.jpg", "scene0706_00/01080.jpg", "scene0706_00/01120.jpg", "scene0706_00/01160.jpg" ], "turn_id": 6, "scan_id": "scene0706_00", "origin_question": "Which two objects are on the same side of you now? \"the white wooden single bed with a white headboard\", \"the rectangular white wooden stand with a drawer\", \"the flat-screen TV with a dark matte surface\". ", "option": [ "the white wooden single bed with a white headboard, the rectangular white wooden stand with a drawer", "the white wooden single bed with a white headboard, the flat-screen TV with a dark matte surface", "the rectangular white wooden stand with a drawer, the flat-screen TV with a dark matte surface" ], "answer": "the white wooden single bed with a white headboard, the rectangular white wooden stand with a drawer", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the white wooden single bed with a white headboard\", \"the rectangular white wooden stand with a drawer\", \"the flat-screen TV with a dark matte surface\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the white wooden single bed with a white headboard, the rectangular white wooden stand with a drawer', 'the white wooden single bed with a white headboard, the flat-screen TV with a dark matte surface', 'the rectangular white wooden stand with a drawer, the flat-screen TV with a dark matte surface'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0706_00/01200.jpg", "scene0706_00/01240.jpg", "scene0706_00/01280.jpg", "scene0706_00/01320.jpg", "scene0706_00/01360.jpg" ], "turn_id": 7, "scan_id": "scene0706_00", "origin_question": "Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0706_00/01400.jpg", "scene0706_00/01440.jpg", "scene0706_00/01480.jpg", "scene0706_00/01520.jpg", "scene0706_00/01560.jpg" ], "turn_id": 8, "scan_id": "scene0706_00", "origin_question": "Is the standalone plant with long, slender green leaves to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the standalone plant with long, slender green leaves to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0706_00/01600.jpg", "scene0706_00/01640.jpg", "scene0706_00/01680.jpg", "scene0706_00/01720.jpg", "scene0706_00/01760.jpg" ], "turn_id": 9, "scan_id": "scene0706_00", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the curtain\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the curtain\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0558_00/00000.jpg", "scene0558_00/00030.jpg", "scene0558_00/00060.jpg", "scene0558_00/00090.jpg" ], "turn_id": 1, "scan_id": "scene0558_00", "origin_question": "Remember, have you seen any table(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any table(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0558_00/00120.jpg", "scene0558_00/00150.jpg", "scene0558_00/00180.jpg", "scene0558_00/00210.jpg" ], "turn_id": 2, "scan_id": "scene0558_00", "origin_question": "In which turn did you see both \"the shelf\" and \"the plant\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the shelf\" and \"the plant\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0558_00/00240.jpg", "scene0558_00/00270.jpg", "scene0558_00/00300.jpg", "scene0558_00/00330.jpg" ], "turn_id": 3, "scan_id": "scene0558_00", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "72.61", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0558_00/00360.jpg", "scene0558_00/00390.jpg", "scene0558_00/00420.jpg", "scene0558_00/00450.jpg" ], "turn_id": 4, "scan_id": "scene0558_00", "origin_question": "Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0558_00/00480.jpg", "scene0558_00/00510.jpg", "scene0558_00/00540.jpg", "scene0558_00/00570.jpg" ], "turn_id": 5, "scan_id": "scene0558_00", "origin_question": "Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "98.18", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0150_00/00000.jpg", "scene0150_00/00030.jpg", "scene0150_00/00060.jpg", "scene0150_00/00090.jpg" ], "turn_id": 1, "scan_id": "scene0150_00", "origin_question": "Remember, how many bottle(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many bottle(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 3 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0150_00/00120.jpg", "scene0150_00/00150.jpg", "scene0150_00/00180.jpg", "scene0150_00/00210.jpg" ], "turn_id": 2, "scan_id": "scene0150_00", "origin_question": "When did you first discover \"the small rectangular blackboard with a white surface\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the small rectangular blackboard with a white surface\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 3 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0150_00/00240.jpg", "scene0150_00/00270.jpg", "scene0150_00/00300.jpg", "scene0150_00/00330.jpg" ], "turn_id": 3, "scan_id": "scene0150_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 3 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0655_02/00000.jpg", "scene0655_02/00030.jpg", "scene0655_02/00060.jpg", "scene0655_02/00090.jpg" ], "turn_id": 1, "scan_id": "scene0655_02", "origin_question": "Remember, have you seen any chair(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any chair(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0655_02/00120.jpg", "scene0655_02/00150.jpg", "scene0655_02/00180.jpg", "scene0655_02/00210.jpg" ], "turn_id": 2, "scan_id": "scene0655_02", "origin_question": "At the end of which turn were both of \"the large rectangular grey table with a red border\" and \"the large rectangular blackboard with a white background\" on your left side?", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the large rectangular grey table with a red border\" and \"the large rectangular blackboard with a white background\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0655_02/00240.jpg", "scene0655_02/00270.jpg", "scene0655_02/00300.jpg", "scene0655_02/00330.jpg" ], "turn_id": 3, "scan_id": "scene0655_02", "origin_question": "When did you first discover \"the large rectangular grey table with a red border\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the large rectangular grey table with a red border\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0655_02/00360.jpg", "scene0655_02/00390.jpg", "scene0655_02/00430.jpg", "scene0655_02/00460.jpg" ], "turn_id": 4, "scan_id": "scene0655_02", "origin_question": "", "option": [], "answer": "", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0594_00/00000.jpg", "scene0594_00/00030.jpg", "scene0594_00/00060.jpg", "scene0594_00/00090.jpg" ], "turn_id": 1, "scan_id": "scene0594_00", "origin_question": "Remember, have you seen any table(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any table(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0594_00/00120.jpg", "scene0594_00/00150.jpg", "scene0594_00/00180.jpg", "scene0594_00/00210.jpg" ], "turn_id": 2, "scan_id": "scene0594_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0594_00/00240.jpg", "scene0594_00/00270.jpg", "scene0594_00/00300.jpg", "scene0594_00/00330.jpg" ], "turn_id": 3, "scan_id": "scene0594_00", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0594_00/00360.jpg", "scene0594_00/00390.jpg", "scene0594_00/00420.jpg", "scene0594_00/00450.jpg" ], "turn_id": 4, "scan_id": "scene0594_00", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0594_00/00480.jpg", "scene0594_00/00510.jpg", "scene0594_00/00540.jpg", "scene0594_00/00570.jpg" ], "turn_id": 5, "scan_id": "scene0594_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0594_00/00600.jpg", "scene0594_00/00630.jpg", "scene0594_00/00660.jpg", "scene0594_00/00690.jpg" ], "turn_id": 6, "scan_id": "scene0594_00", "origin_question": "Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "1mp3d_0004_region17/e8564d8248ba46fbac60fae0e93bad4f_i1_5.jpg", "1mp3d_0004_region17/e8564d8248ba46fbac60fae0e93bad4f_i2_5.jpg", "1mp3d_0004_region17/461f33bfc6de496e982dd04cfa0ad2d7_i1_4.jpg", "1mp3d_0004_region17/461f33bfc6de496e982dd04cfa0ad2d7_i2_4.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0004_region17", "origin_question": "Remember, how many pillow(s) have you seen so far? ", "option": [], "answer": "3", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many pillow(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "1mp3d_0004_region17/bb99864a663040d39ce11f7f1cc3f917_i1_4.jpg", "1mp3d_0004_region17/bb99864a663040d39ce11f7f1cc3f917_i2_4.jpg", "1mp3d_0004_region17/461f33bfc6de496e982dd04cfa0ad2d7_i1_5.jpg", "1mp3d_0004_region17/461f33bfc6de496e982dd04cfa0ad2d7_i2_5.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0004_region17", "origin_question": "In which turn were you closest to \"the tall, gray-brown ceramic vase with a narrow neck\"? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the tall, gray-brown ceramic vase with a narrow neck\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "1mp3d_0004_region17/461f33bfc6de496e982dd04cfa0ad2d7_i1_0.jpg", "1mp3d_0004_region17/461f33bfc6de496e982dd04cfa0ad2d7_i2_0.jpg", "1mp3d_0004_region17/e8564d8248ba46fbac60fae0e93bad4f_i1_2.jpg", "1mp3d_0004_region17/e8564d8248ba46fbac60fae0e93bad4f_i2_2.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0004_region17", "origin_question": "Among these three objects(\"the rectangular fireplace with dark marble surface\",\"the cabinet\",\"the tall, gray-brown ceramic vase with a narrow neck\" ), which one is the closest to you?", "option": [ "the rectangular fireplace with dark marble surface", "the cabinet", "the tall, gray-brown ceramic vase with a narrow neck" ], "answer": "the tall, gray-brown ceramic vase with a narrow neck", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects(\"the rectangular fireplace with dark marble surface\",\"the cabinet\",\"the tall, gray-brown ceramic vase with a narrow neck\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular fireplace with dark marble surface', 'the cabinet', 'the tall, gray-brown ceramic vase with a narrow neck'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0004_region17/e8564d8248ba46fbac60fae0e93bad4f_i1_1.jpg", "1mp3d_0004_region17/e8564d8248ba46fbac60fae0e93bad4f_i2_1.jpg", "1mp3d_0004_region17/e8564d8248ba46fbac60fae0e93bad4f_i1_0.jpg", "1mp3d_0004_region17/e8564d8248ba46fbac60fae0e93bad4f_i2_0.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0004_region17", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the white and blue bed with a beige headboard\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the white and blue bed with a beige headboard\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "1mp3d_0004_region17/4a416bd7e1a24ba99d8845c1b0743cb9_i1_0.jpg", "1mp3d_0004_region17/4a416bd7e1a24ba99d8845c1b0743cb9_i2_0.jpg", "1mp3d_0004_region17/4a416bd7e1a24ba99d8845c1b0743cb9_i1_1.jpg", "1mp3d_0004_region17/4a416bd7e1a24ba99d8845c1b0743cb9_i2_1.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0004_region17", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "1mp3d_0004_region17/41e03824893747c4a334895c8edd6b05_i1_2.jpg", "1mp3d_0004_region17/41e03824893747c4a334895c8edd6b05_i2_2.jpg", "1mp3d_0004_region17/41e03824893747c4a334895c8edd6b05_i1_3.jpg", "1mp3d_0004_region17/41e03824893747c4a334895c8edd6b05_i2_3.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0004_region17", "origin_question": "When did you first discover \"the tv\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the tv\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "1mp3d_0004_region17/41e03824893747c4a334895c8edd6b05_i1_4.jpg", "1mp3d_0004_region17/41e03824893747c4a334895c8edd6b05_i2_4.jpg", "1mp3d_0004_region17/41e03824893747c4a334895c8edd6b05_i1_5.jpg", "1mp3d_0004_region17/41e03824893747c4a334895c8edd6b05_i2_5.jpg" ], "turn_id": 7, "scan_id": "1mp3d_0004_region17", "origin_question": "Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0203_00/00000.jpg", "scene0203_00/00040.jpg", "scene0203_00/00080.jpg", "scene0203_00/00120.jpg" ], "turn_id": 1, "scan_id": "scene0203_00", "origin_question": "Remember, have you seen any paper(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any paper(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0203_00/00160.jpg", "scene0203_00/00200.jpg", "scene0203_00/00240.jpg", "scene0203_00/00280.jpg" ], "turn_id": 2, "scan_id": "scene0203_00", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the standard cylindrical plastic bottle with a blue label\", Y: \"the black round chair with four thin legs\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "farther away from both X and Y", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the standard cylindrical plastic bottle with a blue label\", Y: \"the black round chair with four thin legs\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0203_00/00320.jpg", "scene0203_00/00360.jpg", "scene0203_00/00400.jpg", "scene0203_00/00440.jpg" ], "turn_id": 3, "scan_id": "scene0203_00", "origin_question": "Which direction is the rectangular book with a white cover and blue designs in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the rectangular book with a white cover and blue designs in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0203_00/00480.jpg", "scene0203_00/00520.jpg", "scene0203_00/00560.jpg", "scene0203_00/00600.jpg" ], "turn_id": 4, "scan_id": "scene0203_00", "origin_question": "In which turn were you closest to \"the standard cylindrical plastic bottle with a blue label\"? ", "option": [], "answer": "2", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the standard cylindrical plastic bottle with a blue label\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0203_00/00640.jpg", "scene0203_00/00680.jpg", "scene0203_00/00720.jpg", "scene0203_00/00760.jpg" ], "turn_id": 5, "scan_id": "scene0203_00", "origin_question": "Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the standard cylindrical plastic bottle with a blue label\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the standard cylindrical plastic bottle with a blue label\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0203_00/00800.jpg", "scene0203_00/00840.jpg", "scene0203_00/00880.jpg", "scene0203_00/00920.jpg" ], "turn_id": 6, "scan_id": "scene0203_00", "origin_question": "In which turn did you see both \"the light brown rectangular wooden rack\" and \"the rectangular dark photo frame holding a picture\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the light brown rectangular wooden rack\" and \"the rectangular dark photo frame holding a picture\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0203_00/00960.jpg", "scene0203_00/01000.jpg", "scene0203_00/01040.jpg", "scene0203_00/01080.jpg" ], "turn_id": 7, "scan_id": "scene0203_00", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "1.33", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0203_00/01120.jpg", "scene0203_00/01160.jpg", "scene0203_00/01200.jpg", "scene0203_00/01240.jpg" ], "turn_id": 8, "scan_id": "scene0203_00", "origin_question": "At the end of which turn were both of \"the standard cylindrical plastic bottle with a blue label\" and \"the rectangular dark photo frame holding a picture\" on your right side?", "option": [], "answer": "4", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the standard cylindrical plastic bottle with a blue label\" and \"the rectangular dark photo frame holding a picture\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0203_00/01280.jpg", "scene0203_00/01320.jpg", "scene0203_00/01360.jpg", "scene0203_00/01400.jpg" ], "turn_id": 9, "scan_id": "scene0203_00", "origin_question": "Which direction is the light brown rectangular wooden rack in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 9 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the light brown rectangular wooden rack in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0653_00/00000.jpg", "scene0653_00/00040.jpg", "scene0653_00/00080.jpg", "scene0653_00/00120.jpg", "scene0653_00/00160.jpg" ], "turn_id": 1, "scan_id": "scene0653_00", "origin_question": "Remember, how many chair(s) have you seen so far? ", "option": [], "answer": "3", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many chair(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0653_00/00200.jpg", "scene0653_00/00240.jpg", "scene0653_00/00280.jpg", "scene0653_00/00320.jpg", "scene0653_00/00360.jpg" ], "turn_id": 2, "scan_id": "scene0653_00", "origin_question": "At the end of which turn were both of \"the white rectangular cardboard box with a handle\" and \"the vibrant orange hardcover book with a black and white cover\" on your right side?", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the white rectangular cardboard box with a handle\" and \"the vibrant orange hardcover book with a black and white cover\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0653_00/00400.jpg", "scene0653_00/00440.jpg", "scene0653_00/00480.jpg", "scene0653_00/00520.jpg", "scene0653_00/00560.jpg" ], "turn_id": 3, "scan_id": "scene0653_00", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the stylish black plastic keyboard with a standard layout\"; \"the vibrant orange hardcover book with a black and white cover\"; \"the rectangular board with rounded corners\" ", "option": [ "the stylish black plastic keyboard with a standard layout", "the vibrant orange hardcover book with a black and white cover", "the rectangular board with rounded corners" ], "answer": "the rectangular board with rounded corners", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the stylish black plastic keyboard with a standard layout\"; \"the vibrant orange hardcover book with a black and white cover\"; \"the rectangular board with rounded corners\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the stylish black plastic keyboard with a standard layout', 'the vibrant orange hardcover book with a black and white cover', 'the rectangular board with rounded corners'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0653_00/00600.jpg", "scene0653_00/00640.jpg", "scene0653_00/00680.jpg", "scene0653_00/00720.jpg", "scene0653_00/00760.jpg" ], "turn_id": 4, "scan_id": "scene0653_00", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the rectangular black plastic bin with a protruding lid\"; \"the rectangular black and white wrap\"; \"the vertical white matte divider made of laminated material\" ", "option": [ "the rectangular black plastic bin with a protruding lid", "the rectangular black and white wrap", "the vertical white matte divider made of laminated material" ], "answer": "the rectangular black plastic bin with a protruding lid", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the rectangular black plastic bin with a protruding lid\"; \"the rectangular black and white wrap\"; \"the vertical white matte divider made of laminated material\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular black plastic bin with a protruding lid', 'the rectangular black and white wrap', 'the vertical white matte divider made of laminated material'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0653_00/00800.jpg", "scene0653_00/00840.jpg", "scene0653_00/00880.jpg", "scene0653_00/00920.jpg", "scene0653_00/00960.jpg" ], "turn_id": 5, "scan_id": "scene0653_00", "origin_question": "Which direction is the rectangular black plastic bin with a protruding lid in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the rectangular black plastic bin with a protruding lid in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0653_00/01000.jpg", "scene0653_00/01040.jpg", "scene0653_00/01080.jpg", "scene0653_00/01120.jpg", "scene0653_00/01160.jpg" ], "turn_id": 6, "scan_id": "scene0653_00", "origin_question": "What will be the first-time appearance order of the following categories: backpack, bin, computer?", "option": [ "computer, backpack, bin", "bin, backpack, computer", "bin, computer, backpack", "backpack, bin, computer" ], "answer": "computer, backpack, bin", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: backpack, bin, computer? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['computer, backpack, bin', 'bin, backpack, computer', 'bin, computer, backpack', 'backpack, bin, computer'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0653_00/01200.jpg", "scene0653_00/01240.jpg", "scene0653_00/01280.jpg", "scene0653_00/01320.jpg", "scene0653_00/01360.jpg" ], "turn_id": 7, "scan_id": "scene0653_00", "origin_question": "Remember, have you seen any sheet(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any sheet(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0653_00/01400.jpg", "scene0653_00/01440.jpg", "scene0653_00/01480.jpg", "scene0653_00/01520.jpg", "scene0653_00/01560.jpg" ], "turn_id": 8, "scan_id": "scene0653_00", "origin_question": "In which turn were you closest to \"the modern black wireless mouse with a sleek design\"? ", "option": [], "answer": "4", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the modern black wireless mouse with a sleek design\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0653_00/01600.jpg", "scene0653_00/01640.jpg", "scene0653_00/01680.jpg", "scene0653_00/01720.jpg", "scene0653_00/01760.jpg" ], "turn_id": 9, "scan_id": "scene0653_00", "origin_question": "What will be the first-time appearance order of the following categories: backpack, bin, cup?", "option": [ "cup, backpack, bin", "backpack, bin, cup", "bin, backpack, cup", "cup, bin, backpack" ], "answer": "backpack, bin, cup", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: backpack, bin, cup? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['cup, backpack, bin', 'backpack, bin, cup', 'bin, backpack, cup', 'cup, bin, backpack'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "scene0653_00/01800.jpg", "scene0653_00/01840.jpg", "scene0653_00/01880.jpg", "scene0653_00/01920.jpg", "scene0653_00/01960.jpg" ], "turn_id": 10, "scan_id": "scene0653_00", "origin_question": "Among these three objects(\"the modern black wireless mouse with a sleek design\",\"the rectangular black and white wrap\",\"the ceramic cup with a white exterior and red interior\" ), which one is the farthest from you?", "option": [ "the modern black wireless mouse with a sleek design", "the rectangular black and white wrap", "the ceramic cup with a white exterior and red interior" ], "answer": "the modern black wireless mouse with a sleek design", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the modern black wireless mouse with a sleek design\",\"the rectangular black and white wrap\",\"the ceramic cup with a white exterior and red interior\" ), which one is the farthest from you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the modern black wireless mouse with a sleek design', 'the rectangular black and white wrap', 'the ceramic cup with a white exterior and red interior'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0057_region26/a94e4a04e9844299a87fa2776ba442e3_i1_5.jpg", "1mp3d_0057_region26/a94e4a04e9844299a87fa2776ba442e3_i2_5.jpg", "1mp3d_0057_region26/deaca0e4393847c5921593c4aa7392cc_i1_5.jpg", "1mp3d_0057_region26/deaca0e4393847c5921593c4aa7392cc_i2_5.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0057_region26", "origin_question": "Remember, have you seen any fruit(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any fruit(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0057_region26/a94e4a04e9844299a87fa2776ba442e3_i1_1.jpg", "1mp3d_0057_region26/a94e4a04e9844299a87fa2776ba442e3_i2_1.jpg", "1mp3d_0057_region26/a94e4a04e9844299a87fa2776ba442e3_i1_3.jpg", "1mp3d_0057_region26/a94e4a04e9844299a87fa2776ba442e3_i2_3.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0057_region26", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0057_region26/a94e4a04e9844299a87fa2776ba442e3_i1_4.jpg", "1mp3d_0057_region26/a94e4a04e9844299a87fa2776ba442e3_i2_4.jpg", "1mp3d_0057_region26/a94e4a04e9844299a87fa2776ba442e3_i1_0.jpg", "1mp3d_0057_region26/a94e4a04e9844299a87fa2776ba442e3_i2_0.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0057_region26", "origin_question": "Remember, have you seen any stairs(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any stairs(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0057_region26/b2b6f9bc6773421a90f965ba369a4786_i2_1.jpg", "1mp3d_0057_region26/b2b6f9bc6773421a90f965ba369a4786_i2_3.jpg", "1mp3d_0057_region26/b2b6f9bc6773421a90f965ba369a4786_i2_5.jpg", "1mp3d_0057_region26/b2b6f9bc6773421a90f965ba369a4786_i2_0.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0057_region26", "origin_question": "", "option": [], "answer": "", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0342_00/00000.jpg", "scene0342_00/00030.jpg", "scene0342_00/00060.jpg", "scene0342_00/00090.jpg" ], "turn_id": 1, "scan_id": "scene0342_00", "origin_question": "Remember, have you seen any drum(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any drum(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0342_00/00120.jpg", "scene0342_00/00150.jpg", "scene0342_00/00180.jpg", "scene0342_00/00210.jpg" ], "turn_id": 2, "scan_id": "scene0342_00", "origin_question": "In which turn were you farthest from \"the small black rectangular plastic table\"? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the small black rectangular plastic table\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0342_00/00240.jpg", "scene0342_00/00270.jpg", "scene0342_00/00300.jpg", "scene0342_00/00330.jpg" ], "turn_id": 3, "scan_id": "scene0342_00", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the medium-sized red and black backpack with unique shape\" from your position?", "option": [], "answer": "55.71", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the medium-sized red and black backpack with unique shape\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0342_00/00360.jpg", "scene0342_00/00390.jpg", "scene0342_00/00420.jpg", "scene0342_00/00450.jpg" ], "turn_id": 4, "scan_id": "scene0342_00", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "3.18", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0342_00/00480.jpg", "scene0342_00/00510.jpg", "scene0342_00/00540.jpg", "scene0342_00/00570.jpg" ], "turn_id": 5, "scan_id": "scene0342_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0356_00/00000.jpg", "scene0356_00/00040.jpg", "scene0356_00/00080.jpg", "scene0356_00/00120.jpg" ], "turn_id": 1, "scan_id": "scene0356_00", "origin_question": "Remember, have you seen any chair(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any chair(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0356_00/00160.jpg", "scene0356_00/00200.jpg", "scene0356_00/00240.jpg", "scene0356_00/00280.jpg" ], "turn_id": 2, "scan_id": "scene0356_00", "origin_question": "In which turn did you see both \"the light-colored wooden shelf with two rectangular levels\" and \"the black chair with a curved backrest\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the light-colored wooden shelf with two rectangular levels\" and \"the black chair with a curved backrest\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0356_00/00320.jpg", "scene0356_00/00360.jpg", "scene0356_00/00400.jpg", "scene0356_00/00440.jpg" ], "turn_id": 3, "scan_id": "scene0356_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0356_00/00480.jpg", "scene0356_00/00520.jpg", "scene0356_00/00560.jpg", "scene0356_00/00600.jpg" ], "turn_id": 4, "scan_id": "scene0356_00", "origin_question": "Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "106.92", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0356_00/00640.jpg", "scene0356_00/00680.jpg", "scene0356_00/00720.jpg", "scene0356_00/00760.jpg" ], "turn_id": 5, "scan_id": "scene0356_00", "origin_question": "Which direction is the radiator in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the radiator in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0356_00/00800.jpg", "scene0356_00/00840.jpg", "scene0356_00/00880.jpg", "scene0356_00/00920.jpg" ], "turn_id": 6, "scan_id": "scene0356_00", "origin_question": "Which direction is the light brown wooden dresser with three drawers in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the light brown wooden dresser with three drawers in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0356_00/00960.jpg", "scene0356_00/01000.jpg", "scene0356_00/01040.jpg", "scene0356_00/01080.jpg" ], "turn_id": 7, "scan_id": "scene0356_00", "origin_question": "At the end of which turn were both of \"the light-colored wooden shelf with two rectangular levels\" and \"the light brown wooden dresser with three drawers\" on your right side?", "option": [], "answer": "3", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the light-colored wooden shelf with two rectangular levels\" and \"the light brown wooden dresser with three drawers\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0356_00/01120.jpg", "scene0356_00/01160.jpg", "scene0356_00/01200.jpg", "scene0356_00/01240.jpg" ], "turn_id": 8, "scan_id": "scene0356_00", "origin_question": "Compared to your position at the end of the 5 turn, are you now closer or farther away from \"the light brown rectangular wooden desk\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 5 turn, are you now closer or farther away from \"the light brown rectangular wooden desk\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0329_00/00000.jpg", "scene0329_00/00040.jpg", "scene0329_00/00080.jpg", "scene0329_00/00120.jpg" ], "turn_id": 1, "scan_id": "scene0329_00", "origin_question": "Remember, have you seen any ball(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any ball(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0329_00/00160.jpg", "scene0329_00/00200.jpg", "scene0329_00/00240.jpg", "scene0329_00/00280.jpg" ], "turn_id": 2, "scan_id": "scene0329_00", "origin_question": "Remember, have you seen any bag(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any bag(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0329_00/00320.jpg", "scene0329_00/00360.jpg", "scene0329_00/00400.jpg", "scene0329_00/00440.jpg" ], "turn_id": 3, "scan_id": "scene0329_00", "origin_question": "Which direction is the square, grey table made of sturdy material in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the square, grey table made of sturdy material in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0329_00/00480.jpg", "scene0329_00/00520.jpg", "scene0329_00/00560.jpg", "scene0329_00/00600.jpg" ], "turn_id": 4, "scan_id": "scene0329_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0329_00/00640.jpg", "scene0329_00/00680.jpg", "scene0329_00/00720.jpg", "scene0329_00/00760.jpg" ], "turn_id": 5, "scan_id": "scene0329_00", "origin_question": "When did you first discover \"the cylindrical red fire extinguisher with a metal top\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the cylindrical red fire extinguisher with a metal top\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0329_00/00800.jpg", "scene0329_00/00840.jpg", "scene0329_00/00880.jpg", "scene0329_00/00920.jpg" ], "turn_id": 6, "scan_id": "scene0329_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0329_00/00960.jpg", "scene0329_00/01000.jpg", "scene0329_00/01040.jpg", "scene0329_00/01080.jpg" ], "turn_id": 7, "scan_id": "scene0329_00", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the rectangular, mostly white blackboard with colorful markings\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the rectangular, mostly white blackboard with colorful markings\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0329_00/01120.jpg", "scene0329_00/01160.jpg", "scene0329_00/01200.jpg", "scene0329_00/01240.jpg" ], "turn_id": 8, "scan_id": "scene0329_00", "origin_question": "Please recall \"the rectangular, mostly white blackboard with colorful markings\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "4.78", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the rectangular, mostly white blackboard with colorful markings\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0329_00/01280.jpg", "scene0329_00/01320.jpg", "scene0329_00/01360.jpg", "scene0329_00/01400.jpg" ], "turn_id": 9, "scan_id": "scene0329_00", "origin_question": "Which two objects are on the same side of you now? \"the black rectangular bag made of leather and fabric\", \"the large rectangular white board made of wood\", \"the rectangular black computer tower made of metal\". ", "option": [ "the black rectangular bag made of leather and fabric, the large rectangular white board made of wood", "the black rectangular bag made of leather and fabric, the rectangular black computer tower made of metal", "the large rectangular white board made of wood, the rectangular black computer tower made of metal" ], "answer": "the large rectangular white board made of wood, the rectangular black computer tower made of metal", "user_message": "For the 9 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the black rectangular bag made of leather and fabric\", \"the large rectangular white board made of wood\", \"the rectangular black computer tower made of metal\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the black rectangular bag made of leather and fabric, the large rectangular white board made of wood', 'the black rectangular bag made of leather and fabric, the rectangular black computer tower made of metal', 'the large rectangular white board made of wood, the rectangular black computer tower made of metal'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0684_01/00000.jpg", "scene0684_01/00030.jpg", "scene0684_01/00060.jpg", "scene0684_01/00090.jpg" ], "turn_id": 1, "scan_id": "scene0684_01", "origin_question": "Remember, have you seen any bag(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any bag(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0684_01/00120.jpg", "scene0684_01/00150.jpg", "scene0684_01/00180.jpg", "scene0684_01/00210.jpg" ], "turn_id": 2, "scan_id": "scene0684_01", "origin_question": "When did you first discover \"the window\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the window\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0684_01/00240.jpg", "scene0684_01/00270.jpg", "scene0684_01/00300.jpg", "scene0684_01/00330.jpg" ], "turn_id": 3, "scan_id": "scene0684_01", "origin_question": "Please recall \"the dark-colored modern digital telephone with a touchscreen\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.05", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the dark-colored modern digital telephone with a touchscreen\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0684_01/00380.jpg", "scene0684_01/00410.jpg", "scene0684_01/00440.jpg", "scene0684_01/00470.jpg" ], "turn_id": 4, "scan_id": "scene0684_01", "origin_question": "Which two objects are on the same side of you now? \"the rectangular light-colored wooden table\", \"the dark-colored modern digital telephone with a touchscreen\", \"the rectangular white kitchen cabinet made of sturdy material\". ", "option": [ "the rectangular light-colored wooden table, the dark-colored modern digital telephone with a touchscreen", "the rectangular light-colored wooden table, the rectangular white kitchen cabinet made of sturdy material", "the dark-colored modern digital telephone with a touchscreen, the rectangular white kitchen cabinet made of sturdy material" ], "answer": "the rectangular light-colored wooden table, the dark-colored modern digital telephone with a touchscreen", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the rectangular light-colored wooden table\", \"the dark-colored modern digital telephone with a touchscreen\", \"the rectangular white kitchen cabinet made of sturdy material\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular light-colored wooden table, the dark-colored modern digital telephone with a touchscreen', 'the rectangular light-colored wooden table, the rectangular white kitchen cabinet made of sturdy material', 'the dark-colored modern digital telephone with a touchscreen, the rectangular white kitchen cabinet made of sturdy material'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0004_region21/a085b1aee4be45b6982a1f7b601654eb_i1_0.jpg", "1mp3d_0004_region21/a085b1aee4be45b6982a1f7b601654eb_i2_0.jpg", "1mp3d_0004_region21/126dc91ccd544962bac6d80bded42afa_i1_2.jpg", "1mp3d_0004_region21/126dc91ccd544962bac6d80bded42afa_i2_2.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0004_region21", "origin_question": "Remember, have you seen any picture(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any picture(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0004_region21/126dc91ccd544962bac6d80bded42afa_i1_3.jpg", "1mp3d_0004_region21/126dc91ccd544962bac6d80bded42afa_i2_3.jpg", "1mp3d_0004_region21/a085b1aee4be45b6982a1f7b601654eb_i1_1.jpg", "1mp3d_0004_region21/a085b1aee4be45b6982a1f7b601654eb_i2_1.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0004_region21", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the round, white ceramic flowerpot with a glossy surface\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the round, white ceramic flowerpot with a glossy surface\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "1mp3d_0004_region21/7080fd35b3d54b679b01796cf657144c_i1_5.jpg", "1mp3d_0004_region21/7080fd35b3d54b679b01796cf657144c_i2_5.jpg", "1mp3d_0004_region21/4a7c0e0d49dc4275b7d1aaa2db7cb015_i1_5.jpg", "1mp3d_0004_region21/4a7c0e0d49dc4275b7d1aaa2db7cb015_i2_5.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0004_region21", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular glass window with a white frame\", Y: \"the round, white ceramic flowerpot with a glossy surface\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to X and farther away from Y", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular glass window with a white frame\", Y: \"the round, white ceramic flowerpot with a glossy surface\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "1mp3d_0004_region21/7080fd35b3d54b679b01796cf657144c_i1_0.jpg", "1mp3d_0004_region21/7080fd35b3d54b679b01796cf657144c_i2_0.jpg", "1mp3d_0004_region21/4a7c0e0d49dc4275b7d1aaa2db7cb015_i1_1.jpg", "1mp3d_0004_region21/4a7c0e0d49dc4275b7d1aaa2db7cb015_i2_1.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0004_region21", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "1mp3d_0004_region21/4a7c0e0d49dc4275b7d1aaa2db7cb015_i1_2.jpg", "1mp3d_0004_region21/4a7c0e0d49dc4275b7d1aaa2db7cb015_i2_2.jpg", "1mp3d_0004_region21/4dc7d1e0b20348be87e07114a1d66580_i1_5.jpg", "1mp3d_0004_region21/4dc7d1e0b20348be87e07114a1d66580_i2_5.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0004_region21", "origin_question": "Which direction is the round, white ceramic flowerpot with a glossy surface in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the round, white ceramic flowerpot with a glossy surface in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "1mp3d_0004_region21/692d012ba3474464918840df990756cd_i1_2.jpg", "1mp3d_0004_region21/692d012ba3474464918840df990756cd_i2_2.jpg", "1mp3d_0004_region21/692d012ba3474464918840df990756cd_i1_3.jpg", "1mp3d_0004_region21/692d012ba3474464918840df990756cd_i2_3.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0004_region21", "origin_question": "When did you first discover \"the large, circular blue and brown globe\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the large, circular blue and brown globe\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "1mp3d_0004_region21/692d012ba3474464918840df990756cd_i1_4.jpg", "1mp3d_0004_region21/692d012ba3474464918840df990756cd_i2_4.jpg", "1mp3d_0004_region21/4dc7d1e0b20348be87e07114a1d66580_i1_0.jpg", "1mp3d_0004_region21/4dc7d1e0b20348be87e07114a1d66580_i2_0.jpg" ], "turn_id": 7, "scan_id": "1mp3d_0004_region21", "origin_question": "In which turn did you see both \"the large, circular blue and brown globe\" and \"the medium-sized lamp with a white cylindrical lampshade\" simultaneously? ", "option": [], "answer": "6", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the large, circular blue and brown globe\" and \"the medium-sized lamp with a white cylindrical lampshade\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0088_01/00000.jpg", "scene0088_01/00040.jpg", "scene0088_01/00080.jpg", "scene0088_01/00120.jpg" ], "turn_id": 1, "scan_id": "scene0088_01", "origin_question": "Remember, have you seen any bathtub(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any bathtub(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0088_01/00160.jpg", "scene0088_01/00200.jpg", "scene0088_01/00240.jpg", "scene0088_01/00280.jpg" ], "turn_id": 2, "scan_id": "scene0088_01", "origin_question": "In which turn did you see both \"the massive slate blackboard with a white surface\" and \"the door\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the massive slate blackboard with a white surface\" and \"the door\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0088_01/00320.jpg", "scene0088_01/00360.jpg", "scene0088_01/00400.jpg", "scene0088_01/00440.jpg" ], "turn_id": 3, "scan_id": "scene0088_01", "origin_question": "In which turn were you closest to \"the massive slate blackboard with a white surface\"? ", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the massive slate blackboard with a white surface\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0088_01/00480.jpg", "scene0088_01/00520.jpg", "scene0088_01/00560.jpg", "scene0088_01/00600.jpg" ], "turn_id": 4, "scan_id": "scene0088_01", "origin_question": "By the end of the 3 turn, were you getting closer to or farther away from these two objects: X: \"the massive slate blackboard with a white surface\", Y: \"the eraser\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "farther away from both X and Y", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 3 turn, were you getting closer to or farther away from these two objects: X: \"the massive slate blackboard with a white surface\", Y: \"the eraser\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "scene0088_01/00640.jpg", "scene0088_01/00680.jpg", "scene0088_01/00720.jpg", "scene0088_01/00760.jpg" ], "turn_id": 5, "scan_id": "scene0088_01", "origin_question": "Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Right", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0088_01/00800.jpg", "scene0088_01/00840.jpg", "scene0088_01/00880.jpg", "scene0088_01/00920.jpg" ], "turn_id": 6, "scan_id": "scene0088_01", "origin_question": "At the end of which turn were both of \"the massive slate blackboard with a white surface\" and \"the eraser\" on your left side?", "option": [], "answer": "2", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the massive slate blackboard with a white surface\" and \"the eraser\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47331331/Validation_47331331_47331331_180735.312.png", "Validation_47331331/Validation_47331331_47331331_180736.811.png", "Validation_47331331/Validation_47331331_47331331_180738.310.png", "Validation_47331331/Validation_47331331_47331331_180739.810.png" ], "turn_id": 1, "scan_id": "Validation_47331331", "origin_question": "Remember, have you seen any album(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any album(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "Validation_47331331/Validation_47331331_47331331_180741.309.png", "Validation_47331331/Validation_47331331_47331331_180742.809.png", "Validation_47331331/Validation_47331331_47331331_180744.308.png", "Validation_47331331/Validation_47331331_47331331_180745.807.png" ], "turn_id": 2, "scan_id": "Validation_47331331", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the gray rectangular metal bed with subtle curves\"; \"the soft, round brown toy bear\"; \"the light beige fleece blanket with the plush texture\" ", "option": [ "the gray rectangular metal bed with subtle curves", "the soft, round brown toy bear", "the light beige fleece blanket with the plush texture" ], "answer": "the soft, round brown toy bear", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the gray rectangular metal bed with subtle curves\"; \"the soft, round brown toy bear\"; \"the light beige fleece blanket with the plush texture\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the gray rectangular metal bed with subtle curves', 'the soft, round brown toy bear', 'the light beige fleece blanket with the plush texture'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47331331/Validation_47331331_47331331_180747.307.png", "Validation_47331331/Validation_47331331_47331331_180748.806.png", "Validation_47331331/Validation_47331331_47331331_180750.306.png", "Validation_47331331/Validation_47331331_47331331_180751.805.png" ], "turn_id": 3, "scan_id": "Validation_47331331", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the large rectangular yellow pillow with soft cotton fabric\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the large rectangular yellow pillow with soft cotton fabric\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_47331331/Validation_47331331_47331331_180753.304.png", "Validation_47331331/Validation_47331331_47331331_180754.804.png", "Validation_47331331/Validation_47331331_47331331_180756.303.png", "Validation_47331331/Validation_47331331_47331331_180757.802.png" ], "turn_id": 4, "scan_id": "Validation_47331331", "origin_question": "Which two objects are on the same side of you now? \"the gray rectangular metal bed with subtle curves\", \"the large rectangular yellow pillow with soft cotton fabric\", \"the compact white printer with black accents\". ", "option": [ "the gray rectangular metal bed with subtle curves, the large rectangular yellow pillow with soft cotton fabric", "the gray rectangular metal bed with subtle curves, the compact white printer with black accents", "the large rectangular yellow pillow with soft cotton fabric, the compact white printer with black accents" ], "answer": "the gray rectangular metal bed with subtle curves, the large rectangular yellow pillow with soft cotton fabric", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the gray rectangular metal bed with subtle curves\", \"the large rectangular yellow pillow with soft cotton fabric\", \"the compact white printer with black accents\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the gray rectangular metal bed with subtle curves, the large rectangular yellow pillow with soft cotton fabric', 'the gray rectangular metal bed with subtle curves, the compact white printer with black accents', 'the large rectangular yellow pillow with soft cotton fabric, the compact white printer with black accents'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_47331331/Validation_47331331_47331331_180759.302.png", "Validation_47331331/Validation_47331331_47331331_180760.801.png", "Validation_47331331/Validation_47331331_47331331_180762.301.png", "Validation_47331331/Validation_47331331_47331331_180763.800.png" ], "turn_id": 5, "scan_id": "Validation_47331331", "origin_question": "When did you first discover \"the large rectangular yellow pillow with soft cotton fabric\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the large rectangular yellow pillow with soft cotton fabric\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47331331/Validation_47331331_47331331_180765.299.png", "Validation_47331331/Validation_47331331_47331331_180766.815.png", "Validation_47331331/Validation_47331331_47331331_180768.315.png", "Validation_47331331/Validation_47331331_47331331_180769.814.png" ], "turn_id": 6, "scan_id": "Validation_47331331", "origin_question": "Which direction is the light beige fleece blanket with the plush texture in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the light beige fleece blanket with the plush texture in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_47331331/Validation_47331331_47331331_180771.314.png", "Validation_47331331/Validation_47331331_47331331_180772.813.png", "Validation_47331331/Validation_47331331_47331331_180774.312.png", "Validation_47331331/Validation_47331331_47331331_180775.812.png" ], "turn_id": 7, "scan_id": "Validation_47331331", "origin_question": "At the end of which turn were both of \"the light beige fleece blanket with the plush texture\" and \"the sleek, dark wooden case with the smooth finish\" on your left side?", "option": [], "answer": "1", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the light beige fleece blanket with the plush texture\" and \"the sleek, dark wooden case with the smooth finish\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "Validation_42899679/Validation_42899679_42899679_790358.038.png", "Validation_42899679/Validation_42899679_42899679_790360.020.png", "Validation_42899679/Validation_42899679_42899679_790362.019.png", "Validation_42899679/Validation_42899679_42899679_790364.019.png", "Validation_42899679/Validation_42899679_42899679_790366.018.png" ], "turn_id": 1, "scan_id": "Validation_42899679", "origin_question": "Remember, how many pillow(s) have you seen so far? ", "option": [], "answer": "3", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many pillow(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_42899679/Validation_42899679_42899679_790368.017.png", "Validation_42899679/Validation_42899679_42899679_790370.016.png", "Validation_42899679/Validation_42899679_42899679_790372.015.png", "Validation_42899679/Validation_42899679_42899679_790374.015.png", "Validation_42899679/Validation_42899679_42899679_790376.014.png" ], "turn_id": 2, "scan_id": "Validation_42899679", "origin_question": "Which direction is the rectangular brown cardboard box in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the rectangular brown cardboard box in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_42899679/Validation_42899679_42899679_790378.013.png", "Validation_42899679/Validation_42899679_42899679_790380.012.png", "Validation_42899679/Validation_42899679_42899679_790382.011.png", "Validation_42899679/Validation_42899679_42899679_790384.010.png", "Validation_42899679/Validation_42899679_42899679_790386.010.png" ], "turn_id": 3, "scan_id": "Validation_42899679", "origin_question": "Which direction is the compact navy blue soft fabric stool in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the compact navy blue soft fabric stool in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_42899679/Validation_42899679_42899679_790388.009.png", "Validation_42899679/Validation_42899679_42899679_790390.008.png", "Validation_42899679/Validation_42899679_42899679_790392.007.png", "Validation_42899679/Validation_42899679_42899679_790394.023.png", "Validation_42899679/Validation_42899679_42899679_790396.022.png" ], "turn_id": 4, "scan_id": "Validation_42899679", "origin_question": "Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the tall plant with the dense canopy of deep green leaves\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the tall plant with the dense canopy of deep green leaves\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_42899679/Validation_42899679_42899679_790398.021.png", "Validation_42899679/Validation_42899679_42899679_790400.020.png", "Validation_42899679/Validation_42899679_42899679_790402.020.png", "Validation_42899679/Validation_42899679_42899679_790404.019.png", "Validation_42899679/Validation_42899679_42899679_790406.018.png" ], "turn_id": 5, "scan_id": "Validation_42899679", "origin_question": "", "option": [], "answer": "", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_42899679/Validation_42899679_42899679_790408.017.png", "Validation_42899679/Validation_42899679_42899679_790410.016.png", "Validation_42899679/Validation_42899679_42899679_790412.016.png", "Validation_42899679/Validation_42899679_42899679_790414.015.png", "Validation_42899679/Validation_42899679_42899679_790416.014.png" ], "turn_id": 6, "scan_id": "Validation_42899679", "origin_question": "", "option": [], "answer": "", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_42899679/Validation_42899679_42899679_790418.013.png", "Validation_42899679/Validation_42899679_42899679_790420.012.png", "Validation_42899679/Validation_42899679_42899679_790422.012.png", "Validation_42899679/Validation_42899679_42899679_790424.011.png", "Validation_42899679/Validation_42899679_42899679_790426.010.png" ], "turn_id": 7, "scan_id": "Validation_42899679", "origin_question": "Assuming the direction you are facing at the end of the turn 6 is forward, did you move a certain distance forward or backward from that position?", "option": [ "Forward", "Backward" ], "answer": "Forward", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 6 is forward, did you move a certain distance forward or backward from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Forward', 'Backward'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_42899679/Validation_42899679_42899679_790428.009.png", "Validation_42899679/Validation_42899679_42899679_790430.008.png", "Validation_42899679/Validation_42899679_42899679_790432.007.png", "Validation_42899679/Validation_42899679_42899679_790434.007.png", "Validation_42899679/Validation_42899679_42899679_790436.006.png" ], "turn_id": 8, "scan_id": "Validation_42899679", "origin_question": "Please recall \"the versatile grey or navy shoe with breathable fabric\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "4.14", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the versatile grey or navy shoe with breathable fabric\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_42899679/Validation_42899679_42899679_790438.022.png", "Validation_42899679/Validation_42899679_42899679_790440.021.png", "Validation_42899679/Validation_42899679_42899679_790442.020.png", "Validation_42899679/Validation_42899679_42899679_790444.019.png", "Validation_42899679/Validation_42899679_42899679_790446.018.png" ], "turn_id": 9, "scan_id": "Validation_42899679", "origin_question": "Is the rectangular brown cardboard box to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the rectangular brown cardboard box to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_42899679/Validation_42899679_42899679_790448.017.png", "Validation_42899679/Validation_42899679_42899679_790450.017.png", "Validation_42899679/Validation_42899679_42899679_790452.016.png", "Validation_42899679/Validation_42899679_42899679_790454.015.png", "Validation_42899679/Validation_42899679_42899679_790456.014.png" ], "turn_id": 10, "scan_id": "Validation_42899679", "origin_question": "Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Right", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_45261121/Validation_45261121_45261121_3415.076.png", "Validation_45261121/Validation_45261121_45261121_3417.059.png", "Validation_45261121/Validation_45261121_45261121_3419.058.png", "Validation_45261121/Validation_45261121_45261121_3421.057.png" ], "turn_id": 1, "scan_id": "Validation_45261121", "origin_question": "Remember, have you seen any lamp(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any lamp(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_45261121/Validation_45261121_45261121_3423.056.png", "Validation_45261121/Validation_45261121_45261121_3425.055.png", "Validation_45261121/Validation_45261121_45261121_3427.055.png", "Validation_45261121/Validation_45261121_45261121_3429.070.png" ], "turn_id": 2, "scan_id": "Validation_45261121", "origin_question": "In which turn were you farthest from \"the sleek white compact refrigerator with the rectangular shape\"? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the sleek white compact refrigerator with the rectangular shape\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_45261121/Validation_45261121_45261121_3431.070.png", "Validation_45261121/Validation_45261121_45261121_3433.069.png", "Validation_45261121/Validation_45261121_45261121_3435.068.png", "Validation_45261121/Validation_45261121_45261121_3437.067.png" ], "turn_id": 3, "scan_id": "Validation_45261121", "origin_question": "When did you first discover \"the spacious rectangular table with the rich brown top\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the spacious rectangular table with the rich brown top\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "Validation_45261121/Validation_45261121_45261121_3439.066.png", "Validation_45261121/Validation_45261121_45261121_3441.066.png", "Validation_45261121/Validation_45261121_45261121_3443.065.png", "Validation_45261121/Validation_45261121_45261121_3445.064.png" ], "turn_id": 4, "scan_id": "Validation_45261121", "origin_question": "Among these three objects(\"the sleek black and stainless steel rectangular oven\",\"the modern black rectangular monitor with the plastic frame\",\"the wooden stool with the warm brown seat and metal base\" ), which one is the closest to you?", "option": [ "the sleek black and stainless steel rectangular oven", "the modern black rectangular monitor with the plastic frame", "the wooden stool with the warm brown seat and metal base" ], "answer": "the wooden stool with the warm brown seat and metal base", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects(\"the sleek black and stainless steel rectangular oven\",\"the modern black rectangular monitor with the plastic frame\",\"the wooden stool with the warm brown seat and metal base\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the sleek black and stainless steel rectangular oven', 'the modern black rectangular monitor with the plastic frame', 'the wooden stool with the warm brown seat and metal base'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_45261121/Validation_45261121_45261121_3447.063.png", "Validation_45261121/Validation_45261121_45261121_3449.062.png", "Validation_45261121/Validation_45261121_45261121_3451.061.png", "Validation_45261121/Validation_45261121_45261121_3453.061.png" ], "turn_id": 5, "scan_id": "Validation_45261121", "origin_question": "What will be the first-time appearance order of the following categories: chair, monitor, refrigerator?", "option": [ "refrigerator, chair, monitor", "monitor, chair, refrigerator", "chair, monitor, refrigerator", "monitor, refrigerator, chair" ], "answer": "refrigerator, chair, monitor", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: chair, monitor, refrigerator? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['refrigerator, chair, monitor', 'monitor, chair, refrigerator', 'chair, monitor, refrigerator', 'monitor, refrigerator, chair'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_45261121/Validation_45261121_45261121_3455.060.png", "Validation_45261121/Validation_45261121_45261121_3457.059.png", "Validation_45261121/Validation_45261121_45261121_3459.058.png", "Validation_45261121/Validation_45261121_45261121_3461.057.png" ], "turn_id": 6, "scan_id": "Validation_45261121", "origin_question": "When did you first discover \"the sleek black and stainless steel rectangular oven\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the sleek black and stainless steel rectangular oven\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_45261121/Validation_45261121_45261121_3463.056.png", "Validation_45261121/Validation_45261121_45261121_3465.056.png", "Validation_45261121/Validation_45261121_45261121_3467.055.png", "Validation_45261121/Validation_45261121_45261121_3469.071.png" ], "turn_id": 7, "scan_id": "Validation_45261121", "origin_question": "Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the large, modern black stove with the glossy finish\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the large, modern black stove with the glossy finish\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42899688/Validation_42899688_42899688_789312.524.png", "Validation_42899688/Validation_42899688_42899688_789314.507.png", "Validation_42899688/Validation_42899688_42899688_789316.506.png", "Validation_42899688/Validation_42899688_42899688_789318.505.png" ], "turn_id": 1, "scan_id": "Validation_42899688", "origin_question": "Remember, have you seen any table(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any table(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_42899688/Validation_42899688_42899688_789320.504.png", "Validation_42899688/Validation_42899688_42899688_789322.503.png", "Validation_42899688/Validation_42899688_42899688_789324.503.png", "Validation_42899688/Validation_42899688_42899688_789326.502.png" ], "turn_id": 2, "scan_id": "Validation_42899688", "origin_question": "In which turn were you farthest from \"the bed\"? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the bed\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_42899688/Validation_42899688_42899688_789328.501.png", "Validation_42899688/Validation_42899688_42899688_789330.500.png", "Validation_42899688/Validation_42899688_42899688_789332.499.png", "Validation_42899688/Validation_42899688_42899688_789334.499.png" ], "turn_id": 3, "scan_id": "Validation_42899688", "origin_question": "When did you first discover \"the compact, cube-shaped light gray table with shelves\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the compact, cube-shaped light gray table with shelves\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_42899688/Validation_42899688_42899688_789336.514.png", "Validation_42899688/Validation_42899688_42899688_789338.514.png", "Validation_42899688/Validation_42899688_42899688_789340.513.png", "Validation_42899688/Validation_42899688_42899688_789342.512.png" ], "turn_id": 4, "scan_id": "Validation_42899688", "origin_question": "In which turn did you see both \"the compact, cube-shaped light gray table with shelves\" and \"the bed\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the compact, cube-shaped light gray table with shelves\" and \"the bed\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_42899688/Validation_42899688_42899688_789344.511.png", "Validation_42899688/Validation_42899688_42899688_789346.510.png", "Validation_42899688/Validation_42899688_42899688_789348.510.png", "Validation_42899688/Validation_42899688_42899688_789350.509.png" ], "turn_id": 5, "scan_id": "Validation_42899688", "origin_question": "How far is your current position from where you were at the end of the turn 4? (in meters)", "option": [], "answer": "1.19", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 4? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_42899688/Validation_42899688_42899688_789352.508.png", "Validation_42899688/Validation_42899688_42899688_789354.507.png", "Validation_42899688/Validation_42899688_42899688_789356.506.png", "Validation_42899688/Validation_42899688_42899688_789358.505.png" ], "turn_id": 6, "scan_id": "Validation_42899688", "origin_question": "", "option": [], "answer": "", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_45663164/Validation_45663164_45663164_65931.835.png", "Validation_45663164/Validation_45663164_45663164_65933.817.png", "Validation_45663164/Validation_45663164_45663164_65935.816.png", "Validation_45663164/Validation_45663164_45663164_65937.815.png", "Validation_45663164/Validation_45663164_45663164_65939.815.png" ], "turn_id": 1, "scan_id": "Validation_45663164", "origin_question": "Remember, have you seen any backpack(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any backpack(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_45663164/Validation_45663164_45663164_65941.814.png", "Validation_45663164/Validation_45663164_45663164_65943.813.png", "Validation_45663164/Validation_45663164_45663164_65945.812.png", "Validation_45663164/Validation_45663164_45663164_65947.828.png", "Validation_45663164/Validation_45663164_45663164_65949.827.png" ], "turn_id": 2, "scan_id": "Validation_45663164", "origin_question": "Please recall \"the geometric metal chandelier with the sleek finish\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.69", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the geometric metal chandelier with the sleek finish\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_45663164/Validation_45663164_45663164_65951.827.png", "Validation_45663164/Validation_45663164_45663164_65953.826.png", "Validation_45663164/Validation_45663164_45663164_65955.825.png", "Validation_45663164/Validation_45663164_45663164_65957.824.png", "Validation_45663164/Validation_45663164_45663164_65959.823.png" ], "turn_id": 3, "scan_id": "Validation_45663164", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the sleek, medium-sized sink\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the sleek, medium-sized sink\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_45663164/Validation_45663164_45663164_65961.823.png", "Validation_45663164/Validation_45663164_45663164_65963.822.png", "Validation_45663164/Validation_45663164_45663164_65965.821.png", "Validation_45663164/Validation_45663164_45663164_65967.820.png", "Validation_45663164/Validation_45663164_45663164_65969.819.png" ], "turn_id": 4, "scan_id": "Validation_45663164", "origin_question": "In which turn did you see both \"the sleek, white, stainless steel refrigerator\" and \"the modern black and silver oven with stainless steel design\" simultaneously? ", "option": [], "answer": "4", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the sleek, white, stainless steel refrigerator\" and \"the modern black and silver oven with stainless steel design\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_45663164/Validation_45663164_45663164_65971.819.png", "Validation_45663164/Validation_45663164_45663164_65973.818.png", "Validation_45663164/Validation_45663164_45663164_65975.817.png", "Validation_45663164/Validation_45663164_45663164_65977.816.png", "Validation_45663164/Validation_45663164_45663164_65979.816.png" ], "turn_id": 5, "scan_id": "Validation_45663164", "origin_question": "Please recall \"the matte black monitor with the sleek modern design\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.29", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the matte black monitor with the sleek modern design\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_45663164/Validation_45663164_45663164_65981.815.png", "Validation_45663164/Validation_45663164_45663164_65983.814.png", "Validation_45663164/Validation_45663164_45663164_65985.813.png", "Validation_45663164/Validation_45663164_45663164_65987.812.png", "Validation_45663164/Validation_45663164_45663164_65989.828.png" ], "turn_id": 6, "scan_id": "Validation_45663164", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the sleek, medium-sized sink\" from your position?", "option": [], "answer": "115.14", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the sleek, medium-sized sink\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "Validation_45663164/Validation_45663164_45663164_65991.827.png", "Validation_45663164/Validation_45663164_45663164_65993.827.png", "Validation_45663164/Validation_45663164_45663164_65995.826.png", "Validation_45663164/Validation_45663164_45663164_65997.825.png", "Validation_45663164/Validation_45663164_45663164_65999.824.png" ], "turn_id": 7, "scan_id": "Validation_45663164", "origin_question": "Remember, how many window(s) have you seen so far? ", "option": [], "answer": "2", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many window(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_45663164/Validation_45663164_45663164_66001.824.png", "Validation_45663164/Validation_45663164_45663164_66003.823.png", "Validation_45663164/Validation_45663164_45663164_66005.822.png", "Validation_45663164/Validation_45663164_45663164_66007.821.png", "Validation_45663164/Validation_45663164_45663164_66009.820.png" ], "turn_id": 8, "scan_id": "Validation_45663164", "origin_question": "What will be the first-time appearance order of the following categories: picture, sink, stove?", "option": [ "sink, picture, stove", "picture, sink, stove", "sink, stove, picture", "stove, picture, sink" ], "answer": "picture, sink, stove", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: picture, sink, stove? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['sink, picture, stove', 'picture, sink, stove', 'sink, stove, picture', 'stove, picture, sink'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_45663164/Validation_45663164_45663164_66011.820.png", "Validation_45663164/Validation_45663164_45663164_66013.819.png", "Validation_45663164/Validation_45663164_45663164_66015.818.png", "Validation_45663164/Validation_45663164_45663164_66017.817.png", "Validation_45663164/Validation_45663164_45663164_66019.816.png" ], "turn_id": 9, "scan_id": "Validation_45663164", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the sleek, medium-sized sink\" from your position?", "option": [], "answer": "119.91", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the sleek, medium-sized sink\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_45663164/Validation_45663164_45663164_66021.816.png", "Validation_45663164/Validation_45663164_45663164_66023.815.png", "Validation_45663164/Validation_45663164_45663164_66025.814.png", "Validation_45663164/Validation_45663164_45663164_66027.813.png", "Validation_45663164/Validation_45663164_45663164_66029.812.png" ], "turn_id": 10, "scan_id": "Validation_45663164", "origin_question": "In which turn were you farthest from \"the sleek cylindrical stainless steel bin with the dark lid\"? ", "option": [], "answer": "1", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the sleek cylindrical stainless steel bin with the dark lid\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42897504/Validation_42897504_42897504_269844.486.png", "Validation_42897504/Validation_42897504_42897504_269846.486.png", "Validation_42897504/Validation_42897504_42897504_269848.485.png", "Validation_42897504/Validation_42897504_42897504_269850.484.png" ], "turn_id": 1, "scan_id": "Validation_42897504", "origin_question": "Remember, have you seen any bowl(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any bowl(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_42897504/Validation_42897504_42897504_269852.483.png", "Validation_42897504/Validation_42897504_42897504_269854.482.png", "Validation_42897504/Validation_42897504_42897504_269856.482.png", "Validation_42897504/Validation_42897504_42897504_269858.481.png" ], "turn_id": 2, "scan_id": "Validation_42897504", "origin_question": "Is the large, round wooden bowl with the polished finish to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the large, round wooden bowl with the polished finish to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_42897504/Validation_42897504_42897504_269860.480.png", "Validation_42897504/Validation_42897504_42897504_269862.479.png", "Validation_42897504/Validation_42897504_42897504_269864.478.png", "Validation_42897504/Validation_42897504_42897504_269866.494.png" ], "turn_id": 3, "scan_id": "Validation_42897504", "origin_question": "Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "141.37", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_42897504/Validation_42897504_42897504_269868.493.png", "Validation_42897504/Validation_42897504_42897504_269870.493.png", "Validation_42897504/Validation_42897504_42897504_269872.492.png", "Validation_42897504/Validation_42897504_42897504_269874.491.png" ], "turn_id": 4, "scan_id": "Validation_42897504", "origin_question": "In which turn were you closest to \"the modern glass door with the metallic frame\"? ", "option": [], "answer": "1", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the modern glass door with the metallic frame\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_42897504/Validation_42897504_42897504_269876.490.png", "Validation_42897504/Validation_42897504_42897504_269878.489.png", "Validation_42897504/Validation_42897504_42897504_269880.488.png", "Validation_42897504/Validation_42897504_42897504_269882.488.png" ], "turn_id": 5, "scan_id": "Validation_42897504", "origin_question": "Which direction is the modern stainless steel range hood with the sleek design in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the modern stainless steel range hood with the sleek design in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_42897504/Validation_42897504_42897504_269884.487.png", "Validation_42897504/Validation_42897504_42897504_269886.486.png", "Validation_42897504/Validation_42897504_42897504_269888.485.png", "Validation_42897504/Validation_42897504_42897504_269890.484.png" ], "turn_id": 6, "scan_id": "Validation_42897504", "origin_question": "", "option": [], "answer": "", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_42897504/Validation_42897504_42897504_269892.484.png", "Validation_42897504/Validation_42897504_42897504_269894.483.png", "Validation_42897504/Validation_42897504_42897504_269896.482.png", "Validation_42897504/Validation_42897504_42897504_269898.481.png" ], "turn_id": 7, "scan_id": "Validation_42897504", "origin_question": "What will be the first-time appearance order of the following categories: oven, range hood, sink?", "option": [ "range hood, oven, sink", "range hood, sink, oven", "oven, range hood, sink", "oven, sink, range hood" ], "answer": "range hood, oven, sink", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: oven, range hood, sink? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['range hood, oven, sink', 'range hood, sink, oven', 'oven, range hood, sink', 'oven, sink, range hood'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47895554/Validation_47895554_47895554_457753.743.png", "Validation_47895554/Validation_47895554_47895554_457755.243.png", "Validation_47895554/Validation_47895554_47895554_457756.742.png", "Validation_47895554/Validation_47895554_47895554_457758.242.png" ], "turn_id": 1, "scan_id": "Validation_47895554", "origin_question": "Remember, have you seen any countertop(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any countertop(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47895554/Validation_47895554_47895554_457759.741.png", "Validation_47895554/Validation_47895554_47895554_457761.257.png", "Validation_47895554/Validation_47895554_47895554_457762.757.png", "Validation_47895554/Validation_47895554_47895554_457764.256.png" ], "turn_id": 2, "scan_id": "Validation_47895554", "origin_question": "Remember, have you seen any sink(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any sink(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47895554/Validation_47895554_47895554_457765.755.png", "Validation_47895554/Validation_47895554_47895554_457767.255.png", "Validation_47895554/Validation_47895554_47895554_457768.754.png", "Validation_47895554/Validation_47895554_47895554_457770.253.png" ], "turn_id": 3, "scan_id": "Validation_47895554", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the sink\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the sink\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47895554/Validation_47895554_47895554_457771.753.png", "Validation_47895554/Validation_47895554_47895554_457773.252.png", "Validation_47895554/Validation_47895554_47895554_457774.752.png", "Validation_47895554/Validation_47895554_47895554_457776.251.png" ], "turn_id": 4, "scan_id": "Validation_47895554", "origin_question": "In which turn were you closest to \"the sink\"? ", "option": [], "answer": "2", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the sink\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_47895554/Validation_47895554_47895554_457777.750.png", "Validation_47895554/Validation_47895554_47895554_457779.250.png", "Validation_47895554/Validation_47895554_47895554_457780.749.png", "Validation_47895554/Validation_47895554_47895554_457782.249.png" ], "turn_id": 5, "scan_id": "Validation_47895554", "origin_question": "Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_47895554/Validation_47895554_47895554_457783.748.png", "Validation_47895554/Validation_47895554_47895554_457785.247.png", "Validation_47895554/Validation_47895554_47895554_457786.747.png", "Validation_47895554/Validation_47895554_47895554_457788.246.png" ], "turn_id": 6, "scan_id": "Validation_47895554", "origin_question": "Please recall \"the large red cylindrical plastic basket\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.92", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the large red cylindrical plastic basket\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47429922/Validation_47429922_47429922_136616.259.png", "Validation_47429922/Validation_47429922_47429922_136618.242.png", "Validation_47429922/Validation_47429922_47429922_136620.241.png", "Validation_47429922/Validation_47429922_47429922_136622.240.png" ], "turn_id": 1, "scan_id": "Validation_47429922", "origin_question": "Remember, have you seen any door(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any door(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_47429922/Validation_47429922_47429922_136624.256.png", "Validation_47429922/Validation_47429922_47429922_136626.255.png", "Validation_47429922/Validation_47429922_47429922_136628.254.png", "Validation_47429922/Validation_47429922_47429922_136630.254.png" ], "turn_id": 2, "scan_id": "Validation_47429922", "origin_question": "In which turn did you see both \"the sleek black rectangular monitor with the plastic composition\" and \"the tall, glossy white wooden cabinet with decorative knobs\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the sleek black rectangular monitor with the plastic composition\" and \"the tall, glossy white wooden cabinet with decorative knobs\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47429922/Validation_47429922_47429922_136632.253.png", "Validation_47429922/Validation_47429922_47429922_136634.252.png", "Validation_47429922/Validation_47429922_47429922_136636.251.png", "Validation_47429922/Validation_47429922_47429922_136638.250.png" ], "turn_id": 3, "scan_id": "Validation_47429922", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the light-colored upholstered couch with the wooden frame\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the light-colored upholstered couch with the wooden frame\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47429922/Validation_47429922_47429922_136640.250.png", "Validation_47429922/Validation_47429922_47429922_136642.249.png", "Validation_47429922/Validation_47429922_47429922_136644.248.png", "Validation_47429922/Validation_47429922_47429922_136646.247.png" ], "turn_id": 4, "scan_id": "Validation_47429922", "origin_question": "Which direction is the door in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the door in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_47429922/Validation_47429922_47429922_136648.246.png", "Validation_47429922/Validation_47429922_47429922_136650.245.png", "Validation_47429922/Validation_47429922_47429922_136652.245.png", "Validation_47429922/Validation_47429922_47429922_136654.244.png" ], "turn_id": 5, "scan_id": "Validation_47429922", "origin_question": "Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_47429922/Validation_47429922_47429922_136656.243.png", "Validation_47429922/Validation_47429922_47429922_136658.242.png", "Validation_47429922/Validation_47429922_47429922_136660.241.png", "Validation_47429922/Validation_47429922_47429922_136662.240.png" ], "turn_id": 6, "scan_id": "Validation_47429922", "origin_question": "Which two objects are on the same side of you now? \"the door\", \"the sleek black rectangular monitor with the plastic composition\", \"the plush pink cylindrical stool with wooden legs\". ", "option": [ "the door, the sleek black rectangular monitor with the plastic composition", "the door, the plush pink cylindrical stool with wooden legs", "the sleek black rectangular monitor with the plastic composition, the plush pink cylindrical stool with wooden legs" ], "answer": "the sleek black rectangular monitor with the plastic composition, the plush pink cylindrical stool with wooden legs", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the door\", \"the sleek black rectangular monitor with the plastic composition\", \"the plush pink cylindrical stool with wooden legs\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the door, the sleek black rectangular monitor with the plastic composition', 'the door, the plush pink cylindrical stool with wooden legs', 'the sleek black rectangular monitor with the plastic composition, the plush pink cylindrical stool with wooden legs'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_47429922/Validation_47429922_47429922_136664.256.png", "Validation_47429922/Validation_47429922_47429922_136666.255.png", "Validation_47429922/Validation_47429922_47429922_136668.255.png", "Validation_47429922/Validation_47429922_47429922_136670.254.png" ], "turn_id": 7, "scan_id": "Validation_47429922", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "117.57", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_41142280/Validation_41142280_41142280_3823.005.png", "Validation_41142280/Validation_41142280_41142280_3824.971.png", "Validation_41142280/Validation_41142280_41142280_3826.970.png", "Validation_41142280/Validation_41142280_41142280_3828.986.png" ], "turn_id": 1, "scan_id": "Validation_41142280", "origin_question": "Remember, have you seen any mouse(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any mouse(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_41142280/Validation_41142280_41142280_3830.985.png", "Validation_41142280/Validation_41142280_41142280_3832.984.png", "Validation_41142280/Validation_41142280_41142280_3834.983.png", "Validation_41142280/Validation_41142280_41142280_3836.983.png" ], "turn_id": 2, "scan_id": "Validation_41142280", "origin_question": "In which turn did you see both \"the sleek white porcelain pedestal sink with the curved basin\" and \"the light brown rectangular wooden door\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the sleek white porcelain pedestal sink with the curved basin\" and \"the light brown rectangular wooden door\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_41142280/Validation_41142280_41142280_3838.982.png", "Validation_41142280/Validation_41142280_41142280_3840.981.png", "Validation_41142280/Validation_41142280_41142280_3842.980.png", "Validation_41142280/Validation_41142280_41142280_3844.979.png" ], "turn_id": 3, "scan_id": "Validation_41142280", "origin_question": "What will be the first-time appearance order of the following categories: bathtub, door, toilet?", "option": [ "bathtub, door, toilet", "door, toilet, bathtub", "door, bathtub, toilet", "toilet, door, bathtub" ], "answer": "door, bathtub, toilet", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bathtub, door, toilet? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['bathtub, door, toilet', 'door, toilet, bathtub', 'door, bathtub, toilet', 'toilet, door, bathtub'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_41142280/Validation_41142280_41142280_3846.978.png", "Validation_41142280/Validation_41142280_41142280_3848.977.png", "Validation_41142280/Validation_41142280_41142280_3850.977.png", "Validation_41142280/Validation_41142280_41142280_3852.976.png" ], "turn_id": 4, "scan_id": "Validation_41142280", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "125.78", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_41142280/Validation_41142280_41142280_3854.975.png", "Validation_41142280/Validation_41142280_41142280_3856.974.png", "Validation_41142280/Validation_41142280_41142280_3858.973.png", "Validation_41142280/Validation_41142280_41142280_3860.973.png" ], "turn_id": 5, "scan_id": "Validation_41142280", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the sleek white porcelain pedestal sink with the curved basin\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the sleek white porcelain pedestal sink with the curved basin\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_41142280/Validation_41142280_41142280_3862.972.png", "Validation_41142280/Validation_41142280_41142280_3864.971.png", "Validation_41142280/Validation_41142280_41142280_3866.970.png", "Validation_41142280/Validation_41142280_41142280_3868.986.png" ], "turn_id": 6, "scan_id": "Validation_41142280", "origin_question": "What will be the first-time appearance order of the following categories: bathtub, mirror, toilet?", "option": [ "toilet, mirror, bathtub", "bathtub, toilet, mirror", "toilet, bathtub, mirror", "mirror, bathtub, toilet" ], "answer": "mirror, bathtub, toilet", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bathtub, mirror, toilet? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['toilet, mirror, bathtub', 'bathtub, toilet, mirror', 'toilet, bathtub, mirror', 'mirror, bathtub, toilet'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_41142280/Validation_41142280_41142280_3870.985.png", "Validation_41142280/Validation_41142280_41142280_3872.984.png", "Validation_41142280/Validation_41142280_41142280_3874.983.png", "Validation_41142280/Validation_41142280_41142280_3876.983.png" ], "turn_id": 7, "scan_id": "Validation_41142280", "origin_question": "Is the light brown rectangular wooden door to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the light brown rectangular wooden door to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47331972/Validation_47331972_47331972_168617.171.png", "Validation_47331972/Validation_47331972_47331972_168619.171.png", "Validation_47331972/Validation_47331972_47331972_168621.170.png", "Validation_47331972/Validation_47331972_47331972_168623.169.png", "Validation_47331972/Validation_47331972_47331972_168625.168.png" ], "turn_id": 1, "scan_id": "Validation_47331972", "origin_question": "", "option": [], "answer": "", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_47331972/Validation_47331972_47331972_168627.167.png", "Validation_47331972/Validation_47331972_47331972_168629.166.png", "Validation_47331972/Validation_47331972_47331972_168631.166.png", "Validation_47331972/Validation_47331972_47331972_168633.182.png", "Validation_47331972/Validation_47331972_47331972_168635.181.png" ], "turn_id": 2, "scan_id": "Validation_47331972", "origin_question": "At the end of which turn were both of \"the large, earthy brown pillow\" and \"the cylindrical grey light with the textured surface\" on your right side?", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the large, earthy brown pillow\" and \"the cylindrical grey light with the textured surface\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47331972/Validation_47331972_47331972_168637.180.png", "Validation_47331972/Validation_47331972_47331972_168639.179.png", "Validation_47331972/Validation_47331972_47331972_168641.178.png", "Validation_47331972/Validation_47331972_47331972_168643.178.png", "Validation_47331972/Validation_47331972_47331972_168645.177.png" ], "turn_id": 3, "scan_id": "Validation_47331972", "origin_question": "Which direction is the cylindrical grey light with the textured surface in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the cylindrical grey light with the textured surface in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_47331972/Validation_47331972_47331972_168647.176.png", "Validation_47331972/Validation_47331972_47331972_168649.175.png", "Validation_47331972/Validation_47331972_47331972_168651.174.png", "Validation_47331972/Validation_47331972_47331972_168653.174.png", "Validation_47331972/Validation_47331972_47331972_168655.173.png" ], "turn_id": 4, "scan_id": "Validation_47331972", "origin_question": "When did you first discover \"the large, earthy brown pillow\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the large, earthy brown pillow\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_47331972/Validation_47331972_47331972_168657.172.png", "Validation_47331972/Validation_47331972_47331972_168659.171.png", "Validation_47331972/Validation_47331972_47331972_168661.170.png", "Validation_47331972/Validation_47331972_47331972_168663.169.png", "Validation_47331972/Validation_47331972_47331972_168665.169.png" ], "turn_id": 5, "scan_id": "Validation_47331972", "origin_question": "When did you first discover \"the sleek stainless steel dishwasher with the modern design\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the sleek stainless steel dishwasher with the modern design\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_47331972/Validation_47331972_47331972_168667.168.png", "Validation_47331972/Validation_47331972_47331972_168669.167.png", "Validation_47331972/Validation_47331972_47331972_168671.166.png", "Validation_47331972/Validation_47331972_47331972_168673.166.png", "Validation_47331972/Validation_47331972_47331972_168675.181.png" ], "turn_id": 6, "scan_id": "Validation_47331972", "origin_question": "Please recall \"the cylindrical grey light with the textured surface\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "3.6", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the cylindrical grey light with the textured surface\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_47331972/Validation_47331972_47331972_168677.180.png", "Validation_47331972/Validation_47331972_47331972_168679.180.png", "Validation_47331972/Validation_47331972_47331972_168681.179.png", "Validation_47331972/Validation_47331972_47331972_168683.178.png", "Validation_47331972/Validation_47331972_47331972_168685.177.png" ], "turn_id": 7, "scan_id": "Validation_47331972", "origin_question": "Assuming the direction you are facing at the end of the turn 4 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Right", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 4 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47331972/Validation_47331972_47331972_168687.177.png", "Validation_47331972/Validation_47331972_47331972_168689.176.png", "Validation_47331972/Validation_47331972_47331972_168691.175.png", "Validation_47331972/Validation_47331972_47331972_168693.174.png", "Validation_47331972/Validation_47331972_47331972_168695.173.png" ], "turn_id": 8, "scan_id": "Validation_47331972", "origin_question": "Which direction is the cylindrical grey light with the textured surface in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the cylindrical grey light with the textured surface in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47331972/Validation_47331972_47331972_168697.172.png", "Validation_47331972/Validation_47331972_47331972_168699.172.png", "Validation_47331972/Validation_47331972_47331972_168701.171.png", "Validation_47331972/Validation_47331972_47331972_168703.170.png", "Validation_47331972/Validation_47331972_47331972_168705.169.png" ], "turn_id": 9, "scan_id": "Validation_47331972", "origin_question": "Which direction is the large, glossy black rectangular monitor in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the large, glossy black rectangular monitor in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47331972/Validation_47331972_47331972_168707.168.png", "Validation_47331972/Validation_47331972_47331972_168709.168.png", "Validation_47331972/Validation_47331972_47331972_168711.167.png", "Validation_47331972/Validation_47331972_47331972_168713.166.png", "Validation_47331972/Validation_47331972_47331972_168715.165.png" ], "turn_id": 10, "scan_id": "Validation_47331972", "origin_question": "Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the vintage-style black cast iron fireplace\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the vintage-style black cast iron fireplace\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_48458663/Validation_48458663_48458663_8312.877.png", "Validation_48458663/Validation_48458663_48458663_8314.360.png", "Validation_48458663/Validation_48458663_48458663_8315.859.png", "Validation_48458663/Validation_48458663_48458663_8317.359.png" ], "turn_id": 1, "scan_id": "Validation_48458663", "origin_question": "Remember, have you seen any shelf(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any shelf(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "Validation_48458663/Validation_48458663_48458663_8318.858.png", "Validation_48458663/Validation_48458663_48458663_8320.357.png", "Validation_48458663/Validation_48458663_48458663_8321.857.png", "Validation_48458663/Validation_48458663_48458663_8323.356.png" ], "turn_id": 2, "scan_id": "Validation_48458663", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the black padded clothes with the durable material\"; \"the white collapsible bucket with mesh details\"; \"the bed\" ", "option": [ "the black padded clothes with the durable material", "the white collapsible bucket with mesh details", "the bed" ], "answer": "the white collapsible bucket with mesh details", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the black padded clothes with the durable material\"; \"the white collapsible bucket with mesh details\"; \"the bed\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the black padded clothes with the durable material', 'the white collapsible bucket with mesh details', 'the bed'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_48458663/Validation_48458663_48458663_8324.855.png", "Validation_48458663/Validation_48458663_48458663_8326.355.png", "Validation_48458663/Validation_48458663_48458663_8327.854.png", "Validation_48458663/Validation_48458663_48458663_8329.354.png" ], "turn_id": 3, "scan_id": "Validation_48458663", "origin_question": "Which two objects are on the same side of you now? \"the dark gray upholstered chair with ergonomic backrest\", \"the colorful rectangular vintage poster with warm tones\", \"the white collapsible bucket with mesh details\". ", "option": [ "the dark gray upholstered chair with ergonomic backrest, the colorful rectangular vintage poster with warm tones", "the dark gray upholstered chair with ergonomic backrest, the white collapsible bucket with mesh details", "the colorful rectangular vintage poster with warm tones, the white collapsible bucket with mesh details" ], "answer": "the dark gray upholstered chair with ergonomic backrest, the colorful rectangular vintage poster with warm tones", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the dark gray upholstered chair with ergonomic backrest\", \"the colorful rectangular vintage poster with warm tones\", \"the white collapsible bucket with mesh details\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the dark gray upholstered chair with ergonomic backrest, the colorful rectangular vintage poster with warm tones', 'the dark gray upholstered chair with ergonomic backrest, the white collapsible bucket with mesh details', 'the colorful rectangular vintage poster with warm tones, the white collapsible bucket with mesh details'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_48458663/Validation_48458663_48458663_8330.853.png", "Validation_48458663/Validation_48458663_48458663_8332.352.png", "Validation_48458663/Validation_48458663_48458663_8333.852.png", "Validation_48458663/Validation_48458663_48458663_8335.351.png" ], "turn_id": 4, "scan_id": "Validation_48458663", "origin_question": "Please recall \"the colorful rectangular vintage poster with warm tones\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.99", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the colorful rectangular vintage poster with warm tones\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_48458663/Validation_48458663_48458663_8336.851.png", "Validation_48458663/Validation_48458663_48458663_8338.350.png", "Validation_48458663/Validation_48458663_48458663_8339.849.png", "Validation_48458663/Validation_48458663_48458663_8341.349.png" ], "turn_id": 5, "scan_id": "Validation_48458663", "origin_question": "Please recall \"the modern rectangular white table with the flat top\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.45", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the modern rectangular white table with the flat top\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_44358452/Validation_44358452_44358452_43500.602.png", "Validation_44358452/Validation_44358452_44358452_43502.584.png", "Validation_44358452/Validation_44358452_44358452_43504.583.png", "Validation_44358452/Validation_44358452_44358452_43506.582.png", "Validation_44358452/Validation_44358452_44358452_43508.598.png" ], "turn_id": 1, "scan_id": "Validation_44358452", "origin_question": "Remember, have you seen any washing machine(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any washing machine(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_44358452/Validation_44358452_44358452_43510.598.png", "Validation_44358452/Validation_44358452_44358452_43512.597.png", "Validation_44358452/Validation_44358452_44358452_43514.596.png", "Validation_44358452/Validation_44358452_44358452_43516.595.png", "Validation_44358452/Validation_44358452_44358452_43518.594.png" ], "turn_id": 2, "scan_id": "Validation_44358452", "origin_question": "When did you first discover \"the sleek white washing machine with user-friendly controls\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the sleek white washing machine with user-friendly controls\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_44358452/Validation_44358452_44358452_43520.593.png", "Validation_44358452/Validation_44358452_44358452_43522.593.png", "Validation_44358452/Validation_44358452_44358452_43524.592.png", "Validation_44358452/Validation_44358452_44358452_43526.591.png", "Validation_44358452/Validation_44358452_44358452_43528.590.png" ], "turn_id": 3, "scan_id": "Validation_44358452", "origin_question": "", "option": [], "answer": "", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_44358452/Validation_44358452_44358452_43530.589.png", "Validation_44358452/Validation_44358452_44358452_43532.588.png", "Validation_44358452/Validation_44358452_44358452_43534.588.png", "Validation_44358452/Validation_44358452_44358452_43536.587.png", "Validation_44358452/Validation_44358452_44358452_43538.586.png" ], "turn_id": 4, "scan_id": "Validation_44358452", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the compact stainless steel oven with the sleek silver exterior\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the compact stainless steel oven with the sleek silver exterior\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_44358452/Validation_44358452_44358452_43540.585.png", "Validation_44358452/Validation_44358452_44358452_43542.584.png", "Validation_44358452/Validation_44358452_44358452_43544.584.png", "Validation_44358452/Validation_44358452_44358452_43546.583.png", "Validation_44358452/Validation_44358452_44358452_43548.599.png" ], "turn_id": 5, "scan_id": "Validation_44358452", "origin_question": "Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "84.68", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_44358452/Validation_44358452_44358452_43550.598.png", "Validation_44358452/Validation_44358452_44358452_43552.597.png", "Validation_44358452/Validation_44358452_44358452_43554.596.png", "Validation_44358452/Validation_44358452_44358452_43556.595.png", "Validation_44358452/Validation_44358452_44358452_43558.594.png" ], "turn_id": 6, "scan_id": "Validation_44358452", "origin_question": "Which direction is the compact stainless steel oven with the sleek silver exterior in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the compact stainless steel oven with the sleek silver exterior in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_44358452/Validation_44358452_44358452_43560.594.png", "Validation_44358452/Validation_44358452_44358452_43562.593.png", "Validation_44358452/Validation_44358452_44358452_43564.592.png", "Validation_44358452/Validation_44358452_44358452_43566.591.png", "Validation_44358452/Validation_44358452_44358452_43568.590.png" ], "turn_id": 7, "scan_id": "Validation_44358452", "origin_question": "Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the compact stainless steel oven with the sleek silver exterior\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the compact stainless steel oven with the sleek silver exterior\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_44358452/Validation_44358452_44358452_43570.589.png", "Validation_44358452/Validation_44358452_44358452_43572.588.png", "Validation_44358452/Validation_44358452_44358452_43574.588.png", "Validation_44358452/Validation_44358452_44358452_43576.587.png", "Validation_44358452/Validation_44358452_44358452_43578.586.png" ], "turn_id": 8, "scan_id": "Validation_44358452", "origin_question": "Which direction is the compact stainless steel oven with the sleek silver exterior in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the compact stainless steel oven with the sleek silver exterior in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_44358452/Validation_44358452_44358452_43580.585.png", "Validation_44358452/Validation_44358452_44358452_43582.584.png", "Validation_44358452/Validation_44358452_44358452_43584.584.png", "Validation_44358452/Validation_44358452_44358452_43586.583.png", "Validation_44358452/Validation_44358452_44358452_43588.582.png" ], "turn_id": 9, "scan_id": "Validation_44358452", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "1.82", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47430034/Validation_47430034_47430034_19453.887.png", "Validation_47430034/Validation_47430034_47430034_19455.387.png", "Validation_47430034/Validation_47430034_47430034_19456.886.png", "Validation_47430034/Validation_47430034_47430034_19458.386.png" ], "turn_id": 1, "scan_id": "Validation_47430034", "origin_question": "Remember, have you seen any plunger(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any plunger(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47430034/Validation_47430034_47430034_19459.885.png", "Validation_47430034/Validation_47430034_47430034_19461.384.png", "Validation_47430034/Validation_47430034_47430034_19462.884.png", "Validation_47430034/Validation_47430034_47430034_19464.383.png" ], "turn_id": 2, "scan_id": "Validation_47430034", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47430034/Validation_47430034_47430034_19465.882.png", "Validation_47430034/Validation_47430034_47430034_19467.382.png", "Validation_47430034/Validation_47430034_47430034_19468.881.png", "Validation_47430034/Validation_47430034_47430034_19470.381.png" ], "turn_id": 3, "scan_id": "Validation_47430034", "origin_question": "", "option": [], "answer": "", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_47430034/Validation_47430034_47430034_19471.880.png", "Validation_47430034/Validation_47430034_47430034_19473.379.png", "Validation_47430034/Validation_47430034_47430034_19474.879.png", "Validation_47430034/Validation_47430034_47430034_19476.378.png" ], "turn_id": 4, "scan_id": "Validation_47430034", "origin_question": "Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Right", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_47430034/Validation_47430034_47430034_19477.878.png", "Validation_47430034/Validation_47430034_47430034_19479.377.png", "Validation_47430034/Validation_47430034_47430034_19480.876.png", "Validation_47430034/Validation_47430034_47430034_19482.376.png" ], "turn_id": 5, "scan_id": "Validation_47430034", "origin_question": "Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_47430034/Validation_47430034_47430034_19483.875.png", "Validation_47430034/Validation_47430034_47430034_19485.375.png", "Validation_47430034/Validation_47430034_47430034_19486.874.png", "Validation_47430034/Validation_47430034_47430034_19488.373.png" ], "turn_id": 6, "scan_id": "Validation_47430034", "origin_question": "Is the rectangular cotton pillow with vibrant colors to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the rectangular cotton pillow with vibrant colors to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "Validation_41159538/Validation_41159538_41159538_39669.653.png", "Validation_41159538/Validation_41159538_41159538_39671.635.png", "Validation_41159538/Validation_41159538_41159538_39673.635.png", "Validation_41159538/Validation_41159538_41159538_39675.634.png", "Validation_41159538/Validation_41159538_41159538_39677.633.png" ], "turn_id": 1, "scan_id": "Validation_41159538", "origin_question": "Remember, how many table(s) have you seen so far? ", "option": [], "answer": "4", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many table(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_41159538/Validation_41159538_41159538_39679.632.png", "Validation_41159538/Validation_41159538_41159538_39681.631.png", "Validation_41159538/Validation_41159538_41159538_39683.630.png", "Validation_41159538/Validation_41159538_41159538_39685.630.png", "Validation_41159538/Validation_41159538_41159538_39687.629.png" ], "turn_id": 2, "scan_id": "Validation_41159538", "origin_question": "Which direction is the sleek black large monitor with glossy screen in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the sleek black large monitor with glossy screen in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_41159538/Validation_41159538_41159538_39689.628.png", "Validation_41159538/Validation_41159538_41159538_39691.627.png", "Validation_41159538/Validation_41159538_41159538_39693.626.png", "Validation_41159538/Validation_41159538_41159538_39695.625.png", "Validation_41159538/Validation_41159538_41159538_39697.625.png" ], "turn_id": 3, "scan_id": "Validation_41159538", "origin_question": "At the end of which turn were both of \"the carpet\" and \"the modern black laptop with the sleek rectangular shape\" on your right side?", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the carpet\" and \"the modern black laptop with the sleek rectangular shape\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_41159538/Validation_41159538_41159538_39699.624.png", "Validation_41159538/Validation_41159538_41159538_39701.623.png", "Validation_41159538/Validation_41159538_41159538_39703.639.png", "Validation_41159538/Validation_41159538_41159538_39705.638.png", "Validation_41159538/Validation_41159538_41159538_39707.637.png" ], "turn_id": 4, "scan_id": "Validation_41159538", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the slim off-white cabinet with smooth surfaces\", Y: \"the folder\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to X and farther away from Y", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the slim off-white cabinet with smooth surfaces\", Y: \"the folder\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_41159538/Validation_41159538_41159538_39709.636.png", "Validation_41159538/Validation_41159538_41159538_39711.635.png", "Validation_41159538/Validation_41159538_41159538_39713.635.png", "Validation_41159538/Validation_41159538_41159538_39715.634.png", "Validation_41159538/Validation_41159538_41159538_39717.633.png" ], "turn_id": 5, "scan_id": "Validation_41159538", "origin_question": "Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_41159538/Validation_41159538_41159538_39719.632.png", "Validation_41159538/Validation_41159538_41159538_39721.631.png", "Validation_41159538/Validation_41159538_41159538_39723.630.png", "Validation_41159538/Validation_41159538_41159538_39725.630.png", "Validation_41159538/Validation_41159538_41159538_39727.629.png" ], "turn_id": 6, "scan_id": "Validation_41159538", "origin_question": "In which turn did you see both \"the sleek black large monitor with glossy screen\" and \"the sleek medium brown wooden door with the minimalist design\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the sleek black large monitor with glossy screen\" and \"the sleek medium brown wooden door with the minimalist design\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_41159538/Validation_41159538_41159538_39729.628.png", "Validation_41159538/Validation_41159538_41159538_39731.627.png", "Validation_41159538/Validation_41159538_41159538_39733.626.png", "Validation_41159538/Validation_41159538_41159538_39735.626.png", "Validation_41159538/Validation_41159538_41159538_39737.625.png" ], "turn_id": 7, "scan_id": "Validation_41159538", "origin_question": "Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_41159538/Validation_41159538_41159538_39739.624.png", "Validation_41159538/Validation_41159538_41159538_39741.623.png", "Validation_41159538/Validation_41159538_41159538_39743.622.png", "Validation_41159538/Validation_41159538_41159538_39745.638.png", "Validation_41159538/Validation_41159538_41159538_39747.637.png" ], "turn_id": 8, "scan_id": "Validation_41159538", "origin_question": "Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "132.84", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_41159538/Validation_41159538_41159538_39749.636.png", "Validation_41159538/Validation_41159538_41159538_39751.635.png", "Validation_41159538/Validation_41159538_41159538_39753.635.png", "Validation_41159538/Validation_41159538_41159538_39755.634.png", "Validation_41159538/Validation_41159538_41159538_39757.633.png" ], "turn_id": 9, "scan_id": "Validation_41159538", "origin_question": "Assuming the direction you are facing at the end of the turn 4 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Right", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 4 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_41159538/Validation_41159538_41159538_39759.632.png", "Validation_41159538/Validation_41159538_41159538_39761.631.png", "Validation_41159538/Validation_41159538_41159538_39763.631.png", "Validation_41159538/Validation_41159538_41159538_39765.630.png", "Validation_41159538/Validation_41159538_41159538_39767.629.png" ], "turn_id": 10, "scan_id": "Validation_41159538", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the slim off-white cabinet with smooth surfaces\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the slim off-white cabinet with smooth surfaces\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47430002/Validation_47430002_47430002_17167.776.png", "Validation_47430002/Validation_47430002_47430002_17169.259.png", "Validation_47430002/Validation_47430002_47430002_17170.758.png", "Validation_47430002/Validation_47430002_47430002_17172.258.png" ], "turn_id": 1, "scan_id": "Validation_47430002", "origin_question": "", "option": [], "answer": "", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47430002/Validation_47430002_47430002_17173.757.png", "Validation_47430002/Validation_47430002_47430002_17175.256.png", "Validation_47430002/Validation_47430002_47430002_17176.756.png", "Validation_47430002/Validation_47430002_47430002_17178.255.png" ], "turn_id": 2, "scan_id": "Validation_47430002", "origin_question": "Remember, have you seen any cabinet(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any cabinet(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_47430002/Validation_47430002_47430002_17179.755.png", "Validation_47430002/Validation_47430002_47430002_17181.254.png", "Validation_47430002/Validation_47430002_47430002_17182.753.png", "Validation_47430002/Validation_47430002_47430002_17184.253.png" ], "turn_id": 3, "scan_id": "Validation_47430002", "origin_question": "How far is your current position from where you were at the end of the turn 1? (in meters)", "option": [], "answer": "2.44", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 1? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47430002/Validation_47430002_47430002_17185.752.png", "Validation_47430002/Validation_47430002_47430002_17187.251.png", "Validation_47430002/Validation_47430002_47430002_17188.751.png", "Validation_47430002/Validation_47430002_47430002_17190.250.png" ], "turn_id": 4, "scan_id": "Validation_47430002", "origin_question": "Remember, have you seen any bed(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any bed(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47430002/Validation_47430002_47430002_17191.750.png", "Validation_47430002/Validation_47430002_47430002_17193.249.png", "Validation_47430002/Validation_47430002_47430002_17194.748.png", "Validation_47430002/Validation_47430002_47430002_17196.248.png" ], "turn_id": 5, "scan_id": "Validation_47430002", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the large white vertical rectangle window\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the large white vertical rectangle window\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_47430002/Validation_47430002_47430002_17197.747.png", "Validation_47430002/Validation_47430002_47430002_17199.247.png", "Validation_47430002/Validation_47430002_47430002_17200.746.png", "Validation_47430002/Validation_47430002_47430002_17202.245.png" ], "turn_id": 6, "scan_id": "Validation_47430002", "origin_question": "What will be the first-time appearance order of the following categories: bed, cabinet, frame?", "option": [ "cabinet, bed, frame", "cabinet, frame, bed", "frame, cabinet, bed", "bed, frame, cabinet" ], "answer": "cabinet, bed, frame", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bed, cabinet, frame? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['cabinet, bed, frame', 'cabinet, frame, bed', 'frame, cabinet, bed', 'bed, frame, cabinet'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42445028/Validation_42445028_42445028_49602.716.png", "Validation_42445028/Validation_42445028_42445028_49604.698.png", "Validation_42445028/Validation_42445028_42445028_49606.697.png", "Validation_42445028/Validation_42445028_42445028_49608.697.png" ], "turn_id": 1, "scan_id": "Validation_42445028", "origin_question": "Remember, have you seen any radiator(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any radiator(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_42445028/Validation_42445028_42445028_49610.696.png", "Validation_42445028/Validation_42445028_42445028_49612.695.png", "Validation_42445028/Validation_42445028_42445028_49614.694.png", "Validation_42445028/Validation_42445028_42445028_49616.693.png" ], "turn_id": 2, "scan_id": "Validation_42445028", "origin_question": "At the end of which turn were both of \"the bed\" and \"the white low-top shoe made of mesh and leather\" on your right side?", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the bed\" and \"the white low-top shoe made of mesh and leather\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_42445028/Validation_42445028_42445028_49618.693.png", "Validation_42445028/Validation_42445028_42445028_49620.692.png", "Validation_42445028/Validation_42445028_42445028_49622.691.png", "Validation_42445028/Validation_42445028_42445028_49624.690.png" ], "turn_id": 3, "scan_id": "Validation_42445028", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "29.46", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_42445028/Validation_42445028_42445028_49626.690.png", "Validation_42445028/Validation_42445028_42445028_49628.689.png", "Validation_42445028/Validation_42445028_42445028_49630.688.png", "Validation_42445028/Validation_42445028_42445028_49632.687.png" ], "turn_id": 4, "scan_id": "Validation_42445028", "origin_question": "Please recall \"the sleek white rectangular radiator mounted on the wall\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.32", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the sleek white rectangular radiator mounted on the wall\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_42445028/Validation_42445028_42445028_49634.703.png", "Validation_42445028/Validation_42445028_42445028_49636.702.png", "Validation_42445028/Validation_42445028_42445028_49638.701.png", "Validation_42445028/Validation_42445028_42445028_49640.701.png" ], "turn_id": 5, "scan_id": "Validation_42445028", "origin_question": "Which direction is the sleek black rectangular monitor with the thin bezel in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the sleek black rectangular monitor with the thin bezel in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_42445028/Validation_42445028_42445028_49642.700.png", "Validation_42445028/Validation_42445028_42445028_49644.699.png", "Validation_42445028/Validation_42445028_42445028_49646.698.png", "Validation_42445028/Validation_42445028_42445028_49648.697.png" ], "turn_id": 6, "scan_id": "Validation_42445028", "origin_question": "Which two objects are on the same side of you now? \"the white low-top shoe made of mesh and leather\", \"the dark gray rectangular picture made of synthetic material\", \"the modern white modular shelf with square compartments\". ", "option": [ "the white low-top shoe made of mesh and leather, the dark gray rectangular picture made of synthetic material", "the white low-top shoe made of mesh and leather, the modern white modular shelf with square compartments", "the dark gray rectangular picture made of synthetic material, the modern white modular shelf with square compartments" ], "answer": "the white low-top shoe made of mesh and leather, the dark gray rectangular picture made of synthetic material", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the white low-top shoe made of mesh and leather\", \"the dark gray rectangular picture made of synthetic material\", \"the modern white modular shelf with square compartments\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the white low-top shoe made of mesh and leather, the dark gray rectangular picture made of synthetic material', 'the white low-top shoe made of mesh and leather, the modern white modular shelf with square compartments', 'the dark gray rectangular picture made of synthetic material, the modern white modular shelf with square compartments'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_42445028/Validation_42445028_42445028_49650.697.png", "Validation_42445028/Validation_42445028_42445028_49652.696.png", "Validation_42445028/Validation_42445028_42445028_49654.695.png", "Validation_42445028/Validation_42445028_42445028_49656.694.png" ], "turn_id": 7, "scan_id": "Validation_42445028", "origin_question": "Which direction is the modern white modular shelf with square compartments in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the modern white modular shelf with square compartments in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_42445028/Validation_42445028_42445028_49658.694.png", "Validation_42445028/Validation_42445028_42445028_49660.693.png", "Validation_42445028/Validation_42445028_42445028_49662.692.png", "Validation_42445028/Validation_42445028_42445028_49664.691.png" ], "turn_id": 8, "scan_id": "Validation_42445028", "origin_question": "Please recall \"the dark gray rectangular picture made of synthetic material\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.07", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the dark gray rectangular picture made of synthetic material\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42897528/Validation_42897528_42897528_273902.102.png", "Validation_42897528/Validation_42897528_42897528_273903.601.png", "Validation_42897528/Validation_42897528_42897528_273905.101.png", "Validation_42897528/Validation_42897528_42897528_273906.600.png" ], "turn_id": 1, "scan_id": "Validation_42897528", "origin_question": "Remember, have you seen any bin(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any bin(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_42897528/Validation_42897528_42897528_273908.099.png", "Validation_42897528/Validation_42897528_42897528_273909.599.png", "Validation_42897528/Validation_42897528_42897528_273911.098.png", "Validation_42897528/Validation_42897528_42897528_273912.597.png" ], "turn_id": 2, "scan_id": "Validation_42897528", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_42897528/Validation_42897528_42897528_273914.097.png", "Validation_42897528/Validation_42897528_42897528_273915.596.png", "Validation_42897528/Validation_42897528_42897528_273917.096.png", "Validation_42897528/Validation_42897528_42897528_273918.595.png" ], "turn_id": 3, "scan_id": "Validation_42897528", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_42897528/Validation_42897528_42897528_273920.095.png", "Validation_42897528/Validation_42897528_42897528_273921.594.png", "Validation_42897528/Validation_42897528_42897528_273923.093.png", "Validation_42897528/Validation_42897528_42897528_273924.593.png" ], "turn_id": 4, "scan_id": "Validation_42897528", "origin_question": "Is the sleek, silver stainless steel kettle with the rounded shape to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the sleek, silver stainless steel kettle with the rounded shape to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_42897528/Validation_42897528_42897528_273926.092.png", "Validation_42897528/Validation_42897528_42897528_273927.591.png", "Validation_42897528/Validation_42897528_42897528_273929.091.png", "Validation_42897528/Validation_42897528_42897528_273930.590.png" ], "turn_id": 5, "scan_id": "Validation_42897528", "origin_question": "Which two objects are on the same side of you now? \"the sleek, silver stainless steel kettle with the rounded shape\", \"the stylish black medium-sized oven with the glass door\", \"the sleek rectangular sink\". ", "option": [ "the sleek, silver stainless steel kettle with the rounded shape, the stylish black medium-sized oven with the glass door", "the sleek, silver stainless steel kettle with the rounded shape, the sleek rectangular sink", "the stylish black medium-sized oven with the glass door, the sleek rectangular sink" ], "answer": "the stylish black medium-sized oven with the glass door, the sleek rectangular sink", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the sleek, silver stainless steel kettle with the rounded shape\", \"the stylish black medium-sized oven with the glass door\", \"the sleek rectangular sink\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the sleek, silver stainless steel kettle with the rounded shape, the stylish black medium-sized oven with the glass door', 'the sleek, silver stainless steel kettle with the rounded shape, the sleek rectangular sink', 'the stylish black medium-sized oven with the glass door, the sleek rectangular sink'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_42897528/Validation_42897528_42897528_273932.106.png", "Validation_42897528/Validation_42897528_42897528_273933.606.png", "Validation_42897528/Validation_42897528_42897528_273935.105.png", "Validation_42897528/Validation_42897528_42897528_273936.604.png" ], "turn_id": 6, "scan_id": "Validation_42897528", "origin_question": "Which direction is the sleek, silver stainless steel kettle with the rounded shape in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the sleek, silver stainless steel kettle with the rounded shape in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_42897528/Validation_42897528_42897528_273938.104.png", "Validation_42897528/Validation_42897528_42897528_273939.603.png", "Validation_42897528/Validation_42897528_42897528_273941.102.png", "Validation_42897528/Validation_42897528_42897528_273942.602.png" ], "turn_id": 7, "scan_id": "Validation_42897528", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the stylish black medium-sized oven with the glass door\" from your position?", "option": [], "answer": "103.49", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the stylish black medium-sized oven with the glass door\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42444946/Validation_42444946_42444946_223607.897.png", "Validation_42444946/Validation_42444946_42444946_223609.880.png", "Validation_42444946/Validation_42444946_42444946_223611.879.png", "Validation_42444946/Validation_42444946_42444946_223614.878.png", "Validation_42444946/Validation_42444946_42444946_223616.877.png" ], "turn_id": 1, "scan_id": "Validation_42444946", "origin_question": "Remember, have you seen any rope(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any rope(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_42444946/Validation_42444946_42444946_223618.876.png", "Validation_42444946/Validation_42444946_42444946_223620.875.png", "Validation_42444946/Validation_42444946_42444946_223622.874.png", "Validation_42444946/Validation_42444946_42444946_223624.873.png", "Validation_42444946/Validation_42444946_42444946_223626.873.png" ], "turn_id": 2, "scan_id": "Validation_42444946", "origin_question": "When did you first discover \"the dark blue upholstered couch with armrests\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the dark blue upholstered couch with armrests\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_42444946/Validation_42444946_42444946_223628.872.png", "Validation_42444946/Validation_42444946_42444946_223630.871.png", "Validation_42444946/Validation_42444946_42444946_223632.870.png", "Validation_42444946/Validation_42444946_42444946_223634.869.png", "Validation_42444946/Validation_42444946_42444946_223636.869.png" ], "turn_id": 3, "scan_id": "Validation_42444946", "origin_question": "Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Right", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_42444946/Validation_42444946_42444946_223638.868.png", "Validation_42444946/Validation_42444946_42444946_223640.867.png", "Validation_42444946/Validation_42444946_42444946_223642.866.png", "Validation_42444946/Validation_42444946_42444946_223644.865.png", "Validation_42444946/Validation_42444946_42444946_223646.864.png" ], "turn_id": 4, "scan_id": "Validation_42444946", "origin_question": "Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "73.45", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_42444946/Validation_42444946_42444946_223648.880.png", "Validation_42444946/Validation_42444946_42444946_223650.880.png", "Validation_42444946/Validation_42444946_42444946_223652.879.png", "Validation_42444946/Validation_42444946_42444946_223654.878.png", "Validation_42444946/Validation_42444946_42444946_223656.877.png" ], "turn_id": 5, "scan_id": "Validation_42444946", "origin_question": "Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the dark blue upholstered couch with armrests\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the dark blue upholstered couch with armrests\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_42444946/Validation_42444946_42444946_223658.876.png", "Validation_42444946/Validation_42444946_42444946_223660.875.png", "Validation_42444946/Validation_42444946_42444946_223662.875.png", "Validation_42444946/Validation_42444946_42444946_223664.874.png", "Validation_42444946/Validation_42444946_42444946_223666.873.png" ], "turn_id": 6, "scan_id": "Validation_42444946", "origin_question": "What will be the first-time appearance order of the following categories: couch, door, window?", "option": [ "door, couch, window", "window, couch, door", "door, window, couch", "couch, window, door" ], "answer": "door, couch, window", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: couch, door, window? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['door, couch, window', 'window, couch, door', 'door, window, couch', 'couch, window, door'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_42444946/Validation_42444946_42444946_223668.872.png", "Validation_42444946/Validation_42444946_42444946_223670.871.png", "Validation_42444946/Validation_42444946_42444946_223672.871.png", "Validation_42444946/Validation_42444946_42444946_223674.870.png", "Validation_42444946/Validation_42444946_42444946_223676.869.png" ], "turn_id": 7, "scan_id": "Validation_42444946", "origin_question": "By the end of the 3 turn, were you getting closer to or farther away from these two objects: X: \"the sturdy wooden chair with the natural brown finish\", Y: \"the medium-sized round wooden table with rich brown color\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 3 turn, were you getting closer to or farther away from these two objects: X: \"the sturdy wooden chair with the natural brown finish\", Y: \"the medium-sized round wooden table with rich brown color\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_42444946/Validation_42444946_42444946_223678.868.png", "Validation_42444946/Validation_42444946_42444946_223680.867.png", "Validation_42444946/Validation_42444946_42444946_223682.867.png", "Validation_42444946/Validation_42444946_42444946_223684.866.png", "Validation_42444946/Validation_42444946_42444946_223686.865.png" ], "turn_id": 8, "scan_id": "Validation_42444946", "origin_question": "Please recall \"the classic rectangular wooden window with the cream-colored frame\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.94", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the classic rectangular wooden window with the cream-colored frame\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_42444946/Validation_42444946_42444946_223688.864.png", "Validation_42444946/Validation_42444946_42444946_223690.863.png", "Validation_42444946/Validation_42444946_42444946_223692.879.png", "Validation_42444946/Validation_42444946_42444946_223694.878.png", "Validation_42444946/Validation_42444946_42444946_223696.877.png" ], "turn_id": 9, "scan_id": "Validation_42444946", "origin_question": "Assuming the direction you are facing at the end of the turn 5 is forward, did you move a certain distance forward or backward from that position?", "option": [ "Forward", "Backward" ], "answer": "Backward", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 5 is forward, did you move a certain distance forward or backward from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Forward', 'Backward'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47204563/Validation_47204563_47204563_4033.452.png", "Validation_47204563/Validation_47204563_47204563_4034.952.png", "Validation_47204563/Validation_47204563_47204563_4036.451.png", "Validation_47204563/Validation_47204563_47204563_4037.950.png" ], "turn_id": 1, "scan_id": "Validation_47204563", "origin_question": "Remember, have you seen any door(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any door(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_47204563/Validation_47204563_47204563_4039.450.png", "Validation_47204563/Validation_47204563_47204563_4040.949.png", "Validation_47204563/Validation_47204563_47204563_4042.449.png", "Validation_47204563/Validation_47204563_47204563_4043.948.png" ], "turn_id": 2, "scan_id": "Validation_47204563", "origin_question": "Which two objects are on the same side of you now? \"the medium-sized wooden shelf with the warm brown finish\", \"the collection of colorful hardcover books\", \"the warm brown wooden door with the sleek metallic handle\". ", "option": [ "the medium-sized wooden shelf with the warm brown finish, the collection of colorful hardcover books", "the medium-sized wooden shelf with the warm brown finish, the warm brown wooden door with the sleek metallic handle", "the collection of colorful hardcover books, the warm brown wooden door with the sleek metallic handle" ], "answer": "the medium-sized wooden shelf with the warm brown finish, the collection of colorful hardcover books", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the medium-sized wooden shelf with the warm brown finish\", \"the collection of colorful hardcover books\", \"the warm brown wooden door with the sleek metallic handle\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the medium-sized wooden shelf with the warm brown finish, the collection of colorful hardcover books', 'the medium-sized wooden shelf with the warm brown finish, the warm brown wooden door with the sleek metallic handle', 'the collection of colorful hardcover books, the warm brown wooden door with the sleek metallic handle'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47204563/Validation_47204563_47204563_4045.447.png", "Validation_47204563/Validation_47204563_47204563_4046.947.png", "Validation_47204563/Validation_47204563_47204563_4048.446.png", "Validation_47204563/Validation_47204563_47204563_4049.945.png" ], "turn_id": 3, "scan_id": "Validation_47204563", "origin_question": "Which direction is the sleek, rectangular silver monitor with the curved screen in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the sleek, rectangular silver monitor with the curved screen in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_47204563/Validation_47204563_47204563_4051.445.png", "Validation_47204563/Validation_47204563_47204563_4052.944.png", "Validation_47204563/Validation_47204563_47204563_4054.444.png", "Validation_47204563/Validation_47204563_47204563_4055.943.png" ], "turn_id": 4, "scan_id": "Validation_47204563", "origin_question": "What will be the first-time appearance order of the following categories: chair, door, monitor?", "option": [ "door, chair, monitor", "chair, door, monitor", "chair, monitor, door", "monitor, chair, door" ], "answer": "door, chair, monitor", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: chair, door, monitor? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['door, chair, monitor', 'chair, door, monitor', 'chair, monitor, door', 'monitor, chair, door'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47204563/Validation_47204563_47204563_4057.442.png", "Validation_47204563/Validation_47204563_47204563_4058.942.png", "Validation_47204563/Validation_47204563_47204563_4060.441.png", "Validation_47204563/Validation_47204563_47204563_4061.941.png" ], "turn_id": 5, "scan_id": "Validation_47204563", "origin_question": "Which direction is the sleek, rectangular silver monitor with the curved screen in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the sleek, rectangular silver monitor with the curved screen in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_48018386/Validation_48018386_48018386_390420.956.png", "Validation_48018386/Validation_48018386_48018386_390422.439.png", "Validation_48018386/Validation_48018386_48018386_390423.938.png", "Validation_48018386/Validation_48018386_48018386_390425.438.png" ], "turn_id": 1, "scan_id": "Validation_48018386", "origin_question": "", "option": [], "answer": "", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_48018386/Validation_48018386_48018386_390426.937.png", "Validation_48018386/Validation_48018386_48018386_390428.436.png", "Validation_48018386/Validation_48018386_48018386_390429.936.png", "Validation_48018386/Validation_48018386_48018386_390431.435.png" ], "turn_id": 2, "scan_id": "Validation_48018386", "origin_question": "In which turn did you see both \"the modern, sleek white toilet with an oval shape\" and \"the glossy white rectangular basin made of ceramic\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the modern, sleek white toilet with an oval shape\" and \"the glossy white rectangular basin made of ceramic\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_48018386/Validation_48018386_48018386_390432.934.png", "Validation_48018386/Validation_48018386_48018386_390434.434.png", "Validation_48018386/Validation_48018386_48018386_390435.933.png", "Validation_48018386/Validation_48018386_48018386_390437.449.png" ], "turn_id": 3, "scan_id": "Validation_48018386", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the sleek, oval glossy white sink\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the sleek, oval glossy white sink\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_48018386/Validation_48018386_48018386_390438.949.png", "Validation_48018386/Validation_48018386_48018386_390440.448.png", "Validation_48018386/Validation_48018386_48018386_390441.947.png", "Validation_48018386/Validation_48018386_48018386_390443.447.png" ], "turn_id": 4, "scan_id": "Validation_48018386", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "129.86", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_48018386/Validation_48018386_48018386_390444.946.png", "Validation_48018386/Validation_48018386_48018386_390446.445.png", "Validation_48018386/Validation_48018386_48018386_390447.945.png", "Validation_48018386/Validation_48018386_48018386_390449.444.png" ], "turn_id": 5, "scan_id": "Validation_48018386", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_48018386/Validation_48018386_48018386_390450.944.png", "Validation_48018386/Validation_48018386_48018386_390452.443.png", "Validation_48018386/Validation_48018386_48018386_390453.942.png", "Validation_48018386/Validation_48018386_48018386_390455.442.png" ], "turn_id": 6, "scan_id": "Validation_48018386", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "1.42", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_48018386/Validation_48018386_48018386_390456.941.png", "Validation_48018386/Validation_48018386_48018386_390458.441.png", "Validation_48018386/Validation_48018386_48018386_390459.940.png", "Validation_48018386/Validation_48018386_48018386_390461.439.png" ], "turn_id": 7, "scan_id": "Validation_48018386", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the modern rectangular mirror with the neutral metal frame\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the modern rectangular mirror with the neutral metal frame\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47331322/Validation_47331322_47331322_179115.348.png", "Validation_47331322/Validation_47331322_47331322_179117.330.png", "Validation_47331322/Validation_47331322_47331322_179119.329.png", "Validation_47331322/Validation_47331322_47331322_179121.328.png" ], "turn_id": 1, "scan_id": "Validation_47331322", "origin_question": "Remember, have you seen any tv(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any tv(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47331322/Validation_47331322_47331322_179123.328.png", "Validation_47331322/Validation_47331322_47331322_179125.327.png", "Validation_47331322/Validation_47331322_47331322_179127.326.png", "Validation_47331322/Validation_47331322_47331322_179129.325.png" ], "turn_id": 2, "scan_id": "Validation_47331322", "origin_question": "Which direction is the sleek black rectangular stove with touch controls in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the sleek black rectangular stove with touch controls in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47331322/Validation_47331322_47331322_179131.341.png", "Validation_47331322/Validation_47331322_47331322_179133.340.png", "Validation_47331322/Validation_47331322_47331322_179135.339.png", "Validation_47331322/Validation_47331322_47331322_179137.339.png" ], "turn_id": 3, "scan_id": "Validation_47331322", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the modern stainless steel microwave with the glass door\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the modern stainless steel microwave with the glass door\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_47331322/Validation_47331322_47331322_179139.338.png", "Validation_47331322/Validation_47331322_47331322_179141.337.png", "Validation_47331322/Validation_47331322_47331322_179143.336.png", "Validation_47331322/Validation_47331322_47331322_179145.335.png" ], "turn_id": 4, "scan_id": "Validation_47331322", "origin_question": "At the end of which turn were both of \"the modern stainless steel microwave with the glass door\" and \"the sleek, rectangular oven\" on your left side?", "option": [], "answer": "4", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the modern stainless steel microwave with the glass door\" and \"the sleek, rectangular oven\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47331322/Validation_47331322_47331322_179147.334.png", "Validation_47331322/Validation_47331322_47331322_179149.334.png", "Validation_47331322/Validation_47331322_47331322_179151.333.png", "Validation_47331322/Validation_47331322_47331322_179153.332.png" ], "turn_id": 5, "scan_id": "Validation_47331322", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the sleek black rectangular stove with touch controls\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the sleek black rectangular stove with touch controls\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_47331322/Validation_47331322_47331322_179155.331.png", "Validation_47331322/Validation_47331322_47331322_179157.330.png", "Validation_47331322/Validation_47331322_47331322_179159.330.png", "Validation_47331322/Validation_47331322_47331322_179161.329.png" ], "turn_id": 6, "scan_id": "Validation_47331322", "origin_question": "At the end of which turn were both of \"the sleek black rectangular stove with touch controls\" and \"the modern stainless steel microwave with the glass door\" on your right side?", "option": [], "answer": "2", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the sleek black rectangular stove with touch controls\" and \"the modern stainless steel microwave with the glass door\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_47331322/Validation_47331322_47331322_179163.328.png", "Validation_47331322/Validation_47331322_47331322_179165.327.png", "Validation_47331322/Validation_47331322_47331322_179167.326.png", "Validation_47331322/Validation_47331322_47331322_179169.326.png" ], "turn_id": 7, "scan_id": "Validation_47331322", "origin_question": "By the end of the 6 turn, were you getting closer to or farther away from these two objects: X: \"the sleek black rectangular stove with touch controls\", Y: \"the sleek, rectangular oven\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to X and farther away from Y", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 6 turn, were you getting closer to or farther away from these two objects: X: \"the sleek black rectangular stove with touch controls\", Y: \"the sleek, rectangular oven\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_45261631/Validation_45261631_45261631_56027.422.png", "Validation_45261631/Validation_45261631_45261631_56029.404.png", "Validation_45261631/Validation_45261631_45261631_56031.404.png", "Validation_45261631/Validation_45261631_45261631_56033.403.png", "Validation_45261631/Validation_45261631_45261631_56035.402.png" ], "turn_id": 1, "scan_id": "Validation_45261631", "origin_question": "Remember, have you seen any bed(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any bed(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_45261631/Validation_45261631_45261631_56037.401.png", "Validation_45261631/Validation_45261631_45261631_56039.400.png", "Validation_45261631/Validation_45261631_45261631_56041.400.png", "Validation_45261631/Validation_45261631_45261631_56043.399.png", "Validation_45261631/Validation_45261631_45261631_56045.398.png" ], "turn_id": 2, "scan_id": "Validation_45261631", "origin_question": "Which direction is the sleek, geometric metallic chandelier in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the sleek, geometric metallic chandelier in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_45261631/Validation_45261631_45261631_56047.397.png", "Validation_45261631/Validation_45261631_45261631_56049.397.png", "Validation_45261631/Validation_45261631_45261631_56051.396.png", "Validation_45261631/Validation_45261631_45261631_56053.395.png", "Validation_45261631/Validation_45261631_45261631_56055.394.png" ], "turn_id": 3, "scan_id": "Validation_45261631", "origin_question": "Is the sleek, geometric metallic chandelier to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the sleek, geometric metallic chandelier to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_45261631/Validation_45261631_45261631_56057.393.png", "Validation_45261631/Validation_45261631_45261631_56059.393.png", "Validation_45261631/Validation_45261631_45261631_56061.392.png", "Validation_45261631/Validation_45261631_45261631_56063.391.png", "Validation_45261631/Validation_45261631_45261631_56065.390.png" ], "turn_id": 4, "scan_id": "Validation_45261631", "origin_question": "Is the radiator to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the radiator to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_45261631/Validation_45261631_45261631_56067.389.png", "Validation_45261631/Validation_45261631_45261631_56069.389.png", "Validation_45261631/Validation_45261631_45261631_56071.404.png", "Validation_45261631/Validation_45261631_45261631_56073.404.png", "Validation_45261631/Validation_45261631_45261631_56075.403.png" ], "turn_id": 5, "scan_id": "Validation_45261631", "origin_question": "Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the sleek, geometric metallic chandelier\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the sleek, geometric metallic chandelier\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_45261631/Validation_45261631_45261631_56077.402.png", "Validation_45261631/Validation_45261631_45261631_56079.401.png", "Validation_45261631/Validation_45261631_45261631_56081.401.png", "Validation_45261631/Validation_45261631_45261631_56083.400.png", "Validation_45261631/Validation_45261631_45261631_56085.399.png" ], "turn_id": 6, "scan_id": "Validation_45261631", "origin_question": "Which two objects are on the same side of you now? \"the rectangular, light-colored plush pillow\", \"the sleek, geometric metallic chandelier\", \"the window\". ", "option": [ "the rectangular, light-colored plush pillow, the sleek, geometric metallic chandelier", "the rectangular, light-colored plush pillow, the window", "the sleek, geometric metallic chandelier, the window" ], "answer": "the rectangular, light-colored plush pillow, the sleek, geometric metallic chandelier", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the rectangular, light-colored plush pillow\", \"the sleek, geometric metallic chandelier\", \"the window\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular, light-colored plush pillow, the sleek, geometric metallic chandelier', 'the rectangular, light-colored plush pillow, the window', 'the sleek, geometric metallic chandelier, the window'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_45261631/Validation_45261631_45261631_56087.398.png", "Validation_45261631/Validation_45261631_45261631_56089.397.png", "Validation_45261631/Validation_45261631_45261631_56091.397.png", "Validation_45261631/Validation_45261631_45261631_56093.396.png", "Validation_45261631/Validation_45261631_45261631_56095.395.png" ], "turn_id": 7, "scan_id": "Validation_45261631", "origin_question": "Is the window to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the window to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_45261631/Validation_45261631_45261631_56097.394.png", "Validation_45261631/Validation_45261631_45261631_56099.393.png", "Validation_45261631/Validation_45261631_45261631_56101.393.png", "Validation_45261631/Validation_45261631_45261631_56103.392.png", "Validation_45261631/Validation_45261631_45261631_56105.391.png" ], "turn_id": 8, "scan_id": "Validation_45261631", "origin_question": "Compared to your position at the end of the 5 turn, are you now closer or farther away from \"the window\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 5 turn, are you now closer or farther away from \"the window\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "Validation_47895547/Validation_47895547_47895547_455767.633.png", "Validation_47895547/Validation_47895547_47895547_455769.599.png", "Validation_47895547/Validation_47895547_47895547_455771.598.png", "Validation_47895547/Validation_47895547_47895547_455773.597.png" ], "turn_id": 1, "scan_id": "Validation_47895547", "origin_question": "Remember, how many chair(s) have you seen so far? ", "option": [], "answer": "2", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many chair(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47895547/Validation_47895547_47895547_455775.596.png", "Validation_47895547/Validation_47895547_47895547_455777.612.png", "Validation_47895547/Validation_47895547_47895547_455779.612.png", "Validation_47895547/Validation_47895547_47895547_455781.611.png" ], "turn_id": 2, "scan_id": "Validation_47895547", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the medium-sized white metal radiator\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the medium-sized white metal radiator\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_47895547/Validation_47895547_47895547_455783.610.png", "Validation_47895547/Validation_47895547_47895547_455785.609.png", "Validation_47895547/Validation_47895547_47895547_455787.608.png", "Validation_47895547/Validation_47895547_47895547_455789.607.png" ], "turn_id": 3, "scan_id": "Validation_47895547", "origin_question": "Please recall \"the stainless steel, sleek, modern refrigerator\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.81", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the stainless steel, sleek, modern refrigerator\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_47895547/Validation_47895547_47895547_455791.607.png", "Validation_47895547/Validation_47895547_47895547_455793.606.png", "Validation_47895547/Validation_47895547_47895547_455795.605.png", "Validation_47895547/Validation_47895547_47895547_455797.604.png" ], "turn_id": 4, "scan_id": "Validation_47895547", "origin_question": "Please recall \"the sleek rectangular window with the white pvc frame\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.33", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the sleek rectangular window with the white pvc frame\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_47895547/Validation_47895547_47895547_455799.604.png", "Validation_47895547/Validation_47895547_47895547_455801.603.png", "Validation_47895547/Validation_47895547_47895547_455803.602.png", "Validation_47895547/Validation_47895547_47895547_455805.601.png" ], "turn_id": 5, "scan_id": "Validation_47895547", "origin_question": "Please recall \"the compact stainless steel stove with black burner grates\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.63", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the compact stainless steel stove with black burner grates\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "Validation_47895547/Validation_47895547_47895547_455807.600.png", "Validation_47895547/Validation_47895547_47895547_455809.599.png", "Validation_47895547/Validation_47895547_47895547_455811.599.png", "Validation_47895547/Validation_47895547_47895547_455813.598.png" ], "turn_id": 6, "scan_id": "Validation_47895547", "origin_question": "Among these three objects(\"the oven\",\"the medium-sized white metal radiator\",\"the sleek rectangular window with the white pvc frame\" ), which one is the closest to you?", "option": [ "the oven", "the medium-sized white metal radiator", "the sleek rectangular window with the white pvc frame" ], "answer": "the oven", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects(\"the oven\",\"the medium-sized white metal radiator\",\"the sleek rectangular window with the white pvc frame\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the oven', 'the medium-sized white metal radiator', 'the sleek rectangular window with the white pvc frame'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_47895547/Validation_47895547_47895547_455815.597.png", "Validation_47895547/Validation_47895547_47895547_455817.596.png", "Validation_47895547/Validation_47895547_47895547_455819.612.png", "Validation_47895547/Validation_47895547_47895547_455821.611.png" ], "turn_id": 7, "scan_id": "Validation_47895547", "origin_question": "Please recall \"the stainless steel range hood with the sleek, angular shape\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.04", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the stainless steel range hood with the sleek, angular shape\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42899617/Validation_42899617_42899617_228927.391.png", "Validation_42899617/Validation_42899617_42899617_228929.374.png", "Validation_42899617/Validation_42899617_42899617_228931.373.png", "Validation_42899617/Validation_42899617_42899617_228933.372.png" ], "turn_id": 1, "scan_id": "Validation_42899617", "origin_question": "Remember, have you seen any door(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any door(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_42899617/Validation_42899617_42899617_228935.371.png", "Validation_42899617/Validation_42899617_42899617_228937.370.png", "Validation_42899617/Validation_42899617_42899617_228939.370.png", "Validation_42899617/Validation_42899617_42899617_228941.369.png" ], "turn_id": 2, "scan_id": "Validation_42899617", "origin_question": "Which two objects are on the same side of you now? \"the sleek white rectangular ceramic sink with rounded corners\", \"standard-sized white toilet paper rolls stacked vertically\", \"the bathtub\". ", "option": [ "the sleek white rectangular ceramic sink with rounded corners, standard-sized white toilet paper rolls stacked vertically", "the sleek white rectangular ceramic sink with rounded corners, the bathtub", "standard-sized white toilet paper rolls stacked vertically, the bathtub" ], "answer": "the sleek white rectangular ceramic sink with rounded corners, standard-sized white toilet paper rolls stacked vertically", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the sleek white rectangular ceramic sink with rounded corners\", \"standard-sized white toilet paper rolls stacked vertically\", \"the bathtub\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the sleek white rectangular ceramic sink with rounded corners, standard-sized white toilet paper rolls stacked vertically', 'the sleek white rectangular ceramic sink with rounded corners, the bathtub', 'standard-sized white toilet paper rolls stacked vertically, the bathtub'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_42899617/Validation_42899617_42899617_228943.368.png", "Validation_42899617/Validation_42899617_42899617_228945.367.png", "Validation_42899617/Validation_42899617_42899617_228947.366.png", "Validation_42899617/Validation_42899617_42899617_228949.366.png" ], "turn_id": 3, "scan_id": "Validation_42899617", "origin_question": "Is the sleek white rectangular ceramic sink with rounded corners to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the sleek white rectangular ceramic sink with rounded corners to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_42899617/Validation_42899617_42899617_228951.365.png", "Validation_42899617/Validation_42899617_42899617_228953.364.png", "Validation_42899617/Validation_42899617_42899617_228955.363.png", "Validation_42899617/Validation_42899617_42899617_228957.362.png" ], "turn_id": 4, "scan_id": "Validation_42899617", "origin_question": "Please recall \"standard-sized white toilet paper rolls stacked vertically\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.85", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"standard-sized white toilet paper rolls stacked vertically\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_42899617/Validation_42899617_42899617_228959.362.png", "Validation_42899617/Validation_42899617_42899617_228961.361.png", "Validation_42899617/Validation_42899617_42899617_228963.360.png", "Validation_42899617/Validation_42899617_42899617_228965.359.png" ], "turn_id": 5, "scan_id": "Validation_42899617", "origin_question": "In which turn did you see both \"standard-sized white toilet paper rolls stacked vertically\" and \"the sleek, vertical stainless steel ladder\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"standard-sized white toilet paper rolls stacked vertically\" and \"the sleek, vertical stainless steel ladder\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_42899617/Validation_42899617_42899617_228967.359.png", "Validation_42899617/Validation_42899617_42899617_228969.374.png", "Validation_42899617/Validation_42899617_42899617_228971.374.png", "Validation_42899617/Validation_42899617_42899617_228973.373.png" ], "turn_id": 6, "scan_id": "Validation_42899617", "origin_question": "Please recall \"the toilet\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.5", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the toilet\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_42899617/Validation_42899617_42899617_228975.372.png", "Validation_42899617/Validation_42899617_42899617_228977.371.png", "Validation_42899617/Validation_42899617_42899617_228979.370.png", "Validation_42899617/Validation_42899617_42899617_228981.370.png" ], "turn_id": 7, "scan_id": "Validation_42899617", "origin_question": "Which two objects are on the same side of you now? \"the sleek, vertical stainless steel ladder\", \"the sleek white rectangular door with the brass handle\", \"the toilet\". ", "option": [ "the sleek, vertical stainless steel ladder, the sleek white rectangular door with the brass handle", "the sleek, vertical stainless steel ladder, the toilet", "the sleek white rectangular door with the brass handle, the toilet" ], "answer": "the sleek, vertical stainless steel ladder, the sleek white rectangular door with the brass handle", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the sleek, vertical stainless steel ladder\", \"the sleek white rectangular door with the brass handle\", \"the toilet\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the sleek, vertical stainless steel ladder, the sleek white rectangular door with the brass handle', 'the sleek, vertical stainless steel ladder, the toilet', 'the sleek white rectangular door with the brass handle, the toilet'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "Validation_47430470/Validation_47430470_47430470_311316.953.png", "Validation_47430470/Validation_47430470_47430470_311318.935.png", "Validation_47430470/Validation_47430470_47430470_311320.934.png", "Validation_47430470/Validation_47430470_47430470_311322.934.png", "Validation_47430470/Validation_47430470_47430470_311324.950.png" ], "turn_id": 1, "scan_id": "Validation_47430470", "origin_question": "Remember, how many oven(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many oven(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_47430470/Validation_47430470_47430470_311326.949.png", "Validation_47430470/Validation_47430470_47430470_311328.948.png", "Validation_47430470/Validation_47430470_47430470_311330.947.png", "Validation_47430470/Validation_47430470_47430470_311332.946.png", "Validation_47430470/Validation_47430470_47430470_311334.945.png" ], "turn_id": 2, "scan_id": "Validation_47430470", "origin_question": "When did you first discover \"the tall, rectangular wood-framed window\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the tall, rectangular wood-framed window\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_47430470/Validation_47430470_47430470_311336.945.png", "Validation_47430470/Validation_47430470_47430470_311338.944.png", "Validation_47430470/Validation_47430470_47430470_311340.943.png", "Validation_47430470/Validation_47430470_47430470_311342.942.png", "Validation_47430470/Validation_47430470_47430470_311344.942.png" ], "turn_id": 3, "scan_id": "Validation_47430470", "origin_question": "How far is your current position from where you were at the end of the turn 1? (in meters)", "option": [], "answer": "3.6", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 1? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_47430470/Validation_47430470_47430470_311346.941.png", "Validation_47430470/Validation_47430470_47430470_311348.940.png", "Validation_47430470/Validation_47430470_47430470_311350.939.png", "Validation_47430470/Validation_47430470_47430470_311352.938.png", "Validation_47430470/Validation_47430470_47430470_311354.937.png" ], "turn_id": 4, "scan_id": "Validation_47430470", "origin_question": "What will be the first-time appearance order of the following categories: oven, sink, table?", "option": [ "table, oven, sink", "table, sink, oven", "sink, table, oven", "oven, table, sink" ], "answer": "table, oven, sink", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: oven, sink, table? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['table, oven, sink', 'table, sink, oven', 'sink, table, oven', 'oven, table, sink'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_47430470/Validation_47430470_47430470_311356.937.png", "Validation_47430470/Validation_47430470_47430470_311358.936.png", "Validation_47430470/Validation_47430470_47430470_311360.935.png", "Validation_47430470/Validation_47430470_47430470_311362.934.png", "Validation_47430470/Validation_47430470_47430470_311364.950.png" ], "turn_id": 5, "scan_id": "Validation_47430470", "origin_question": "By the end of the 3 turn, were you getting closer to or farther away from these two objects: X: \"the rich brown curtain made of thick material\", Y: \"the compact beige plastic stool with circular holes\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "farther away from both X and Y", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 3 turn, were you getting closer to or farther away from these two objects: X: \"the rich brown curtain made of thick material\", Y: \"the compact beige plastic stool with circular holes\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_47430470/Validation_47430470_47430470_311366.949.png", "Validation_47430470/Validation_47430470_47430470_311368.948.png", "Validation_47430470/Validation_47430470_47430470_311370.948.png", "Validation_47430470/Validation_47430470_47430470_311372.947.png", "Validation_47430470/Validation_47430470_47430470_311374.946.png" ], "turn_id": 6, "scan_id": "Validation_47430470", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the tall, stainless steel refrigerator with colorful magnets\" from your position?", "option": [], "answer": "171.02", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the tall, stainless steel refrigerator with colorful magnets\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_47430470/Validation_47430470_47430470_311376.945.png", "Validation_47430470/Validation_47430470_47430470_311378.944.png", "Validation_47430470/Validation_47430470_47430470_311380.943.png", "Validation_47430470/Validation_47430470_47430470_311382.943.png", "Validation_47430470/Validation_47430470_47430470_311384.942.png" ], "turn_id": 7, "scan_id": "Validation_47430470", "origin_question": "Is the tall, stainless steel refrigerator with colorful magnets to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the tall, stainless steel refrigerator with colorful magnets to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_47430470/Validation_47430470_47430470_311386.941.png", "Validation_47430470/Validation_47430470_47430470_311388.940.png", "Validation_47430470/Validation_47430470_47430470_311390.939.png", "Validation_47430470/Validation_47430470_47430470_311392.939.png", "Validation_47430470/Validation_47430470_47430470_311394.938.png" ], "turn_id": 8, "scan_id": "Validation_47430470", "origin_question": "When did you first discover \"the sleek cylindrical glass flowerpot\"? (index of the turn)", "option": [], "answer": "8", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the sleek cylindrical glass flowerpot\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_47430470/Validation_47430470_47430470_311396.937.png", "Validation_47430470/Validation_47430470_47430470_311398.936.png", "Validation_47430470/Validation_47430470_47430470_311400.935.png", "Validation_47430470/Validation_47430470_47430470_311402.934.png", "Validation_47430470/Validation_47430470_47430470_311404.934.png" ], "turn_id": 9, "scan_id": "Validation_47430470", "origin_question": "By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the rich brown curtain made of thick material\", Y: \"the tall, stainless steel refrigerator with colorful magnets\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to X and farther away from Y", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the rich brown curtain made of thick material\", Y: \"the tall, stainless steel refrigerator with colorful magnets\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_47430470/Validation_47430470_47430470_311406.950.png", "Validation_47430470/Validation_47430470_47430470_311408.949.png", "Validation_47430470/Validation_47430470_47430470_311410.948.png", "Validation_47430470/Validation_47430470_47430470_311412.947.png", "Validation_47430470/Validation_47430470_47430470_311414.946.png" ], "turn_id": 10, "scan_id": "Validation_47430470", "origin_question": "By the end of the 6 turn, were you getting closer to or farther away from these two objects: X: \"the compact beige plastic stool with circular holes\", Y: \"the sleek cylindrical glass flowerpot\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to Y and farther away from X", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 6 turn, were you getting closer to or farther away from these two objects: X: \"the compact beige plastic stool with circular holes\", Y: \"the sleek cylindrical glass flowerpot\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47895777/Validation_47895777_47895777_386611.486.png", "Validation_47895777/Validation_47895777_47895777_386613.468.png", "Validation_47895777/Validation_47895777_47895777_386615.468.png", "Validation_47895777/Validation_47895777_47895777_386617.467.png" ], "turn_id": 1, "scan_id": "Validation_47895777", "origin_question": "Remember, have you seen any soap dish(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any soap dish(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47895777/Validation_47895777_47895777_386619.466.png", "Validation_47895777/Validation_47895777_47895777_386621.465.png", "Validation_47895777/Validation_47895777_47895777_386623.464.png", "Validation_47895777/Validation_47895777_47895777_386625.464.png" ], "turn_id": 2, "scan_id": "Validation_47895777", "origin_question": "In which turn were you farthest from \"the medium-sized rectangular gray and white striped pillow\"? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the medium-sized rectangular gray and white striped pillow\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47895777/Validation_47895777_47895777_386627.463.png", "Validation_47895777/Validation_47895777_47895777_386629.462.png", "Validation_47895777/Validation_47895777_47895777_386631.461.png", "Validation_47895777/Validation_47895777_47895777_386633.477.png" ], "turn_id": 3, "scan_id": "Validation_47895777", "origin_question": "In which turn were you closest to \"the light rectangular picture frame with small photographs\"? ", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the light rectangular picture frame with small photographs\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_47895777/Validation_47895777_47895777_386635.476.png", "Validation_47895777/Validation_47895777_47895777_386637.475.png", "Validation_47895777/Validation_47895777_47895777_386639.474.png", "Validation_47895777/Validation_47895777_47895777_386641.474.png" ], "turn_id": 4, "scan_id": "Validation_47895777", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the large, sleek black monitor with plastic and metallic construction\" from your position?", "option": [], "answer": "152.95", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the large, sleek black monitor with plastic and metallic construction\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_47895777/Validation_47895777_47895777_386643.473.png", "Validation_47895777/Validation_47895777_47895777_386645.472.png", "Validation_47895777/Validation_47895777_47895777_386647.471.png", "Validation_47895777/Validation_47895777_47895777_386649.470.png" ], "turn_id": 5, "scan_id": "Validation_47895777", "origin_question": "Assuming the direction you are facing at the end of the turn 3 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Right", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 3 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "Validation_47895777/Validation_47895777_47895777_386651.470.png", "Validation_47895777/Validation_47895777_47895777_386653.469.png", "Validation_47895777/Validation_47895777_47895777_386655.468.png", "Validation_47895777/Validation_47895777_47895777_386657.467.png" ], "turn_id": 6, "scan_id": "Validation_47895777", "origin_question": "Among these three objects(\"the light rectangular picture frame with small photographs\",\"the tall light with the cream shade and steel base\",\"the medium, cylindrical ceramic vase in soft white\" ), which one is the closest to you?", "option": [ "the light rectangular picture frame with small photographs", "the tall light with the cream shade and steel base", "the medium, cylindrical ceramic vase in soft white" ], "answer": "the medium, cylindrical ceramic vase in soft white", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects(\"the light rectangular picture frame with small photographs\",\"the tall light with the cream shade and steel base\",\"the medium, cylindrical ceramic vase in soft white\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the light rectangular picture frame with small photographs', 'the tall light with the cream shade and steel base', 'the medium, cylindrical ceramic vase in soft white'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_47895777/Validation_47895777_47895777_386659.466.png", "Validation_47895777/Validation_47895777_47895777_386661.465.png", "Validation_47895777/Validation_47895777_47895777_386663.465.png", "Validation_47895777/Validation_47895777_47895777_386665.464.png" ], "turn_id": 7, "scan_id": "Validation_47895777", "origin_question": "Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance forward or backward from that position?", "option": [ "Forward", "Backward" ], "answer": "Backward", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance forward or backward from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Forward', 'Backward'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47895777/Validation_47895777_47895777_386667.463.png", "Validation_47895777/Validation_47895777_47895777_386669.462.png", "Validation_47895777/Validation_47895777_47895777_386671.461.png", "Validation_47895777/Validation_47895777_47895777_386673.477.png" ], "turn_id": 8, "scan_id": "Validation_47895777", "origin_question": "Which direction is the light brown rectangular drop-leaf table with curved edges in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the light brown rectangular drop-leaf table with curved edges in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "Validation_45662942/Validation_45662942_45662942_22882.675.png", "Validation_45662942/Validation_45662942_45662942_22884.657.png", "Validation_45662942/Validation_45662942_45662942_22886.656.png", "Validation_45662942/Validation_45662942_45662942_22888.656.png" ], "turn_id": 1, "scan_id": "Validation_45662942", "origin_question": "Remember, how many pillow(s) have you seen so far? ", "option": [], "answer": "4", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many pillow(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_45662942/Validation_45662942_45662942_22890.671.png", "Validation_45662942/Validation_45662942_45662942_22892.671.png", "Validation_45662942/Validation_45662942_45662942_22894.670.png", "Validation_45662942/Validation_45662942_45662942_22896.669.png" ], "turn_id": 2, "scan_id": "Validation_45662942", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the queen or king-sized bed with the light beige bedspread\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the queen or king-sized bed with the light beige bedspread\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_45662942/Validation_45662942_45662942_22898.668.png", "Validation_45662942/Validation_45662942_45662942_22900.667.png", "Validation_45662942/Validation_45662942_45662942_22902.666.png", "Validation_45662942/Validation_45662942_45662942_22904.666.png" ], "turn_id": 3, "scan_id": "Validation_45662942", "origin_question": "Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Right", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_45662942/Validation_45662942_45662942_22906.665.png", "Validation_45662942/Validation_45662942_45662942_22908.664.png", "Validation_45662942/Validation_45662942_45662942_22910.663.png", "Validation_45662942/Validation_45662942_45662942_22912.662.png" ], "turn_id": 4, "scan_id": "Validation_45662942", "origin_question": "By the end of the 3 turn, were you getting closer to or farther away from these two objects: X: \"the queen or king-sized bed with the light beige bedspread\", Y: \"the neutral-colored wooden door with panel design\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "farther away from both X and Y", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 3 turn, were you getting closer to or farther away from these two objects: X: \"the queen or king-sized bed with the light beige bedspread\", Y: \"the neutral-colored wooden door with panel design\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_45662942/Validation_45662942_45662942_22914.661.png", "Validation_45662942/Validation_45662942_45662942_22916.661.png", "Validation_45662942/Validation_45662942_45662942_22918.660.png", "Validation_45662942/Validation_45662942_45662942_22920.659.png" ], "turn_id": 5, "scan_id": "Validation_45662942", "origin_question": "Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "108.44", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "Validation_45662942/Validation_45662942_45662942_22922.658.png", "Validation_45662942/Validation_45662942_45662942_22924.657.png", "Validation_45662942/Validation_45662942_45662942_22926.657.png", "Validation_45662942/Validation_45662942_45662942_22928.656.png" ], "turn_id": 6, "scan_id": "Validation_45662942", "origin_question": "Remember, how many pillow(s) have you seen so far? ", "option": [], "answer": "4", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many pillow(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_45260914/Validation_45260914_45260914_1608.911.png", "Validation_45260914/Validation_45260914_45260914_1610.894.png", "Validation_45260914/Validation_45260914_45260914_1612.893.png", "Validation_45260914/Validation_45260914_45260914_1614.892.png", "Validation_45260914/Validation_45260914_45260914_1616.892.png" ], "turn_id": 1, "scan_id": "Validation_45260914", "origin_question": "Remember, have you seen any couch(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any couch(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_45260914/Validation_45260914_45260914_1618.891.png", "Validation_45260914/Validation_45260914_45260914_1620.890.png", "Validation_45260914/Validation_45260914_45260914_1622.889.png", "Validation_45260914/Validation_45260914_45260914_1624.889.png", "Validation_45260914/Validation_45260914_45260914_1626.888.png" ], "turn_id": 2, "scan_id": "Validation_45260914", "origin_question": "In which turn were you farthest from \"the tall, bushy plant with lush green foliage\"? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the tall, bushy plant with lush green foliage\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_45260914/Validation_45260914_45260914_1628.904.png", "Validation_45260914/Validation_45260914_45260914_1630.903.png", "Validation_45260914/Validation_45260914_45260914_1632.902.png", "Validation_45260914/Validation_45260914_45260914_1634.901.png", "Validation_45260914/Validation_45260914_45260914_1636.900.png" ], "turn_id": 3, "scan_id": "Validation_45260914", "origin_question": "", "option": [], "answer": "", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_45260914/Validation_45260914_45260914_1638.900.png", "Validation_45260914/Validation_45260914_45260914_1640.899.png", "Validation_45260914/Validation_45260914_45260914_1642.898.png", "Validation_45260914/Validation_45260914_45260914_1644.897.png", "Validation_45260914/Validation_45260914_45260914_1646.897.png" ], "turn_id": 4, "scan_id": "Validation_45260914", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_45260914/Validation_45260914_45260914_1648.896.png", "Validation_45260914/Validation_45260914_45260914_1650.895.png", "Validation_45260914/Validation_45260914_45260914_1652.894.png", "Validation_45260914/Validation_45260914_45260914_1654.893.png", "Validation_45260914/Validation_45260914_45260914_1656.892.png" ], "turn_id": 5, "scan_id": "Validation_45260914", "origin_question": "", "option": [], "answer": "", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_45260914/Validation_45260914_45260914_1658.892.png", "Validation_45260914/Validation_45260914_45260914_1660.891.png", "Validation_45260914/Validation_45260914_45260914_1662.890.png", "Validation_45260914/Validation_45260914_45260914_1664.889.png", "Validation_45260914/Validation_45260914_45260914_1666.889.png" ], "turn_id": 6, "scan_id": "Validation_45260914", "origin_question": "When did you first discover \"the tall, bushy plant with lush green foliage\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the tall, bushy plant with lush green foliage\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_45260914/Validation_45260914_45260914_1668.888.png", "Validation_45260914/Validation_45260914_45260914_1670.904.png", "Validation_45260914/Validation_45260914_45260914_1672.903.png", "Validation_45260914/Validation_45260914_45260914_1674.902.png", "Validation_45260914/Validation_45260914_45260914_1676.901.png" ], "turn_id": 7, "scan_id": "Validation_45260914", "origin_question": "Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "141.58", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_45260914/Validation_45260914_45260914_1678.900.png", "Validation_45260914/Validation_45260914_45260914_1680.900.png", "Validation_45260914/Validation_45260914_45260914_1682.899.png", "Validation_45260914/Validation_45260914_45260914_1684.898.png", "Validation_45260914/Validation_45260914_45260914_1686.897.png" ], "turn_id": 8, "scan_id": "Validation_45260914", "origin_question": "What will be the first-time appearance order of the following categories: couch, fireplace, monitor?", "option": [ "fireplace, monitor, couch", "monitor, fireplace, couch", "couch, monitor, fireplace", "couch, fireplace, monitor" ], "answer": "couch, monitor, fireplace", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: couch, fireplace, monitor? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['fireplace, monitor, couch', 'monitor, fireplace, couch', 'couch, monitor, fireplace', 'couch, fireplace, monitor'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_45260914/Validation_45260914_45260914_1688.897.png", "Validation_45260914/Validation_45260914_45260914_1690.896.png", "Validation_45260914/Validation_45260914_45260914_1692.895.png", "Validation_45260914/Validation_45260914_45260914_1694.894.png", "Validation_45260914/Validation_45260914_45260914_1696.893.png" ], "turn_id": 9, "scan_id": "Validation_45260914", "origin_question": "Assuming the direction you are facing at the end of the turn 7 is forward, did you move a certain distance forward or backward from that position?", "option": [ "Forward", "Backward" ], "answer": "Backward", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 7 is forward, did you move a certain distance forward or backward from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Forward', 'Backward'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_45260914/Validation_45260914_45260914_1698.893.png", "Validation_45260914/Validation_45260914_45260914_1700.892.png", "Validation_45260914/Validation_45260914_45260914_1702.891.png", "Validation_45260914/Validation_45260914_45260914_1704.890.png", "Validation_45260914/Validation_45260914_45260914_1706.889.png" ], "turn_id": 10, "scan_id": "Validation_45260914", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the medium-sized lamp with the cylindrical cream shade\", Y: \"the medium-sized rectangular monitor with the glossy black frame\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to Y and farther away from X", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the medium-sized lamp with the cylindrical cream shade\", Y: \"the medium-sized rectangular monitor with the glossy black frame\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47115543/Validation_47115543_47115543_15621.632.png", "Validation_47115543/Validation_47115543_47115543_15623.614.png", "Validation_47115543/Validation_47115543_47115543_15625.614.png", "Validation_47115543/Validation_47115543_47115543_15627.613.png" ], "turn_id": 1, "scan_id": "Validation_47115543", "origin_question": "Remember, have you seen any range hood(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any range hood(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47115543/Validation_47115543_47115543_15629.612.png", "Validation_47115543/Validation_47115543_47115543_15631.611.png", "Validation_47115543/Validation_47115543_47115543_15633.610.png", "Validation_47115543/Validation_47115543_47115543_15635.626.png" ], "turn_id": 2, "scan_id": "Validation_47115543", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47115543/Validation_47115543_47115543_15637.626.png", "Validation_47115543/Validation_47115543_47115543_15639.625.png", "Validation_47115543/Validation_47115543_47115543_15641.624.png", "Validation_47115543/Validation_47115543_47115543_15643.623.png" ], "turn_id": 3, "scan_id": "Validation_47115543", "origin_question": "In which turn were you closest to \"the light rectangular paneled door\"? ", "option": [], "answer": "2", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the light rectangular paneled door\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_47115543/Validation_47115543_47115543_15645.622.png", "Validation_47115543/Validation_47115543_47115543_15647.621.png", "Validation_47115543/Validation_47115543_47115543_15649.621.png", "Validation_47115543/Validation_47115543_47115543_15651.620.png" ], "turn_id": 4, "scan_id": "Validation_47115543", "origin_question": "Is the light rectangular paneled door to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the light rectangular paneled door to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_47115543/Validation_47115543_47115543_15653.619.png", "Validation_47115543/Validation_47115543_47115543_15655.618.png", "Validation_47115543/Validation_47115543_47115543_15657.618.png", "Validation_47115543/Validation_47115543_47115543_15659.617.png" ], "turn_id": 5, "scan_id": "Validation_47115543", "origin_question": "What will be the first-time appearance order of the following categories: door, table, window?", "option": [ "table, window, door", "table, door, window", "door, window, table", "door, table, window" ], "answer": "door, table, window", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: door, table, window? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['table, window, door', 'table, door, window', 'door, window, table', 'door, table, window'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_47115543/Validation_47115543_47115543_15661.616.png", "Validation_47115543/Validation_47115543_47115543_15663.615.png", "Validation_47115543/Validation_47115543_47115543_15665.614.png", "Validation_47115543/Validation_47115543_47115543_15667.614.png" ], "turn_id": 6, "scan_id": "Validation_47115543", "origin_question": "Using your orientation at the end of turn 5 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "54.5", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 5 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47115543/Validation_47115543_47115543_15669.613.png", "Validation_47115543/Validation_47115543_47115543_15671.612.png", "Validation_47115543/Validation_47115543_47115543_15673.611.png", "Validation_47115543/Validation_47115543_47115543_15675.611.png" ], "turn_id": 7, "scan_id": "Validation_47115543", "origin_question": "In which turn were you closest to \"an antique wooden box with the rustic finish\"? ", "option": [], "answer": "7", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"an antique wooden box with the rustic finish\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_47115543/Validation_47115543_47115543_15677.626.png", "Validation_47115543/Validation_47115543_47115543_15679.626.png", "Validation_47115543/Validation_47115543_47115543_15681.625.png", "Validation_47115543/Validation_47115543_47115543_15683.624.png" ], "turn_id": 8, "scan_id": "Validation_47115543", "origin_question": "Using your orientation at the end of turn 7 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "100.79", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 7 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_45261594/Validation_45261594_45261594_78029.211.png", "Validation_45261594/Validation_45261594_45261594_78030.694.png", "Validation_45261594/Validation_45261594_45261594_78032.210.png", "Validation_45261594/Validation_45261594_45261594_78033.709.png" ], "turn_id": 1, "scan_id": "Validation_45261594", "origin_question": "Remember, have you seen any sink(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any sink(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_45261594/Validation_45261594_45261594_78035.208.png", "Validation_45261594/Validation_45261594_45261594_78036.708.png", "Validation_45261594/Validation_45261594_45261594_78038.207.png", "Validation_45261594/Validation_45261594_45261594_78039.707.png" ], "turn_id": 2, "scan_id": "Validation_45261594", "origin_question": "Is the rectangular white-framed window providing light to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the rectangular white-framed window providing light to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_45261594/Validation_45261594_45261594_78041.206.png", "Validation_45261594/Validation_45261594_45261594_78042.705.png", "Validation_45261594/Validation_45261594_45261594_78044.205.png", "Validation_45261594/Validation_45261594_45261594_78045.704.png" ], "turn_id": 3, "scan_id": "Validation_45261594", "origin_question": "Please recall \"the compact, wall-mounted white ceramic sink\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.97", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the compact, wall-mounted white ceramic sink\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_45261594/Validation_45261594_45261594_78047.204.png", "Validation_45261594/Validation_45261594_45261594_78048.703.png", "Validation_45261594/Validation_45261594_45261594_78050.202.png", "Validation_45261594/Validation_45261594_45261594_78051.702.png" ], "turn_id": 4, "scan_id": "Validation_45261594", "origin_question": "Please recall \"the sleek, rounded white ceramic toilet\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.97", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the sleek, rounded white ceramic toilet\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_45261594/Validation_45261594_45261594_78053.201.png", "Validation_45261594/Validation_45261594_45261594_78054.700.png", "Validation_45261594/Validation_45261594_45261594_78056.200.png", "Validation_45261594/Validation_45261594_45261594_78057.699.png" ], "turn_id": 5, "scan_id": "Validation_45261594", "origin_question": "At the end of which turn were both of \"the sleek rectangular white acrylic bathtub\" and \"the rectangular white-framed window providing light\" on your right side?", "option": [], "answer": "3", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the sleek rectangular white acrylic bathtub\" and \"the rectangular white-framed window providing light\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42445021/Validation_42445021_42445021_48880.276.png", "Validation_42445021/Validation_42445021_42445021_48881.759.png", "Validation_42445021/Validation_42445021_42445021_48883.258.png", "Validation_42445021/Validation_42445021_42445021_48884.758.png" ], "turn_id": 1, "scan_id": "Validation_42445021", "origin_question": "Remember, have you seen any soap dish(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any soap dish(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "Validation_42445021/Validation_42445021_42445021_48886.257.png", "Validation_42445021/Validation_42445021_42445021_48887.757.png", "Validation_42445021/Validation_42445021_42445021_48889.256.png", "Validation_42445021/Validation_42445021_42445021_48890.755.png" ], "turn_id": 2, "scan_id": "Validation_42445021", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the large rectangular white aluminum window\"; \"the stylish neutral-colored rectangular couch with patterns\"; \"the door\" ", "option": [ "the large rectangular white aluminum window", "the stylish neutral-colored rectangular couch with patterns", "the door" ], "answer": "the stylish neutral-colored rectangular couch with patterns", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the large rectangular white aluminum window\"; \"the stylish neutral-colored rectangular couch with patterns\"; \"the door\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the large rectangular white aluminum window', 'the stylish neutral-colored rectangular couch with patterns', 'the door'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_42445021/Validation_42445021_42445021_48892.255.png", "Validation_42445021/Validation_42445021_42445021_48893.754.png", "Validation_42445021/Validation_42445021_42445021_48895.270.png", "Validation_42445021/Validation_42445021_42445021_48896.770.png" ], "turn_id": 3, "scan_id": "Validation_42445021", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "93.54", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_42445021/Validation_42445021_42445021_48898.269.png", "Validation_42445021/Validation_42445021_42445021_48899.768.png", "Validation_42445021/Validation_42445021_42445021_48901.268.png", "Validation_42445021/Validation_42445021_42445021_48902.767.png" ], "turn_id": 4, "scan_id": "Validation_42445021", "origin_question": "Is the large rectangular white aluminum window to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the large rectangular white aluminum window to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_42445021/Validation_42445021_42445021_48904.267.png", "Validation_42445021/Validation_42445021_42445021_48905.766.png", "Validation_42445021/Validation_42445021_42445021_48907.266.png", "Validation_42445021/Validation_42445021_42445021_48908.765.png" ], "turn_id": 5, "scan_id": "Validation_42445021", "origin_question": "In which turn did you see both \"the large square pillow with the floral pattern\" and \"the sleek black all-in-one printer with the glossy finish\" simultaneously? ", "option": [], "answer": "5", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the large square pillow with the floral pattern\" and \"the sleek black all-in-one printer with the glossy finish\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_42445021/Validation_42445021_42445021_48910.264.png", "Validation_42445021/Validation_42445021_42445021_48911.764.png", "Validation_42445021/Validation_42445021_42445021_48913.263.png", "Validation_42445021/Validation_42445021_42445021_48914.762.png" ], "turn_id": 6, "scan_id": "Validation_42445021", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the stylish neutral-colored rectangular couch with patterns\" from your position?", "option": [], "answer": "135.98", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the stylish neutral-colored rectangular couch with patterns\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_42445021/Validation_42445021_42445021_48916.262.png", "Validation_42445021/Validation_42445021_42445021_48917.761.png", "Validation_42445021/Validation_42445021_42445021_48919.261.png", "Validation_42445021/Validation_42445021_42445021_48920.760.png" ], "turn_id": 7, "scan_id": "Validation_42445021", "origin_question": "Which two objects are on the same side of you now? \"the door\", \"the large square pillow with the floral pattern\", \"the sleek black all-in-one printer with the glossy finish\". ", "option": [ "the door, the large square pillow with the floral pattern", "the door, the sleek black all-in-one printer with the glossy finish", "the large square pillow with the floral pattern, the sleek black all-in-one printer with the glossy finish" ], "answer": "the large square pillow with the floral pattern, the sleek black all-in-one printer with the glossy finish", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the door\", \"the large square pillow with the floral pattern\", \"the sleek black all-in-one printer with the glossy finish\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the door, the large square pillow with the floral pattern', 'the door, the sleek black all-in-one printer with the glossy finish', 'the large square pillow with the floral pattern, the sleek black all-in-one printer with the glossy finish'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "Validation_42445021/Validation_42445021_42445021_48922.260.png", "Validation_42445021/Validation_42445021_42445021_48923.759.png", "Validation_42445021/Validation_42445021_42445021_48925.258.png", "Validation_42445021/Validation_42445021_42445021_48926.758.png" ], "turn_id": 8, "scan_id": "Validation_42445021", "origin_question": "Among these three objects(\"the large rectangular white aluminum window\",\"the large square pillow with the floral pattern\",\"the sleek white rectangular metal radiator\" ), which one is the closest to you?", "option": [ "the large rectangular white aluminum window", "the large square pillow with the floral pattern", "the sleek white rectangular metal radiator" ], "answer": "the large square pillow with the floral pattern", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects(\"the large rectangular white aluminum window\",\"the large square pillow with the floral pattern\",\"the sleek white rectangular metal radiator\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the large rectangular white aluminum window', 'the large square pillow with the floral pattern', 'the sleek white rectangular metal radiator'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42899731/Validation_42899731_42899731_800928.182.png", "Validation_42899731/Validation_42899731_42899731_800930.165.png", "Validation_42899731/Validation_42899731_42899731_800932.164.png", "Validation_42899731/Validation_42899731_42899731_800934.180.png", "Validation_42899731/Validation_42899731_42899731_800936.179.png" ], "turn_id": 1, "scan_id": "Validation_42899731", "origin_question": "Remember, have you seen any bed(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any bed(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_42899731/Validation_42899731_42899731_800938.178.png", "Validation_42899731/Validation_42899731_42899731_800940.177.png", "Validation_42899731/Validation_42899731_42899731_800942.176.png", "Validation_42899731/Validation_42899731_42899731_800944.176.png", "Validation_42899731/Validation_42899731_42899731_800946.175.png" ], "turn_id": 2, "scan_id": "Validation_42899731", "origin_question": "In which turn were you closest to \"the modern bed with the dark taupe bedspread and headboard\"? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the modern bed with the dark taupe bedspread and headboard\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_42899731/Validation_42899731_42899731_800948.174.png", "Validation_42899731/Validation_42899731_42899731_800950.173.png", "Validation_42899731/Validation_42899731_42899731_800952.172.png", "Validation_42899731/Validation_42899731_42899731_800954.172.png", "Validation_42899731/Validation_42899731_42899731_800956.171.png" ], "turn_id": 3, "scan_id": "Validation_42899731", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "61.15", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_42899731/Validation_42899731_42899731_800958.170.png", "Validation_42899731/Validation_42899731_42899731_800960.169.png", "Validation_42899731/Validation_42899731_42899731_800962.168.png", "Validation_42899731/Validation_42899731_42899731_800964.167.png", "Validation_42899731/Validation_42899731_42899731_800966.167.png" ], "turn_id": 4, "scan_id": "Validation_42899731", "origin_question": "Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the set of pristine white cotton towels\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the set of pristine white cotton towels\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_42899731/Validation_42899731_42899731_800968.166.png", "Validation_42899731/Validation_42899731_42899731_800970.165.png", "Validation_42899731/Validation_42899731_42899731_800972.164.png", "Validation_42899731/Validation_42899731_42899731_800974.163.png", "Validation_42899731/Validation_42899731_42899731_800976.179.png" ], "turn_id": 5, "scan_id": "Validation_42899731", "origin_question": "How far is your current position from where you were at the end of the turn 3? (in meters)", "option": [], "answer": "2.57", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 3? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_42899731/Validation_42899731_42899731_800978.178.png", "Validation_42899731/Validation_42899731_42899731_800980.177.png", "Validation_42899731/Validation_42899731_42899731_800982.177.png", "Validation_42899731/Validation_42899731_42899731_800984.176.png", "Validation_42899731/Validation_42899731_42899731_800986.175.png" ], "turn_id": 6, "scan_id": "Validation_42899731", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "127.46", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_42899731/Validation_42899731_42899731_800988.174.png", "Validation_42899731/Validation_42899731_42899731_800990.173.png", "Validation_42899731/Validation_42899731_42899731_800992.172.png", "Validation_42899731/Validation_42899731_42899731_800994.172.png", "Validation_42899731/Validation_42899731_42899731_800996.171.png" ], "turn_id": 7, "scan_id": "Validation_42899731", "origin_question": "", "option": [], "answer": "", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_42899731/Validation_42899731_42899731_800998.170.png", "Validation_42899731/Validation_42899731_42899731_801000.169.png", "Validation_42899731/Validation_42899731_42899731_801002.168.png", "Validation_42899731/Validation_42899731_42899731_801004.167.png", "Validation_42899731/Validation_42899731_42899731_801006.167.png" ], "turn_id": 8, "scan_id": "Validation_42899731", "origin_question": "At the end of which turn were both of \"the modern bed with the dark taupe bedspread and headboard\" and \"the sink\" on your left side?", "option": [], "answer": "2", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the modern bed with the dark taupe bedspread and headboard\" and \"the sink\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_42899731/Validation_42899731_42899731_801008.166.png", "Validation_42899731/Validation_42899731_42899731_801010.165.png", "Validation_42899731/Validation_42899731_42899731_801012.164.png", "Validation_42899731/Validation_42899731_42899731_801014.163.png", "Validation_42899731/Validation_42899731_42899731_801016.179.png" ], "turn_id": 9, "scan_id": "Validation_42899731", "origin_question": "Which direction is the sink in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the sink in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42898554/Validation_42898554_42898554_752206.170.png", "Validation_42898554/Validation_42898554_42898554_752208.153.png", "Validation_42898554/Validation_42898554_42898554_752210.152.png", "Validation_42898554/Validation_42898554_42898554_752212.151.png", "Validation_42898554/Validation_42898554_42898554_752214.150.png" ], "turn_id": 1, "scan_id": "Validation_42898554", "origin_question": "Remember, have you seen any cabinet(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any cabinet(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_42898554/Validation_42898554_42898554_752216.149.png", "Validation_42898554/Validation_42898554_42898554_752218.148.png", "Validation_42898554/Validation_42898554_42898554_752220.148.png", "Validation_42898554/Validation_42898554_42898554_752222.147.png", "Validation_42898554/Validation_42898554_42898554_752224.146.png" ], "turn_id": 2, "scan_id": "Validation_42898554", "origin_question": "When did you first discover \"the bright blue reusable bag made of durable material\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the bright blue reusable bag made of durable material\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_42898554/Validation_42898554_42898554_752226.145.png", "Validation_42898554/Validation_42898554_42898554_752228.144.png", "Validation_42898554/Validation_42898554_42898554_752230.144.png", "Validation_42898554/Validation_42898554_42898554_752232.143.png", "Validation_42898554/Validation_42898554_42898554_752234.142.png" ], "turn_id": 3, "scan_id": "Validation_42898554", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_42898554/Validation_42898554_42898554_752236.141.png", "Validation_42898554/Validation_42898554_42898554_752238.140.png", "Validation_42898554/Validation_42898554_42898554_752240.139.png", "Validation_42898554/Validation_42898554_42898554_752242.139.png", "Validation_42898554/Validation_42898554_42898554_752244.154.png" ], "turn_id": 4, "scan_id": "Validation_42898554", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "56.13", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_42898554/Validation_42898554_42898554_752246.154.png", "Validation_42898554/Validation_42898554_42898554_752248.153.png", "Validation_42898554/Validation_42898554_42898554_752250.152.png", "Validation_42898554/Validation_42898554_42898554_752252.151.png", "Validation_42898554/Validation_42898554_42898554_752254.150.png" ], "turn_id": 5, "scan_id": "Validation_42898554", "origin_question": "In which turn were you closest to \"the bright blue reusable bag made of durable material\"? ", "option": [], "answer": "5", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the bright blue reusable bag made of durable material\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_42898554/Validation_42898554_42898554_752256.149.png", "Validation_42898554/Validation_42898554_42898554_752258.149.png", "Validation_42898554/Validation_42898554_42898554_752260.148.png", "Validation_42898554/Validation_42898554_42898554_752262.147.png", "Validation_42898554/Validation_42898554_42898554_752264.146.png" ], "turn_id": 6, "scan_id": "Validation_42898554", "origin_question": "Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "79.87", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_42898554/Validation_42898554_42898554_752266.145.png", "Validation_42898554/Validation_42898554_42898554_752268.144.png", "Validation_42898554/Validation_42898554_42898554_752270.144.png", "Validation_42898554/Validation_42898554_42898554_752272.143.png", "Validation_42898554/Validation_42898554_42898554_752274.142.png" ], "turn_id": 7, "scan_id": "Validation_42898554", "origin_question": "", "option": [], "answer": "", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_42898554/Validation_42898554_42898554_752276.141.png", "Validation_42898554/Validation_42898554_42898554_752278.140.png", "Validation_42898554/Validation_42898554_42898554_752280.139.png", "Validation_42898554/Validation_42898554_42898554_752282.139.png", "Validation_42898554/Validation_42898554_42898554_752284.138.png" ], "turn_id": 8, "scan_id": "Validation_42898554", "origin_question": "At the end of which turn were both of \"the bright blue reusable bag made of durable material\" and \"the medium-sized red glossy box with handles\" on your right side?", "option": [], "answer": "8", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the bright blue reusable bag made of durable material\" and \"the medium-sized red glossy box with handles\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_42898554/Validation_42898554_42898554_752286.154.png", "Validation_42898554/Validation_42898554_42898554_752288.153.png", "Validation_42898554/Validation_42898554_42898554_752290.152.png", "Validation_42898554/Validation_42898554_42898554_752292.151.png", "Validation_42898554/Validation_42898554_42898554_752294.150.png" ], "turn_id": 9, "scan_id": "Validation_42898554", "origin_question": "Is the medium-sized red glossy box with handles to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the medium-sized red glossy box with handles to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_42898554/Validation_42898554_42898554_752296.149.png", "Validation_42898554/Validation_42898554_42898554_752298.149.png", "Validation_42898554/Validation_42898554_42898554_752300.148.png", "Validation_42898554/Validation_42898554_42898554_752302.147.png", "Validation_42898554/Validation_42898554_42898554_752304.146.png" ], "turn_id": 10, "scan_id": "Validation_42898554", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_45261181/Validation_45261181_45261181_2827.052.png", "Validation_45261181/Validation_45261181_45261181_2828.535.png", "Validation_45261181/Validation_45261181_45261181_2830.034.png", "Validation_45261181/Validation_45261181_45261181_2831.534.png" ], "turn_id": 1, "scan_id": "Validation_45261181", "origin_question": "Remember, have you seen any sink(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any sink(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_45261181/Validation_45261181_45261181_2833.033.png", "Validation_45261181/Validation_45261181_45261181_2834.532.png", "Validation_45261181/Validation_45261181_45261181_2836.032.png", "Validation_45261181/Validation_45261181_45261181_2837.531.png" ], "turn_id": 2, "scan_id": "Validation_45261181", "origin_question": "Which direction is the compact cabinet with an off-white finish and woven drawers in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the compact cabinet with an off-white finish and woven drawers in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_45261181/Validation_45261181_45261181_2839.031.png", "Validation_45261181/Validation_45261181_45261181_2840.530.png", "Validation_45261181/Validation_45261181_45261181_2842.030.png", "Validation_45261181/Validation_45261181_45261181_2843.529.png" ], "turn_id": 3, "scan_id": "Validation_45261181", "origin_question": "Is the compact cabinet with an off-white finish and woven drawers to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the compact cabinet with an off-white finish and woven drawers to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_45261181/Validation_45261181_45261181_2845.028.png", "Validation_45261181/Validation_45261181_45261181_2846.528.png", "Validation_45261181/Validation_45261181_45261181_2848.027.png", "Validation_45261181/Validation_45261181_45261181_2849.526.png" ], "turn_id": 4, "scan_id": "Validation_45261181", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the compact cabinet with an off-white finish and woven drawers\" from your position?", "option": [], "answer": "156.26", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the compact cabinet with an off-white finish and woven drawers\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_45261181/Validation_45261181_45261181_2851.026.png", "Validation_45261181/Validation_45261181_45261181_2852.525.png", "Validation_45261181/Validation_45261181_45261181_2854.025.png", "Validation_45261181/Validation_45261181_45261181_2855.524.png" ], "turn_id": 5, "scan_id": "Validation_45261181", "origin_question": "Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "115.02", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_45261181/Validation_45261181_45261181_2857.024.png", "Validation_45261181/Validation_45261181_45261181_2858.540.png", "Validation_45261181/Validation_45261181_45261181_2860.039.png", "Validation_45261181/Validation_45261181_45261181_2861.539.png" ], "turn_id": 6, "scan_id": "Validation_45261181", "origin_question": "Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the sleek, elongated oval-shaped toilet with the glossy white finish\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the sleek, elongated oval-shaped toilet with the glossy white finish\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_44358536/Validation_44358536_44358536_53517.093.png", "Validation_44358536/Validation_44358536_44358536_53518.576.png", "Validation_44358536/Validation_44358536_44358536_53520.075.png", "Validation_44358536/Validation_44358536_44358536_53521.575.png" ], "turn_id": 1, "scan_id": "Validation_44358536", "origin_question": "Remember, have you seen any bathtub(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any bathtub(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_44358536/Validation_44358536_44358536_53523.074.png", "Validation_44358536/Validation_44358536_44358536_53524.573.png", "Validation_44358536/Validation_44358536_44358536_53526.073.png", "Validation_44358536/Validation_44358536_44358536_53527.572.png" ], "turn_id": 2, "scan_id": "Validation_44358536", "origin_question": "In which turn were you farthest from \"the sleek white pedestal sink with the curved basin\"? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the sleek white pedestal sink with the curved basin\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_44358536/Validation_44358536_44358536_53529.071.png", "Validation_44358536/Validation_44358536_44358536_53530.571.png", "Validation_44358536/Validation_44358536_44358536_53532.070.png", "Validation_44358536/Validation_44358536_44358536_53533.570.png" ], "turn_id": 3, "scan_id": "Validation_44358536", "origin_question": "In which turn were you closest to \"the rectangular pink bag made of lightweight fabric\"? ", "option": [], "answer": "2", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the rectangular pink bag made of lightweight fabric\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_44358536/Validation_44358536_44358536_53535.069.png", "Validation_44358536/Validation_44358536_44358536_53536.568.png", "Validation_44358536/Validation_44358536_44358536_53538.068.png", "Validation_44358536/Validation_44358536_44358536_53539.567.png" ], "turn_id": 4, "scan_id": "Validation_44358536", "origin_question": "Is the compact, white cylindrical bin to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the compact, white cylindrical bin to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47331314/Validation_47331314_47331314_177638.283.png", "Validation_47331314/Validation_47331314_47331314_177640.266.png", "Validation_47331314/Validation_47331314_47331314_177642.265.png", "Validation_47331314/Validation_47331314_47331314_177644.264.png" ], "turn_id": 1, "scan_id": "Validation_47331314", "origin_question": "Remember, have you seen any dvd(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any dvd(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47331314/Validation_47331314_47331314_177646.280.png", "Validation_47331314/Validation_47331314_47331314_177648.279.png", "Validation_47331314/Validation_47331314_47331314_177650.278.png", "Validation_47331314/Validation_47331314_47331314_177652.278.png" ], "turn_id": 2, "scan_id": "Validation_47331314", "origin_question": "Remember, have you seen any radiator(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any radiator(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47331314/Validation_47331314_47331314_177654.277.png", "Validation_47331314/Validation_47331314_47331314_177656.276.png", "Validation_47331314/Validation_47331314_47331314_177658.275.png", "Validation_47331314/Validation_47331314_47331314_177660.274.png" ], "turn_id": 3, "scan_id": "Validation_47331314", "origin_question": "", "option": [], "answer": "", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_47331314/Validation_47331314_47331314_177662.274.png", "Validation_47331314/Validation_47331314_47331314_177664.273.png", "Validation_47331314/Validation_47331314_47331314_177666.272.png", "Validation_47331314/Validation_47331314_47331314_177668.271.png" ], "turn_id": 4, "scan_id": "Validation_47331314", "origin_question": "What will be the first-time appearance order of the following categories: door, monitor, picture?", "option": [ "monitor, door, picture", "picture, door, monitor", "door, monitor, picture", "picture, monitor, door" ], "answer": "picture, door, monitor", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: door, monitor, picture? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['monitor, door, picture', 'picture, door, monitor', 'door, monitor, picture', 'picture, monitor, door'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47331314/Validation_47331314_47331314_177670.270.png", "Validation_47331314/Validation_47331314_47331314_177672.270.png", "Validation_47331314/Validation_47331314_47331314_177674.269.png", "Validation_47331314/Validation_47331314_47331314_177676.268.png" ], "turn_id": 5, "scan_id": "Validation_47331314", "origin_question": "Which direction is the large, sleek monitor with the black plastic frame in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the large, sleek monitor with the black plastic frame in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47331314/Validation_47331314_47331314_177678.267.png", "Validation_47331314/Validation_47331314_47331314_177680.266.png", "Validation_47331314/Validation_47331314_47331314_177682.266.png", "Validation_47331314/Validation_47331314_47331314_177684.265.png" ], "turn_id": 6, "scan_id": "Validation_47331314", "origin_question": "In which turn were you closest to \"the rectangular brown hardwood door with the polished finish\"? ", "option": [], "answer": "1", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the rectangular brown hardwood door with the polished finish\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42898822/Validation_42898822_42898822_168205.592.png", "Validation_42898822/Validation_42898822_42898822_168207.574.png", "Validation_42898822/Validation_42898822_42898822_168209.573.png", "Validation_42898822/Validation_42898822_42898822_168211.573.png" ], "turn_id": 1, "scan_id": "Validation_42898822", "origin_question": "Remember, have you seen any bathtub(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any bathtub(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_42898822/Validation_42898822_42898822_168213.572.png", "Validation_42898822/Validation_42898822_42898822_168215.571.png", "Validation_42898822/Validation_42898822_42898822_168217.587.png", "Validation_42898822/Validation_42898822_42898822_168219.586.png" ], "turn_id": 2, "scan_id": "Validation_42898822", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the compact wooden cabinet with the warm brown finish\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the compact wooden cabinet with the warm brown finish\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_42898822/Validation_42898822_42898822_168221.585.png", "Validation_42898822/Validation_42898822_42898822_168223.584.png", "Validation_42898822/Validation_42898822_42898822_168225.584.png", "Validation_42898822/Validation_42898822_42898822_168227.583.png" ], "turn_id": 3, "scan_id": "Validation_42898822", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_42898822/Validation_42898822_42898822_168229.582.png", "Validation_42898822/Validation_42898822_42898822_168231.581.png", "Validation_42898822/Validation_42898822_42898822_168233.580.png", "Validation_42898822/Validation_42898822_42898822_168235.579.png" ], "turn_id": 4, "scan_id": "Validation_42898822", "origin_question": "Please recall \"the compact wooden cabinet with the warm brown finish\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.54", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the compact wooden cabinet with the warm brown finish\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "Validation_42898822/Validation_42898822_42898822_168237.579.png", "Validation_42898822/Validation_42898822_42898822_168239.578.png", "Validation_42898822/Validation_42898822_42898822_168241.577.png", "Validation_42898822/Validation_42898822_42898822_168243.576.png" ], "turn_id": 5, "scan_id": "Validation_42898822", "origin_question": "Among these three objects(\"the smooth, white acrylic bathtub with an oval shape\",\"the vibrant red absorbent cotton towel\",\"the lush green plant with dense, oval leaves\" ), which one is the closest to you?", "option": [ "the smooth, white acrylic bathtub with an oval shape", "the vibrant red absorbent cotton towel", "the lush green plant with dense, oval leaves" ], "answer": "the smooth, white acrylic bathtub with an oval shape", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects(\"the smooth, white acrylic bathtub with an oval shape\",\"the vibrant red absorbent cotton towel\",\"the lush green plant with dense, oval leaves\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the smooth, white acrylic bathtub with an oval shape', 'the vibrant red absorbent cotton towel', 'the lush green plant with dense, oval leaves'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_42898822/Validation_42898822_42898822_168245.575.png", "Validation_42898822/Validation_42898822_42898822_168247.574.png", "Validation_42898822/Validation_42898822_42898822_168249.574.png", "Validation_42898822/Validation_42898822_42898822_168251.573.png" ], "turn_id": 6, "scan_id": "Validation_42898822", "origin_question": "In which turn were you closest to \"the lush green plant with dense, oval leaves\"? ", "option": [], "answer": "4", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the lush green plant with dense, oval leaves\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_42898822/Validation_42898822_42898822_168253.572.png", "Validation_42898822/Validation_42898822_42898822_168255.571.png", "Validation_42898822/Validation_42898822_42898822_168257.587.png", "Validation_42898822/Validation_42898822_42898822_168259.586.png" ], "turn_id": 7, "scan_id": "Validation_42898822", "origin_question": "In which turn were you closest to \"an off-white rounded porcelain toilet\"? ", "option": [], "answer": "7", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"an off-white rounded porcelain toilet\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_42898822/Validation_42898822_42898822_168261.585.png", "Validation_42898822/Validation_42898822_42898822_168263.584.png", "Validation_42898822/Validation_42898822_42898822_168265.584.png", "Validation_42898822/Validation_42898822_42898822_168267.583.png" ], "turn_id": 8, "scan_id": "Validation_42898822", "origin_question": "Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "108.13", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_45261587/Validation_45261587_45261587_78100.615.png", "Validation_45261587/Validation_45261587_45261587_78102.597.png", "Validation_45261587/Validation_45261587_45261587_78104.596.png", "Validation_45261587/Validation_45261587_45261587_78106.595.png" ], "turn_id": 1, "scan_id": "Validation_45261587", "origin_question": "Remember, have you seen any sink(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any sink(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_45261587/Validation_45261587_45261587_78108.595.png", "Validation_45261587/Validation_45261587_45261587_78110.594.png", "Validation_45261587/Validation_45261587_45261587_78112.593.png", "Validation_45261587/Validation_45261587_45261587_78114.609.png" ], "turn_id": 2, "scan_id": "Validation_45261587", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the rectangular window with the sleek white frame\" from your position?", "option": [], "answer": "71.87", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the rectangular window with the sleek white frame\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_45261587/Validation_45261587_45261587_78116.608.png", "Validation_45261587/Validation_45261587_45261587_78118.607.png", "Validation_45261587/Validation_45261587_45261587_78120.606.png", "Validation_45261587/Validation_45261587_45261587_78122.606.png" ], "turn_id": 3, "scan_id": "Validation_45261587", "origin_question": "In which turn did you see both \"the sleek modern white rectangular bathtub\" and \"the door\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the sleek modern white rectangular bathtub\" and \"the door\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_45261587/Validation_45261587_45261587_78124.605.png", "Validation_45261587/Validation_45261587_45261587_78126.604.png", "Validation_45261587/Validation_45261587_45261587_78128.603.png", "Validation_45261587/Validation_45261587_45261587_78130.602.png" ], "turn_id": 4, "scan_id": "Validation_45261587", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_45261587/Validation_45261587_45261587_78132.601.png", "Validation_45261587/Validation_45261587_45261587_78134.601.png", "Validation_45261587/Validation_45261587_45261587_78136.600.png", "Validation_45261587/Validation_45261587_45261587_78138.599.png" ], "turn_id": 5, "scan_id": "Validation_45261587", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular ceramic sink with rounded edges\", Y: \"the door\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to X and farther away from Y", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular ceramic sink with rounded edges\", Y: \"the door\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_45261587/Validation_45261587_45261587_78140.598.png", "Validation_45261587/Validation_45261587_45261587_78142.597.png", "Validation_45261587/Validation_45261587_45261587_78144.596.png", "Validation_45261587/Validation_45261587_45261587_78146.596.png" ], "turn_id": 6, "scan_id": "Validation_45261587", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the rectangular ceramic sink with rounded edges\" from your position?", "option": [], "answer": "86.24", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the rectangular ceramic sink with rounded edges\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_45261587/Validation_45261587_45261587_78148.595.png", "Validation_45261587/Validation_45261587_45261587_78150.594.png", "Validation_45261587/Validation_45261587_45261587_78152.593.png", "Validation_45261587/Validation_45261587_45261587_78154.609.png" ], "turn_id": 7, "scan_id": "Validation_45261587", "origin_question": "", "option": [], "answer": "", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_45261587/Validation_45261587_45261587_78156.608.png", "Validation_45261587/Validation_45261587_45261587_78158.607.png", "Validation_45261587/Validation_45261587_45261587_78160.606.png", "Validation_45261587/Validation_45261587_45261587_78162.606.png" ], "turn_id": 8, "scan_id": "Validation_45261587", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the door\" from your position?", "option": [], "answer": "179.0", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the door\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47429914/Validation_47429914_47429914_135998.597.png", "Validation_47429914/Validation_47429914_47429914_136000.079.png", "Validation_47429914/Validation_47429914_47429914_136001.595.png", "Validation_47429914/Validation_47429914_47429914_136003.095.png" ], "turn_id": 1, "scan_id": "Validation_47429914", "origin_question": "Remember, have you seen any door(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any door(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 3 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47429914/Validation_47429914_47429914_136004.594.png", "Validation_47429914/Validation_47429914_47429914_136006.093.png", "Validation_47429914/Validation_47429914_47429914_136007.593.png", "Validation_47429914/Validation_47429914_47429914_136009.092.png" ], "turn_id": 2, "scan_id": "Validation_47429914", "origin_question": "Remember, have you seen any cabinet(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any cabinet(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 3 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_47429914/Validation_47429914_47429914_136010.592.png", "Validation_47429914/Validation_47429914_47429914_136012.091.png", "Validation_47429914/Validation_47429914_47429914_136013.590.png", "Validation_47429914/Validation_47429914_47429914_136015.090.png" ], "turn_id": 3, "scan_id": "Validation_47429914", "origin_question": "When did you first discover \"the bathtub\"? (index of the turn)", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the bathtub\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 3 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "Validation_48458465/Validation_48458465_48458465_110483.371.png", "Validation_48458465/Validation_48458465_48458465_110484.854.png", "Validation_48458465/Validation_48458465_48458465_110486.353.png", "Validation_48458465/Validation_48458465_48458465_110487.852.png" ], "turn_id": 1, "scan_id": "Validation_48458465", "origin_question": "Remember, how many couch(s) have you seen so far? ", "option": [], "answer": "2", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many couch(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_48458465/Validation_48458465_48458465_110489.352.png", "Validation_48458465/Validation_48458465_48458465_110490.851.png", "Validation_48458465/Validation_48458465_48458465_110492.351.png", "Validation_48458465/Validation_48458465_48458465_110493.850.png" ], "turn_id": 2, "scan_id": "Validation_48458465", "origin_question": "Which two objects are on the same side of you now? \"the contemporary wall-mounted lamp with the frosted glass shade\", \"the round white ceramic plate with the minimalist design\", \"the smooth medium brown wooden door\". ", "option": [ "the contemporary wall-mounted lamp with the frosted glass shade, the round white ceramic plate with the minimalist design", "the contemporary wall-mounted lamp with the frosted glass shade, the smooth medium brown wooden door", "the round white ceramic plate with the minimalist design, the smooth medium brown wooden door" ], "answer": "the round white ceramic plate with the minimalist design, the smooth medium brown wooden door", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the contemporary wall-mounted lamp with the frosted glass shade\", \"the round white ceramic plate with the minimalist design\", \"the smooth medium brown wooden door\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the contemporary wall-mounted lamp with the frosted glass shade, the round white ceramic plate with the minimalist design', 'the contemporary wall-mounted lamp with the frosted glass shade, the smooth medium brown wooden door', 'the round white ceramic plate with the minimalist design, the smooth medium brown wooden door'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_48458465/Validation_48458465_48458465_110495.349.png", "Validation_48458465/Validation_48458465_48458465_110496.849.png", "Validation_48458465/Validation_48458465_48458465_110498.348.png", "Validation_48458465/Validation_48458465_48458465_110499.848.png" ], "turn_id": 3, "scan_id": "Validation_48458465", "origin_question": "Which two objects are on the same side of you now? \"the table\", \"varied-shaped colorful books with diverse covers\", \"the contemporary wall-mounted lamp with the frosted glass shade\". ", "option": [ "the table, varied-shaped colorful books with diverse covers", "the table, the contemporary wall-mounted lamp with the frosted glass shade", "varied-shaped colorful books with diverse covers, the contemporary wall-mounted lamp with the frosted glass shade" ], "answer": "the table, varied-shaped colorful books with diverse covers", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the table\", \"varied-shaped colorful books with diverse covers\", \"the contemporary wall-mounted lamp with the frosted glass shade\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the table, varied-shaped colorful books with diverse covers', 'the table, the contemporary wall-mounted lamp with the frosted glass shade', 'varied-shaped colorful books with diverse covers, the contemporary wall-mounted lamp with the frosted glass shade'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_48458465/Validation_48458465_48458465_110501.347.png", "Validation_48458465/Validation_48458465_48458465_110502.846.png", "Validation_48458465/Validation_48458465_48458465_110504.346.png", "Validation_48458465/Validation_48458465_48458465_110505.845.png" ], "turn_id": 4, "scan_id": "Validation_48458465", "origin_question": "Please recall \"the large rectangular window with the white frame\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.97", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the large rectangular window with the white frame\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_48458465/Validation_48458465_48458465_110507.345.png", "Validation_48458465/Validation_48458465_48458465_110508.844.png", "Validation_48458465/Validation_48458465_48458465_110510.344.png", "Validation_48458465/Validation_48458465_48458465_110511.843.png" ], "turn_id": 5, "scan_id": "Validation_48458465", "origin_question": "Which direction is the glossy white ceramic teapot with the bulbous shape in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the glossy white ceramic teapot with the bulbous shape in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47334107/Validation_47334107_47334107_93454.803.png", "Validation_47334107/Validation_47334107_47334107_93456.802.png", "Validation_47334107/Validation_47334107_47334107_93458.802.png", "Validation_47334107/Validation_47334107_47334107_93460.801.png" ], "turn_id": 1, "scan_id": "Validation_47334107", "origin_question": "", "option": [], "answer": "", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47334107/Validation_47334107_47334107_93462.800.png", "Validation_47334107/Validation_47334107_47334107_93464.799.png", "Validation_47334107/Validation_47334107_47334107_93466.798.png", "Validation_47334107/Validation_47334107_47334107_93468.798.png" ], "turn_id": 2, "scan_id": "Validation_47334107", "origin_question": "Remember, have you seen any soap(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any soap(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_47334107/Validation_47334107_47334107_93470.797.png", "Validation_47334107/Validation_47334107_47334107_93472.796.png", "Validation_47334107/Validation_47334107_47334107_93474.795.png", "Validation_47334107/Validation_47334107_47334107_93476.794.png" ], "turn_id": 3, "scan_id": "Validation_47334107", "origin_question": "At the end of which turn were both of \"the sleek, rounded white ceramic toilet\" and \"the medium-sized plastic bucket with the smooth surface\" on your left side?", "option": [], "answer": "2", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the sleek, rounded white ceramic toilet\" and \"the medium-sized plastic bucket with the smooth surface\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47334107/Validation_47334107_47334107_93478.793.png", "Validation_47334107/Validation_47334107_47334107_93480.793.png", "Validation_47334107/Validation_47334107_47334107_93482.792.png", "Validation_47334107/Validation_47334107_47334107_93484.791.png" ], "turn_id": 4, "scan_id": "Validation_47334107", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the rounded glossy white ceramic flowerpot\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the rounded glossy white ceramic flowerpot\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_47334107/Validation_47334107_47334107_93486.790.png", "Validation_47334107/Validation_47334107_47334107_93488.789.png", "Validation_47334107/Validation_47334107_47334107_93490.789.png", "Validation_47334107/Validation_47334107_47334107_93492.788.png" ], "turn_id": 5, "scan_id": "Validation_47334107", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the rounded glossy white ceramic flowerpot\" from your position?", "option": [], "answer": "109.45", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the rounded glossy white ceramic flowerpot\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47334107/Validation_47334107_47334107_93494.787.png", "Validation_47334107/Validation_47334107_47334107_93496.803.png", "Validation_47334107/Validation_47334107_47334107_93498.802.png", "Validation_47334107/Validation_47334107_47334107_93500.801.png" ], "turn_id": 6, "scan_id": "Validation_47334107", "origin_question": "In which turn were you closest to \"the medium-sized plastic bucket with the smooth surface\"? ", "option": [], "answer": "4", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the medium-sized plastic bucket with the smooth surface\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_47334107/Validation_47334107_47334107_93502.800.png", "Validation_47334107/Validation_47334107_47334107_93504.800.png", "Validation_47334107/Validation_47334107_47334107_93506.799.png", "Validation_47334107/Validation_47334107_47334107_93508.798.png" ], "turn_id": 7, "scan_id": "Validation_47334107", "origin_question": "When did you first discover \"the white rectangular front-loading washing machine\"? (index of the turn)", "option": [], "answer": "3", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the white rectangular front-loading washing machine\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_47334107/Validation_47334107_47334107_93510.797.png", "Validation_47334107/Validation_47334107_47334107_93512.796.png", "Validation_47334107/Validation_47334107_47334107_93514.796.png", "Validation_47334107/Validation_47334107_47334107_93516.795.png" ], "turn_id": 8, "scan_id": "Validation_47334107", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the white rectangular front-loading washing machine\" from your position?", "option": [], "answer": "103.37", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the white rectangular front-loading washing machine\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42446533/Validation_42446533_42446533_200395.314.png", "Validation_42446533/Validation_42446533_42446533_200396.797.png", "Validation_42446533/Validation_42446533_42446533_200398.296.png", "Validation_42446533/Validation_42446533_42446533_200399.796.png" ], "turn_id": 1, "scan_id": "Validation_42446533", "origin_question": "Remember, have you seen any bin(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any bin(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_42446533/Validation_42446533_42446533_200401.295.png", "Validation_42446533/Validation_42446533_42446533_200402.794.png", "Validation_42446533/Validation_42446533_42446533_200404.294.png", "Validation_42446533/Validation_42446533_42446533_200405.793.png" ], "turn_id": 2, "scan_id": "Validation_42446533", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the sleek, modern light-colored shelf\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the sleek, modern light-colored shelf\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_42446533/Validation_42446533_42446533_200407.292.png", "Validation_42446533/Validation_42446533_42446533_200408.792.png", "Validation_42446533/Validation_42446533_42446533_200410.291.png", "Validation_42446533/Validation_42446533_42446533_200411.791.png" ], "turn_id": 3, "scan_id": "Validation_42446533", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the sleek, round glossy white sink\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the sleek, round glossy white sink\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_42446533/Validation_42446533_42446533_200413.307.png", "Validation_42446533/Validation_42446533_42446533_200414.806.png", "Validation_42446533/Validation_42446533_42446533_200416.305.png", "Validation_42446533/Validation_42446533_42446533_200417.805.png" ], "turn_id": 4, "scan_id": "Validation_42446533", "origin_question": "Which two objects are on the same side of you now? \"the sleek, round glossy white sink\", \"the sleek, modern light-colored shelf\", \"the sleek white rectangular door with the metal handle\". ", "option": [ "the sleek, round glossy white sink, the sleek, modern light-colored shelf", "the sleek, round glossy white sink, the sleek white rectangular door with the metal handle", "the sleek, modern light-colored shelf, the sleek white rectangular door with the metal handle" ], "answer": "the sleek, modern light-colored shelf, the sleek white rectangular door with the metal handle", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the sleek, round glossy white sink\", \"the sleek, modern light-colored shelf\", \"the sleek white rectangular door with the metal handle\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the sleek, round glossy white sink, the sleek, modern light-colored shelf', 'the sleek, round glossy white sink, the sleek white rectangular door with the metal handle', 'the sleek, modern light-colored shelf, the sleek white rectangular door with the metal handle'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_42446533/Validation_42446533_42446533_200419.304.png", "Validation_42446533/Validation_42446533_42446533_200420.804.png", "Validation_42446533/Validation_42446533_42446533_200422.303.png", "Validation_42446533/Validation_42446533_42446533_200423.802.png" ], "turn_id": 5, "scan_id": "Validation_42446533", "origin_question": "", "option": [], "answer": "", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_42446533/Validation_42446533_42446533_200425.302.png", "Validation_42446533/Validation_42446533_42446533_200426.801.png", "Validation_42446533/Validation_42446533_42446533_200428.300.png", "Validation_42446533/Validation_42446533_42446533_200429.800.png" ], "turn_id": 6, "scan_id": "Validation_42446533", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_48458427/Validation_48458427_48458427_116505.369.png", "Validation_48458427/Validation_48458427_48458427_116507.351.png", "Validation_48458427/Validation_48458427_48458427_116509.350.png", "Validation_48458427/Validation_48458427_48458427_116511.349.png" ], "turn_id": 1, "scan_id": "Validation_48458427", "origin_question": "Remember, have you seen any sink(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any sink(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_48458427/Validation_48458427_48458427_116513.349.png", "Validation_48458427/Validation_48458427_48458427_116515.348.png", "Validation_48458427/Validation_48458427_48458427_116517.347.png", "Validation_48458427/Validation_48458427_48458427_116519.346.png" ], "turn_id": 2, "scan_id": "Validation_48458427", "origin_question": "Which direction is the sleek, cylindrical gray flowerpot made of ceramic in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the sleek, cylindrical gray flowerpot made of ceramic in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_48458427/Validation_48458427_48458427_116521.345.png", "Validation_48458427/Validation_48458427_48458427_116523.345.png", "Validation_48458427/Validation_48458427_48458427_116525.344.png", "Validation_48458427/Validation_48458427_48458427_116527.343.png" ], "turn_id": 3, "scan_id": "Validation_48458427", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the compact rectangular bin in dark grey or black\" from your position?", "option": [], "answer": "99.87", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the compact rectangular bin in dark grey or black\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_48458427/Validation_48458427_48458427_116529.342.png", "Validation_48458427/Validation_48458427_48458427_116531.341.png", "Validation_48458427/Validation_48458427_48458427_116533.340.png", "Validation_48458427/Validation_48458427_48458427_116535.340.png" ], "turn_id": 4, "scan_id": "Validation_48458427", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the small dome-shaped olive green decoration\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the small dome-shaped olive green decoration\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_48458427/Validation_48458427_48458427_116537.339.png", "Validation_48458427/Validation_48458427_48458427_116539.338.png", "Validation_48458427/Validation_48458427_48458427_116541.337.png", "Validation_48458427/Validation_48458427_48458427_116543.336.png" ], "turn_id": 5, "scan_id": "Validation_48458427", "origin_question": "In which turn did you see both \"the small dome-shaped olive green decoration\" and \"the compact rectangular bin in dark grey or black\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the small dome-shaped olive green decoration\" and \"the compact rectangular bin in dark grey or black\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_48458427/Validation_48458427_48458427_116545.352.png", "Validation_48458427/Validation_48458427_48458427_116547.351.png", "Validation_48458427/Validation_48458427_48458427_116549.350.png", "Validation_48458427/Validation_48458427_48458427_116551.349.png" ], "turn_id": 6, "scan_id": "Validation_48458427", "origin_question": "Which two objects are on the same side of you now? \"the bathtub\", \"the sleek, modern white ceramic toilet\", \"the small dome-shaped olive green decoration\". ", "option": [ "the bathtub, the sleek, modern white ceramic toilet", "the bathtub, the small dome-shaped olive green decoration", "the sleek, modern white ceramic toilet, the small dome-shaped olive green decoration" ], "answer": "the sleek, modern white ceramic toilet, the small dome-shaped olive green decoration", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the bathtub\", \"the sleek, modern white ceramic toilet\", \"the small dome-shaped olive green decoration\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the bathtub, the sleek, modern white ceramic toilet', 'the bathtub, the small dome-shaped olive green decoration', 'the sleek, modern white ceramic toilet, the small dome-shaped olive green decoration'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47115550/Validation_47115550_47115550_15773.938.png", "Validation_47115550/Validation_47115550_47115550_15775.921.png", "Validation_47115550/Validation_47115550_47115550_15777.920.png", "Validation_47115550/Validation_47115550_47115550_15779.919.png", "Validation_47115550/Validation_47115550_47115550_15781.918.png" ], "turn_id": 1, "scan_id": "Validation_47115550", "origin_question": "Remember, have you seen any door(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any door(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47115550/Validation_47115550_47115550_15783.918.png", "Validation_47115550/Validation_47115550_47115550_15785.917.png", "Validation_47115550/Validation_47115550_47115550_15787.916.png", "Validation_47115550/Validation_47115550_47115550_15789.915.png", "Validation_47115550/Validation_47115550_47115550_15791.914.png" ], "turn_id": 2, "scan_id": "Validation_47115550", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the sturdy rectangular wooden door with frosted glass insets\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the sturdy rectangular wooden door with frosted glass insets\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_47115550/Validation_47115550_47115550_15793.914.png", "Validation_47115550/Validation_47115550_47115550_15795.913.png", "Validation_47115550/Validation_47115550_47115550_15797.912.png", "Validation_47115550/Validation_47115550_47115550_15799.911.png", "Validation_47115550/Validation_47115550_47115550_15801.910.png" ], "turn_id": 3, "scan_id": "Validation_47115550", "origin_question": "What will be the first-time appearance order of the following categories: mat, monitor, window?", "option": [ "mat, monitor, window", "window, monitor, mat", "window, mat, monitor", "monitor, mat, window" ], "answer": "mat, monitor, window", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: mat, monitor, window? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['mat, monitor, window', 'window, monitor, mat', 'window, mat, monitor', 'monitor, mat, window'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47115550/Validation_47115550_47115550_15803.926.png", "Validation_47115550/Validation_47115550_47115550_15805.926.png", "Validation_47115550/Validation_47115550_47115550_15807.925.png", "Validation_47115550/Validation_47115550_47115550_15809.924.png", "Validation_47115550/Validation_47115550_47115550_15811.923.png" ], "turn_id": 4, "scan_id": "Validation_47115550", "origin_question": "", "option": [], "answer": "", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47115550/Validation_47115550_47115550_15813.922.png", "Validation_47115550/Validation_47115550_47115550_15815.922.png", "Validation_47115550/Validation_47115550_47115550_15817.921.png", "Validation_47115550/Validation_47115550_47115550_15819.920.png", "Validation_47115550/Validation_47115550_47115550_15821.919.png" ], "turn_id": 5, "scan_id": "Validation_47115550", "origin_question": "In which turn were you closest to \"the large, light brown wooden box with grain patterns\"? ", "option": [], "answer": "5", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the large, light brown wooden box with grain patterns\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_47115550/Validation_47115550_47115550_15823.918.png", "Validation_47115550/Validation_47115550_47115550_15825.918.png", "Validation_47115550/Validation_47115550_47115550_15827.917.png", "Validation_47115550/Validation_47115550_47115550_15829.916.png", "Validation_47115550/Validation_47115550_47115550_15831.915.png" ], "turn_id": 6, "scan_id": "Validation_47115550", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the large rectangular window with the white wooden frame\", Y: \"the large, light brown wooden box with grain patterns\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to X and farther away from Y", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the large rectangular window with the white wooden frame\", Y: \"the large, light brown wooden box with grain patterns\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_47115550/Validation_47115550_47115550_15833.915.png", "Validation_47115550/Validation_47115550_47115550_15835.914.png", "Validation_47115550/Validation_47115550_47115550_15837.913.png", "Validation_47115550/Validation_47115550_47115550_15839.912.png", "Validation_47115550/Validation_47115550_47115550_15841.911.png" ], "turn_id": 7, "scan_id": "Validation_47115550", "origin_question": "In which turn did you see both \"the large rectangular window with the white wooden frame\" and \"the large, light brown wooden box with grain patterns\" simultaneously? ", "option": [], "answer": "4", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the large rectangular window with the white wooden frame\" and \"the large, light brown wooden box with grain patterns\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47115550/Validation_47115550_47115550_15843.910.png", "Validation_47115550/Validation_47115550_47115550_15845.926.png", "Validation_47115550/Validation_47115550_47115550_15847.926.png", "Validation_47115550/Validation_47115550_47115550_15849.925.png", "Validation_47115550/Validation_47115550_47115550_15851.924.png" ], "turn_id": 8, "scan_id": "Validation_47115550", "origin_question": "Which direction is the large rectangular window with the white wooden frame in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the large rectangular window with the white wooden frame in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47115550/Validation_47115550_47115550_15853.923.png", "Validation_47115550/Validation_47115550_47115550_15855.922.png", "Validation_47115550/Validation_47115550_47115550_15857.922.png", "Validation_47115550/Validation_47115550_47115550_15859.921.png", "Validation_47115550/Validation_47115550_47115550_15861.920.png" ], "turn_id": 9, "scan_id": "Validation_47115550", "origin_question": "Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the large, light brown wooden box with grain patterns\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the large, light brown wooden box with grain patterns\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47115550/Validation_47115550_47115550_15863.919.png", "Validation_47115550/Validation_47115550_47115550_15865.918.png", "Validation_47115550/Validation_47115550_47115550_15867.918.png", "Validation_47115550/Validation_47115550_47115550_15869.917.png", "Validation_47115550/Validation_47115550_47115550_15871.916.png" ], "turn_id": 10, "scan_id": "Validation_47115550", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the large rectangular window with the white wooden frame\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the large rectangular window with the white wooden frame\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47430479/Validation_47430479_47430479_310685.476.png", "Validation_47430479/Validation_47430479_47430479_310687.458.png", "Validation_47430479/Validation_47430479_47430479_310689.457.png", "Validation_47430479/Validation_47430479_47430479_310691.473.png" ], "turn_id": 1, "scan_id": "Validation_47430479", "origin_question": "Remember, have you seen any rack(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any rack(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47430479/Validation_47430479_47430479_310693.472.png", "Validation_47430479/Validation_47430479_47430479_310695.471.png", "Validation_47430479/Validation_47430479_47430479_310697.471.png", "Validation_47430479/Validation_47430479_47430479_310699.470.png" ], "turn_id": 2, "scan_id": "Validation_47430479", "origin_question": "In which turn were you farthest from \"the luxurious deep purple bed with quilted bedding\"? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the luxurious deep purple bed with quilted bedding\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_47430479/Validation_47430479_47430479_310701.469.png", "Validation_47430479/Validation_47430479_47430479_310703.468.png", "Validation_47430479/Validation_47430479_47430479_310705.467.png", "Validation_47430479/Validation_47430479_47430479_310707.466.png" ], "turn_id": 3, "scan_id": "Validation_47430479", "origin_question": "In which turn did you see both \"the modern rectangular window with the brown wooden frame\" and \"the rectangular purple pillow made of soft fabric\" simultaneously? ", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the modern rectangular window with the brown wooden frame\" and \"the rectangular purple pillow made of soft fabric\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_47430479/Validation_47430479_47430479_310709.466.png", "Validation_47430479/Validation_47430479_47430479_310711.465.png", "Validation_47430479/Validation_47430479_47430479_310713.464.png", "Validation_47430479/Validation_47430479_47430479_310715.463.png" ], "turn_id": 4, "scan_id": "Validation_47430479", "origin_question": "In which turn did you see both \"the neutral light brown rectangular wooden stool\" and \"the collection of colorful cotton and polyester clothes\" simultaneously? ", "option": [], "answer": "4", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the neutral light brown rectangular wooden stool\" and \"the collection of colorful cotton and polyester clothes\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_47430479/Validation_47430479_47430479_310717.463.png", "Validation_47430479/Validation_47430479_47430479_310719.462.png", "Validation_47430479/Validation_47430479_47430479_310721.461.png", "Validation_47430479/Validation_47430479_47430479_310723.460.png" ], "turn_id": 5, "scan_id": "Validation_47430479", "origin_question": "When did you first discover \"the modern rectangular window with the brown wooden frame\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the modern rectangular window with the brown wooden frame\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "Validation_47430479/Validation_47430479_47430479_310725.459.png", "Validation_47430479/Validation_47430479_47430479_310727.458.png", "Validation_47430479/Validation_47430479_47430479_310729.458.png", "Validation_47430479/Validation_47430479_47430479_310731.457.png" ], "turn_id": 6, "scan_id": "Validation_47430479", "origin_question": "Among these three objects(\"the modern rectangular window with the brown wooden frame\",\"the rectangular purple pillow made of soft fabric\",\"the collection of colorful cotton and polyester clothes\" ), which one is the farthest from you?", "option": [ "the modern rectangular window with the brown wooden frame", "the rectangular purple pillow made of soft fabric", "the collection of colorful cotton and polyester clothes" ], "answer": "the collection of colorful cotton and polyester clothes", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects(\"the modern rectangular window with the brown wooden frame\",\"the rectangular purple pillow made of soft fabric\",\"the collection of colorful cotton and polyester clothes\" ), which one is the farthest from you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the modern rectangular window with the brown wooden frame', 'the rectangular purple pillow made of soft fabric', 'the collection of colorful cotton and polyester clothes'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47430479/Validation_47430479_47430479_310733.473.png", "Validation_47430479/Validation_47430479_47430479_310735.472.png", "Validation_47430479/Validation_47430479_47430479_310737.471.png", "Validation_47430479/Validation_47430479_47430479_310739.470.png" ], "turn_id": 7, "scan_id": "Validation_47430479", "origin_question": "Which direction is the neutral light brown rectangular wooden stool in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the neutral light brown rectangular wooden stool in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47331653/Validation_47331653_47331653_22985.347.png", "Validation_47331653/Validation_47331653_47331653_22987.329.png", "Validation_47331653/Validation_47331653_47331653_22989.329.png", "Validation_47331653/Validation_47331653_47331653_22991.328.png", "Validation_47331653/Validation_47331653_47331653_22993.327.png" ], "turn_id": 1, "scan_id": "Validation_47331653", "origin_question": "Remember, have you seen any cabinet(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any cabinet(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "Validation_47331653/Validation_47331653_47331653_22995.326.png", "Validation_47331653/Validation_47331653_47331653_22997.325.png", "Validation_47331653/Validation_47331653_47331653_22999.324.png", "Validation_47331653/Validation_47331653_47331653_23001.324.png", "Validation_47331653/Validation_47331653_47331653_23003.323.png" ], "turn_id": 2, "scan_id": "Validation_47331653", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the round white ceiling light fixture made of metal and glass\"; \"the sleek silver and black oven\"; \"the sleek white stove with the rectangular shape and four burners\" ", "option": [ "the round white ceiling light fixture made of metal and glass", "the sleek silver and black oven", "the sleek white stove with the rectangular shape and four burners" ], "answer": "the round white ceiling light fixture made of metal and glass", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the round white ceiling light fixture made of metal and glass\"; \"the sleek silver and black oven\"; \"the sleek white stove with the rectangular shape and four burners\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the round white ceiling light fixture made of metal and glass', 'the sleek silver and black oven', 'the sleek white stove with the rectangular shape and four burners'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47331653/Validation_47331653_47331653_23005.322.png", "Validation_47331653/Validation_47331653_47331653_23007.321.png", "Validation_47331653/Validation_47331653_47331653_23009.337.png", "Validation_47331653/Validation_47331653_47331653_23011.336.png", "Validation_47331653/Validation_47331653_47331653_23013.336.png" ], "turn_id": 3, "scan_id": "Validation_47331653", "origin_question": "In which turn were you closest to \"the sleek silver and black oven\"? ", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the sleek silver and black oven\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "Validation_47331653/Validation_47331653_47331653_23015.335.png", "Validation_47331653/Validation_47331653_47331653_23017.334.png", "Validation_47331653/Validation_47331653_47331653_23019.333.png", "Validation_47331653/Validation_47331653_47331653_23021.332.png", "Validation_47331653/Validation_47331653_47331653_23023.331.png" ], "turn_id": 4, "scan_id": "Validation_47331653", "origin_question": "Among these three objects(\"the sleek silver and black oven\",\"the sleek white stove with the rectangular shape and four burners\",\"the compact, vivid orange plastic stool\" ), which one is the farthest from you?", "option": [ "the sleek silver and black oven", "the sleek white stove with the rectangular shape and four burners", "the compact, vivid orange plastic stool" ], "answer": "the compact, vivid orange plastic stool", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the sleek silver and black oven\",\"the sleek white stove with the rectangular shape and four burners\",\"the compact, vivid orange plastic stool\" ), which one is the farthest from you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the sleek silver and black oven', 'the sleek white stove with the rectangular shape and four burners', 'the compact, vivid orange plastic stool'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "Validation_47331653/Validation_47331653_47331653_23025.331.png", "Validation_47331653/Validation_47331653_47331653_23027.330.png", "Validation_47331653/Validation_47331653_47331653_23029.329.png", "Validation_47331653/Validation_47331653_47331653_23031.328.png", "Validation_47331653/Validation_47331653_47331653_23033.327.png" ], "turn_id": 5, "scan_id": "Validation_47331653", "origin_question": "Among these three objects(\"the sleek white stove with the rectangular shape and four burners\",\"the compact, vivid orange plastic stool\",\"the round white ceiling light fixture made of metal and glass\" ), which one is the closest to you?", "option": [ "the sleek white stove with the rectangular shape and four burners", "the compact, vivid orange plastic stool", "the round white ceiling light fixture made of metal and glass" ], "answer": "the round white ceiling light fixture made of metal and glass", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the sleek white stove with the rectangular shape and four burners\",\"the compact, vivid orange plastic stool\",\"the round white ceiling light fixture made of metal and glass\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the sleek white stove with the rectangular shape and four burners', 'the compact, vivid orange plastic stool', 'the round white ceiling light fixture made of metal and glass'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47331653/Validation_47331653_47331653_23035.326.png", "Validation_47331653/Validation_47331653_47331653_23037.326.png", "Validation_47331653/Validation_47331653_47331653_23039.325.png", "Validation_47331653/Validation_47331653_47331653_23041.324.png", "Validation_47331653/Validation_47331653_47331653_23043.323.png" ], "turn_id": 6, "scan_id": "Validation_47331653", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the large round wicker basket with an elegant handle\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the large round wicker basket with an elegant handle\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_47331653/Validation_47331653_47331653_23045.322.png", "Validation_47331653/Validation_47331653_47331653_23047.322.png", "Validation_47331653/Validation_47331653_47331653_23049.337.png", "Validation_47331653/Validation_47331653_47331653_23051.337.png", "Validation_47331653/Validation_47331653_47331653_23053.336.png" ], "turn_id": 7, "scan_id": "Validation_47331653", "origin_question": "Please recall \"the large round wicker basket with an elegant handle\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.49", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the large round wicker basket with an elegant handle\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_47331653/Validation_47331653_47331653_23055.335.png", "Validation_47331653/Validation_47331653_47331653_23057.334.png", "Validation_47331653/Validation_47331653_47331653_23059.333.png", "Validation_47331653/Validation_47331653_47331653_23061.333.png", "Validation_47331653/Validation_47331653_47331653_23063.332.png" ], "turn_id": 8, "scan_id": "Validation_47331653", "origin_question": "At the end of which turn were both of \"the round white ceiling light fixture made of metal and glass\" and \"the bright yellow rubber glove for practical use\" on your left side?", "option": [], "answer": "7", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the round white ceiling light fixture made of metal and glass\" and \"the bright yellow rubber glove for practical use\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_47331653/Validation_47331653_47331653_23065.331.png", "Validation_47331653/Validation_47331653_47331653_23067.330.png", "Validation_47331653/Validation_47331653_47331653_23069.329.png", "Validation_47331653/Validation_47331653_47331653_23071.328.png", "Validation_47331653/Validation_47331653_47331653_23073.328.png" ], "turn_id": 9, "scan_id": "Validation_47331653", "origin_question": "Is the sleek silver and black oven to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the sleek silver and black oven to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_47331653/Validation_47331653_47331653_23075.327.png", "Validation_47331653/Validation_47331653_47331653_23077.326.png", "Validation_47331653/Validation_47331653_47331653_23079.325.png", "Validation_47331653/Validation_47331653_47331653_23081.324.png", "Validation_47331653/Validation_47331653_47331653_23083.323.png" ], "turn_id": 10, "scan_id": "Validation_47331653", "origin_question": "Is the sleek silver and black oven to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the sleek silver and black oven to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47334234/Validation_47334234_47334234_109620.199.png", "Validation_47334234/Validation_47334234_47334234_109622.182.png", "Validation_47334234/Validation_47334234_47334234_109624.181.png", "Validation_47334234/Validation_47334234_47334234_109626.180.png", "Validation_47334234/Validation_47334234_47334234_109628.179.png" ], "turn_id": 1, "scan_id": "Validation_47334234", "origin_question": "Remember, have you seen any door(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any door(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47334234/Validation_47334234_47334234_109630.178.png", "Validation_47334234/Validation_47334234_47334234_109632.178.png", "Validation_47334234/Validation_47334234_47334234_109634.177.png", "Validation_47334234/Validation_47334234_47334234_109636.176.png", "Validation_47334234/Validation_47334234_47334234_109638.175.png" ], "turn_id": 2, "scan_id": "Validation_47334234", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47334234/Validation_47334234_47334234_109640.174.png", "Validation_47334234/Validation_47334234_47334234_109642.174.png", "Validation_47334234/Validation_47334234_47334234_109644.173.png", "Validation_47334234/Validation_47334234_47334234_109646.172.png", "Validation_47334234/Validation_47334234_47334234_109648.171.png" ], "turn_id": 3, "scan_id": "Validation_47334234", "origin_question": "Which direction is the neutral brown square wooden stool in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the neutral brown square wooden stool in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47334234/Validation_47334234_47334234_109650.170.png", "Validation_47334234/Validation_47334234_47334234_109652.169.png", "Validation_47334234/Validation_47334234_47334234_109654.169.png", "Validation_47334234/Validation_47334234_47334234_109656.185.png", "Validation_47334234/Validation_47334234_47334234_109658.184.png" ], "turn_id": 4, "scan_id": "Validation_47334234", "origin_question": "Which direction is the light-colored wooden door with frosted glass panels in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the light-colored wooden door with frosted glass panels in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_47334234/Validation_47334234_47334234_109660.183.png", "Validation_47334234/Validation_47334234_47334234_109662.182.png", "Validation_47334234/Validation_47334234_47334234_109664.181.png", "Validation_47334234/Validation_47334234_47334234_109666.180.png", "Validation_47334234/Validation_47334234_47334234_109668.180.png" ], "turn_id": 5, "scan_id": "Validation_47334234", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "1.28", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_47334234/Validation_47334234_47334234_109670.179.png", "Validation_47334234/Validation_47334234_47334234_109672.178.png", "Validation_47334234/Validation_47334234_47334234_109674.177.png", "Validation_47334234/Validation_47334234_47334234_109676.176.png", "Validation_47334234/Validation_47334234_47334234_109678.175.png" ], "turn_id": 6, "scan_id": "Validation_47334234", "origin_question": "At the end of which turn were both of \"the light-colored wooden door with frosted glass panels\" and \"the dark brown wooden chair with the light beige cushion\" on your left side?", "option": [], "answer": "6", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the light-colored wooden door with frosted glass panels\" and \"the dark brown wooden chair with the light beige cushion\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_47334234/Validation_47334234_47334234_109680.175.png", "Validation_47334234/Validation_47334234_47334234_109682.174.png", "Validation_47334234/Validation_47334234_47334234_109684.173.png", "Validation_47334234/Validation_47334234_47334234_109686.172.png", "Validation_47334234/Validation_47334234_47334234_109688.171.png" ], "turn_id": 7, "scan_id": "Validation_47334234", "origin_question": "Is the neutral brown square wooden stool to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the neutral brown square wooden stool to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47334234/Validation_47334234_47334234_109690.171.png", "Validation_47334234/Validation_47334234_47334234_109692.170.png", "Validation_47334234/Validation_47334234_47334234_109694.169.png", "Validation_47334234/Validation_47334234_47334234_109696.168.png", "Validation_47334234/Validation_47334234_47334234_109698.184.png" ], "turn_id": 8, "scan_id": "Validation_47334234", "origin_question": "Compared to your position at the end of the 6 turn, are you now closer or farther away from \"the tall, dark wooden cabinet with glass-paneled doors\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 6 turn, are you now closer or farther away from \"the tall, dark wooden cabinet with glass-paneled doors\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_47334234/Validation_47334234_47334234_109700.183.png", "Validation_47334234/Validation_47334234_47334234_109702.182.png", "Validation_47334234/Validation_47334234_47334234_109704.182.png", "Validation_47334234/Validation_47334234_47334234_109706.181.png", "Validation_47334234/Validation_47334234_47334234_109708.180.png" ], "turn_id": 9, "scan_id": "Validation_47334234", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the dark brown wooden chair with the light beige cushion\", Y: \"the large, elegant bottle with earthy patterns\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the dark brown wooden chair with the light beige cushion\", Y: \"the large, elegant bottle with earthy patterns\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47334234/Validation_47334234_47334234_109710.179.png", "Validation_47334234/Validation_47334234_47334234_109712.178.png", "Validation_47334234/Validation_47334234_47334234_109714.178.png", "Validation_47334234/Validation_47334234_47334234_109716.177.png", "Validation_47334234/Validation_47334234_47334234_109718.176.png" ], "turn_id": 10, "scan_id": "Validation_47334234", "origin_question": "Which direction is the large, elegant bottle with earthy patterns in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the large, elegant bottle with earthy patterns in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_44358435/Validation_44358435_44358435_42412.803.png", "Validation_44358435/Validation_44358435_44358435_42414.802.png", "Validation_44358435/Validation_44358435_44358435_42416.802.png", "Validation_44358435/Validation_44358435_44358435_42418.801.png", "Validation_44358435/Validation_44358435_44358435_42420.800.png" ], "turn_id": 1, "scan_id": "Validation_44358435", "origin_question": "Remember, have you seen any chair(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any chair(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_44358435/Validation_44358435_44358435_42422.799.png", "Validation_44358435/Validation_44358435_44358435_42424.798.png", "Validation_44358435/Validation_44358435_44358435_42426.797.png", "Validation_44358435/Validation_44358435_44358435_42428.797.png", "Validation_44358435/Validation_44358435_44358435_42430.796.png" ], "turn_id": 2, "scan_id": "Validation_44358435", "origin_question": "In which turn were you farthest from \"the sleek, modern gray couch with the simple silhouette\"? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the sleek, modern gray couch with the simple silhouette\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_44358435/Validation_44358435_44358435_42432.795.png", "Validation_44358435/Validation_44358435_44358435_42434.794.png", "Validation_44358435/Validation_44358435_44358435_42436.793.png", "Validation_44358435/Validation_44358435_44358435_42438.792.png", "Validation_44358435/Validation_44358435_44358435_42440.792.png" ], "turn_id": 3, "scan_id": "Validation_44358435", "origin_question": "In which turn did you see both \"the sleek, modern gray couch with the simple silhouette\" and \"the modern silver industrial metal shelf\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the sleek, modern gray couch with the simple silhouette\" and \"the modern silver industrial metal shelf\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_44358435/Validation_44358435_44358435_42442.791.png", "Validation_44358435/Validation_44358435_44358435_42444.790.png", "Validation_44358435/Validation_44358435_44358435_42446.789.png", "Validation_44358435/Validation_44358435_44358435_42448.788.png", "Validation_44358435/Validation_44358435_44358435_42450.787.png" ], "turn_id": 4, "scan_id": "Validation_44358435", "origin_question": "", "option": [], "answer": "", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_44358435/Validation_44358435_44358435_42452.787.png", "Validation_44358435/Validation_44358435_44358435_42454.786.png", "Validation_44358435/Validation_44358435_44358435_42456.802.png", "Validation_44358435/Validation_44358435_44358435_42458.801.png", "Validation_44358435/Validation_44358435_44358435_42460.800.png" ], "turn_id": 5, "scan_id": "Validation_44358435", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_44358435/Validation_44358435_44358435_42462.799.png", "Validation_44358435/Validation_44358435_44358435_42464.798.png", "Validation_44358435/Validation_44358435_44358435_42466.798.png", "Validation_44358435/Validation_44358435_44358435_42468.797.png", "Validation_44358435/Validation_44358435_44358435_42470.796.png" ], "turn_id": 6, "scan_id": "Validation_44358435", "origin_question": "When did you first discover \"the modern silver industrial metal shelf\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the modern silver industrial metal shelf\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_44358435/Validation_44358435_44358435_42472.795.png", "Validation_44358435/Validation_44358435_44358435_42474.794.png", "Validation_44358435/Validation_44358435_44358435_42476.793.png", "Validation_44358435/Validation_44358435_44358435_42478.793.png", "Validation_44358435/Validation_44358435_44358435_42480.792.png" ], "turn_id": 7, "scan_id": "Validation_44358435", "origin_question": "Which direction is the cylindrical green bin with the metallic lid in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the cylindrical green bin with the metallic lid in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_44358435/Validation_44358435_44358435_42482.791.png", "Validation_44358435/Validation_44358435_44358435_42484.790.png", "Validation_44358435/Validation_44358435_44358435_42486.789.png", "Validation_44358435/Validation_44358435_44358435_42488.788.png", "Validation_44358435/Validation_44358435_44358435_42490.788.png" ], "turn_id": 8, "scan_id": "Validation_44358435", "origin_question": "Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_44358435/Validation_44358435_44358435_42492.787.png", "Validation_44358435/Validation_44358435_44358435_42494.786.png", "Validation_44358435/Validation_44358435_44358435_42496.802.png", "Validation_44358435/Validation_44358435_44358435_42498.801.png", "Validation_44358435/Validation_44358435_44358435_42500.800.png" ], "turn_id": 9, "scan_id": "Validation_44358435", "origin_question": "Which direction is the cylindrical green bin with the metallic lid in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the cylindrical green bin with the metallic lid in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_44358435/Validation_44358435_44358435_42502.799.png", "Validation_44358435/Validation_44358435_44358435_42504.798.png", "Validation_44358435/Validation_44358435_44358435_42506.798.png", "Validation_44358435/Validation_44358435_44358435_42508.797.png", "Validation_44358435/Validation_44358435_44358435_42510.796.png" ], "turn_id": 10, "scan_id": "Validation_44358435", "origin_question": "What will be the first-time appearance order of the following categories: bin, couch, shelf?", "option": [ "couch, shelf, bin", "bin, couch, shelf", "shelf, couch, bin", "couch, bin, shelf" ], "answer": "couch, shelf, bin", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bin, couch, shelf? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['couch, shelf, bin', 'bin, couch, shelf', 'shelf, couch, bin', 'couch, bin, shelf'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47331990/Validation_47331990_47331990_170480.953.png", "Validation_47331990/Validation_47331990_47331990_170482.952.png", "Validation_47331990/Validation_47331990_47331990_170484.951.png", "Validation_47331990/Validation_47331990_47331990_170486.951.png", "Validation_47331990/Validation_47331990_47331990_170488.950.png" ], "turn_id": 1, "scan_id": "Validation_47331990", "origin_question": "Remember, have you seen any footstool(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any footstool(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47331990/Validation_47331990_47331990_170490.949.png", "Validation_47331990/Validation_47331990_47331990_170492.948.png", "Validation_47331990/Validation_47331990_47331990_170494.964.png", "Validation_47331990/Validation_47331990_47331990_170496.963.png", "Validation_47331990/Validation_47331990_47331990_170498.962.png" ], "turn_id": 2, "scan_id": "Validation_47331990", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_47331990/Validation_47331990_47331990_170500.962.png", "Validation_47331990/Validation_47331990_47331990_170502.961.png", "Validation_47331990/Validation_47331990_47331990_170504.960.png", "Validation_47331990/Validation_47331990_47331990_170506.959.png", "Validation_47331990/Validation_47331990_47331990_170508.958.png" ], "turn_id": 3, "scan_id": "Validation_47331990", "origin_question": "How far is your current position from where you were at the end of the turn 1? (in meters)", "option": [], "answer": "3.5", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 1? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_47331990/Validation_47331990_47331990_170510.958.png", "Validation_47331990/Validation_47331990_47331990_170512.957.png", "Validation_47331990/Validation_47331990_47331990_170514.956.png", "Validation_47331990/Validation_47331990_47331990_170516.955.png", "Validation_47331990/Validation_47331990_47331990_170518.954.png" ], "turn_id": 4, "scan_id": "Validation_47331990", "origin_question": "Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance forward or backward from that position?", "option": [ "Forward", "Backward" ], "answer": "Forward", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance forward or backward from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Forward', 'Backward'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_47331990/Validation_47331990_47331990_170520.954.png", "Validation_47331990/Validation_47331990_47331990_170522.953.png", "Validation_47331990/Validation_47331990_47331990_170524.952.png", "Validation_47331990/Validation_47331990_47331990_170526.951.png", "Validation_47331990/Validation_47331990_47331990_170528.950.png" ], "turn_id": 5, "scan_id": "Validation_47331990", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "1.1", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_47331990/Validation_47331990_47331990_170530.949.png", "Validation_47331990/Validation_47331990_47331990_170532.949.png", "Validation_47331990/Validation_47331990_47331990_170534.948.png", "Validation_47331990/Validation_47331990_47331990_170536.964.png", "Validation_47331990/Validation_47331990_47331990_170538.963.png" ], "turn_id": 6, "scan_id": "Validation_47331990", "origin_question": "In which turn did you see both \"the plush brown bedspread with white polkthe dots\" and \"the vibrant purple round flowerpot made of plastic\" simultaneously? ", "option": [], "answer": "6", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the plush brown bedspread with white polkthe dots\" and \"the vibrant purple round flowerpot made of plastic\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47331990/Validation_47331990_47331990_170540.962.png", "Validation_47331990/Validation_47331990_47331990_170542.961.png", "Validation_47331990/Validation_47331990_47331990_170544.961.png", "Validation_47331990/Validation_47331990_47331990_170546.960.png", "Validation_47331990/Validation_47331990_47331990_170548.959.png" ], "turn_id": 7, "scan_id": "Validation_47331990", "origin_question": "", "option": [], "answer": "", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47331990/Validation_47331990_47331990_170550.958.png", "Validation_47331990/Validation_47331990_47331990_170552.957.png", "Validation_47331990/Validation_47331990_47331990_170554.957.png", "Validation_47331990/Validation_47331990_47331990_170556.956.png", "Validation_47331990/Validation_47331990_47331990_170558.955.png" ], "turn_id": 8, "scan_id": "Validation_47331990", "origin_question": "Which direction is the modern interior door with frosted glass panels in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the modern interior door with frosted glass panels in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_47331990/Validation_47331990_47331990_170560.954.png", "Validation_47331990/Validation_47331990_47331990_170562.953.png", "Validation_47331990/Validation_47331990_47331990_170564.953.png", "Validation_47331990/Validation_47331990_47331990_170566.952.png", "Validation_47331990/Validation_47331990_47331990_170568.951.png" ], "turn_id": 9, "scan_id": "Validation_47331990", "origin_question": "How far is your current position from where you were at the end of the turn 1? (in meters)", "option": [], "answer": "2.26", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 1? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_47331990/Validation_47331990_47331990_170570.950.png", "Validation_47331990/Validation_47331990_47331990_170572.949.png", "Validation_47331990/Validation_47331990_47331990_170574.948.png", "Validation_47331990/Validation_47331990_47331990_170576.964.png", "Validation_47331990/Validation_47331990_47331990_170578.964.png" ], "turn_id": 10, "scan_id": "Validation_47331990", "origin_question": "Using your orientation at the end of turn 8 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 8 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42897521/Validation_42897521_42897521_273954.897.png", "Validation_42897521/Validation_42897521_42897521_273956.896.png", "Validation_42897521/Validation_42897521_42897521_273958.895.png", "Validation_42897521/Validation_42897521_42897521_273960.894.png" ], "turn_id": 1, "scan_id": "Validation_42897521", "origin_question": "Remember, have you seen any stove(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any stove(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_42897521/Validation_42897521_42897521_273962.894.png", "Validation_42897521/Validation_42897521_42897521_273964.893.png", "Validation_42897521/Validation_42897521_42897521_273966.892.png", "Validation_42897521/Validation_42897521_42897521_273968.891.png" ], "turn_id": 2, "scan_id": "Validation_42897521", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_42897521/Validation_42897521_42897521_273970.890.png", "Validation_42897521/Validation_42897521_42897521_273972.906.png", "Validation_42897521/Validation_42897521_42897521_273974.905.png", "Validation_42897521/Validation_42897521_42897521_273976.905.png" ], "turn_id": 3, "scan_id": "Validation_42897521", "origin_question": "", "option": [], "answer": "", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_42897521/Validation_42897521_42897521_273978.904.png", "Validation_42897521/Validation_42897521_42897521_273980.903.png", "Validation_42897521/Validation_42897521_42897521_273982.902.png", "Validation_42897521/Validation_42897521_42897521_273984.901.png" ], "turn_id": 4, "scan_id": "Validation_42897521", "origin_question": "In which turn were you closest to \"the refrigerator\"? ", "option": [], "answer": "1", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the refrigerator\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_42897521/Validation_42897521_42897521_273986.900.png", "Validation_42897521/Validation_42897521_42897521_273988.899.png", "Validation_42897521/Validation_42897521_42897521_273990.899.png", "Validation_42897521/Validation_42897521_42897521_273992.898.png" ], "turn_id": 5, "scan_id": "Validation_42897521", "origin_question": "Which direction is the stylish black metal range hood with the glossy finish in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the stylish black metal range hood with the glossy finish in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_42897521/Validation_42897521_42897521_273994.897.png", "Validation_42897521/Validation_42897521_42897521_273996.896.png", "Validation_42897521/Validation_42897521_42897521_273998.895.png", "Validation_42897521/Validation_42897521_42897521_274000.895.png" ], "turn_id": 6, "scan_id": "Validation_42897521", "origin_question": "What will be the first-time appearance order of the following categories: kettle, refrigerator, window?", "option": [ "window, kettle, refrigerator", "refrigerator, window, kettle", "kettle, refrigerator, window", "window, refrigerator, kettle" ], "answer": "window, refrigerator, kettle", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: kettle, refrigerator, window? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['window, kettle, refrigerator', 'refrigerator, window, kettle', 'kettle, refrigerator, window', 'window, refrigerator, kettle'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_42897521/Validation_42897521_42897521_274002.894.png", "Validation_42897521/Validation_42897521_42897521_274004.893.png", "Validation_42897521/Validation_42897521_42897521_274006.892.png", "Validation_42897521/Validation_42897521_42897521_274008.891.png" ], "turn_id": 7, "scan_id": "Validation_42897521", "origin_question": "By the end of the 4 turn, were you getting closer to or farther away from these two objects: X: \"the refrigerator\", Y: \"the cylindrical stainless steel bin with the black bag\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 4 turn, were you getting closer to or farther away from these two objects: X: \"the refrigerator\", Y: \"the cylindrical stainless steel bin with the black bag\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47332911/Validation_47332911_47332911_12259.850.png", "Validation_47332911/Validation_47332911_47332911_12261.832.png", "Validation_47332911/Validation_47332911_47332911_12263.832.png", "Validation_47332911/Validation_47332911_47332911_12265.831.png", "Validation_47332911/Validation_47332911_47332911_12267.830.png" ], "turn_id": 1, "scan_id": "Validation_47332911", "origin_question": "Remember, have you seen any bathtub(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any bathtub(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_47332911/Validation_47332911_47332911_12269.829.png", "Validation_47332911/Validation_47332911_47332911_12271.828.png", "Validation_47332911/Validation_47332911_47332911_12273.827.png", "Validation_47332911/Validation_47332911_47332911_12275.827.png", "Validation_47332911/Validation_47332911_47332911_12277.826.png" ], "turn_id": 2, "scan_id": "Validation_47332911", "origin_question": "In which turn did you see both \"an off-white, rectangular cabinet made of medium-density fiberboard\" and \"the leaning rectangular metal shelf with the sleek finish\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"an off-white, rectangular cabinet made of medium-density fiberboard\" and \"the leaning rectangular metal shelf with the sleek finish\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_47332911/Validation_47332911_47332911_12279.825.png", "Validation_47332911/Validation_47332911_47332911_12281.824.png", "Validation_47332911/Validation_47332911_47332911_12283.824.png", "Validation_47332911/Validation_47332911_47332911_12285.839.png", "Validation_47332911/Validation_47332911_47332911_12287.838.png" ], "turn_id": 3, "scan_id": "Validation_47332911", "origin_question": "What will be the first-time appearance order of the following categories: bathtub, shelf, washing machine?", "option": [ "washing machine, bathtub, shelf", "bathtub, shelf, washing machine", "shelf, bathtub, washing machine", "bathtub, washing machine, shelf" ], "answer": "bathtub, shelf, washing machine", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bathtub, shelf, washing machine? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['washing machine, bathtub, shelf', 'bathtub, shelf, washing machine', 'shelf, bathtub, washing machine', 'bathtub, washing machine, shelf'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_47332911/Validation_47332911_47332911_12289.838.png", "Validation_47332911/Validation_47332911_47332911_12291.837.png", "Validation_47332911/Validation_47332911_47332911_12293.836.png", "Validation_47332911/Validation_47332911_47332911_12295.835.png", "Validation_47332911/Validation_47332911_47332911_12297.834.png" ], "turn_id": 4, "scan_id": "Validation_47332911", "origin_question": "When did you first discover \"the sleek rectangular white bathtub made of durable materials\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the sleek rectangular white bathtub made of durable materials\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47332911/Validation_47332911_47332911_12299.834.png", "Validation_47332911/Validation_47332911_47332911_12301.833.png", "Validation_47332911/Validation_47332911_47332911_12303.832.png", "Validation_47332911/Validation_47332911_47332911_12305.831.png", "Validation_47332911/Validation_47332911_47332911_12307.830.png" ], "turn_id": 5, "scan_id": "Validation_47332911", "origin_question": "In which turn were you farthest from \"the sleek rectangular white bathtub made of durable materials\"? ", "option": [], "answer": "1", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the sleek rectangular white bathtub made of durable materials\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_47332911/Validation_47332911_47332911_12309.829.png", "Validation_47332911/Validation_47332911_47332911_12311.829.png", "Validation_47332911/Validation_47332911_47332911_12313.828.png", "Validation_47332911/Validation_47332911_47332911_12315.827.png", "Validation_47332911/Validation_47332911_47332911_12317.826.png" ], "turn_id": 6, "scan_id": "Validation_47332911", "origin_question": "What will be the first-time appearance order of the following categories: bathtub, cabinet, washing machine?", "option": [ "bathtub, cabinet, washing machine", "cabinet, washing machine, bathtub", "bathtub, washing machine, cabinet", "washing machine, cabinet, bathtub" ], "answer": "bathtub, cabinet, washing machine", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bathtub, cabinet, washing machine? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['bathtub, cabinet, washing machine', 'cabinet, washing machine, bathtub', 'bathtub, washing machine, cabinet', 'washing machine, cabinet, bathtub'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47332911/Validation_47332911_47332911_12319.826.png", "Validation_47332911/Validation_47332911_47332911_12321.825.png", "Validation_47332911/Validation_47332911_47332911_12323.824.png", "Validation_47332911/Validation_47332911_47332911_12325.823.png", "Validation_47332911/Validation_47332911_47332911_12327.839.png" ], "turn_id": 7, "scan_id": "Validation_47332911", "origin_question": "In which turn were you closest to \"the leaning rectangular metal shelf with the sleek finish\"? ", "option": [], "answer": "6", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the leaning rectangular metal shelf with the sleek finish\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_47332911/Validation_47332911_47332911_12329.838.png", "Validation_47332911/Validation_47332911_47332911_12331.837.png", "Validation_47332911/Validation_47332911_47332911_12333.836.png", "Validation_47332911/Validation_47332911_47332911_12335.836.png", "Validation_47332911/Validation_47332911_47332911_12337.835.png" ], "turn_id": 8, "scan_id": "Validation_47332911", "origin_question": "Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_47332911/Validation_47332911_47332911_12339.834.png", "Validation_47332911/Validation_47332911_47332911_12341.833.png", "Validation_47332911/Validation_47332911_47332911_12343.832.png", "Validation_47332911/Validation_47332911_47332911_12345.832.png", "Validation_47332911/Validation_47332911_47332911_12347.831.png" ], "turn_id": 9, "scan_id": "Validation_47332911", "origin_question": "Assuming the direction you are facing at the end of the turn 8 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Left", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 8 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_47332911/Validation_47332911_47332911_12349.830.png", "Validation_47332911/Validation_47332911_47332911_12351.829.png", "Validation_47332911/Validation_47332911_47332911_12353.828.png", "Validation_47332911/Validation_47332911_47332911_12355.827.png", "Validation_47332911/Validation_47332911_47332911_12357.827.png" ], "turn_id": 10, "scan_id": "Validation_47332911", "origin_question": "Assuming the direction you are facing at the end of the turn 6 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Left", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 6 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_45261128/Validation_45261128_45261128_3490.178.png", "Validation_45261128/Validation_45261128_45261128_3492.161.png", "Validation_45261128/Validation_45261128_45261128_3494.160.png", "Validation_45261128/Validation_45261128_45261128_3496.159.png" ], "turn_id": 1, "scan_id": "Validation_45261128", "origin_question": "Remember, have you seen any refrigerator(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any refrigerator(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_45261128/Validation_45261128_45261128_3498.159.png", "Validation_45261128/Validation_45261128_45261128_3500.158.png", "Validation_45261128/Validation_45261128_45261128_3502.157.png", "Validation_45261128/Validation_45261128_45261128_3504.156.png" ], "turn_id": 2, "scan_id": "Validation_45261128", "origin_question": "When did you first discover \"the sleek black metal shelf with circular openings\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the sleek black metal shelf with circular openings\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_45261128/Validation_45261128_45261128_3506.155.png", "Validation_45261128/Validation_45261128_45261128_3508.154.png", "Validation_45261128/Validation_45261128_45261128_3510.154.png", "Validation_45261128/Validation_45261128_45261128_3512.169.png" ], "turn_id": 3, "scan_id": "Validation_45261128", "origin_question": "What will be the first-time appearance order of the following categories: cabinet, refrigerator, shelf?", "option": [ "refrigerator, shelf, cabinet", "cabinet, shelf, refrigerator", "shelf, refrigerator, cabinet", "refrigerator, cabinet, shelf" ], "answer": "refrigerator, shelf, cabinet", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: cabinet, refrigerator, shelf? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['refrigerator, shelf, cabinet', 'cabinet, shelf, refrigerator', 'shelf, refrigerator, cabinet', 'refrigerator, cabinet, shelf'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_45261128/Validation_45261128_45261128_3514.169.png", "Validation_45261128/Validation_45261128_45261128_3516.168.png", "Validation_45261128/Validation_45261128_45261128_3518.167.png", "Validation_45261128/Validation_45261128_45261128_3520.166.png" ], "turn_id": 4, "scan_id": "Validation_45261128", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_45261128/Validation_45261128_45261128_3522.165.png", "Validation_45261128/Validation_45261128_45261128_3524.164.png", "Validation_45261128/Validation_45261128_45261128_3526.164.png", "Validation_45261128/Validation_45261128_45261128_3528.163.png" ], "turn_id": 5, "scan_id": "Validation_45261128", "origin_question": "Please recall \"the modern rectangular silver microwave with the glass door\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.22", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the modern rectangular silver microwave with the glass door\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_45261128/Validation_45261128_45261128_3530.162.png", "Validation_45261128/Validation_45261128_45261128_3532.161.png", "Validation_45261128/Validation_45261128_45261128_3534.160.png", "Validation_45261128/Validation_45261128_45261128_3536.160.png" ], "turn_id": 6, "scan_id": "Validation_45261128", "origin_question": "Which two objects are on the same side of you now? \"the modern rectangular silver microwave with the glass door\", \"the sleek, black glass-ceramic stove\", \"the warm wooden table with the sleek white structure\". ", "option": [ "the modern rectangular silver microwave with the glass door, the sleek, black glass-ceramic stove", "the modern rectangular silver microwave with the glass door, the warm wooden table with the sleek white structure", "the sleek, black glass-ceramic stove, the warm wooden table with the sleek white structure" ], "answer": "the modern rectangular silver microwave with the glass door, the sleek, black glass-ceramic stove", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the modern rectangular silver microwave with the glass door\", \"the sleek, black glass-ceramic stove\", \"the warm wooden table with the sleek white structure\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the modern rectangular silver microwave with the glass door, the sleek, black glass-ceramic stove', 'the modern rectangular silver microwave with the glass door, the warm wooden table with the sleek white structure', 'the sleek, black glass-ceramic stove, the warm wooden table with the sleek white structure'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_45261128/Validation_45261128_45261128_3538.159.png", "Validation_45261128/Validation_45261128_45261128_3540.158.png", "Validation_45261128/Validation_45261128_45261128_3542.157.png", "Validation_45261128/Validation_45261128_45261128_3544.156.png" ], "turn_id": 7, "scan_id": "Validation_45261128", "origin_question": "At the end of which turn were both of \"the creamy off-white vintage refrigerator with rounded corners\" and \"the black stainless steel oven with the large glass window\" on your right side?", "option": [], "answer": "7", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the creamy off-white vintage refrigerator with rounded corners\" and \"the black stainless steel oven with the large glass window\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_45261128/Validation_45261128_45261128_3546.155.png", "Validation_45261128/Validation_45261128_45261128_3548.155.png", "Validation_45261128/Validation_45261128_45261128_3550.154.png", "Validation_45261128/Validation_45261128_45261128_3552.169.png" ], "turn_id": 8, "scan_id": "Validation_45261128", "origin_question": "Is the creamy off-white vintage refrigerator with rounded corners to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the creamy off-white vintage refrigerator with rounded corners to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "Validation_45261128/Validation_45261128_45261128_3554.169.png", "Validation_45261128/Validation_45261128_45261128_3556.168.png", "Validation_45261128/Validation_45261128_45261128_3558.167.png", "Validation_45261128/Validation_45261128_45261128_3560.166.png" ], "turn_id": 9, "scan_id": "Validation_45261128", "origin_question": "Among these three objects(\"the sleek, black glass-ceramic stove\",\"the sleek, medium-sized metallic gray machine\",\"the compact, round neutral-toned stool\" ), which one is the closest to you?", "option": [ "the sleek, black glass-ceramic stove", "the sleek, medium-sized metallic gray machine", "the compact, round neutral-toned stool" ], "answer": "the compact, round neutral-toned stool", "user_message": "For the 9 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects(\"the sleek, black glass-ceramic stove\",\"the sleek, medium-sized metallic gray machine\",\"the compact, round neutral-toned stool\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the sleek, black glass-ceramic stove', 'the sleek, medium-sized metallic gray machine', 'the compact, round neutral-toned stool'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47333923/Validation_47333923_47333923_83598.520.png", "Validation_47333923/Validation_47333923_47333923_83600.503.png", "Validation_47333923/Validation_47333923_47333923_83602.502.png", "Validation_47333923/Validation_47333923_47333923_83604.501.png" ], "turn_id": 1, "scan_id": "Validation_47333923", "origin_question": "Remember, have you seen any hat(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any hat(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_47333923/Validation_47333923_47333923_83606.500.png", "Validation_47333923/Validation_47333923_47333923_83608.499.png", "Validation_47333923/Validation_47333923_47333923_83610.515.png", "Validation_47333923/Validation_47333923_47333923_83612.514.png" ], "turn_id": 2, "scan_id": "Validation_47333923", "origin_question": "At the end of which turn were both of \"the sleek rectangular white washing machine\" and \"the modern rectangular white sink\" on your right side?", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the sleek rectangular white washing machine\" and \"the modern rectangular white sink\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47333923/Validation_47333923_47333923_83614.514.png", "Validation_47333923/Validation_47333923_47333923_83616.513.png", "Validation_47333923/Validation_47333923_47333923_83618.512.png", "Validation_47333923/Validation_47333923_47333923_83620.511.png" ], "turn_id": 3, "scan_id": "Validation_47333923", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the modern rectangular white sink\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the modern rectangular white sink\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_47333923/Validation_47333923_47333923_83622.510.png", "Validation_47333923/Validation_47333923_47333923_83624.509.png", "Validation_47333923/Validation_47333923_47333923_83626.509.png", "Validation_47333923/Validation_47333923_47333923_83628.508.png" ], "turn_id": 4, "scan_id": "Validation_47333923", "origin_question": "Is the sleek rectangular white washing machine to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the sleek rectangular white washing machine to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_47333923/Validation_47333923_47333923_83630.507.png", "Validation_47333923/Validation_47333923_47333923_83632.506.png", "Validation_47333923/Validation_47333923_47333923_83634.505.png", "Validation_47333923/Validation_47333923_47333923_83636.504.png" ], "turn_id": 5, "scan_id": "Validation_47333923", "origin_question": "At the end of which turn were both of \"the sleek rectangular white washing machine\" and \"the modern rectangular white sink\" on your right side?", "option": [], "answer": "2", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the sleek rectangular white washing machine\" and \"the modern rectangular white sink\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_47333923/Validation_47333923_47333923_83638.504.png", "Validation_47333923/Validation_47333923_47333923_83640.503.png", "Validation_47333923/Validation_47333923_47333923_83642.502.png", "Validation_47333923/Validation_47333923_47333923_83644.501.png" ], "turn_id": 6, "scan_id": "Validation_47333923", "origin_question": "Which two objects are on the same side of you now? \"the modern rectangular white sink\", \"the circular deep blue plastic basin\", \"the cylindrical white plastic bucket with the lid\". ", "option": [ "the modern rectangular white sink, the circular deep blue plastic basin", "the modern rectangular white sink, the cylindrical white plastic bucket with the lid", "the circular deep blue plastic basin, the cylindrical white plastic bucket with the lid" ], "answer": "the modern rectangular white sink, the circular deep blue plastic basin", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the modern rectangular white sink\", \"the circular deep blue plastic basin\", \"the cylindrical white plastic bucket with the lid\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the modern rectangular white sink, the circular deep blue plastic basin', 'the modern rectangular white sink, the cylindrical white plastic bucket with the lid', 'the circular deep blue plastic basin, the cylindrical white plastic bucket with the lid'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_47333923/Validation_47333923_47333923_83646.500.png", "Validation_47333923/Validation_47333923_47333923_83648.500.png", "Validation_47333923/Validation_47333923_47333923_83650.499.png", "Validation_47333923/Validation_47333923_47333923_83652.515.png" ], "turn_id": 7, "scan_id": "Validation_47333923", "origin_question": "Please recall \"the medium-sized light brown wicker basket\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.9", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the medium-sized light brown wicker basket\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47332918/Validation_47332918_47332918_13316.901.png", "Validation_47332918/Validation_47332918_47332918_13318.867.png", "Validation_47332918/Validation_47332918_47332918_13320.866.png", "Validation_47332918/Validation_47332918_47332918_13322.865.png" ], "turn_id": 1, "scan_id": "Validation_47332918", "origin_question": "Remember, have you seen any monitor(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any monitor(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_47332918/Validation_47332918_47332918_13324.864.png", "Validation_47332918/Validation_47332918_47332918_13326.864.png", "Validation_47332918/Validation_47332918_47332918_13328.863.png", "Validation_47332918/Validation_47332918_47332918_13330.879.png" ], "turn_id": 2, "scan_id": "Validation_47332918", "origin_question": "When did you first discover \"the rectangular bed with the light pink bedsheet\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the rectangular bed with the light pink bedsheet\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_47332918/Validation_47332918_47332918_13332.878.png", "Validation_47332918/Validation_47332918_47332918_13334.877.png", "Validation_47332918/Validation_47332918_47332918_13336.876.png", "Validation_47332918/Validation_47332918_47332918_13338.875.png" ], "turn_id": 3, "scan_id": "Validation_47332918", "origin_question": "Which two objects are on the same side of you now? \"the modern circular light fixture with frosted glass\", \"the sleek, rectangular metal radiator with vertical fins\", \"the polished wooden chair with the classic brown hue\". ", "option": [ "the modern circular light fixture with frosted glass, the sleek, rectangular metal radiator with vertical fins", "the modern circular light fixture with frosted glass, the polished wooden chair with the classic brown hue", "the sleek, rectangular metal radiator with vertical fins, the polished wooden chair with the classic brown hue" ], "answer": "the modern circular light fixture with frosted glass, the sleek, rectangular metal radiator with vertical fins", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the modern circular light fixture with frosted glass\", \"the sleek, rectangular metal radiator with vertical fins\", \"the polished wooden chair with the classic brown hue\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the modern circular light fixture with frosted glass, the sleek, rectangular metal radiator with vertical fins', 'the modern circular light fixture with frosted glass, the polished wooden chair with the classic brown hue', 'the sleek, rectangular metal radiator with vertical fins, the polished wooden chair with the classic brown hue'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_47332918/Validation_47332918_47332918_13340.875.png", "Validation_47332918/Validation_47332918_47332918_13342.874.png", "Validation_47332918/Validation_47332918_47332918_13344.873.png", "Validation_47332918/Validation_47332918_47332918_13346.872.png" ], "turn_id": 4, "scan_id": "Validation_47332918", "origin_question": "Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "104.13", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47332918/Validation_47332918_47332918_13348.871.png", "Validation_47332918/Validation_47332918_47332918_13350.871.png", "Validation_47332918/Validation_47332918_47332918_13352.870.png", "Validation_47332918/Validation_47332918_47332918_13354.869.png" ], "turn_id": 5, "scan_id": "Validation_47332918", "origin_question": "In which turn were you farthest from \"the modern circular light fixture with frosted glass\"? ", "option": [], "answer": "1", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the modern circular light fixture with frosted glass\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_47332918/Validation_47332918_47332918_13356.868.png", "Validation_47332918/Validation_47332918_47332918_13358.867.png", "Validation_47332918/Validation_47332918_47332918_13360.867.png", "Validation_47332918/Validation_47332918_47332918_13362.866.png" ], "turn_id": 6, "scan_id": "Validation_47332918", "origin_question": "Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_47332918/Validation_47332918_47332918_13364.865.png", "Validation_47332918/Validation_47332918_47332918_13366.864.png", "Validation_47332918/Validation_47332918_47332918_13368.863.png", "Validation_47332918/Validation_47332918_47332918_13370.862.png" ], "turn_id": 7, "scan_id": "Validation_47332918", "origin_question": "In which turn did you see both \"the modern circular light fixture with frosted glass\" and \"the rectangular bed with the light pink bedsheet\" simultaneously? ", "option": [], "answer": "3", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the modern circular light fixture with frosted glass\" and \"the rectangular bed with the light pink bedsheet\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_47332918/Validation_47332918_47332918_13372.878.png", "Validation_47332918/Validation_47332918_47332918_13374.878.png", "Validation_47332918/Validation_47332918_47332918_13376.877.png", "Validation_47332918/Validation_47332918_47332918_13378.876.png" ], "turn_id": 8, "scan_id": "Validation_47332918", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42899714/Validation_42899714_42899714_796905.619.png", "Validation_42899714/Validation_42899714_42899714_796907.602.png", "Validation_42899714/Validation_42899714_42899714_796909.601.png", "Validation_42899714/Validation_42899714_42899714_796911.600.png", "Validation_42899714/Validation_42899714_42899714_796913.599.png" ], "turn_id": 1, "scan_id": "Validation_42899714", "origin_question": "Remember, have you seen any couch(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any couch(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_42899714/Validation_42899714_42899714_796915.598.png", "Validation_42899714/Validation_42899714_42899714_796917.598.png", "Validation_42899714/Validation_42899714_42899714_796919.597.png", "Validation_42899714/Validation_42899714_42899714_796921.596.png", "Validation_42899714/Validation_42899714_42899714_796923.595.png" ], "turn_id": 2, "scan_id": "Validation_42899714", "origin_question": "In which turn were you closest to \"the cabinet\"? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the cabinet\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_42899714/Validation_42899714_42899714_796925.594.png", "Validation_42899714/Validation_42899714_42899714_796927.593.png", "Validation_42899714/Validation_42899714_42899714_796929.593.png", "Validation_42899714/Validation_42899714_42899714_796931.592.png", "Validation_42899714/Validation_42899714_42899714_796933.608.png" ], "turn_id": 3, "scan_id": "Validation_42899714", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_42899714/Validation_42899714_42899714_796935.607.png", "Validation_42899714/Validation_42899714_42899714_796937.606.png", "Validation_42899714/Validation_42899714_42899714_796939.605.png", "Validation_42899714/Validation_42899714_42899714_796941.604.png", "Validation_42899714/Validation_42899714_42899714_796943.603.png" ], "turn_id": 4, "scan_id": "Validation_42899714", "origin_question": "How far is your current position from where you were at the end of the turn 1? (in meters)", "option": [], "answer": "1.23", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 1? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_42899714/Validation_42899714_42899714_796945.603.png", "Validation_42899714/Validation_42899714_42899714_796947.602.png", "Validation_42899714/Validation_42899714_42899714_796949.601.png", "Validation_42899714/Validation_42899714_42899714_796951.600.png", "Validation_42899714/Validation_42899714_42899714_796953.599.png" ], "turn_id": 5, "scan_id": "Validation_42899714", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_42899714/Validation_42899714_42899714_796955.598.png", "Validation_42899714/Validation_42899714_42899714_796957.598.png", "Validation_42899714/Validation_42899714_42899714_796959.597.png", "Validation_42899714/Validation_42899714_42899714_796961.596.png", "Validation_42899714/Validation_42899714_42899714_796963.595.png" ], "turn_id": 6, "scan_id": "Validation_42899714", "origin_question": "What will be the first-time appearance order of the following categories: bin, cabinet, couch?", "option": [ "cabinet, bin, couch", "bin, cabinet, couch", "cabinet, couch, bin", "couch, cabinet, bin" ], "answer": "couch, cabinet, bin", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bin, cabinet, couch? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['cabinet, bin, couch', 'bin, cabinet, couch', 'cabinet, couch, bin', 'couch, cabinet, bin'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_42899714/Validation_42899714_42899714_796965.594.png", "Validation_42899714/Validation_42899714_42899714_796967.593.png", "Validation_42899714/Validation_42899714_42899714_796969.593.png", "Validation_42899714/Validation_42899714_42899714_796971.592.png", "Validation_42899714/Validation_42899714_42899714_796973.591.png" ], "turn_id": 7, "scan_id": "Validation_42899714", "origin_question": "Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Right", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_42899714/Validation_42899714_42899714_796975.590.png", "Validation_42899714/Validation_42899714_42899714_796977.606.png", "Validation_42899714/Validation_42899714_42899714_796979.605.png", "Validation_42899714/Validation_42899714_42899714_796981.604.png", "Validation_42899714/Validation_42899714_42899714_796983.603.png" ], "turn_id": 8, "scan_id": "Validation_42899714", "origin_question": "What will be the first-time appearance order of the following categories: bin, cabinet, microwave?", "option": [ "cabinet, microwave, bin", "microwave, cabinet, bin", "microwave, bin, cabinet", "cabinet, bin, microwave" ], "answer": "cabinet, bin, microwave", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bin, cabinet, microwave? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['cabinet, microwave, bin', 'microwave, cabinet, bin', 'microwave, bin, cabinet', 'cabinet, bin, microwave'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_42899714/Validation_42899714_42899714_796985.603.png", "Validation_42899714/Validation_42899714_42899714_796987.602.png", "Validation_42899714/Validation_42899714_42899714_796989.601.png", "Validation_42899714/Validation_42899714_42899714_796991.600.png", "Validation_42899714/Validation_42899714_42899714_796993.599.png" ], "turn_id": 9, "scan_id": "Validation_42899714", "origin_question": "", "option": [], "answer": "", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "Validation_42899714/Validation_42899714_42899714_796995.599.png", "Validation_42899714/Validation_42899714_42899714_796997.598.png", "Validation_42899714/Validation_42899714_42899714_796999.597.png", "Validation_42899714/Validation_42899714_42899714_797001.596.png", "Validation_42899714/Validation_42899714_42899714_797003.595.png" ], "turn_id": 10, "scan_id": "Validation_42899714", "origin_question": "Which two objects are on the same side of you now? \"the compact white microwave with the black window\", \"the medium-sized rectangular stainless steel sink\", \"the stove\". ", "option": [ "the compact white microwave with the black window, the medium-sized rectangular stainless steel sink", "the compact white microwave with the black window, the stove", "the medium-sized rectangular stainless steel sink, the stove" ], "answer": "the medium-sized rectangular stainless steel sink, the stove", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the compact white microwave with the black window\", \"the medium-sized rectangular stainless steel sink\", \"the stove\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the compact white microwave with the black window, the medium-sized rectangular stainless steel sink', 'the compact white microwave with the black window, the stove', 'the medium-sized rectangular stainless steel sink, the stove'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47429998/Validation_47429998_47429998_17243.362.png", "Validation_47429998/Validation_47429998_47429998_17244.845.png", "Validation_47429998/Validation_47429998_47429998_17246.344.png", "Validation_47429998/Validation_47429998_47429998_17247.843.png" ], "turn_id": 1, "scan_id": "Validation_47429998", "origin_question": "Remember, have you seen any light(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any light(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47429998/Validation_47429998_47429998_17249.343.png", "Validation_47429998/Validation_47429998_47429998_17250.859.png", "Validation_47429998/Validation_47429998_47429998_17252.358.png", "Validation_47429998/Validation_47429998_47429998_17253.857.png" ], "turn_id": 2, "scan_id": "Validation_47429998", "origin_question": "In which turn were you farthest from \"the teardrop-shaped light fixture with the muted white finish\"? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the teardrop-shaped light fixture with the muted white finish\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47429998/Validation_47429998_47429998_17255.357.png", "Validation_47429998/Validation_47429998_47429998_17256.856.png", "Validation_47429998/Validation_47429998_47429998_17258.356.png", "Validation_47429998/Validation_47429998_47429998_17259.855.png" ], "turn_id": 3, "scan_id": "Validation_47429998", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the sleek white rectangular window frame\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the sleek white rectangular window frame\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "Validation_47429998/Validation_47429998_47429998_17261.354.png", "Validation_47429998/Validation_47429998_47429998_17262.854.png", "Validation_47429998/Validation_47429998_47429998_17264.353.png", "Validation_47429998/Validation_47429998_47429998_17265.853.png" ], "turn_id": 4, "scan_id": "Validation_47429998", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the carpet\"; \"the sleek white rectangular window frame\"; \"the teardrop-shaped light fixture with the muted white finish\" ", "option": [ "the carpet", "the sleek white rectangular window frame", "the teardrop-shaped light fixture with the muted white finish" ], "answer": "the carpet", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the carpet\"; \"the sleek white rectangular window frame\"; \"the teardrop-shaped light fixture with the muted white finish\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the carpet', 'the sleek white rectangular window frame', 'the teardrop-shaped light fixture with the muted white finish'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47429998/Validation_47429998_47429998_17267.352.png", "Validation_47429998/Validation_47429998_47429998_17268.851.png", "Validation_47429998/Validation_47429998_47429998_17270.351.png", "Validation_47429998/Validation_47429998_47429998_17271.850.png" ], "turn_id": 5, "scan_id": "Validation_47429998", "origin_question": "Remember, have you seen any cabinet(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any cabinet(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_47429998/Validation_47429998_47429998_17273.350.png", "Validation_47429998/Validation_47429998_47429998_17274.849.png", "Validation_47429998/Validation_47429998_47429998_17276.348.png", "Validation_47429998/Validation_47429998_47429998_17277.848.png" ], "turn_id": 6, "scan_id": "Validation_47429998", "origin_question": "Which direction is the sleek, muted gold metal shelf in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the sleek, muted gold metal shelf in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_45261179/Validation_45261179_45261179_2871.651.png", "Validation_45261179/Validation_45261179_45261179_2873.134.png", "Validation_45261179/Validation_45261179_45261179_2874.633.png", "Validation_45261179/Validation_45261179_45261179_2876.133.png" ], "turn_id": 1, "scan_id": "Validation_45261179", "origin_question": "Remember, have you seen any sink(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any sink(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_45261179/Validation_45261179_45261179_2877.632.png", "Validation_45261179/Validation_45261179_45261179_2879.131.png", "Validation_45261179/Validation_45261179_45261179_2880.631.png", "Validation_45261179/Validation_45261179_45261179_2882.130.png" ], "turn_id": 2, "scan_id": "Validation_45261179", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_45261179/Validation_45261179_45261179_2883.630.png", "Validation_45261179/Validation_45261179_45261179_2885.129.png", "Validation_45261179/Validation_45261179_45261179_2886.628.png", "Validation_45261179/Validation_45261179_45261179_2888.128.png" ], "turn_id": 3, "scan_id": "Validation_45261179", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "107.6", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_45261179/Validation_45261179_45261179_2889.627.png", "Validation_45261179/Validation_45261179_45261179_2891.127.png", "Validation_45261179/Validation_45261179_45261179_2892.626.png", "Validation_45261179/Validation_45261179_45261179_2894.126.png" ], "turn_id": 4, "scan_id": "Validation_45261179", "origin_question": "At the end of which turn were both of \"the compact, glossy white toilet with the rounded design\" and \"the sleek rectangular white ceramic sink\" on your right side?", "option": [], "answer": "2", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the compact, glossy white toilet with the rounded design\" and \"the sleek rectangular white ceramic sink\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_45261179/Validation_45261179_45261179_2895.625.png", "Validation_45261179/Validation_45261179_45261179_2897.124.png", "Validation_45261179/Validation_45261179_45261179_2898.640.png", "Validation_45261179/Validation_45261179_45261179_2900.140.png" ], "turn_id": 5, "scan_id": "Validation_45261179", "origin_question": "At the end of which turn were both of \"the compact, glossy white toilet with the rounded design\" and \"the sleek rectangular white ceramic sink\" on your right side?", "option": [], "answer": "2", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the compact, glossy white toilet with the rounded design\" and \"the sleek rectangular white ceramic sink\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_45261179/Validation_45261179_45261179_2901.639.png", "Validation_45261179/Validation_45261179_45261179_2903.139.png", "Validation_45261179/Validation_45261179_45261179_2904.638.png", "Validation_45261179/Validation_45261179_45261179_2906.137.png" ], "turn_id": 6, "scan_id": "Validation_45261179", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the compact, glossy white toilet with the rounded design\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the compact, glossy white toilet with the rounded design\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42898869/Validation_42898869_42898869_197774.836.png", "Validation_42898869/Validation_42898869_42898869_197776.835.png", "Validation_42898869/Validation_42898869_42898869_197778.834.png", "Validation_42898869/Validation_42898869_42898869_197780.833.png", "Validation_42898869/Validation_42898869_42898869_197782.832.png" ], "turn_id": 1, "scan_id": "Validation_42898869", "origin_question": "Remember, have you seen any package(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any package(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_42898869/Validation_42898869_42898869_197784.832.png", "Validation_42898869/Validation_42898869_42898869_197786.831.png", "Validation_42898869/Validation_42898869_42898869_197788.830.png", "Validation_42898869/Validation_42898869_42898869_197790.829.png", "Validation_42898869/Validation_42898869_42898869_197792.828.png" ], "turn_id": 2, "scan_id": "Validation_42898869", "origin_question": "When did you first discover \"the sleek black rectangular widescreen monitor\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the sleek black rectangular widescreen monitor\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_42898869/Validation_42898869_42898869_197794.828.png", "Validation_42898869/Validation_42898869_42898869_197796.827.png", "Validation_42898869/Validation_42898869_42898869_197798.826.png", "Validation_42898869/Validation_42898869_42898869_197800.825.png", "Validation_42898869/Validation_42898869_42898869_197802.824.png" ], "turn_id": 3, "scan_id": "Validation_42898869", "origin_question": "When did you first discover \"the sleek rectangular shelf with the gray cabinet\"? (index of the turn)", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the sleek rectangular shelf with the gray cabinet\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_42898869/Validation_42898869_42898869_197804.824.png", "Validation_42898869/Validation_42898869_42898869_197806.840.png", "Validation_42898869/Validation_42898869_42898869_197808.839.png", "Validation_42898869/Validation_42898869_42898869_197810.838.png", "Validation_42898869/Validation_42898869_42898869_197812.837.png" ], "turn_id": 4, "scan_id": "Validation_42898869", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_42898869/Validation_42898869_42898869_197814.836.png", "Validation_42898869/Validation_42898869_42898869_197816.836.png", "Validation_42898869/Validation_42898869_42898869_197818.835.png", "Validation_42898869/Validation_42898869_42898869_197820.834.png", "Validation_42898869/Validation_42898869_42898869_197822.833.png" ], "turn_id": 5, "scan_id": "Validation_42898869", "origin_question": "Is the sleek black rectangular widescreen monitor to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the sleek black rectangular widescreen monitor to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_42898869/Validation_42898869_42898869_197824.832.png", "Validation_42898869/Validation_42898869_42898869_197826.832.png", "Validation_42898869/Validation_42898869_42898869_197828.831.png", "Validation_42898869/Validation_42898869_42898869_197830.830.png", "Validation_42898869/Validation_42898869_42898869_197832.829.png" ], "turn_id": 6, "scan_id": "Validation_42898869", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the sleek black rectangular widescreen monitor\", Y: \"the sleek rectangular shelf with the gray cabinet\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the sleek black rectangular widescreen monitor\", Y: \"the sleek rectangular shelf with the gray cabinet\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_42898869/Validation_42898869_42898869_197834.829.png", "Validation_42898869/Validation_42898869_42898869_197836.828.png", "Validation_42898869/Validation_42898869_42898869_197838.827.png", "Validation_42898869/Validation_42898869_42898869_197840.826.png", "Validation_42898869/Validation_42898869_42898869_197842.825.png" ], "turn_id": 7, "scan_id": "Validation_42898869", "origin_question": "Please recall \"the cylindrical light emitting warm yellow glow\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.55", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the cylindrical light emitting warm yellow glow\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "Validation_42898869/Validation_42898869_42898869_197844.824.png", "Validation_42898869/Validation_42898869_42898869_197846.824.png", "Validation_42898869/Validation_42898869_42898869_197848.840.png", "Validation_42898869/Validation_42898869_42898869_197850.839.png", "Validation_42898869/Validation_42898869_42898869_197852.838.png" ], "turn_id": 8, "scan_id": "Validation_42898869", "origin_question": "Remember, how many couch(s) have you seen so far? ", "option": [], "answer": "2", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many couch(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47895735/Validation_47895735_47895735_384594.639.png", "Validation_47895735/Validation_47895735_47895735_384596.621.png", "Validation_47895735/Validation_47895735_47895735_384598.620.png", "Validation_47895735/Validation_47895735_47895735_384600.619.png" ], "turn_id": 1, "scan_id": "Validation_47895735", "origin_question": "Remember, have you seen any flowerpot(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any flowerpot(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47895735/Validation_47895735_47895735_384602.618.png", "Validation_47895735/Validation_47895735_47895735_384604.618.png", "Validation_47895735/Validation_47895735_47895735_384606.617.png", "Validation_47895735/Validation_47895735_47895735_384608.633.png" ], "turn_id": 2, "scan_id": "Validation_47895735", "origin_question": "Remember, have you seen any shelf(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any shelf(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47895735/Validation_47895735_47895735_384610.632.png", "Validation_47895735/Validation_47895735_47895735_384612.631.png", "Validation_47895735/Validation_47895735_47895735_384614.630.png", "Validation_47895735/Validation_47895735_47895735_384616.629.png" ], "turn_id": 3, "scan_id": "Validation_47895735", "origin_question": "", "option": [], "answer": "", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_47895735/Validation_47895735_47895735_384618.629.png", "Validation_47895735/Validation_47895735_47895735_384620.628.png", "Validation_47895735/Validation_47895735_47895735_384622.627.png", "Validation_47895735/Validation_47895735_47895735_384624.626.png" ], "turn_id": 4, "scan_id": "Validation_47895735", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47895735/Validation_47895735_47895735_384626.625.png", "Validation_47895735/Validation_47895735_47895735_384628.624.png", "Validation_47895735/Validation_47895735_47895735_384630.624.png", "Validation_47895735/Validation_47895735_47895735_384632.623.png" ], "turn_id": 5, "scan_id": "Validation_47895735", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the deep green plush velvet cushion\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the deep green plush velvet cushion\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47895735/Validation_47895735_47895735_384634.622.png", "Validation_47895735/Validation_47895735_47895735_384636.621.png", "Validation_47895735/Validation_47895735_47895735_384638.620.png", "Validation_47895735/Validation_47895735_47895735_384640.619.png" ], "turn_id": 6, "scan_id": "Validation_47895735", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the deep green plush velvet cushion\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the deep green plush velvet cushion\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42444955/Validation_42444955_42444955_225046.283.png", "Validation_42444955/Validation_42444955_42444955_225048.266.png", "Validation_42444955/Validation_42444955_42444955_225050.265.png", "Validation_42444955/Validation_42444955_42444955_225052.264.png" ], "turn_id": 1, "scan_id": "Validation_42444955", "origin_question": "Remember, have you seen any notebook(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any notebook(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_42444955/Validation_42444955_42444955_225054.263.png", "Validation_42444955/Validation_42444955_42444955_225056.262.png", "Validation_42444955/Validation_42444955_42444955_225058.261.png", "Validation_42444955/Validation_42444955_42444955_225060.261.png" ], "turn_id": 2, "scan_id": "Validation_42444955", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the sleek rectangular glossy white bathtub\" from your position?", "option": [], "answer": "84.3", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the sleek rectangular glossy white bathtub\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_42444955/Validation_42444955_42444955_225062.260.png", "Validation_42444955/Validation_42444955_42444955_225064.259.png", "Validation_42444955/Validation_42444955_42444955_225066.258.png", "Validation_42444955/Validation_42444955_42444955_225068.257.png" ], "turn_id": 3, "scan_id": "Validation_42444955", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the sleek rectangular glossy white bathtub\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the sleek rectangular glossy white bathtub\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_42444955/Validation_42444955_42444955_225070.256.png", "Validation_42444955/Validation_42444955_42444955_225072.256.png", "Validation_42444955/Validation_42444955_42444955_225074.255.png", "Validation_42444955/Validation_42444955_42444955_225076.254.png" ], "turn_id": 4, "scan_id": "Validation_42444955", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the classic oval white porcelain sink on the pedestal\" from your position?", "option": [], "answer": "93.17", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the classic oval white porcelain sink on the pedestal\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_42444955/Validation_42444955_42444955_225078.253.png", "Validation_42444955/Validation_42444955_42444955_225080.269.png", "Validation_42444955/Validation_42444955_42444955_225082.268.png", "Validation_42444955/Validation_42444955_42444955_225084.267.png" ], "turn_id": 5, "scan_id": "Validation_42444955", "origin_question": "In which turn did you see both \"the sleek rectangular glossy white bathtub\" and \"the classic oval white porcelain sink on the pedestal\" simultaneously? ", "option": [], "answer": "3", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the sleek rectangular glossy white bathtub\" and \"the classic oval white porcelain sink on the pedestal\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_42444955/Validation_42444955_42444955_225086.267.png", "Validation_42444955/Validation_42444955_42444955_225088.266.png", "Validation_42444955/Validation_42444955_42444955_225090.265.png", "Validation_42444955/Validation_42444955_42444955_225092.264.png" ], "turn_id": 6, "scan_id": "Validation_42444955", "origin_question": "When did you first discover \"the classic rounded white porcelain toilet\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the classic rounded white porcelain toilet\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_42444955/Validation_42444955_42444955_225094.263.png", "Validation_42444955/Validation_42444955_42444955_225096.263.png", "Validation_42444955/Validation_42444955_42444955_225098.262.png", "Validation_42444955/Validation_42444955_42444955_225100.261.png" ], "turn_id": 7, "scan_id": "Validation_42444955", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the classic oval white porcelain sink on the pedestal\" from your position?", "option": [], "answer": "107.62", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the classic oval white porcelain sink on the pedestal\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_44358448/Validation_44358448_44358448_42950.825.png", "Validation_44358448/Validation_44358448_44358448_42952.791.png", "Validation_44358448/Validation_44358448_44358448_42954.790.png", "Validation_44358448/Validation_44358448_44358448_42956.789.png" ], "turn_id": 1, "scan_id": "Validation_44358448", "origin_question": "Remember, have you seen any chair(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any chair(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_44358448/Validation_44358448_44358448_42958.789.png", "Validation_44358448/Validation_44358448_44358448_42960.805.png", "Validation_44358448/Validation_44358448_44358448_42962.804.png", "Validation_44358448/Validation_44358448_44358448_42964.803.png" ], "turn_id": 2, "scan_id": "Validation_44358448", "origin_question": "At the end of which turn were both of \"the tall, sleek white wooden cabinet\" and \"the sleek, black metal rack\" on your left side?", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the tall, sleek white wooden cabinet\" and \"the sleek, black metal rack\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "Validation_44358448/Validation_44358448_44358448_42966.802.png", "Validation_44358448/Validation_44358448_44358448_42968.801.png", "Validation_44358448/Validation_44358448_44358448_42970.800.png", "Validation_44358448/Validation_44358448_44358448_42972.800.png" ], "turn_id": 3, "scan_id": "Validation_44358448", "origin_question": "Remember, how many table(s) have you seen so far? ", "option": [], "answer": "2", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many table(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_44358448/Validation_44358448_44358448_42974.799.png", "Validation_44358448/Validation_44358448_44358448_42976.798.png", "Validation_44358448/Validation_44358448_44358448_42978.797.png", "Validation_44358448/Validation_44358448_44358448_42980.796.png" ], "turn_id": 4, "scan_id": "Validation_44358448", "origin_question": "Please recall \"the sleek chair with the curved backrest and dark upholstery\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.67", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the sleek chair with the curved backrest and dark upholstery\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_44358448/Validation_44358448_44358448_42982.795.png", "Validation_44358448/Validation_44358448_44358448_42984.795.png", "Validation_44358448/Validation_44358448_44358448_42986.794.png", "Validation_44358448/Validation_44358448_44358448_42988.793.png" ], "turn_id": 5, "scan_id": "Validation_44358448", "origin_question": "Using your orientation at the end of turn 4 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "114.49", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 4 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_44358448/Validation_44358448_44358448_42990.792.png", "Validation_44358448/Validation_44358448_44358448_42992.791.png", "Validation_44358448/Validation_44358448_44358448_42994.791.png", "Validation_44358448/Validation_44358448_44358448_42996.790.png" ], "turn_id": 6, "scan_id": "Validation_44358448", "origin_question": "In which turn were you farthest from \"the sleek, black metal rack\"? ", "option": [], "answer": "1", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the sleek, black metal rack\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_44358513/Validation_44358513_44358513_52636.725.png", "Validation_44358513/Validation_44358513_44358513_52638.708.png", "Validation_44358513/Validation_44358513_44358513_52640.707.png", "Validation_44358513/Validation_44358513_44358513_52642.706.png" ], "turn_id": 1, "scan_id": "Validation_44358513", "origin_question": "Remember, have you seen any table(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any table(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_44358513/Validation_44358513_44358513_52644.705.png", "Validation_44358513/Validation_44358513_44358513_52646.704.png", "Validation_44358513/Validation_44358513_44358513_52648.704.png", "Validation_44358513/Validation_44358513_44358513_52650.703.png" ], "turn_id": 2, "scan_id": "Validation_44358513", "origin_question": "At the end of which turn were both of \"the sleek rectangular white fireplace made of stone-like material\" and \"the medium-sized indoor plant with vibrant green foliage\" on your left side?", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the sleek rectangular white fireplace made of stone-like material\" and \"the medium-sized indoor plant with vibrant green foliage\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_44358513/Validation_44358513_44358513_52652.719.png", "Validation_44358513/Validation_44358513_44358513_52654.718.png", "Validation_44358513/Validation_44358513_44358513_52656.717.png", "Validation_44358513/Validation_44358513_44358513_52658.716.png" ], "turn_id": 3, "scan_id": "Validation_44358513", "origin_question": "Please recall \"the medium-sized indoor plant with vibrant green foliage\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.73", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the medium-sized indoor plant with vibrant green foliage\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_44358513/Validation_44358513_44358513_52660.715.png", "Validation_44358513/Validation_44358513_44358513_52662.714.png", "Validation_44358513/Validation_44358513_44358513_52664.714.png", "Validation_44358513/Validation_44358513_44358513_52666.713.png" ], "turn_id": 4, "scan_id": "Validation_44358513", "origin_question": "When did you first discover \"the sleek rectangular white fireplace made of stone-like material\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the sleek rectangular white fireplace made of stone-like material\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_44358513/Validation_44358513_44358513_52668.712.png", "Validation_44358513/Validation_44358513_44358513_52670.711.png", "Validation_44358513/Validation_44358513_44358513_52672.710.png", "Validation_44358513/Validation_44358513_44358513_52674.709.png" ], "turn_id": 5, "scan_id": "Validation_44358513", "origin_question": "By the end of the 3 turn, were you getting closer to or farther away from these two objects: X: \"the sleek rectangular white fireplace made of stone-like material\", Y: \"the modern rectangular metal or wood lamp\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to X and farther away from Y", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 3 turn, were you getting closer to or farther away from these two objects: X: \"the sleek rectangular white fireplace made of stone-like material\", Y: \"the modern rectangular metal or wood lamp\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_44358513/Validation_44358513_44358513_52676.709.png", "Validation_44358513/Validation_44358513_44358513_52678.708.png", "Validation_44358513/Validation_44358513_44358513_52680.707.png", "Validation_44358513/Validation_44358513_44358513_52682.706.png" ], "turn_id": 6, "scan_id": "Validation_44358513", "origin_question": "In which turn did you see both \"the medium-sized indoor plant with vibrant green foliage\" and \"the modern rectangular metal or wood lamp\" simultaneously? ", "option": [], "answer": "6", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the medium-sized indoor plant with vibrant green foliage\" and \"the modern rectangular metal or wood lamp\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_44358513/Validation_44358513_44358513_52684.705.png", "Validation_44358513/Validation_44358513_44358513_52686.704.png", "Validation_44358513/Validation_44358513_44358513_52688.704.png", "Validation_44358513/Validation_44358513_44358513_52690.703.png" ], "turn_id": 7, "scan_id": "Validation_44358513", "origin_question": "How far is your current position from where you were at the end of the turn 6? (in meters)", "option": [], "answer": "2.66", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 6? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "Validation_42446103/Validation_42446103_42446103_78053.189.png", "Validation_42446103/Validation_42446103_42446103_78055.188.png", "Validation_42446103/Validation_42446103_42446103_78057.188.png", "Validation_42446103/Validation_42446103_42446103_78059.187.png", "Validation_42446103/Validation_42446103_42446103_78061.186.png" ], "turn_id": 1, "scan_id": "Validation_42446103", "origin_question": "Remember, how many chair(s) have you seen so far? ", "option": [], "answer": "3", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many chair(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_42446103/Validation_42446103_42446103_78063.185.png", "Validation_42446103/Validation_42446103_42446103_78065.184.png", "Validation_42446103/Validation_42446103_42446103_78067.184.png", "Validation_42446103/Validation_42446103_42446103_78069.183.png", "Validation_42446103/Validation_42446103_42446103_78071.182.png" ], "turn_id": 2, "scan_id": "Validation_42446103", "origin_question": "When did you first discover \"the large, sleek black monitor with the flat screen\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the large, sleek black monitor with the flat screen\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_42446103/Validation_42446103_42446103_78073.181.png", "Validation_42446103/Validation_42446103_42446103_78075.181.png", "Validation_42446103/Validation_42446103_42446103_78077.196.png", "Validation_42446103/Validation_42446103_42446103_78079.196.png", "Validation_42446103/Validation_42446103_42446103_78081.195.png" ], "turn_id": 3, "scan_id": "Validation_42446103", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the modern rectangular white ceramic sink\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the modern rectangular white ceramic sink\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_42446103/Validation_42446103_42446103_78083.194.png", "Validation_42446103/Validation_42446103_42446103_78085.193.png", "Validation_42446103/Validation_42446103_42446103_78087.192.png", "Validation_42446103/Validation_42446103_42446103_78089.192.png", "Validation_42446103/Validation_42446103_42446103_78091.191.png" ], "turn_id": 4, "scan_id": "Validation_42446103", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the large, sleek black monitor with the flat screen\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the large, sleek black monitor with the flat screen\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_42446103/Validation_42446103_42446103_78093.190.png", "Validation_42446103/Validation_42446103_42446103_78095.189.png", "Validation_42446103/Validation_42446103_42446103_78097.188.png", "Validation_42446103/Validation_42446103_42446103_78099.188.png", "Validation_42446103/Validation_42446103_42446103_78101.187.png" ], "turn_id": 5, "scan_id": "Validation_42446103", "origin_question": "By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the compact rectangular dark brown wooden stool\", Y: \"the compact cylindrical plastic bin with the white lid\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to Y and farther away from X", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the compact rectangular dark brown wooden stool\", Y: \"the compact cylindrical plastic bin with the white lid\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_42446103/Validation_42446103_42446103_78103.186.png", "Validation_42446103/Validation_42446103_42446103_78105.185.png", "Validation_42446103/Validation_42446103_42446103_78107.185.png", "Validation_42446103/Validation_42446103_42446103_78109.184.png", "Validation_42446103/Validation_42446103_42446103_78111.183.png" ], "turn_id": 6, "scan_id": "Validation_42446103", "origin_question": "Assuming the direction you are facing at the end of the turn 4 is forward, did you move a certain distance forward or backward from that position?", "option": [ "Forward", "Backward" ], "answer": "Forward", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 4 is forward, did you move a certain distance forward or backward from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Forward', 'Backward'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_42446103/Validation_42446103_42446103_78113.182.png", "Validation_42446103/Validation_42446103_42446103_78115.181.png", "Validation_42446103/Validation_42446103_42446103_78117.181.png", "Validation_42446103/Validation_42446103_42446103_78119.196.png", "Validation_42446103/Validation_42446103_42446103_78121.196.png" ], "turn_id": 7, "scan_id": "Validation_42446103", "origin_question": "Which direction is the large, sleek black monitor with the flat screen in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the large, sleek black monitor with the flat screen in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_42446103/Validation_42446103_42446103_78123.195.png", "Validation_42446103/Validation_42446103_42446103_78125.194.png", "Validation_42446103/Validation_42446103_42446103_78127.193.png", "Validation_42446103/Validation_42446103_42446103_78129.192.png", "Validation_42446103/Validation_42446103_42446103_78131.192.png" ], "turn_id": 8, "scan_id": "Validation_42446103", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the compact cylindrical plastic bin with the white lid\" from your position?", "option": [], "answer": "155.59", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the compact cylindrical plastic bin with the white lid\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_42446103/Validation_42446103_42446103_78133.191.png", "Validation_42446103/Validation_42446103_42446103_78135.190.png", "Validation_42446103/Validation_42446103_42446103_78137.189.png", "Validation_42446103/Validation_42446103_42446103_78139.189.png", "Validation_42446103/Validation_42446103_42446103_78141.188.png" ], "turn_id": 9, "scan_id": "Validation_42446103", "origin_question": "Which direction is the compact rectangular dark brown wooden stool in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the compact rectangular dark brown wooden stool in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_42446103/Validation_42446103_42446103_78143.187.png", "Validation_42446103/Validation_42446103_42446103_78145.186.png", "Validation_42446103/Validation_42446103_42446103_78147.185.png", "Validation_42446103/Validation_42446103_42446103_78149.185.png", "Validation_42446103/Validation_42446103_42446103_78151.184.png" ], "turn_id": 10, "scan_id": "Validation_42446103", "origin_question": "At the end of which turn were both of \"the large, sleek black monitor with the flat screen\" and \"the compact rectangular dark brown wooden stool\" on your left side?", "option": [], "answer": "8", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the large, sleek black monitor with the flat screen\" and \"the compact rectangular dark brown wooden stool\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47334240/Validation_47334240_47334240_111160.434.png", "Validation_47334240/Validation_47334240_47334240_111162.417.png", "Validation_47334240/Validation_47334240_47334240_111164.416.png", "Validation_47334240/Validation_47334240_47334240_111166.415.png" ], "turn_id": 1, "scan_id": "Validation_47334240", "origin_question": "Remember, have you seen any microwave(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any microwave(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "Validation_47334240/Validation_47334240_47334240_111168.414.png", "Validation_47334240/Validation_47334240_47334240_111170.430.png", "Validation_47334240/Validation_47334240_47334240_111172.429.png", "Validation_47334240/Validation_47334240_47334240_111174.428.png" ], "turn_id": 2, "scan_id": "Validation_47334240", "origin_question": "When did you first discover \"the multicolored striped cotton towel in rectangular shape\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the multicolored striped cotton towel in rectangular shape\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47334240/Validation_47334240_47334240_111176.428.png", "Validation_47334240/Validation_47334240_47334240_111178.427.png", "Validation_47334240/Validation_47334240_47334240_111180.426.png", "Validation_47334240/Validation_47334240_47334240_111182.425.png" ], "turn_id": 3, "scan_id": "Validation_47334240", "origin_question": "Remember, have you seen any cap(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any cap(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_47334240/Validation_47334240_47334240_111184.424.png", "Validation_47334240/Validation_47334240_47334240_111186.424.png", "Validation_47334240/Validation_47334240_47334240_111188.423.png", "Validation_47334240/Validation_47334240_47334240_111190.422.png" ], "turn_id": 4, "scan_id": "Validation_47334240", "origin_question": "In which turn did you see both \"the contemporary rectangular light cabinet with ample storage\" and \"the multicolored striped cotton towel in rectangular shape\" simultaneously? ", "option": [], "answer": "4", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the contemporary rectangular light cabinet with ample storage\" and \"the multicolored striped cotton towel in rectangular shape\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "Validation_47334240/Validation_47334240_47334240_111192.421.png", "Validation_47334240/Validation_47334240_47334240_111194.420.png", "Validation_47334240/Validation_47334240_47334240_111196.420.png", "Validation_47334240/Validation_47334240_47334240_111198.419.png" ], "turn_id": 5, "scan_id": "Validation_47334240", "origin_question": "Is the bathtub to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the bathtub to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "Validation_47334240/Validation_47334240_47334240_111200.418.png", "Validation_47334240/Validation_47334240_47334240_111202.417.png", "Validation_47334240/Validation_47334240_47334240_111204.416.png", "Validation_47334240/Validation_47334240_47334240_111206.415.png" ], "turn_id": 6, "scan_id": "Validation_47334240", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the modern oval white toilet made of vitreous china\"; \"the white rectangular shelf with parallel horizontal bars\"; \"an oval white ceramic sink on the light mdf vanity\" ", "option": [ "the modern oval white toilet made of vitreous china", "the white rectangular shelf with parallel horizontal bars", "an oval white ceramic sink on the light mdf vanity" ], "answer": "the white rectangular shelf with parallel horizontal bars", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the modern oval white toilet made of vitreous china\"; \"the white rectangular shelf with parallel horizontal bars\"; \"an oval white ceramic sink on the light mdf vanity\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the modern oval white toilet made of vitreous china', 'the white rectangular shelf with parallel horizontal bars', 'an oval white ceramic sink on the light mdf vanity'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "Validation_47334240/Validation_47334240_47334240_111208.415.png", "Validation_47334240/Validation_47334240_47334240_111210.414.png", "Validation_47334240/Validation_47334240_47334240_111212.430.png", "Validation_47334240/Validation_47334240_47334240_111214.429.png" ], "turn_id": 7, "scan_id": "Validation_47334240", "origin_question": "In which turn were you closest to \"an oval white ceramic sink on the light mdf vanity\"? ", "option": [], "answer": "7", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"an oval white ceramic sink on the light mdf vanity\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47334240/Validation_47334240_47334240_111216.428.png", "Validation_47334240/Validation_47334240_47334240_111218.427.png", "Validation_47334240/Validation_47334240_47334240_111220.426.png", "Validation_47334240/Validation_47334240_47334240_111222.426.png" ], "turn_id": 8, "scan_id": "Validation_47334240", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the white rectangular shelf with parallel horizontal bars\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the white rectangular shelf with parallel horizontal bars\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47334240/Validation_47334240_47334240_111224.425.png", "Validation_47334240/Validation_47334240_47334240_111226.424.png", "Validation_47334240/Validation_47334240_47334240_111228.423.png", "Validation_47334240/Validation_47334240_47334240_111230.422.png" ], "turn_id": 9, "scan_id": "Validation_47334240", "origin_question": "Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the white rectangular shelf with parallel horizontal bars\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 9 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the white rectangular shelf with parallel horizontal bars\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_41142278/Validation_41142278_41142278_4002.081.png", "Validation_41142278/Validation_41142278_41142278_4004.080.png", "Validation_41142278/Validation_41142278_41142278_4006.079.png", "Validation_41142278/Validation_41142278_41142278_4008.079.png" ], "turn_id": 1, "scan_id": "Validation_41142278", "origin_question": "Remember, have you seen any towel(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any towel(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_41142278/Validation_41142278_41142278_4010.078.png", "Validation_41142278/Validation_41142278_41142278_4012.077.png", "Validation_41142278/Validation_41142278_41142278_4014.076.png", "Validation_41142278/Validation_41142278_41142278_4016.075.png" ], "turn_id": 2, "scan_id": "Validation_41142278", "origin_question": "Remember, have you seen any toilet(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any toilet(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "Validation_41142278/Validation_41142278_41142278_4018.074.png", "Validation_41142278/Validation_41142278_41142278_4020.074.png", "Validation_41142278/Validation_41142278_41142278_4022.073.png", "Validation_41142278/Validation_41142278_41142278_4024.072.png" ], "turn_id": 3, "scan_id": "Validation_41142278", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the compact, rounded white ceramic toilet\", Y: \"the dark-toned picture with the possibly depicted image\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "farther away from both X and Y", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the compact, rounded white ceramic toilet\", Y: \"the dark-toned picture with the possibly depicted image\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "Validation_41142278/Validation_41142278_41142278_4026.071.png", "Validation_41142278/Validation_41142278_41142278_4028.070.png", "Validation_41142278/Validation_41142278_41142278_4030.069.png", "Validation_41142278/Validation_41142278_41142278_4032.069.png" ], "turn_id": 4, "scan_id": "Validation_41142278", "origin_question": "How far is your current position from where you were at the end of the turn 3? (in meters)", "option": [], "answer": "1.18", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 3? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_41142278/Validation_41142278_41142278_4034.068.png", "Validation_41142278/Validation_41142278_41142278_4036.084.png", "Validation_41142278/Validation_41142278_41142278_4038.083.png", "Validation_41142278/Validation_41142278_41142278_4040.082.png" ], "turn_id": 5, "scan_id": "Validation_41142278", "origin_question": "", "option": [], "answer": "", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_41142278/Validation_41142278_41142278_4042.081.png", "Validation_41142278/Validation_41142278_41142278_4044.080.png", "Validation_41142278/Validation_41142278_41142278_4046.079.png", "Validation_41142278/Validation_41142278_41142278_4048.079.png" ], "turn_id": 6, "scan_id": "Validation_41142278", "origin_question": "Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "137.4", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "Validation_41142278/Validation_41142278_41142278_4050.078.png", "Validation_41142278/Validation_41142278_41142278_4052.077.png", "Validation_41142278/Validation_41142278_41142278_4054.076.png", "Validation_41142278/Validation_41142278_41142278_4056.075.png" ], "turn_id": 7, "scan_id": "Validation_41142278", "origin_question": "In which turn did you see both \"the bathtub\" and \"the compact, rounded white ceramic toilet\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the bathtub\" and \"the compact, rounded white ceramic toilet\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_41142278/Validation_41142278_41142278_4058.075.png", "Validation_41142278/Validation_41142278_41142278_4060.074.png", "Validation_41142278/Validation_41142278_41142278_4062.073.png", "Validation_41142278/Validation_41142278_41142278_4064.072.png" ], "turn_id": 8, "scan_id": "Validation_41142278", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the rectangular, off-white bar of soap\" from your position?", "option": [], "answer": "132.64", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the rectangular, off-white bar of soap\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_42446529/Validation_42446529_42446529_198722.862.png", "Validation_42446529/Validation_42446529_42446529_198724.844.png", "Validation_42446529/Validation_42446529_42446529_198726.843.png", "Validation_42446529/Validation_42446529_42446529_198728.843.png" ], "turn_id": 1, "scan_id": "Validation_42446529", "origin_question": "Remember, have you seen any window(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any window(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_42446529/Validation_42446529_42446529_198730.842.png", "Validation_42446529/Validation_42446529_42446529_198732.841.png", "Validation_42446529/Validation_42446529_42446529_198734.840.png", "Validation_42446529/Validation_42446529_42446529_198736.839.png" ], "turn_id": 2, "scan_id": "Validation_42446529", "origin_question": "At the end of which turn were both of \"the bed\" and \"the rectangular window with the white frame and dark curtains\" on your left side?", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the bed\" and \"the rectangular window with the white frame and dark curtains\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_42446529/Validation_42446529_42446529_198738.839.png", "Validation_42446529/Validation_42446529_42446529_198740.838.png", "Validation_42446529/Validation_42446529_42446529_198742.837.png", "Validation_42446529/Validation_42446529_42446529_198744.836.png" ], "turn_id": 3, "scan_id": "Validation_42446529", "origin_question": "Please recall \"the rectangular window with the white frame and dark curtains\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.01", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the rectangular window with the white frame and dark curtains\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "Validation_42446529/Validation_42446529_42446529_198746.835.png", "Validation_42446529/Validation_42446529_42446529_198748.851.png", "Validation_42446529/Validation_42446529_42446529_198750.850.png", "Validation_42446529/Validation_42446529_42446529_198752.849.png" ], "turn_id": 4, "scan_id": "Validation_42446529", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the soft pastel blue rectangular pillow\" from your position?", "option": [], "answer": "103.34", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the soft pastel blue rectangular pillow\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "Validation_42446529/Validation_42446529_42446529_198754.849.png", "Validation_42446529/Validation_42446529_42446529_198756.848.png", "Validation_42446529/Validation_42446529_42446529_198758.847.png", "Validation_42446529/Validation_42446529_42446529_198760.846.png" ], "turn_id": 5, "scan_id": "Validation_42446529", "origin_question": "What will be the first-time appearance order of the following categories: cabinet, chair, pillow?", "option": [ "chair, pillow, cabinet", "pillow, cabinet, chair", "pillow, chair, cabinet", "cabinet, chair, pillow" ], "answer": "pillow, cabinet, chair", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: cabinet, chair, pillow? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['chair, pillow, cabinet', 'pillow, cabinet, chair', 'pillow, chair, cabinet', 'cabinet, chair, pillow'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_42446529/Validation_42446529_42446529_198762.845.png", "Validation_42446529/Validation_42446529_42446529_198764.845.png", "Validation_42446529/Validation_42446529_42446529_198766.844.png", "Validation_42446529/Validation_42446529_42446529_198768.843.png" ], "turn_id": 6, "scan_id": "Validation_42446529", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the light wood-toned rectangular table with the sleek design\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the light wood-toned rectangular table with the sleek design\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_42446529/Validation_42446529_42446529_198770.842.png", "Validation_42446529/Validation_42446529_42446529_198772.841.png", "Validation_42446529/Validation_42446529_42446529_198774.840.png", "Validation_42446529/Validation_42446529_42446529_198776.840.png" ], "turn_id": 7, "scan_id": "Validation_42446529", "origin_question": "Which direction is the rectangular window with the white frame and dark curtains in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the rectangular window with the white frame and dark curtains in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "Validation_42446529/Validation_42446529_42446529_198778.839.png", "Validation_42446529/Validation_42446529_42446529_198780.838.png", "Validation_42446529/Validation_42446529_42446529_198782.837.png", "Validation_42446529/Validation_42446529_42446529_198784.836.png" ], "turn_id": 8, "scan_id": "Validation_42446529", "origin_question": "Please recall \"the tall stainless steel rack with the sleek silver finish\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.52", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the tall stainless steel rack with the sleek silver finish\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "Validation_42446529/Validation_42446529_42446529_198786.835.png", "Validation_42446529/Validation_42446529_42446529_198788.851.png", "Validation_42446529/Validation_42446529_42446529_198790.850.png", "Validation_42446529/Validation_42446529_42446529_198792.850.png" ], "turn_id": 9, "scan_id": "Validation_42446529", "origin_question": "Which direction is the rectangular window with the white frame and dark curtains in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 9 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the rectangular window with the white frame and dark curtains in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "Validation_47895371/Validation_47895371_47895371_101971.460.png", "Validation_47895371/Validation_47895371_47895371_101973.442.png", "Validation_47895371/Validation_47895371_47895371_101975.441.png", "Validation_47895371/Validation_47895371_47895371_101977.441.png", "Validation_47895371/Validation_47895371_47895371_101979.440.png" ], "turn_id": 1, "scan_id": "Validation_47895371", "origin_question": "Remember, have you seen any mailbox(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any mailbox(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "Validation_47895371/Validation_47895371_47895371_101981.439.png", "Validation_47895371/Validation_47895371_47895371_101983.438.png", "Validation_47895371/Validation_47895371_47895371_101985.437.png", "Validation_47895371/Validation_47895371_47895371_101987.437.png", "Validation_47895371/Validation_47895371_47895371_101989.436.png" ], "turn_id": 2, "scan_id": "Validation_47895371", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "Validation_47895371/Validation_47895371_47895371_101991.435.png", "Validation_47895371/Validation_47895371_47895371_101993.434.png", "Validation_47895371/Validation_47895371_47895371_101995.433.png", "Validation_47895371/Validation_47895371_47895371_101997.432.png", "Validation_47895371/Validation_47895371_47895371_101999.432.png" ], "turn_id": 3, "scan_id": "Validation_47895371", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the sleek rectangular couch with gray and white hues\"; \"the rectangular white metal radiator\"; \"the warm-toned wooden door with frosted glass panels\" ", "option": [ "the sleek rectangular couch with gray and white hues", "the rectangular white metal radiator", "the warm-toned wooden door with frosted glass panels" ], "answer": "the rectangular white metal radiator", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the sleek rectangular couch with gray and white hues\"; \"the rectangular white metal radiator\"; \"the warm-toned wooden door with frosted glass panels\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the sleek rectangular couch with gray and white hues', 'the rectangular white metal radiator', 'the warm-toned wooden door with frosted glass panels'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "Validation_47895371/Validation_47895371_47895371_102001.431.png", "Validation_47895371/Validation_47895371_47895371_102003.430.png", "Validation_47895371/Validation_47895371_47895371_102005.429.png", "Validation_47895371/Validation_47895371_47895371_102007.428.png", "Validation_47895371/Validation_47895371_47895371_102009.428.png" ], "turn_id": 4, "scan_id": "Validation_47895371", "origin_question": "Assuming the direction you are facing at the end of the turn 3 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Left", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 3 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "Validation_47895371/Validation_47895371_47895371_102011.427.png", "Validation_47895371/Validation_47895371_47895371_102013.443.png", "Validation_47895371/Validation_47895371_47895371_102015.442.png", "Validation_47895371/Validation_47895371_47895371_102017.441.png", "Validation_47895371/Validation_47895371_47895371_102019.440.png" ], "turn_id": 5, "scan_id": "Validation_47895371", "origin_question": "Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the rectangular white metal radiator\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the rectangular white metal radiator\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "Validation_47895371/Validation_47895371_47895371_102021.439.png", "Validation_47895371/Validation_47895371_47895371_102023.439.png", "Validation_47895371/Validation_47895371_47895371_102025.438.png", "Validation_47895371/Validation_47895371_47895371_102027.437.png", "Validation_47895371/Validation_47895371_47895371_102029.436.png" ], "turn_id": 6, "scan_id": "Validation_47895371", "origin_question": "At the end of which turn were both of \"the sleek rectangular couch with gray and white hues\" and \"the cylindrical silver-gray bin\" on your left side?", "option": [], "answer": "1", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the sleek rectangular couch with gray and white hues\" and \"the cylindrical silver-gray bin\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_47895371/Validation_47895371_47895371_102031.435.png", "Validation_47895371/Validation_47895371_47895371_102033.435.png", "Validation_47895371/Validation_47895371_47895371_102035.434.png", "Validation_47895371/Validation_47895371_47895371_102037.433.png", "Validation_47895371/Validation_47895371_47895371_102039.432.png" ], "turn_id": 7, "scan_id": "Validation_47895371", "origin_question": "Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "63.46", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "Validation_47895371/Validation_47895371_47895371_102041.431.png", "Validation_47895371/Validation_47895371_47895371_102043.431.png", "Validation_47895371/Validation_47895371_47895371_102045.430.png", "Validation_47895371/Validation_47895371_47895371_102047.429.png", "Validation_47895371/Validation_47895371_47895371_102049.428.png" ], "turn_id": 8, "scan_id": "Validation_47895371", "origin_question": "Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_47895371/Validation_47895371_47895371_102051.427.png", "Validation_47895371/Validation_47895371_47895371_102053.443.png", "Validation_47895371/Validation_47895371_47895371_102055.442.png", "Validation_47895371/Validation_47895371_47895371_102057.442.png", "Validation_47895371/Validation_47895371_47895371_102059.441.png" ], "turn_id": 9, "scan_id": "Validation_47895371", "origin_question": "Using your orientation at the end of turn 7 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "39.83", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 7 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "Validation_47895371/Validation_47895371_47895371_102061.440.png", "Validation_47895371/Validation_47895371_47895371_102063.439.png", "Validation_47895371/Validation_47895371_47895371_102065.438.png", "Validation_47895371/Validation_47895371_47895371_102067.437.png", "Validation_47895371/Validation_47895371_47895371_102069.437.png" ], "turn_id": 10, "scan_id": "Validation_47895371", "origin_question": "Using your orientation at the end of turn 9 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "59.82", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 9 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0500_01/00000.jpg", "scene0500_01/00040.jpg", "scene0500_01/00110.jpg", "scene0500_01/00150.jpg", "scene0500_01/00190.jpg" ], "turn_id": 1, "scan_id": "scene0500_01", "origin_question": "Remember, how many table(s) have you seen so far? ", "option": [], "answer": "2", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many table(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0500_01/00230.jpg", "scene0500_01/00270.jpg", "scene0500_01/00310.jpg", "scene0500_01/00350.jpg", "scene0500_01/00390.jpg" ], "turn_id": 2, "scan_id": "scene0500_01", "origin_question": "In which turn were you farthest from \"the large rectangular blackboard\"? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the large rectangular blackboard\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "scene0500_01/00430.jpg", "scene0500_01/00470.jpg", "scene0500_01/00510.jpg", "scene0500_01/00570.jpg", "scene0500_01/00610.jpg" ], "turn_id": 3, "scan_id": "scene0500_01", "origin_question": "Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Left", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0500_01/00650.jpg", "scene0500_01/00690.jpg", "scene0500_01/00730.jpg", "scene0500_01/00770.jpg", "scene0500_01/00810.jpg" ], "turn_id": 4, "scan_id": "scene0500_01", "origin_question": "", "option": [], "answer": "", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0500_01/00850.jpg", "scene0500_01/00890.jpg", "scene0500_01/00930.jpg", "scene0500_01/00970.jpg", "scene0500_01/01010.jpg" ], "turn_id": 5, "scan_id": "scene0500_01", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "3.38", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0500_01/01050.jpg", "scene0500_01/01090.jpg", "scene0500_01/01130.jpg", "scene0500_01/01170.jpg", "scene0500_01/01210.jpg" ], "turn_id": 6, "scan_id": "scene0500_01", "origin_question": "Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0500_01/01250.jpg", "scene0500_01/01290.jpg", "scene0500_01/01330.jpg", "scene0500_01/01370.jpg", "scene0500_01/01410.jpg" ], "turn_id": 7, "scan_id": "scene0500_01", "origin_question": "What will be the first-time appearance order of the following categories: bin, blackboard, window?", "option": [ "blackboard, bin, window", "bin, window, blackboard", "blackboard, window, bin", "bin, blackboard, window" ], "answer": "blackboard, window, bin", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bin, blackboard, window? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['blackboard, bin, window', 'bin, window, blackboard', 'blackboard, window, bin', 'bin, blackboard, window'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0500_01/01450.jpg", "scene0500_01/01490.jpg", "scene0500_01/01530.jpg", "scene0500_01/01570.jpg", "scene0500_01/01610.jpg" ], "turn_id": 8, "scan_id": "scene0500_01", "origin_question": "Using your orientation at the end of turn 6 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "148.54", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 6 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0030_region4/7720ca137cc7414bae45ad8e534c4dd0_i1_4.jpg", "1mp3d_0030_region4/7720ca137cc7414bae45ad8e534c4dd0_i2_4.jpg", "1mp3d_0030_region4/a358b83b1c6f4ee3bbabb66ac93be11b_i1_4.jpg", "1mp3d_0030_region4/a358b83b1c6f4ee3bbabb66ac93be11b_i2_4.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0030_region4", "origin_question": "Remember, have you seen any mirror(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any mirror(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "1mp3d_0030_region4/a358b83b1c6f4ee3bbabb66ac93be11b_i1_3.jpg", "1mp3d_0030_region4/a358b83b1c6f4ee3bbabb66ac93be11b_i2_3.jpg", "1mp3d_0030_region4/7720ca137cc7414bae45ad8e534c4dd0_i1_3.jpg", "1mp3d_0030_region4/7720ca137cc7414bae45ad8e534c4dd0_i2_3.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0030_region4", "origin_question": "Please recall \"the dark rectangular scale with a display screen\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.11", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the dark rectangular scale with a display screen\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "1mp3d_0030_region4/d3e02cae7b9747529a760845ab75a688_i1_2.jpg", "1mp3d_0030_region4/d3e02cae7b9747529a760845ab75a688_i2_2.jpg", "1mp3d_0030_region4/fdea111fb87245c1bd4fb377b2619c5b_i1_1.jpg", "1mp3d_0030_region4/fdea111fb87245c1bd4fb377b2619c5b_i2_1.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0030_region4", "origin_question": "Among these three objects(\"the rectangular mirror with a white border\",\"the rectangular white lightweight fabric curtain\",\"the moderate-sized white toilet with a round seat\" ), which one is the farthest from you?", "option": [ "the rectangular mirror with a white border", "the rectangular white lightweight fabric curtain", "the moderate-sized white toilet with a round seat" ], "answer": "the rectangular white lightweight fabric curtain", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects(\"the rectangular mirror with a white border\",\"the rectangular white lightweight fabric curtain\",\"the moderate-sized white toilet with a round seat\" ), which one is the farthest from you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular mirror with a white border', 'the rectangular white lightweight fabric curtain', 'the moderate-sized white toilet with a round seat'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "1mp3d_0030_region4/d3e02cae7b9747529a760845ab75a688_i1_3.jpg", "1mp3d_0030_region4/d3e02cae7b9747529a760845ab75a688_i2_3.jpg", "1mp3d_0030_region4/d3e02cae7b9747529a760845ab75a688_i1_4.jpg", "1mp3d_0030_region4/d3e02cae7b9747529a760845ab75a688_i2_4.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0030_region4", "origin_question": "At the end of which turn were both of \"the rectangular white lightweight fabric curtain\" and \"the large classic oval white porcelain bathtub\" on your right side?", "option": [], "answer": "3", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the rectangular white lightweight fabric curtain\" and \"the large classic oval white porcelain bathtub\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "1mp3d_0030_region4/d3e02cae7b9747529a760845ab75a688_i1_5.jpg", "1mp3d_0030_region4/d3e02cae7b9747529a760845ab75a688_i2_5.jpg", "1mp3d_0030_region4/fdea111fb87245c1bd4fb377b2619c5b_i2_5.jpg", "1mp3d_0030_region4/fdea111fb87245c1bd4fb377b2619c5b_i1_4.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0030_region4", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "1.12", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0428_01/00000.jpg", "scene0428_01/00030.jpg", "scene0428_01/00060.jpg", "scene0428_01/00090.jpg" ], "turn_id": 1, "scan_id": "scene0428_01", "origin_question": "Remember, have you seen any bathtub(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any bathtub(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0428_01/00120.jpg", "scene0428_01/00150.jpg", "scene0428_01/00180.jpg", "scene0428_01/00210.jpg" ], "turn_id": 2, "scan_id": "scene0428_01", "origin_question": "In which turn were you farthest from \"the mirror\"? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the mirror\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0428_01/00240.jpg", "scene0428_01/00270.jpg", "scene0428_01/00300.jpg", "scene0428_01/00330.jpg" ], "turn_id": 3, "scan_id": "scene0428_01", "origin_question": "Remember, have you seen any excercise equipment(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any excercise equipment(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0428_01/00360.jpg", "scene0428_01/00390.jpg", "scene0428_01/00420.jpg", "scene0428_01/00450.jpg" ], "turn_id": 4, "scan_id": "scene0428_01", "origin_question": "When did you first discover \"the mirror\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the mirror\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0428_01/00480.jpg", "scene0428_01/00510.jpg", "scene0428_01/00540.jpg", "scene0428_01/00570.jpg" ], "turn_id": 5, "scan_id": "scene0428_01", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0428_01/00600.jpg", "scene0428_01/00630.jpg", "scene0428_01/00660.jpg", "scene0428_01/00690.jpg" ], "turn_id": 6, "scan_id": "scene0428_01", "origin_question": "In which turn were you closest to \"the mirror\"? ", "option": [], "answer": "1", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the mirror\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0448_02/00000.jpg", "scene0448_02/00040.jpg", "scene0448_02/00080.jpg", "scene0448_02/00120.jpg", "scene0448_02/00160.jpg" ], "turn_id": 1, "scan_id": "scene0448_02", "origin_question": "Remember, have you seen any printer(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any printer(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0448_02/00200.jpg", "scene0448_02/00240.jpg", "scene0448_02/00280.jpg", "scene0448_02/00320.jpg", "scene0448_02/00360.jpg" ], "turn_id": 2, "scan_id": "scene0448_02", "origin_question": "In which turn did you see both \"the medium-sized red plastic cylindrical cup\" and \"the cylindrical white bin with a plastic liner\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the medium-sized red plastic cylindrical cup\" and \"the cylindrical white bin with a plastic liner\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0448_02/00400.jpg", "scene0448_02/00440.jpg", "scene0448_02/00490.jpg", "scene0448_02/00530.jpg", "scene0448_02/00570.jpg" ], "turn_id": 3, "scan_id": "scene0448_02", "origin_question": "At the end of which turn were both of \"the round, off-white granite or marble sink\" and \"the soft, white rectangular towel made of cotton\" on your left side?", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the round, off-white granite or marble sink\" and \"the soft, white rectangular towel made of cotton\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0448_02/00610.jpg", "scene0448_02/00650.jpg", "scene0448_02/00690.jpg", "scene0448_02/00730.jpg", "scene0448_02/00770.jpg" ], "turn_id": 4, "scan_id": "scene0448_02", "origin_question": "Is the medium-sized red plastic cylindrical cup to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the medium-sized red plastic cylindrical cup to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0448_02/00810.jpg", "scene0448_02/00850.jpg", "scene0448_02/00890.jpg", "scene0448_02/00930.jpg", "scene0448_02/00970.jpg" ], "turn_id": 5, "scan_id": "scene0448_02", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0448_02/01010.jpg", "scene0448_02/01050.jpg", "scene0448_02/01090.jpg", "scene0448_02/01130.jpg", "scene0448_02/01170.jpg" ], "turn_id": 6, "scan_id": "scene0448_02", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the long, gray and white geometric curtain\"; \"the white, oval-shaped toilet\"; \"the window\" ", "option": [ "the long, gray and white geometric curtain", "the white, oval-shaped toilet", "the window" ], "answer": "the window", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the long, gray and white geometric curtain\"; \"the white, oval-shaped toilet\"; \"the window\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the long, gray and white geometric curtain', 'the white, oval-shaped toilet', 'the window'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0448_02/01210.jpg", "scene0448_02/01250.jpg", "scene0448_02/01290.jpg", "scene0448_02/01330.jpg", "scene0448_02/01370.jpg" ], "turn_id": 7, "scan_id": "scene0448_02", "origin_question": "What will be the first-time appearance order of the following categories: door, sink, towel?", "option": [ "towel, sink, door", "towel, door, sink", "door, towel, sink", "sink, door, towel" ], "answer": "sink, door, towel", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: door, sink, towel? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['towel, sink, door', 'towel, door, sink', 'door, towel, sink', 'sink, door, towel'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0448_02/01410.jpg", "scene0448_02/01450.jpg", "scene0448_02/01490.jpg", "scene0448_02/01530.jpg", "scene0448_02/01570.jpg" ], "turn_id": 8, "scan_id": "scene0448_02", "origin_question": "Using your orientation at the end of turn 7 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 7 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "scene0448_02/01610.jpg", "scene0448_02/01650.jpg", "scene0448_02/01690.jpg", "scene0448_02/01760.jpg", "scene0448_02/01800.jpg" ], "turn_id": 9, "scan_id": "scene0448_02", "origin_question": "Among these three objects(\"the cylindrical blue and white plastic toothpaste tube\",\"the long, gray and white geometric curtain\",\"the classic circular wooden table with curved legs\" ), which one is the closest to you?", "option": [ "the cylindrical blue and white plastic toothpaste tube", "the long, gray and white geometric curtain", "the classic circular wooden table with curved legs" ], "answer": "the long, gray and white geometric curtain", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the cylindrical blue and white plastic toothpaste tube\",\"the long, gray and white geometric curtain\",\"the classic circular wooden table with curved legs\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the cylindrical blue and white plastic toothpaste tube', 'the long, gray and white geometric curtain', 'the classic circular wooden table with curved legs'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0448_02/01840.jpg", "scene0448_02/01880.jpg", "scene0448_02/01920.jpg", "scene0448_02/01960.jpg", "scene0448_02/02000.jpg" ], "turn_id": 10, "scan_id": "scene0448_02", "origin_question": "Is the classic circular wooden table with curved legs to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the classic circular wooden table with curved legs to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0047_region18/5218b81977034823899188bbcef96257_i1_1.jpg", "1mp3d_0047_region18/5218b81977034823899188bbcef96257_i2_1.jpg", "1mp3d_0047_region18/2c65818beeee40af97d3fce96a3db303_i1_5.jpg", "1mp3d_0047_region18/2c65818beeee40af97d3fce96a3db303_i2_5.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0047_region18", "origin_question": "Remember, have you seen any jacket(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any jacket(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0047_region18/d364091f55a84c36ae76f8f626adceb1_i1_2.jpg", "1mp3d_0047_region18/d364091f55a84c36ae76f8f626adceb1_i2_2.jpg", "1mp3d_0047_region18/47ce431b07cc499db2004559398948ed_i1_0.jpg", "1mp3d_0047_region18/47ce431b07cc499db2004559398948ed_i2_0.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0047_region18", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "1mp3d_0047_region18/35f681db472f4a899a4096b76096861c_i2_1.jpg", "1mp3d_0047_region18/35f681db472f4a899a4096b76096861c_i1_0.jpg", "1mp3d_0047_region18/35f681db472f4a899a4096b76096861c_i2_0.jpg", "1mp3d_0047_region18/455c874aa7b949a2bf202a683e1061de_i1_0.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0047_region18", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "1.81", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "1mp3d_0047_region18/455c874aa7b949a2bf202a683e1061de_i2_0.jpg", "1mp3d_0047_region18/b2295a9f6d6449b3901838eb6ba109ff_i1_3.jpg", "1mp3d_0047_region18/b2295a9f6d6449b3901838eb6ba109ff_i2_3.jpg", "1mp3d_0047_region18/e1acc1e4419f42e6ba0e51c42e76aab4_i1_3.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0047_region18", "origin_question": "Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Right", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "1mp3d_0072_region2/88fda52a9c9e4a5cb1774e741ecce8f6_i1_1.jpg", "1mp3d_0072_region2/88fda52a9c9e4a5cb1774e741ecce8f6_i2_1.jpg", "1mp3d_0072_region2/88fda52a9c9e4a5cb1774e741ecce8f6_i1_2.jpg", "1mp3d_0072_region2/88fda52a9c9e4a5cb1774e741ecce8f6_i2_2.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0072_region2", "origin_question": "Remember, how many carpet(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many carpet(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 3 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "1mp3d_0072_region2/5ee3f3c0dac54a1893594c1fde6a427a_i1_0.jpg", "1mp3d_0072_region2/5ee3f3c0dac54a1893594c1fde6a427a_i2_0.jpg", "1mp3d_0072_region2/bed1e142845343c5b9695eabad5d170e_i1_3.jpg", "1mp3d_0072_region2/bed1e142845343c5b9695eabad5d170e_i2_3.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0072_region2", "origin_question": "In which turn did you see both \"the circular black vanity with a wooden base\" and \"the circular, silver stainless steel shower\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the circular black vanity with a wooden base\" and \"the circular, silver stainless steel shower\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 3 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "1mp3d_0072_region2/5ee3f3c0dac54a1893594c1fde6a427a_i1_5.jpg", "1mp3d_0072_region2/5ee3f3c0dac54a1893594c1fde6a427a_i2_5.jpg", "1mp3d_0072_region2/bed1e142845343c5b9695eabad5d170e_i1_2.jpg", "1mp3d_0072_region2/bed1e142845343c5b9695eabad5d170e_i2_2.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0072_region2", "origin_question": "When did you first discover \"the white hand towel made of soft, absorbent material\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the white hand towel made of soft, absorbent material\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 3 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0516_01/00000.jpg", "scene0516_01/00040.jpg", "scene0516_01/00080.jpg", "scene0516_01/00120.jpg", "scene0516_01/00160.jpg" ], "turn_id": 1, "scan_id": "scene0516_01", "origin_question": "Remember, have you seen any printer(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any printer(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0516_01/00200.jpg", "scene0516_01/00240.jpg", "scene0516_01/00280.jpg", "scene0516_01/00320.jpg", "scene0516_01/00360.jpg" ], "turn_id": 2, "scan_id": "scene0516_01", "origin_question": "In which turn did you see both \"the long white thin curtain made of fabric\" and \"the medium-sized blue cylindrical plastic bucket\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the long white thin curtain made of fabric\" and \"the medium-sized blue cylindrical plastic bucket\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0516_01/00400.jpg", "scene0516_01/00440.jpg", "scene0516_01/00480.jpg", "scene0516_01/00520.jpg", "scene0516_01/00560.jpg" ], "turn_id": 3, "scan_id": "scene0516_01", "origin_question": "Which direction is the stairs in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the stairs in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0516_01/00600.jpg", "scene0516_01/00640.jpg", "scene0516_01/00680.jpg", "scene0516_01/00720.jpg", "scene0516_01/00760.jpg" ], "turn_id": 4, "scan_id": "scene0516_01", "origin_question": "When did you first discover \"the rectangular white porcelain sink\"? (index of the turn)", "option": [], "answer": "4", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the rectangular white porcelain sink\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0516_01/00800.jpg", "scene0516_01/00840.jpg", "scene0516_01/00880.jpg", "scene0516_01/00920.jpg", "scene0516_01/00960.jpg" ], "turn_id": 5, "scan_id": "scene0516_01", "origin_question": "What will be the first-time appearance order of the following categories: door, mirror, stairs?", "option": [ "mirror, stairs, door", "stairs, mirror, door", "mirror, door, stairs", "door, stairs, mirror" ], "answer": "door, stairs, mirror", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: door, mirror, stairs? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['mirror, stairs, door', 'stairs, mirror, door', 'mirror, door, stairs', 'door, stairs, mirror'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0516_01/01000.jpg", "scene0516_01/01040.jpg", "scene0516_01/01080.jpg", "scene0516_01/01120.jpg", "scene0516_01/01160.jpg" ], "turn_id": 6, "scan_id": "scene0516_01", "origin_question": "What will be the first-time appearance order of the following categories: door, sink, stairs?", "option": [ "door, sink, stairs", "stairs, sink, door", "stairs, door, sink", "door, stairs, sink" ], "answer": "door, stairs, sink", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: door, sink, stairs? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['door, sink, stairs', 'stairs, sink, door', 'stairs, door, sink', 'door, stairs, sink'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0516_01/01200.jpg", "scene0516_01/01240.jpg", "scene0516_01/01280.jpg", "scene0516_01/01320.jpg", "scene0516_01/01360.jpg" ], "turn_id": 7, "scan_id": "scene0516_01", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the shelf\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the shelf\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0516_01/01400.jpg", "scene0516_01/01440.jpg", "scene0516_01/01480.jpg", "scene0516_01/01520.jpg", "scene0516_01/01560.jpg" ], "turn_id": 8, "scan_id": "scene0516_01", "origin_question": "Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the medium-sized blue cylindrical plastic bucket\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the medium-sized blue cylindrical plastic bucket\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0516_01/01600.jpg", "scene0516_01/01640.jpg", "scene0516_01/01680.jpg", "scene0516_01/01720.jpg", "scene0516_01/01760.jpg" ], "turn_id": 9, "scan_id": "scene0516_01", "origin_question": "Which two objects are on the same side of you now? \"the dark gray metal rail of average length\", \"the medium-sized blue cylindrical plastic bucket\", \"the standard white ceramic toilet with a round oval shape\". ", "option": [ "the dark gray metal rail of average length, the medium-sized blue cylindrical plastic bucket", "the dark gray metal rail of average length, the standard white ceramic toilet with a round oval shape", "the medium-sized blue cylindrical plastic bucket, the standard white ceramic toilet with a round oval shape" ], "answer": "the dark gray metal rail of average length, the medium-sized blue cylindrical plastic bucket", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the dark gray metal rail of average length\", \"the medium-sized blue cylindrical plastic bucket\", \"the standard white ceramic toilet with a round oval shape\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the dark gray metal rail of average length, the medium-sized blue cylindrical plastic bucket', 'the dark gray metal rail of average length, the standard white ceramic toilet with a round oval shape', 'the medium-sized blue cylindrical plastic bucket, the standard white ceramic toilet with a round oval shape'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0516_01/01800.jpg", "scene0516_01/01840.jpg", "scene0516_01/01880.jpg", "scene0516_01/01920.jpg", "scene0516_01/01960.jpg" ], "turn_id": 10, "scan_id": "scene0516_01", "origin_question": "Which direction is the standard white ceramic toilet with a round oval shape in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the standard white ceramic toilet with a round oval shape in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0541_00/00000.jpg", "scene0541_00/00040.jpg", "scene0541_00/00080.jpg", "scene0541_00/00120.jpg" ], "turn_id": 1, "scan_id": "scene0541_00", "origin_question": "Remember, have you seen any towel(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any towel(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0541_00/00160.jpg", "scene0541_00/00200.jpg", "scene0541_00/00240.jpg", "scene0541_00/00280.jpg" ], "turn_id": 2, "scan_id": "scene0541_00", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the mirror\"; \"the rectangular gray and white striped pillow\"; \"the rectangular single bed with a brown and white checkered bedsheet\" ", "option": [ "the mirror", "the rectangular gray and white striped pillow", "the rectangular single bed with a brown and white checkered bedsheet" ], "answer": "the mirror", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the mirror\"; \"the rectangular gray and white striped pillow\"; \"the rectangular single bed with a brown and white checkered bedsheet\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the mirror', 'the rectangular gray and white striped pillow', 'the rectangular single bed with a brown and white checkered bedsheet'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0541_00/00320.jpg", "scene0541_00/00360.jpg", "scene0541_00/00400.jpg", "scene0541_00/00440.jpg" ], "turn_id": 3, "scan_id": "scene0541_00", "origin_question": "Which direction is the modern lamp with a tilted cylindrical shade in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the modern lamp with a tilted cylindrical shade in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0541_00/00480.jpg", "scene0541_00/00520.jpg", "scene0541_00/00560.jpg", "scene0541_00/00600.jpg" ], "turn_id": 4, "scan_id": "scene0541_00", "origin_question": "Which two objects are on the same side of you now? \"the standard-sized hardcover book with a beige cover\", \"the mirror\", \"the rectangular white board with a blue cover\". ", "option": [ "the standard-sized hardcover book with a beige cover, the mirror", "the standard-sized hardcover book with a beige cover, the rectangular white board with a blue cover", "the mirror, the rectangular white board with a blue cover" ], "answer": "the standard-sized hardcover book with a beige cover, the mirror", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the standard-sized hardcover book with a beige cover\", \"the mirror\", \"the rectangular white board with a blue cover\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the standard-sized hardcover book with a beige cover, the mirror', 'the standard-sized hardcover book with a beige cover, the rectangular white board with a blue cover', 'the mirror, the rectangular white board with a blue cover'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0541_00/00640.jpg", "scene0541_00/00680.jpg", "scene0541_00/00720.jpg", "scene0541_00/00760.jpg" ], "turn_id": 5, "scan_id": "scene0541_00", "origin_question": "At the end of which turn were both of \"the jar\" and \"the brown wooden door\" on your right side?", "option": [], "answer": "3", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the jar\" and \"the brown wooden door\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0541_00/00800.jpg", "scene0541_00/00840.jpg", "scene0541_00/00880.jpg", "scene0541_00/00920.jpg" ], "turn_id": 6, "scan_id": "scene0541_00", "origin_question": "What will be the first-time appearance order of the following categories: book, carpet, picture?", "option": [ "book, carpet, picture", "picture, book, carpet", "carpet, picture, book", "carpet, book, picture" ], "answer": "carpet, picture, book", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: book, carpet, picture? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['book, carpet, picture', 'picture, book, carpet', 'carpet, picture, book', 'carpet, book, picture'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0541_00/00960.jpg", "scene0541_00/01000.jpg", "scene0541_00/01040.jpg", "scene0541_00/01080.jpg" ], "turn_id": 7, "scan_id": "scene0541_00", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the jar\" from your position?", "option": [], "answer": "107.25", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the jar\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "scene0541_00/01120.jpg", "scene0541_00/01160.jpg", "scene0541_00/01200.jpg", "scene0541_00/01240.jpg" ], "turn_id": 8, "scan_id": "scene0541_00", "origin_question": "Among these three objects(\"the jar\",\"the mirror\",\"the brown wooden door\" ), which one is the closest to you?", "option": [ "the jar", "the mirror", "the brown wooden door" ], "answer": "the brown wooden door", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects(\"the jar\",\"the mirror\",\"the brown wooden door\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the jar', 'the mirror', 'the brown wooden door'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0610_00/00000.jpg", "scene0610_00/00040.jpg", "scene0610_00/00080.jpg", "scene0610_00/00120.jpg", "scene0610_00/00160.jpg" ], "turn_id": 1, "scan_id": "scene0610_00", "origin_question": "Remember, have you seen any bottle(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any bottle(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0610_00/00200.jpg", "scene0610_00/00240.jpg", "scene0610_00/00280.jpg", "scene0610_00/00320.jpg", "scene0610_00/00360.jpg" ], "turn_id": 2, "scan_id": "scene0610_00", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the fluffy white cotton towel\" from your position?", "option": [], "answer": "166.44", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the fluffy white cotton towel\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0610_00/00400.jpg", "scene0610_00/00440.jpg", "scene0610_00/00480.jpg", "scene0610_00/00520.jpg", "scene0610_00/00940.jpg" ], "turn_id": 3, "scan_id": "scene0610_00", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the door\"; \"the fluffy white cotton towel\"; \"the switch\" ", "option": [ "the door", "the fluffy white cotton towel", "the switch" ], "answer": "the switch", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the door\"; \"the fluffy white cotton towel\"; \"the switch\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the door', 'the fluffy white cotton towel', 'the switch'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0610_00/00980.jpg", "scene0610_00/01020.jpg", "scene0610_00/01060.jpg", "scene0610_00/01100.jpg", "scene0610_00/01140.jpg" ], "turn_id": 4, "scan_id": "scene0610_00", "origin_question": "At the end of which turn were both of \"the standard white rectangular ceramic bathtub\" and \"the dark brown ceramic vanity with a white sink\" on your right side?", "option": [], "answer": "3", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the standard white rectangular ceramic bathtub\" and \"the dark brown ceramic vanity with a white sink\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0610_00/01180.jpg", "scene0610_00/01220.jpg", "scene0610_00/01260.jpg", "scene0610_00/01300.jpg", "scene0610_00/01340.jpg" ], "turn_id": 5, "scan_id": "scene0610_00", "origin_question": "Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "101.54", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0610_00/01380.jpg", "scene0610_00/01420.jpg", "scene0610_00/01460.jpg", "scene0610_00/01500.jpg", "scene0610_00/01540.jpg" ], "turn_id": 6, "scan_id": "scene0610_00", "origin_question": "In which turn did you see both \"the white porcelain toilet in standard size\" and \"the switch\" simultaneously? ", "option": [], "answer": "6", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the white porcelain toilet in standard size\" and \"the switch\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0610_00/01580.jpg", "scene0610_00/01620.jpg", "scene0610_00/01660.jpg", "scene0610_00/01700.jpg", "scene0610_00/01740.jpg" ], "turn_id": 7, "scan_id": "scene0610_00", "origin_question": "Which two objects are on the same side of you now? \"the standard white rectangular ceramic bathtub\", \"the white shower curtain with colorful star patterns\", \"the switch\". ", "option": [ "the standard white rectangular ceramic bathtub, the white shower curtain with colorful star patterns", "the standard white rectangular ceramic bathtub, the switch", "the white shower curtain with colorful star patterns, the switch" ], "answer": "the standard white rectangular ceramic bathtub, the white shower curtain with colorful star patterns", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the standard white rectangular ceramic bathtub\", \"the white shower curtain with colorful star patterns\", \"the switch\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the standard white rectangular ceramic bathtub, the white shower curtain with colorful star patterns', 'the standard white rectangular ceramic bathtub, the switch', 'the white shower curtain with colorful star patterns, the switch'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0610_00/01780.jpg", "scene0610_00/01820.jpg", "scene0610_00/01860.jpg", "scene0610_00/01900.jpg", "scene0610_00/01940.jpg" ], "turn_id": 8, "scan_id": "scene0610_00", "origin_question": "Compared to your position at the end of the 6 turn, are you now closer or farther away from \"the white porcelain toilet in standard size\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 6 turn, are you now closer or farther away from \"the white porcelain toilet in standard size\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0610_00/01980.jpg", "scene0610_00/02020.jpg", "scene0610_00/02060.jpg", "scene0610_00/02100.jpg", "scene0610_00/02140.jpg" ], "turn_id": 9, "scan_id": "scene0610_00", "origin_question": "Please recall \"the white, round, smooth ceramic sink\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.03", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the white, round, smooth ceramic sink\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0610_00/02180.jpg", "scene0610_00/02220.jpg", "scene0610_00/02260.jpg", "scene0610_00/02300.jpg", "scene0610_00/02350.jpg" ], "turn_id": 10, "scan_id": "scene0610_00", "origin_question": "Please recall \"the switch\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.81", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the switch\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0256_00/00000.jpg", "scene0256_00/00040.jpg", "scene0256_00/00080.jpg", "scene0256_00/00120.jpg" ], "turn_id": 1, "scan_id": "scene0256_00", "origin_question": "Remember, have you seen any guitar(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any guitar(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0256_00/00160.jpg", "scene0256_00/00200.jpg", "scene0256_00/00240.jpg", "scene0256_00/00280.jpg" ], "turn_id": 2, "scan_id": "scene0256_00", "origin_question": "At the end of which turn were both of \"the white rectangular smooth laminated table\" and \"the rectangular black tissue box made of smooth material\" on your right side?", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the white rectangular smooth laminated table\" and \"the rectangular black tissue box made of smooth material\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0256_00/00320.jpg", "scene0256_00/00360.jpg", "scene0256_00/00400.jpg", "scene0256_00/00440.jpg" ], "turn_id": 3, "scan_id": "scene0256_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0256_00/00480.jpg", "scene0256_00/00520.jpg", "scene0256_00/00560.jpg", "scene0256_00/00600.jpg" ], "turn_id": 4, "scan_id": "scene0256_00", "origin_question": "Which direction is the chair in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the chair in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0256_00/00640.jpg", "scene0256_00/00680.jpg", "scene0256_00/00720.jpg", "scene0256_00/00760.jpg" ], "turn_id": 5, "scan_id": "scene0256_00", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the rectangular black tissue box made of smooth material\" from your position?", "option": [], "answer": "84.45", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the rectangular black tissue box made of smooth material\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0256_00/00800.jpg", "scene0256_00/00840.jpg", "scene0256_00/00880.jpg", "scene0256_00/00920.jpg" ], "turn_id": 6, "scan_id": "scene0256_00", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the white rectangular smooth laminated table\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the white rectangular smooth laminated table\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0256_00/00960.jpg", "scene0256_00/01000.jpg", "scene0256_00/01040.jpg", "scene0256_00/01080.jpg" ], "turn_id": 7, "scan_id": "scene0256_00", "origin_question": "Using your orientation at the end of turn 6 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 6 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0256_00/01120.jpg", "scene0256_00/01160.jpg", "scene0256_00/01200.jpg", "scene0256_00/01330.jpg" ], "turn_id": 8, "scan_id": "scene0256_00", "origin_question": "Please recall \"the chair\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.05", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the chair\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0256_00/01370.jpg", "scene0256_00/01410.jpg", "scene0256_00/01450.jpg", "scene0256_00/01490.jpg" ], "turn_id": 9, "scan_id": "scene0256_00", "origin_question": "Is the rectangular white cotton pillow with a minimalist design to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 9 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the rectangular white cotton pillow with a minimalist design to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0029_region12/54e499c336a0490d8cf1684049f36040_i1_5.jpg", "1mp3d_0029_region12/54e499c336a0490d8cf1684049f36040_i2_5.jpg", "1mp3d_0029_region12/bc561d6547f44d568a59ddae163df0d8_i1_0.jpg", "1mp3d_0029_region12/bc561d6547f44d568a59ddae163df0d8_i2_0.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0029_region12", "origin_question": "Remember, have you seen any couch(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any couch(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "1mp3d_0029_region12/bc561d6547f44d568a59ddae163df0d8_i1_5.jpg", "1mp3d_0029_region12/bc561d6547f44d568a59ddae163df0d8_i2_5.jpg", "1mp3d_0029_region12/54e499c336a0490d8cf1684049f36040_i1_4.jpg", "1mp3d_0029_region12/54e499c336a0490d8cf1684049f36040_i2_4.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0029_region12", "origin_question": "At the end of which turn were both of \"the white spherical ceramic flowerpot with an irregular opening\" and \"the modern black rectangular TV with a sleek frame\" on your right side?", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the white spherical ceramic flowerpot with an irregular opening\" and \"the modern black rectangular TV with a sleek frame\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "1mp3d_0029_region12/9bbf903d50da4ffd9e5d1fb7c9f4d69b_i1_0.jpg", "1mp3d_0029_region12/9bbf903d50da4ffd9e5d1fb7c9f4d69b_i2_0.jpg", "1mp3d_0029_region12/7c039232353c4d84b5e5800189758af0_i1_5.jpg", "1mp3d_0029_region12/7c039232353c4d84b5e5800189758af0_i2_5.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0029_region12", "origin_question": "Which direction is the modern black rectangular TV with a sleek frame in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the modern black rectangular TV with a sleek frame in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "1mp3d_0029_region12/606f9842b8af48daa28f10df8b69d861_i1_5.jpg", "1mp3d_0029_region12/606f9842b8af48daa28f10df8b69d861_i2_5.jpg", "1mp3d_0029_region12/ad77d6eeca5b492b8fe3317177f4f03f_i1_3.jpg", "1mp3d_0029_region12/ad77d6eeca5b492b8fe3317177f4f03f_i2_3.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0029_region12", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "141.5", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "1mp3d_0029_region12/db2c0754e41d4af0959e1361acc7b6db_i1_5.jpg", "1mp3d_0029_region12/db2c0754e41d4af0959e1361acc7b6db_i2_5.jpg", "1mp3d_0029_region12/c5261853647b41cf82e6af496ba95dcf_i1_3.jpg", "1mp3d_0029_region12/c5261853647b41cf82e6af496ba95dcf_i2_3.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0029_region12", "origin_question": "In which turn were you closest to \"the white spherical ceramic flowerpot with an irregular opening\"? ", "option": [], "answer": "4", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the white spherical ceramic flowerpot with an irregular opening\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "1mp3d_0029_region12/c5261853647b41cf82e6af496ba95dcf_i1_2.jpg", "1mp3d_0029_region12/c5261853647b41cf82e6af496ba95dcf_i2_2.jpg", "1mp3d_0029_region12/db2c0754e41d4af0959e1361acc7b6db_i1_4.jpg", "1mp3d_0029_region12/db2c0754e41d4af0959e1361acc7b6db_i2_4.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0029_region12", "origin_question": "Which direction is the large framed abstract picture with black, white, and yellow in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the large framed abstract picture with black, white, and yellow in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "1mp3d_0029_region12/ad77d6eeca5b492b8fe3317177f4f03f_i1_5.jpg", "1mp3d_0029_region12/ad77d6eeca5b492b8fe3317177f4f03f_i2_5.jpg", "1mp3d_0029_region12/606f9842b8af48daa28f10df8b69d861_i1_1.jpg", "1mp3d_0029_region12/606f9842b8af48daa28f10df8b69d861_i2_1.jpg" ], "turn_id": 7, "scan_id": "1mp3d_0029_region12", "origin_question": "Assuming the direction you are facing at the end of the turn 5 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Left", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 5 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "1mp3d_0029_region12/606f9842b8af48daa28f10df8b69d861_i1_0.jpg", "1mp3d_0029_region12/606f9842b8af48daa28f10df8b69d861_i2_0.jpg", "1mp3d_0029_region12/7c039232353c4d84b5e5800189758af0_i1_1.jpg", "1mp3d_0029_region12/7c039232353c4d84b5e5800189758af0_i2_1.jpg" ], "turn_id": 8, "scan_id": "1mp3d_0029_region12", "origin_question": "Is the white spherical ceramic flowerpot with an irregular opening to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the white spherical ceramic flowerpot with an irregular opening to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0329_01/00000.jpg", "scene0329_01/00040.jpg", "scene0329_01/00080.jpg", "scene0329_01/00120.jpg" ], "turn_id": 1, "scan_id": "scene0329_01", "origin_question": "Remember, have you seen any paper(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any paper(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0329_01/00160.jpg", "scene0329_01/00200.jpg", "scene0329_01/00240.jpg", "scene0329_01/00280.jpg" ], "turn_id": 2, "scan_id": "scene0329_01", "origin_question": "At the end of which turn were both of \"the transparent plastic bottle with a blue label\" and \"the cabinet\" on your right side?", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the transparent plastic bottle with a blue label\" and \"the cabinet\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "scene0329_01/00320.jpg", "scene0329_01/00360.jpg", "scene0329_01/00400.jpg", "scene0329_01/00440.jpg" ], "turn_id": 3, "scan_id": "scene0329_01", "origin_question": "Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Right", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0329_01/00480.jpg", "scene0329_01/00520.jpg", "scene0329_01/00560.jpg", "scene0329_01/00600.jpg" ], "turn_id": 4, "scan_id": "scene0329_01", "origin_question": "Remember, have you seen any clothes dryer(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any clothes dryer(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0329_01/00640.jpg", "scene0329_01/00680.jpg", "scene0329_01/00720.jpg", "scene0329_01/00760.jpg" ], "turn_id": 5, "scan_id": "scene0329_01", "origin_question": "In which turn were you closest to \"the white cube-style shelf with multiple compartments\"? ", "option": [], "answer": "3", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the white cube-style shelf with multiple compartments\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0329_01/00800.jpg", "scene0329_01/00840.jpg", "scene0329_01/00880.jpg", "scene0329_01/00920.jpg" ], "turn_id": 6, "scan_id": "scene0329_01", "origin_question": "Remember, have you seen any blackboard(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any blackboard(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0329_01/00960.jpg", "scene0329_01/01000.jpg", "scene0329_01/01040.jpg", "scene0329_01/01080.jpg" ], "turn_id": 7, "scan_id": "scene0329_01", "origin_question": "When did you first discover \"the cylindrical red and silver fire extinguisher\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the cylindrical red and silver fire extinguisher\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0065_02/00000.jpg", "scene0065_02/00030.jpg", "scene0065_02/00060.jpg", "scene0065_02/00090.jpg" ], "turn_id": 1, "scan_id": "scene0065_02", "origin_question": "Remember, have you seen any plunger(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any plunger(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0065_02/00120.jpg", "scene0065_02/00150.jpg", "scene0065_02/00180.jpg", "scene0065_02/00210.jpg" ], "turn_id": 2, "scan_id": "scene0065_02", "origin_question": "When did you first discover \"the rubber plunger with a wooden handle and red cup\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the rubber plunger with a wooden handle and red cup\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0065_02/00240.jpg", "scene0065_02/00270.jpg", "scene0065_02/00300.jpg", "scene0065_02/00330.jpg" ], "turn_id": 3, "scan_id": "scene0065_02", "origin_question": "Which direction is the medium-sized, white cylindrical toilet paper roll in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the medium-sized, white cylindrical toilet paper roll in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0065_02/00360.jpg", "scene0065_02/00390.jpg", "scene0065_02/00420.jpg", "scene0065_02/00450.jpg" ], "turn_id": 4, "scan_id": "scene0065_02", "origin_question": "Please recall \"the rubber plunger with a wooden handle and red cup\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.2", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the rubber plunger with a wooden handle and red cup\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0065_02/00480.jpg", "scene0065_02/00510.jpg", "scene0065_02/00540.jpg", "scene0065_02/00570.jpg" ], "turn_id": 5, "scan_id": "scene0065_02", "origin_question": "What will be the first-time appearance order of the following categories: cabinet, curtain, switch?", "option": [ "curtain, cabinet, switch", "cabinet, switch, curtain", "cabinet, curtain, switch", "switch, curtain, cabinet" ], "answer": "cabinet, curtain, switch", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: cabinet, curtain, switch? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['curtain, cabinet, switch', 'cabinet, switch, curtain', 'cabinet, curtain, switch', 'switch, curtain, cabinet'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0065_02/00600.jpg", "scene0065_02/00630.jpg", "scene0065_02/00660.jpg", "scene0065_02/00690.jpg" ], "turn_id": 6, "scan_id": "scene0065_02", "origin_question": "By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the light brown rectangular wooden cabinet\", Y: \"then oval-shaped white ceramic sink on a countertop\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the light brown rectangular wooden cabinet\", Y: \"then oval-shaped white ceramic sink on a countertop\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0065_02/00720.jpg", "scene0065_02/00750.jpg", "scene0065_02/00780.jpg", "scene0065_02/00810.jpg" ], "turn_id": 7, "scan_id": "scene0065_02", "origin_question": "By the end of the 3 turn, were you getting closer to or farther away from these two objects: X: \"the rubber plunger with a wooden handle and red cup\", Y: \"the rectangular white plastic switch\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to Y and farther away from X", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 3 turn, were you getting closer to or farther away from these two objects: X: \"the rubber plunger with a wooden handle and red cup\", Y: \"the rectangular white plastic switch\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0040_region62/51ddde8be586471b8cbd18f793fbf2cb_i2_3.jpg", "1mp3d_0040_region62/fd8580274f564f31b5b5f042b2b85aa6_i2_3.jpg", "1mp3d_0040_region62/fd8580274f564f31b5b5f042b2b85aa6_i1_4.jpg", "1mp3d_0040_region62/fd8580274f564f31b5b5f042b2b85aa6_i2_4.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0040_region62", "origin_question": "Remember, have you seen any fence(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any fence(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0040_region62/51ddde8be586471b8cbd18f793fbf2cb_i1_0.jpg", "1mp3d_0040_region62/51ddde8be586471b8cbd18f793fbf2cb_i2_0.jpg", "1mp3d_0040_region62/c4376e6a0d66466ca75505d35e17bc9b_i1_0.jpg", "1mp3d_0040_region62/c4376e6a0d66466ca75505d35e17bc9b_i2_0.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0040_region62", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "1mp3d_0040_region62/51ddde8be586471b8cbd18f793fbf2cb_i1_4.jpg", "1mp3d_0040_region62/51ddde8be586471b8cbd18f793fbf2cb_i2_4.jpg", "1mp3d_0040_region62/51ddde8be586471b8cbd18f793fbf2cb_i1_5.jpg", "1mp3d_0040_region62/51ddde8be586471b8cbd18f793fbf2cb_i2_5.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0040_region62", "origin_question": "How far is your current position from where you were at the end of the turn 1? (in meters)", "option": [], "answer": "1.5", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 1? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "1mp3d_0040_region62/c4376e6a0d66466ca75505d35e17bc9b_i1_5.jpg", "1mp3d_0040_region62/c4376e6a0d66466ca75505d35e17bc9b_i2_5.jpg", "1mp3d_0040_region62/c4376e6a0d66466ca75505d35e17bc9b_i1_4.jpg", "1mp3d_0040_region62/c4376e6a0d66466ca75505d35e17bc9b_i2_4.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0040_region62", "origin_question": "Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Left", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0574_02/00000.jpg", "scene0574_02/00040.jpg", "scene0574_02/00080.jpg", "scene0574_02/00120.jpg", "scene0574_02/00160.jpg" ], "turn_id": 1, "scan_id": "scene0574_02", "origin_question": "Remember, have you seen any door(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any door(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0574_02/00200.jpg", "scene0574_02/00240.jpg", "scene0574_02/00280.jpg", "scene0574_02/00320.jpg", "scene0574_02/00360.jpg" ], "turn_id": 2, "scan_id": "scene0574_02", "origin_question": "When did you first discover \"the mirror\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the mirror\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0574_02/00400.jpg", "scene0574_02/00440.jpg", "scene0574_02/00480.jpg", "scene0574_02/00520.jpg", "scene0574_02/00560.jpg" ], "turn_id": 3, "scan_id": "scene0574_02", "origin_question": "Remember, how many door(s) have you seen so far? ", "option": [], "answer": "2", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many door(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0574_02/00600.jpg", "scene0574_02/00640.jpg", "scene0574_02/00680.jpg", "scene0574_02/00720.jpg", "scene0574_02/00760.jpg" ], "turn_id": 4, "scan_id": "scene0574_02", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the rectangular white plastic bin\"; \"the fluffy white cotton towel\"; \"the rectangular wooden bench with a soft beige cushion\" ", "option": [ "the rectangular white plastic bin", "the fluffy white cotton towel", "the rectangular wooden bench with a soft beige cushion" ], "answer": "the rectangular white plastic bin", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the rectangular white plastic bin\"; \"the fluffy white cotton towel\"; \"the rectangular wooden bench with a soft beige cushion\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular white plastic bin', 'the fluffy white cotton towel', 'the rectangular wooden bench with a soft beige cushion'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0574_02/00800.jpg", "scene0574_02/00840.jpg", "scene0574_02/00880.jpg", "scene0574_02/00920.jpg", "scene0574_02/00960.jpg" ], "turn_id": 5, "scan_id": "scene0574_02", "origin_question": "Please recall \"the rectangular wooden bench with a soft beige cushion\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.77", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the rectangular wooden bench with a soft beige cushion\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0574_02/01000.jpg", "scene0574_02/01040.jpg", "scene0574_02/01080.jpg", "scene0574_02/01120.jpg", "scene0574_02/01160.jpg" ], "turn_id": 6, "scan_id": "scene0574_02", "origin_question": "Which direction is the medium-sized light gray board in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the medium-sized light gray board in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0574_02/01200.jpg", "scene0574_02/01240.jpg", "scene0574_02/01280.jpg", "scene0574_02/01320.jpg", "scene0574_02/01360.jpg" ], "turn_id": 7, "scan_id": "scene0574_02", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "141.46", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0574_02/01400.jpg", "scene0574_02/01440.jpg", "scene0574_02/01480.jpg", "scene0574_02/01520.jpg", "scene0574_02/01560.jpg" ], "turn_id": 8, "scan_id": "scene0574_02", "origin_question": "In which turn were you closest to \"the vanity\"? ", "option": [], "answer": "6", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the vanity\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0651_01/00000.jpg", "scene0651_01/00030.jpg", "scene0651_01/00060.jpg", "scene0651_01/00090.jpg" ], "turn_id": 1, "scan_id": "scene0651_01", "origin_question": "Remember, have you seen any cabinet(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any cabinet(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0651_01/00120.jpg", "scene0651_01/00150.jpg", "scene0651_01/00180.jpg", "scene0651_01/00210.jpg" ], "turn_id": 2, "scan_id": "scene0651_01", "origin_question": "At the end of which turn were both of \"the white double-door refrigerator with a shiny surface\" and \"the compact white electric stove with six burners\" on your right side?", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the white double-door refrigerator with a shiny surface\" and \"the compact white electric stove with six burners\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0651_01/00240.jpg", "scene0651_01/00270.jpg", "scene0651_01/00300.jpg", "scene0651_01/00330.jpg" ], "turn_id": 3, "scan_id": "scene0651_01", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the white rectangular range hood with a slanted top\" from your position?", "option": [], "answer": "51.76", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the white rectangular range hood with a slanted top\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0651_01/00360.jpg", "scene0651_01/00390.jpg", "scene0651_01/00420.jpg", "scene0651_01/00450.jpg" ], "turn_id": 4, "scan_id": "scene0651_01", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the white double-door refrigerator with a shiny surface\" from your position?", "option": [], "answer": "153.86", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the white double-door refrigerator with a shiny surface\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0651_01/00480.jpg", "scene0651_01/00510.jpg", "scene0651_01/00540.jpg", "scene0651_01/00570.jpg" ], "turn_id": 5, "scan_id": "scene0651_01", "origin_question": "By the end of the 3 turn, were you getting closer to or farther away from these two objects: X: \"the compact white electric stove with six burners\", Y: \"the white rectangular range hood with a slanted top\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 3 turn, were you getting closer to or farther away from these two objects: X: \"the compact white electric stove with six burners\", Y: \"the white rectangular range hood with a slanted top\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0651_01/00600.jpg", "scene0651_01/00630.jpg", "scene0651_01/00660.jpg", "scene0651_01/00690.jpg" ], "turn_id": 6, "scan_id": "scene0651_01", "origin_question": "Is the rectangular stainless steel sink with a faucet to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the rectangular stainless steel sink with a faucet to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0651_01/00720.jpg", "scene0651_01/00750.jpg", "scene0651_01/00780.jpg", "scene0651_01/00810.jpg" ], "turn_id": 7, "scan_id": "scene0651_01", "origin_question": "When did you first discover \"the white rectangular range hood with a slanted top\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the white rectangular range hood with a slanted top\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0589_02/00000.jpg", "scene0589_02/00040.jpg", "scene0589_02/00080.jpg", "scene0589_02/00120.jpg", "scene0589_02/00160.jpg" ], "turn_id": 1, "scan_id": "scene0589_02", "origin_question": "Remember, have you seen any bed(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any bed(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0589_02/00200.jpg", "scene0589_02/00240.jpg", "scene0589_02/00280.jpg", "scene0589_02/00320.jpg", "scene0589_02/00360.jpg" ], "turn_id": 2, "scan_id": "scene0589_02", "origin_question": "Is the deep blue fabric folding chair with a soft cushion to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the deep blue fabric folding chair with a soft cushion to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0589_02/00400.jpg", "scene0589_02/00440.jpg", "scene0589_02/00480.jpg", "scene0589_02/00520.jpg", "scene0589_02/00560.jpg" ], "turn_id": 3, "scan_id": "scene0589_02", "origin_question": "What will be the first-time appearance order of the following categories: bed, curtain, radiator?", "option": [ "curtain, bed, radiator", "radiator, curtain, bed", "bed, curtain, radiator", "radiator, bed, curtain" ], "answer": "bed, curtain, radiator", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bed, curtain, radiator? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['curtain, bed, radiator', 'radiator, curtain, bed', 'bed, curtain, radiator', 'radiator, bed, curtain'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "scene0589_02/00600.jpg", "scene0589_02/00640.jpg", "scene0589_02/00680.jpg", "scene0589_02/00720.jpg", "scene0589_02/00760.jpg" ], "turn_id": 4, "scan_id": "scene0589_02", "origin_question": "Among these three objects(\"the medium-sized wooden shelf with a simple design\",\"the rectangular blue bag with green accents\",\"the red classical guitar with a wide neck\" ), which one is the farthest from you?", "option": [ "the medium-sized wooden shelf with a simple design", "the rectangular blue bag with green accents", "the red classical guitar with a wide neck" ], "answer": "the red classical guitar with a wide neck", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the medium-sized wooden shelf with a simple design\",\"the rectangular blue bag with green accents\",\"the red classical guitar with a wide neck\" ), which one is the farthest from you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the medium-sized wooden shelf with a simple design', 'the rectangular blue bag with green accents', 'the red classical guitar with a wide neck'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0589_02/00800.jpg", "scene0589_02/00840.jpg", "scene0589_02/00880.jpg", "scene0589_02/00920.jpg", "scene0589_02/00960.jpg" ], "turn_id": 5, "scan_id": "scene0589_02", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0589_02/01000.jpg", "scene0589_02/01040.jpg", "scene0589_02/01080.jpg", "scene0589_02/01120.jpg", "scene0589_02/01160.jpg" ], "turn_id": 6, "scan_id": "scene0589_02", "origin_question": "Using your orientation at the end of turn 5 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "102.31", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 5 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0589_02/01200.jpg", "scene0589_02/01240.jpg", "scene0589_02/01280.jpg", "scene0589_02/01320.jpg", "scene0589_02/01360.jpg" ], "turn_id": 7, "scan_id": "scene0589_02", "origin_question": "Which two objects are on the same side of you now? \"the rectangular, warm brown wooden stand\", \"the rectangular wooden bed in warm orange tones\", \"the rectangular deep brown picture\". ", "option": [ "the rectangular, warm brown wooden stand, the rectangular wooden bed in warm orange tones", "the rectangular, warm brown wooden stand, the rectangular deep brown picture", "the rectangular wooden bed in warm orange tones, the rectangular deep brown picture" ], "answer": "the rectangular, warm brown wooden stand, the rectangular wooden bed in warm orange tones", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the rectangular, warm brown wooden stand\", \"the rectangular wooden bed in warm orange tones\", \"the rectangular deep brown picture\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular, warm brown wooden stand, the rectangular wooden bed in warm orange tones', 'the rectangular, warm brown wooden stand, the rectangular deep brown picture', 'the rectangular wooden bed in warm orange tones, the rectangular deep brown picture'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0589_02/01400.jpg", "scene0589_02/01440.jpg", "scene0589_02/01480.jpg", "scene0589_02/01520.jpg", "scene0589_02/01560.jpg" ], "turn_id": 8, "scan_id": "scene0589_02", "origin_question": "What will be the first-time appearance order of the following categories: clothes, door, window?", "option": [ "clothes, window, door", "window, clothes, door", "door, window, clothes", "clothes, door, window" ], "answer": "clothes, window, door", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: clothes, door, window? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['clothes, window, door', 'window, clothes, door', 'door, window, clothes', 'clothes, door, window'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0589_02/01600.jpg", "scene0589_02/01640.jpg", "scene0589_02/01680.jpg", "scene0589_02/01720.jpg", "scene0589_02/01760.jpg" ], "turn_id": 9, "scan_id": "scene0589_02", "origin_question": "Compared to your position at the end of the 5 turn, are you now closer or farther away from \"the slender light emitting a soft, warm glow\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 5 turn, are you now closer or farther away from \"the slender light emitting a soft, warm glow\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0589_02/01800.jpg", "scene0589_02/01840.jpg", "scene0589_02/01880.jpg", "scene0589_02/01920.jpg", "scene0589_02/01960.jpg" ], "turn_id": 10, "scan_id": "scene0589_02", "origin_question": "Remember, have you seen any tv(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any tv(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0044_region12/dd251ee8bca44b64a5dd2bb6036091f5_i1_3.jpg", "1mp3d_0044_region12/dd251ee8bca44b64a5dd2bb6036091f5_i2_3.jpg", "1mp3d_0044_region12/dd251ee8bca44b64a5dd2bb6036091f5_i1_4.jpg", "1mp3d_0044_region12/dd251ee8bca44b64a5dd2bb6036091f5_i2_4.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0044_region12", "origin_question": "Remember, have you seen any plant(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any plant(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "1mp3d_0044_region12/a2d600c414544aee8f19f6a83069e3ef_i1_2.jpg", "1mp3d_0044_region12/a2d600c414544aee8f19f6a83069e3ef_i2_2.jpg", "1mp3d_0044_region12/23b4f824dd6a4eab8dac7282a6623257_i1_2.jpg", "1mp3d_0044_region12/23b4f824dd6a4eab8dac7282a6623257_i2_2.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0044_region12", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the thriving, lush plant with rich green leaves\", Y: \"the small brown leather bench with a curved backrest\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "farther away from both X and Y", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the thriving, lush plant with rich green leaves\", Y: \"the small brown leather bench with a curved backrest\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "1mp3d_0044_region12/a2d600c414544aee8f19f6a83069e3ef_i1_0.jpg", "1mp3d_0044_region12/a2d600c414544aee8f19f6a83069e3ef_i2_0.jpg", "1mp3d_0044_region12/932611f1e4db4bd68f80a9dba5b3d55b_i1_4.jpg", "1mp3d_0044_region12/932611f1e4db4bd68f80a9dba5b3d55b_i2_4.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0044_region12", "origin_question": "Which direction is the thriving, lush plant with rich green leaves in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the thriving, lush plant with rich green leaves in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "1mp3d_0044_region12/932611f1e4db4bd68f80a9dba5b3d55b_i1_5.jpg", "1mp3d_0044_region12/932611f1e4db4bd68f80a9dba5b3d55b_i2_5.jpg", "1mp3d_0044_region12/932611f1e4db4bd68f80a9dba5b3d55b_i1_0.jpg", "1mp3d_0044_region12/932611f1e4db4bd68f80a9dba5b3d55b_i2_0.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0044_region12", "origin_question": "Which two objects are on the same side of you now? \"the thriving, lush plant with rich green leaves\", \"the medium round polished wooden table\", \"the medium-sized rectangular wooden door with frosted glass\". ", "option": [ "the thriving, lush plant with rich green leaves, the medium round polished wooden table", "the thriving, lush plant with rich green leaves, the medium-sized rectangular wooden door with frosted glass", "the medium round polished wooden table, the medium-sized rectangular wooden door with frosted glass" ], "answer": "the medium round polished wooden table, the medium-sized rectangular wooden door with frosted glass", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the thriving, lush plant with rich green leaves\", \"the medium round polished wooden table\", \"the medium-sized rectangular wooden door with frosted glass\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the thriving, lush plant with rich green leaves, the medium round polished wooden table', 'the thriving, lush plant with rich green leaves, the medium-sized rectangular wooden door with frosted glass', 'the medium round polished wooden table, the medium-sized rectangular wooden door with frosted glass'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "1mp3d_0044_region12/932611f1e4db4bd68f80a9dba5b3d55b_i1_1.jpg", "1mp3d_0044_region12/932611f1e4db4bd68f80a9dba5b3d55b_i2_1.jpg", "1mp3d_0044_region12/932611f1e4db4bd68f80a9dba5b3d55b_i1_2.jpg", "1mp3d_0044_region12/932611f1e4db4bd68f80a9dba5b3d55b_i2_2.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0044_region12", "origin_question": "Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "1mp3d_0044_region12/0fec8540d8224d2ea27496abea08e0d3_i1_5.jpg", "1mp3d_0044_region12/0fec8540d8224d2ea27496abea08e0d3_i2_5.jpg", "1mp3d_0044_region12/0ff26224c3f04c3a924aa1982fa13976_i1_3.jpg", "1mp3d_0044_region12/0ff26224c3f04c3a924aa1982fa13976_i2_3.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0044_region12", "origin_question": "At the end of which turn were both of \"the thriving, lush plant with rich green leaves\" and \"the medium round polished wooden table\" on your right side?", "option": [], "answer": "3", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the thriving, lush plant with rich green leaves\" and \"the medium round polished wooden table\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "1mp3d_0044_region12/0ff26224c3f04c3a924aa1982fa13976_i1_2.jpg", "1mp3d_0044_region12/0ff26224c3f04c3a924aa1982fa13976_i2_2.jpg", "1mp3d_0044_region12/0fec8540d8224d2ea27496abea08e0d3_i1_4.jpg", "1mp3d_0044_region12/0fec8540d8224d2ea27496abea08e0d3_i2_4.jpg" ], "turn_id": 7, "scan_id": "1mp3d_0044_region12", "origin_question": "Assuming the direction you are facing at the end of the turn 3 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Right", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 3 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "1mp3d_0044_region12/0fec8540d8224d2ea27496abea08e0d3_i1_3.jpg", "1mp3d_0044_region12/0fec8540d8224d2ea27496abea08e0d3_i2_3.jpg", "1mp3d_0044_region12/df4659517bee4f6aa48b36d05bb1972c_i1_1.jpg", "1mp3d_0044_region12/df4659517bee4f6aa48b36d05bb1972c_i2_1.jpg" ], "turn_id": 8, "scan_id": "1mp3d_0044_region12", "origin_question": "In which turn were you closest to \"the thriving, lush plant with rich green leaves\"? ", "option": [], "answer": "1", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the thriving, lush plant with rich green leaves\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0044_region12/df4659517bee4f6aa48b36d05bb1972c_i1_2.jpg", "1mp3d_0044_region12/df4659517bee4f6aa48b36d05bb1972c_i2_2.jpg", "1mp3d_0044_region12/9d7a230cc1e643f5bb3cc0b0f3746284_i1_3.jpg", "1mp3d_0044_region12/9d7a230cc1e643f5bb3cc0b0f3746284_i2_3.jpg" ], "turn_id": 9, "scan_id": "1mp3d_0044_region12", "origin_question": "Compared to your position at the end of the 7 turn, are you now closer or farther away from \"the medium round polished wooden table\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 9 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 7 turn, are you now closer or farther away from \"the medium round polished wooden table\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0704_01/00000.jpg", "scene0704_01/00040.jpg", "scene0704_01/00080.jpg", "scene0704_01/00120.jpg", "scene0704_01/00160.jpg" ], "turn_id": 1, "scan_id": "scene0704_01", "origin_question": "Remember, have you seen any cabinet(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any cabinet(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0704_01/00200.jpg", "scene0704_01/00240.jpg", "scene0704_01/00280.jpg", "scene0704_01/00320.jpg", "scene0704_01/00360.jpg" ], "turn_id": 2, "scan_id": "scene0704_01", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the brown rectangular box made of cardboard\" from your position?", "option": [], "answer": "43.35", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the brown rectangular box made of cardboard\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0704_01/00400.jpg", "scene0704_01/00440.jpg", "scene0704_01/00480.jpg", "scene0704_01/00520.jpg", "scene0704_01/00560.jpg" ], "turn_id": 3, "scan_id": "scene0704_01", "origin_question": "In which turn did you see both \"the small white paper cup with a green emblem\" and \"the large rectangular black plastic shelf for water bottles\" simultaneously? ", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the small white paper cup with a green emblem\" and \"the large rectangular black plastic shelf for water bottles\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0704_01/00600.jpg", "scene0704_01/00640.jpg", "scene0704_01/00680.jpg", "scene0704_01/00720.jpg", "scene0704_01/00760.jpg" ], "turn_id": 4, "scan_id": "scene0704_01", "origin_question": "Is the large rectangular light brown wooden table to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the large rectangular light brown wooden table to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "scene0704_01/00800.jpg", "scene0704_01/00840.jpg", "scene0704_01/00880.jpg", "scene0704_01/00920.jpg", "scene0704_01/00960.jpg" ], "turn_id": 5, "scan_id": "scene0704_01", "origin_question": "Assuming the direction you are facing at the end of the turn 3 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Right", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 3 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0704_01/01000.jpg", "scene0704_01/01040.jpg", "scene0704_01/01080.jpg", "scene0704_01/01120.jpg", "scene0704_01/01160.jpg" ], "turn_id": 6, "scan_id": "scene0704_01", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the brown rectangular box made of cardboard\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the brown rectangular box made of cardboard\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0704_01/01200.jpg", "scene0704_01/01240.jpg", "scene0704_01/01280.jpg", "scene0704_01/01320.jpg", "scene0704_01/01360.jpg" ], "turn_id": 7, "scan_id": "scene0704_01", "origin_question": "By the end of the 3 turn, were you getting closer to or farther away from these two objects: X: \"the small white paper cup with a green emblem\", Y: \"the small rectangular white plastic water cooler\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 3 turn, were you getting closer to or farther away from these two objects: X: \"the small white paper cup with a green emblem\", Y: \"the small rectangular white plastic water cooler\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0704_01/01400.jpg", "scene0704_01/01440.jpg", "scene0704_01/01480.jpg", "scene0704_01/01520.jpg", "scene0704_01/01560.jpg" ], "turn_id": 8, "scan_id": "scene0704_01", "origin_question": "Please recall \"the large gray-white multifunction copier machine\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.43", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the large gray-white multifunction copier machine\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0077_00/00000.jpg", "scene0077_00/00030.jpg", "scene0077_00/00060.jpg", "scene0077_00/00090.jpg" ], "turn_id": 1, "scan_id": "scene0077_00", "origin_question": "Remember, have you seen any fan(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any fan(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0077_00/00120.jpg", "scene0077_00/00150.jpg", "scene0077_00/00180.jpg", "scene0077_00/00210.jpg" ], "turn_id": 2, "scan_id": "scene0077_00", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the large hardcover book with a dark brown cover\" from your position?", "option": [], "answer": "54.08", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the large hardcover book with a dark brown cover\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0077_00/00240.jpg", "scene0077_00/00270.jpg", "scene0077_00/00300.jpg", "scene0077_00/00330.jpg" ], "turn_id": 3, "scan_id": "scene0077_00", "origin_question": "Which direction is the small round white clock with a golden border in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the small round white clock with a golden border in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0077_00/00360.jpg", "scene0077_00/00390.jpg", "scene0077_00/00420.jpg", "scene0077_00/00550.jpg" ], "turn_id": 4, "scan_id": "scene0077_00", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the small round white clock with a golden border\" from your position?", "option": [], "answer": "37.74", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the small round white clock with a golden border\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0474_03/00000.jpg", "scene0474_03/00040.jpg", "scene0474_03/00080.jpg", "scene0474_03/00120.jpg", "scene0474_03/00160.jpg" ], "turn_id": 1, "scan_id": "scene0474_03", "origin_question": "Remember, have you seen any keyboard(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any keyboard(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0474_03/00200.jpg", "scene0474_03/00240.jpg", "scene0474_03/00280.jpg", "scene0474_03/00320.jpg", "scene0474_03/00360.jpg" ], "turn_id": 2, "scan_id": "scene0474_03", "origin_question": "At the end of which turn were both of \"the white plastic keyboard with black letters\" and \"the black rectangular computer tower made of metal\" on your right side?", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the white plastic keyboard with black letters\" and \"the black rectangular computer tower made of metal\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0474_03/00400.jpg", "scene0474_03/00440.jpg", "scene0474_03/00480.jpg", "scene0474_03/00520.jpg", "scene0474_03/00560.jpg" ], "turn_id": 3, "scan_id": "scene0474_03", "origin_question": "In which turn did you see both \"the black cabinet with three drawers and silver handles\" and \"the sleek silver metal laptop with a minimalist design\" simultaneously? ", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the black cabinet with three drawers and silver handles\" and \"the sleek silver metal laptop with a minimalist design\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0474_03/00600.jpg", "scene0474_03/00640.jpg", "scene0474_03/00680.jpg", "scene0474_03/00720.jpg", "scene0474_03/00760.jpg" ], "turn_id": 4, "scan_id": "scene0474_03", "origin_question": "Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "50.73", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0474_03/00800.jpg", "scene0474_03/00840.jpg", "scene0474_03/00880.jpg", "scene0474_03/00920.jpg", "scene0474_03/00960.jpg" ], "turn_id": 5, "scan_id": "scene0474_03", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the couch\"; \"the medium white ceramic cup with a black rim\"; \"the sleek silver metal laptop with a minimalist design\" ", "option": [ "the couch", "the medium white ceramic cup with a black rim", "the sleek silver metal laptop with a minimalist design" ], "answer": "the couch", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the couch\"; \"the medium white ceramic cup with a black rim\"; \"the sleek silver metal laptop with a minimalist design\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the couch', 'the medium white ceramic cup with a black rim', 'the sleek silver metal laptop with a minimalist design'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0474_03/01000.jpg", "scene0474_03/01040.jpg", "scene0474_03/01080.jpg", "scene0474_03/01120.jpg", "scene0474_03/01160.jpg" ], "turn_id": 6, "scan_id": "scene0474_03", "origin_question": "Which direction is the sleek silver metal laptop with a minimalist design in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the sleek silver metal laptop with a minimalist design in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0474_03/01200.jpg", "scene0474_03/01240.jpg", "scene0474_03/01280.jpg", "scene0474_03/01320.jpg", "scene0474_03/01360.jpg" ], "turn_id": 7, "scan_id": "scene0474_03", "origin_question": "Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "145.61", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "scene0474_03/01400.jpg", "scene0474_03/01440.jpg", "scene0474_03/01480.jpg", "scene0474_03/01520.jpg", "scene0474_03/01560.jpg" ], "turn_id": 8, "scan_id": "scene0474_03", "origin_question": "Among these three objects(\"the white plastic keyboard with black letters\",\"the sleek silver metal laptop with a minimalist design\",\"the large black chair with a curved backrest\" ), which one is the closest to you?", "option": [ "the white plastic keyboard with black letters", "the sleek silver metal laptop with a minimalist design", "the large black chair with a curved backrest" ], "answer": "the large black chair with a curved backrest", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the white plastic keyboard with black letters\",\"the sleek silver metal laptop with a minimalist design\",\"the large black chair with a curved backrest\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the white plastic keyboard with black letters', 'the sleek silver metal laptop with a minimalist design', 'the large black chair with a curved backrest'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0474_03/01600.jpg", "scene0474_03/01640.jpg", "scene0474_03/01680.jpg", "scene0474_03/01720.jpg", "scene0474_03/01760.jpg" ], "turn_id": 9, "scan_id": "scene0474_03", "origin_question": "Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "155.63", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0474_03/01800.jpg", "scene0474_03/01840.jpg", "scene0474_03/01880.jpg", "scene0474_03/01920.jpg", "scene0474_03/01960.jpg" ], "turn_id": 10, "scan_id": "scene0474_03", "origin_question": "In which turn did you see both \"the white plastic keyboard with black letters\" and \"the large rectangular canvas picture with black and white images\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the white plastic keyboard with black letters\" and \"the large rectangular canvas picture with black and white images\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0334_02/00000.jpg", "scene0334_02/00040.jpg", "scene0334_02/00080.jpg", "scene0334_02/00120.jpg" ], "turn_id": 1, "scan_id": "scene0334_02", "origin_question": "Remember, have you seen any duct(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any duct(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0334_02/00160.jpg", "scene0334_02/00200.jpg", "scene0334_02/00240.jpg", "scene0334_02/00280.jpg" ], "turn_id": 2, "scan_id": "scene0334_02", "origin_question": "In which turn were you closest to \"the round light brown table with a wooden veneer\"? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the round light brown table with a wooden veneer\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0334_02/00320.jpg", "scene0334_02/00360.jpg", "scene0334_02/00400.jpg", "scene0334_02/00440.jpg" ], "turn_id": 3, "scan_id": "scene0334_02", "origin_question": "", "option": [], "answer": "", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0334_02/00480.jpg", "scene0334_02/00520.jpg", "scene0334_02/00560.jpg", "scene0334_02/00600.jpg" ], "turn_id": 4, "scan_id": "scene0334_02", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "6.43", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "scene0334_02/00640.jpg", "scene0334_02/00680.jpg", "scene0334_02/00720.jpg", "scene0334_02/00760.jpg" ], "turn_id": 5, "scan_id": "scene0334_02", "origin_question": "Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Left", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0334_02/00800.jpg", "scene0334_02/00840.jpg", "scene0334_02/00880.jpg", "scene0334_02/00920.jpg" ], "turn_id": 6, "scan_id": "scene0334_02", "origin_question": "Remember, how many chandelier(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many chandelier(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0040_region47/e0e69f99ad6347b2aaf144db43f53ee5_i1_5.jpg", "1mp3d_0040_region47/e0e69f99ad6347b2aaf144db43f53ee5_i2_5.jpg", "1mp3d_0040_region47/e93d23ba6ebd41469ac71306bc57fb71_i1_4.jpg", "1mp3d_0040_region47/e93d23ba6ebd41469ac71306bc57fb71_i2_4.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0040_region47", "origin_question": "Remember, have you seen any door(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any door(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "1mp3d_0040_region47/bc9804993e6c4e4a8d2b3c9569276946_i1_4.jpg", "1mp3d_0040_region47/bc9804993e6c4e4a8d2b3c9569276946_i2_4.jpg", "1mp3d_0040_region47/bc9804993e6c4e4a8d2b3c9569276946_i1_3.jpg", "1mp3d_0040_region47/bc9804993e6c4e4a8d2b3c9569276946_i2_3.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0040_region47", "origin_question": "When did you first discover \"the cylindrical pillar with a mottled white surface\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the cylindrical pillar with a mottled white surface\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "1mp3d_0040_region47/e93d23ba6ebd41469ac71306bc57fb71_i1_3.jpg", "1mp3d_0040_region47/e93d23ba6ebd41469ac71306bc57fb71_i2_3.jpg", "1mp3d_0040_region47/e0e69f99ad6347b2aaf144db43f53ee5_i1_0.jpg", "1mp3d_0040_region47/e0e69f99ad6347b2aaf144db43f53ee5_i2_0.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0040_region47", "origin_question": "Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Right", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0040_region47/7cb43b9d9c6e4f04bfd9e52ef67d9201_i1_2.jpg", "1mp3d_0040_region47/7cb43b9d9c6e4f04bfd9e52ef67d9201_i2_2.jpg", "1mp3d_0040_region47/e0e69f99ad6347b2aaf144db43f53ee5_i1_2.jpg", "1mp3d_0040_region47/e0e69f99ad6347b2aaf144db43f53ee5_i2_2.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0040_region47", "origin_question": "", "option": [], "answer": "", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0494_00/00000.jpg", "scene0494_00/00030.jpg", "scene0494_00/00060.jpg", "scene0494_00/00090.jpg" ], "turn_id": 1, "scan_id": "scene0494_00", "origin_question": "Remember, have you seen any chair(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any chair(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0494_00/00120.jpg", "scene0494_00/00150.jpg", "scene0494_00/00180.jpg", "scene0494_00/00210.jpg" ], "turn_id": 2, "scan_id": "scene0494_00", "origin_question": "In which turn did you see both \"the rectangular light-colored wooden table with a smooth surface\" and \"the rectangular blackboard with a white surface\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the rectangular light-colored wooden table with a smooth surface\" and \"the rectangular blackboard with a white surface\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0494_00/00240.jpg", "scene0494_00/00270.jpg", "scene0494_00/00300.jpg", "scene0494_00/00330.jpg" ], "turn_id": 3, "scan_id": "scene0494_00", "origin_question": "Remember, have you seen any chair(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any chair(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0494_00/00360.jpg", "scene0494_00/00390.jpg", "scene0494_00/00420.jpg", "scene0494_00/00450.jpg" ], "turn_id": 4, "scan_id": "scene0494_00", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the white rectangular door with a silver handle\"; \"the rectangular blackboard with a white surface\"; \"the rectangular light-colored wooden table with a smooth surface\" ", "option": [ "the white rectangular door with a silver handle", "the rectangular blackboard with a white surface", "the rectangular light-colored wooden table with a smooth surface" ], "answer": "the white rectangular door with a silver handle", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the white rectangular door with a silver handle\"; \"the rectangular blackboard with a white surface\"; \"the rectangular light-colored wooden table with a smooth surface\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the white rectangular door with a silver handle', 'the rectangular blackboard with a white surface', 'the rectangular light-colored wooden table with a smooth surface'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0494_00/00480.jpg", "scene0494_00/00510.jpg", "scene0494_00/00540.jpg", "scene0494_00/00570.jpg" ], "turn_id": 5, "scan_id": "scene0494_00", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "61.2", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0494_00/00600.jpg", "scene0494_00/00630.jpg", "scene0494_00/00660.jpg", "scene0494_00/00690.jpg" ], "turn_id": 6, "scan_id": "scene0494_00", "origin_question": "Using your orientation at the end of turn 5 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "48.91", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 5 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0150_01/00000.jpg", "scene0150_01/00030.jpg", "scene0150_01/00060.jpg", "scene0150_01/00090.jpg" ], "turn_id": 1, "scan_id": "scene0150_01", "origin_question": "Remember, how many table(s) have you seen so far? ", "option": [], "answer": "2", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many table(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 3 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0150_01/00120.jpg", "scene0150_01/00150.jpg", "scene0150_01/00180.jpg", "scene0150_01/00210.jpg" ], "turn_id": 2, "scan_id": "scene0150_01", "origin_question": "Remember, have you seen any soap dish(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any soap dish(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 3 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0150_01/00240.jpg", "scene0150_01/00270.jpg", "scene0150_01/00300.jpg", "scene0150_01/00330.jpg" ], "turn_id": 3, "scan_id": "scene0150_01", "origin_question": "When did you first discover \"the large rectangular blackboard with a white surface\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the large rectangular blackboard with a white surface\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 3 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0044_region5/62580c9256eb4272b118d3e0ea4e13aa_i1_0.jpg", "1mp3d_0044_region5/62580c9256eb4272b118d3e0ea4e13aa_i2_0.jpg", "1mp3d_0044_region5/9c8468e02b584bf496a27c11ed7f9eda_i1_2.jpg", "1mp3d_0044_region5/9c8468e02b584bf496a27c11ed7f9eda_i2_2.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0044_region5", "origin_question": "Remember, have you seen any bench(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any bench(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "1mp3d_0044_region5/9c8468e02b584bf496a27c11ed7f9eda_i1_5.jpg", "1mp3d_0044_region5/9c8468e02b584bf496a27c11ed7f9eda_i2_5.jpg", "1mp3d_0044_region5/62580c9256eb4272b118d3e0ea4e13aa_i1_3.jpg", "1mp3d_0044_region5/62580c9256eb4272b118d3e0ea4e13aa_i2_3.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0044_region5", "origin_question": "When did you first discover \"the bench\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the bench\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "1mp3d_0044_region5/62580c9256eb4272b118d3e0ea4e13aa_i1_2.jpg", "1mp3d_0044_region5/62580c9256eb4272b118d3e0ea4e13aa_i2_2.jpg", "1mp3d_0044_region5/62580c9256eb4272b118d3e0ea4e13aa_i1_4.jpg", "1mp3d_0044_region5/62580c9256eb4272b118d3e0ea4e13aa_i2_4.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0044_region5", "origin_question": "In which turn did you see both \"the rectangular wooden door with a rich brown color\" and \"the bench\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the rectangular wooden door with a rich brown color\" and \"the bench\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "1mp3d_0044_region5/62580c9256eb4272b118d3e0ea4e13aa_i1_5.jpg", "1mp3d_0044_region5/62580c9256eb4272b118d3e0ea4e13aa_i2_5.jpg", "1mp3d_0044_region5/9c8468e02b584bf496a27c11ed7f9eda_i1_1.jpg", "1mp3d_0044_region5/9c8468e02b584bf496a27c11ed7f9eda_i2_1.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0044_region5", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "1mp3d_0044_region5/9c8468e02b584bf496a27c11ed7f9eda_i1_0.jpg", "1mp3d_0044_region5/9c8468e02b584bf496a27c11ed7f9eda_i2_0.jpg", "1mp3d_0044_region5/9c8468e02b584bf496a27c11ed7f9eda_i1_4.jpg", "1mp3d_0044_region5/9c8468e02b584bf496a27c11ed7f9eda_i2_4.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0044_region5", "origin_question": "Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Left", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "1mp3d_0044_region5/87649a1903e348d596618acdc412b644_i1_5.jpg", "1mp3d_0044_region5/87649a1903e348d596618acdc412b644_i2_5.jpg", "1mp3d_0044_region5/87649a1903e348d596618acdc412b644_i1_0.jpg", "1mp3d_0044_region5/87649a1903e348d596618acdc412b644_i2_0.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0044_region5", "origin_question": "Using your orientation at the end of turn 5 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 5 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0692_00/00000.jpg", "scene0692_00/00040.jpg", "scene0692_00/00080.jpg", "scene0692_00/00120.jpg" ], "turn_id": 1, "scan_id": "scene0692_00", "origin_question": "Remember, have you seen any book(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any book(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0692_00/00160.jpg", "scene0692_00/00200.jpg", "scene0692_00/00240.jpg", "scene0692_00/00280.jpg" ], "turn_id": 2, "scan_id": "scene0692_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0692_00/00320.jpg", "scene0692_00/00360.jpg", "scene0692_00/00400.jpg", "scene0692_00/00440.jpg" ], "turn_id": 3, "scan_id": "scene0692_00", "origin_question": "What will be the first-time appearance order of the following categories: bin, book, light?", "option": [ "book, bin, light", "bin, book, light", "bin, light, book", "book, light, bin" ], "answer": "book, bin, light", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bin, book, light? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['book, bin, light', 'bin, book, light', 'bin, light, book', 'book, light, bin'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0692_00/00480.jpg", "scene0692_00/00520.jpg", "scene0692_00/00560.jpg", "scene0692_00/00600.jpg" ], "turn_id": 4, "scan_id": "scene0692_00", "origin_question": "At the end of which turn were both of \"the large rectangular light brown wooden cabinet\" and \"the modern lamp with a cone-shaped white lampshade\" on your right side?", "option": [], "answer": "2", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the large rectangular light brown wooden cabinet\" and \"the modern lamp with a cone-shaped white lampshade\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0692_00/00640.jpg", "scene0692_00/00680.jpg", "scene0692_00/00720.jpg", "scene0692_00/00760.jpg" ], "turn_id": 5, "scan_id": "scene0692_00", "origin_question": "What will be the first-time appearance order of the following categories: bin, light, shelf?", "option": [ "shelf, bin, light", "bin, shelf, light", "light, bin, shelf", "bin, light, shelf" ], "answer": "shelf, bin, light", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bin, light, shelf? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['shelf, bin, light', 'bin, shelf, light', 'light, bin, shelf', 'bin, light, shelf'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0692_00/00800.jpg", "scene0692_00/00840.jpg", "scene0692_00/00880.jpg", "scene0692_00/00920.jpg" ], "turn_id": 6, "scan_id": "scene0692_00", "origin_question": "Remember, how many table(s) have you seen so far? ", "option": [], "answer": "2", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many table(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0692_00/00960.jpg", "scene0692_00/01000.jpg", "scene0692_00/01040.jpg", "scene0692_00/01080.jpg" ], "turn_id": 7, "scan_id": "scene0692_00", "origin_question": "Is the large rectangular light brown wooden cabinet to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the large rectangular light brown wooden cabinet to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0692_00/01120.jpg", "scene0692_00/01160.jpg", "scene0692_00/01200.jpg", "scene0692_00/01240.jpg" ], "turn_id": 8, "scan_id": "scene0692_00", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the tall, slim wooden rack with multiple hooks\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the tall, slim wooden rack with multiple hooks\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0044_region43/15783cee785d4630bb732f4399519e8b_i1_3.jpg", "1mp3d_0044_region43/15783cee785d4630bb732f4399519e8b_i2_3.jpg", "1mp3d_0044_region43/ea01296895574f95a606857c5e02eadf_i1_2.jpg", "1mp3d_0044_region43/ea01296895574f95a606857c5e02eadf_i2_2.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0044_region43", "origin_question": "Remember, have you seen any shelf(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any shelf(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0044_region43/0ff26224c3f04c3a924aa1982fa13976_i1_0.jpg", "1mp3d_0044_region43/0ff26224c3f04c3a924aa1982fa13976_i2_0.jpg", "1mp3d_0044_region43/0ff26224c3f04c3a924aa1982fa13976_i1_5.jpg", "1mp3d_0044_region43/0ff26224c3f04c3a924aa1982fa13976_i2_5.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0044_region43", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0044_region43/0ff26224c3f04c3a924aa1982fa13976_i1_4.jpg", "1mp3d_0044_region43/0ff26224c3f04c3a924aa1982fa13976_i2_4.jpg", "1mp3d_0044_region43/ea01296895574f95a606857c5e02eadf_i1_3.jpg", "1mp3d_0044_region43/ea01296895574f95a606857c5e02eadf_i2_3.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0044_region43", "origin_question": "", "option": [], "answer": "", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "1mp3d_0044_region43/15783cee785d4630bb732f4399519e8b_i1_4.jpg", "1mp3d_0044_region43/15783cee785d4630bb732f4399519e8b_i2_4.jpg", "1mp3d_0044_region43/15783cee785d4630bb732f4399519e8b_i1_5.jpg", "1mp3d_0044_region43/15783cee785d4630bb732f4399519e8b_i2_5.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0044_region43", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "80.4", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "1mp3d_0044_region43/15783cee785d4630bb732f4399519e8b_i1_0.jpg", "1mp3d_0044_region43/15783cee785d4630bb732f4399519e8b_i2_0.jpg", "1mp3d_0044_region43/15783cee785d4630bb732f4399519e8b_i1_1.jpg", "1mp3d_0044_region43/15783cee785d4630bb732f4399519e8b_i2_1.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0044_region43", "origin_question": "Using your orientation at the end of turn 4 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "120.07", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 4 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "1mp3d_0044_region43/ea01296895574f95a606857c5e02eadf_i1_0.jpg", "1mp3d_0044_region43/ea01296895574f95a606857c5e02eadf_i2_0.jpg", "1mp3d_0044_region43/ea01296895574f95a606857c5e02eadf_i1_1.jpg", "1mp3d_0044_region43/ea01296895574f95a606857c5e02eadf_i2_1.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0044_region43", "origin_question": "Using your orientation at the end of turn 4 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "151.68", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 4 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0647_01/00000.jpg", "scene0647_01/00040.jpg", "scene0647_01/00080.jpg", "scene0647_01/00120.jpg" ], "turn_id": 1, "scan_id": "scene0647_01", "origin_question": "Remember, have you seen any stool(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any stool(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0647_01/00160.jpg", "scene0647_01/00200.jpg", "scene0647_01/00240.jpg", "scene0647_01/00280.jpg" ], "turn_id": 2, "scan_id": "scene0647_01", "origin_question": "When did you first discover \"the wooden chair with vertical bars on the backrest\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the wooden chair with vertical bars on the backrest\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0647_01/00320.jpg", "scene0647_01/00360.jpg", "scene0647_01/00400.jpg", "scene0647_01/00440.jpg" ], "turn_id": 3, "scan_id": "scene0647_01", "origin_question": "How far is your current position from where you were at the end of the turn 1? (in meters)", "option": [], "answer": "4.12", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 1? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0647_01/00480.jpg", "scene0647_01/00520.jpg", "scene0647_01/00560.jpg", "scene0647_01/00600.jpg" ], "turn_id": 4, "scan_id": "scene0647_01", "origin_question": "Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0647_01/00640.jpg", "scene0647_01/00680.jpg", "scene0647_01/00720.jpg", "scene0647_01/00760.jpg" ], "turn_id": 5, "scan_id": "scene0647_01", "origin_question": "Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "142.73", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0647_01/00800.jpg", "scene0647_01/00840.jpg", "scene0647_01/00880.jpg", "scene0647_01/00920.jpg" ], "turn_id": 6, "scan_id": "scene0647_01", "origin_question": "In which turn did you see both \"the gray and black camouflage backpack\" and \"the medium-sized black flat-screen TV\" simultaneously? ", "option": [], "answer": "5", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the gray and black camouflage backpack\" and \"the medium-sized black flat-screen TV\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0647_01/00960.jpg", "scene0647_01/01000.jpg", "scene0647_01/01040.jpg", "scene0647_01/01080.jpg" ], "turn_id": 7, "scan_id": "scene0647_01", "origin_question": "Which direction is the medium-sized black flat-screen TV in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the medium-sized black flat-screen TV in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0226_01/00000.jpg", "scene0226_01/00040.jpg", "scene0226_01/00080.jpg", "scene0226_01/00120.jpg" ], "turn_id": 1, "scan_id": "scene0226_01", "origin_question": "Remember, have you seen any tool(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any tool(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0226_01/00160.jpg", "scene0226_01/00200.jpg", "scene0226_01/00240.jpg", "scene0226_01/00280.jpg" ], "turn_id": 2, "scan_id": "scene0226_01", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the vibrant picture with a red-orange sky and silhouettes\" from your position?", "option": [], "answer": "84.68", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the vibrant picture with a red-orange sky and silhouettes\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0226_01/00320.jpg", "scene0226_01/00360.jpg", "scene0226_01/00400.jpg", "scene0226_01/00440.jpg" ], "turn_id": 3, "scan_id": "scene0226_01", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the large vertical beige board with a wooden frame\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the large vertical beige board with a wooden frame\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0226_01/00480.jpg", "scene0226_01/00520.jpg", "scene0226_01/00560.jpg", "scene0226_01/00600.jpg" ], "turn_id": 4, "scan_id": "scene0226_01", "origin_question": "Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the vibrant picture with a red-orange sky and silhouettes\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the vibrant picture with a red-orange sky and silhouettes\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0226_01/00640.jpg", "scene0226_01/00680.jpg", "scene0226_01/00720.jpg", "scene0226_01/00760.jpg" ], "turn_id": 5, "scan_id": "scene0226_01", "origin_question": "Is the rectangular, blue hardcover book with a minimalist design to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the rectangular, blue hardcover book with a minimalist design to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0226_01/00800.jpg", "scene0226_01/00840.jpg", "scene0226_01/00880.jpg", "scene0226_01/00920.jpg" ], "turn_id": 6, "scan_id": "scene0226_01", "origin_question": "Please recall \"the simple white metal stool with a round seat\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.79", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the simple white metal stool with a round seat\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0307_02/00000.jpg", "scene0307_02/00040.jpg", "scene0307_02/00080.jpg", "scene0307_02/00120.jpg", "scene0307_02/00160.jpg" ], "turn_id": 1, "scan_id": "scene0307_02", "origin_question": "Remember, have you seen any dustpan(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any dustpan(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0307_02/00200.jpg", "scene0307_02/00240.jpg", "scene0307_02/00280.jpg", "scene0307_02/00320.jpg", "scene0307_02/00360.jpg" ], "turn_id": 2, "scan_id": "scene0307_02", "origin_question": "Which two objects are on the same side of you now? \"the white rectangular metal washing machine\", \"the rectangular green and white box of detergent\", \"the medium-sized white stainless steel refrigerator\". ", "option": [ "the white rectangular metal washing machine, the rectangular green and white box of detergent", "the white rectangular metal washing machine, the medium-sized white stainless steel refrigerator", "the rectangular green and white box of detergent, the medium-sized white stainless steel refrigerator" ], "answer": "the white rectangular metal washing machine, the rectangular green and white box of detergent", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the white rectangular metal washing machine\", \"the rectangular green and white box of detergent\", \"the medium-sized white stainless steel refrigerator\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the white rectangular metal washing machine, the rectangular green and white box of detergent', 'the white rectangular metal washing machine, the medium-sized white stainless steel refrigerator', 'the rectangular green and white box of detergent, the medium-sized white stainless steel refrigerator'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "scene0307_02/00400.jpg", "scene0307_02/00440.jpg", "scene0307_02/00480.jpg", "scene0307_02/00520.jpg", "scene0307_02/00560.jpg" ], "turn_id": 3, "scan_id": "scene0307_02", "origin_question": "Among these three objects(\"the rectangular glass window with a white frame\",\"the medium-sized white stainless steel refrigerator\",\"the table\" ), which one is the closest to you?", "option": [ "the rectangular glass window with a white frame", "the medium-sized white stainless steel refrigerator", "the table" ], "answer": "the rectangular glass window with a white frame", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the rectangular glass window with a white frame\",\"the medium-sized white stainless steel refrigerator\",\"the table\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular glass window with a white frame', 'the medium-sized white stainless steel refrigerator', 'the table'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0307_02/00600.jpg", "scene0307_02/00640.jpg", "scene0307_02/00680.jpg", "scene0307_02/00720.jpg", "scene0307_02/00760.jpg" ], "turn_id": 4, "scan_id": "scene0307_02", "origin_question": "Which two objects are on the same side of you now? \"the medium-sized white stainless steel refrigerator\", \"the clothes\", \"the radiator\". ", "option": [ "the medium-sized white stainless steel refrigerator, the clothes", "the medium-sized white stainless steel refrigerator, the radiator", "the clothes, the radiator" ], "answer": "the clothes, the radiator", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the medium-sized white stainless steel refrigerator\", \"the clothes\", \"the radiator\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the medium-sized white stainless steel refrigerator, the clothes', 'the medium-sized white stainless steel refrigerator, the radiator', 'the clothes, the radiator'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0307_02/00800.jpg", "scene0307_02/00840.jpg", "scene0307_02/00880.jpg", "scene0307_02/00920.jpg", "scene0307_02/00960.jpg" ], "turn_id": 5, "scan_id": "scene0307_02", "origin_question": "", "option": [], "answer": "", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0307_02/01000.jpg", "scene0307_02/01040.jpg", "scene0307_02/01080.jpg", "scene0307_02/01120.jpg", "scene0307_02/01160.jpg" ], "turn_id": 6, "scan_id": "scene0307_02", "origin_question": "In which turn did you see both \"the shower\" and \"the radiator\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the shower\" and \"the radiator\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0307_02/01200.jpg", "scene0307_02/01240.jpg", "scene0307_02/01280.jpg", "scene0307_02/01320.jpg", "scene0307_02/01360.jpg" ], "turn_id": 7, "scan_id": "scene0307_02", "origin_question": "Which direction is the white rectangular metal washing machine in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the white rectangular metal washing machine in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0307_02/01400.jpg", "scene0307_02/01440.jpg", "scene0307_02/01480.jpg", "scene0307_02/01520.jpg", "scene0307_02/01560.jpg" ], "turn_id": 8, "scan_id": "scene0307_02", "origin_question": "Please recall \"the radiator\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "4.76", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the radiator\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0307_02/01600.jpg", "scene0307_02/01640.jpg", "scene0307_02/01680.jpg", "scene0307_02/01720.jpg", "scene0307_02/01760.jpg" ], "turn_id": 9, "scan_id": "scene0307_02", "origin_question": "Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0307_02/01800.jpg", "scene0307_02/01840.jpg", "scene0307_02/01880.jpg", "scene0307_02/01920.jpg", "scene0307_02/01960.jpg" ], "turn_id": 10, "scan_id": "scene0307_02", "origin_question": "Using your orientation at the end of turn 9 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 9 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0421_02/00000.jpg", "scene0421_02/00040.jpg", "scene0421_02/00080.jpg", "scene0421_02/00120.jpg", "scene0421_02/00160.jpg" ], "turn_id": 1, "scan_id": "scene0421_02", "origin_question": "Remember, how many pillar(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many pillar(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0421_02/00200.jpg", "scene0421_02/00240.jpg", "scene0421_02/00280.jpg", "scene0421_02/00320.jpg", "scene0421_02/00360.jpg" ], "turn_id": 2, "scan_id": "scene0421_02", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0421_02/00400.jpg", "scene0421_02/00440.jpg", "scene0421_02/00480.jpg", "scene0421_02/00520.jpg", "scene0421_02/00560.jpg" ], "turn_id": 3, "scan_id": "scene0421_02", "origin_question": "In which turn did you see both \"the white front-loading washing machine with a stainless steel drum\" and \"the clothes dryer\" simultaneously? ", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the white front-loading washing machine with a stainless steel drum\" and \"the clothes dryer\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0421_02/00600.jpg", "scene0421_02/00640.jpg", "scene0421_02/00680.jpg", "scene0421_02/00720.jpg", "scene0421_02/00760.jpg" ], "turn_id": 4, "scan_id": "scene0421_02", "origin_question": "Is the rectangular white wooden door to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the rectangular white wooden door to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0421_02/00800.jpg", "scene0421_02/00840.jpg", "scene0421_02/00880.jpg", "scene0421_02/00920.jpg", "scene0421_02/00960.jpg" ], "turn_id": 5, "scan_id": "scene0421_02", "origin_question": "Please recall \"the rectangular white wooden door\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.11", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the rectangular white wooden door\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0421_02/01000.jpg", "scene0421_02/01040.jpg", "scene0421_02/01080.jpg", "scene0421_02/01120.jpg", "scene0421_02/01160.jpg" ], "turn_id": 6, "scan_id": "scene0421_02", "origin_question": "Is the round brown wicker container with a sturdy handle to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the round brown wicker container with a sturdy handle to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0421_02/01200.jpg", "scene0421_02/01240.jpg", "scene0421_02/01280.jpg", "scene0421_02/01320.jpg", "scene0421_02/01360.jpg" ], "turn_id": 7, "scan_id": "scene0421_02", "origin_question": "Using your orientation at the end of turn 4 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "92.68", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 4 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0421_02/01400.jpg", "scene0421_02/01440.jpg", "scene0421_02/01480.jpg", "scene0421_02/01520.jpg", "scene0421_02/01560.jpg" ], "turn_id": 8, "scan_id": "scene0421_02", "origin_question": "", "option": [], "answer": "", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0421_02/01600.jpg", "scene0421_02/01640.jpg", "scene0421_02/01680.jpg", "scene0421_02/01720.jpg", "scene0421_02/01760.jpg" ], "turn_id": 9, "scan_id": "scene0421_02", "origin_question": "Which direction is the clothes dryer in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the clothes dryer in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0421_02/01800.jpg", "scene0421_02/01840.jpg", "scene0421_02/01880.jpg", "scene0421_02/01920.jpg", "scene0421_02/01960.jpg" ], "turn_id": 10, "scan_id": "scene0421_02", "origin_question": "Which two objects are on the same side of you now? \"the rectangular white wooden door\", \"the round brown hamper made of sturdy material\", \"the clothes dryer\". ", "option": [ "the rectangular white wooden door, the round brown hamper made of sturdy material", "the rectangular white wooden door, the clothes dryer", "the round brown hamper made of sturdy material, the clothes dryer" ], "answer": "the rectangular white wooden door, the round brown hamper made of sturdy material", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the rectangular white wooden door\", \"the round brown hamper made of sturdy material\", \"the clothes dryer\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular white wooden door, the round brown hamper made of sturdy material', 'the rectangular white wooden door, the clothes dryer', 'the round brown hamper made of sturdy material, the clothes dryer'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0011_01/00000.jpg", "scene0011_01/00040.jpg", "scene0011_01/00080.jpg", "scene0011_01/00120.jpg", "scene0011_01/00160.jpg" ], "turn_id": 1, "scan_id": "scene0011_01", "origin_question": "Remember, have you seen any chair(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any chair(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0011_01/00200.jpg", "scene0011_01/00240.jpg", "scene0011_01/00280.jpg", "scene0011_01/00320.jpg", "scene0011_01/00360.jpg" ], "turn_id": 2, "scan_id": "scene0011_01", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the small rectangular gray plastic bin with a lid\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the small rectangular gray plastic bin with a lid\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0011_01/00400.jpg", "scene0011_01/00440.jpg", "scene0011_01/00480.jpg", "scene0011_01/00520.jpg", "scene0011_01/00560.jpg" ], "turn_id": 3, "scan_id": "scene0011_01", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0011_01/00600.jpg", "scene0011_01/00640.jpg", "scene0011_01/00680.jpg", "scene0011_01/00720.jpg", "scene0011_01/00760.jpg" ], "turn_id": 4, "scan_id": "scene0011_01", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "2.68", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0011_01/00800.jpg", "scene0011_01/00840.jpg", "scene0011_01/00880.jpg", "scene0011_01/00920.jpg", "scene0011_01/00960.jpg" ], "turn_id": 5, "scan_id": "scene0011_01", "origin_question": "In which turn did you see both \"the smooth, shiny granite or quartz counter\" and \"the stainless steel modern refrigerator with a smooth surface\" simultaneously? ", "option": [], "answer": "5", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the smooth, shiny granite or quartz counter\" and \"the stainless steel modern refrigerator with a smooth surface\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0011_01/01000.jpg", "scene0011_01/01040.jpg", "scene0011_01/01080.jpg", "scene0011_01/01120.jpg", "scene0011_01/01160.jpg" ], "turn_id": 6, "scan_id": "scene0011_01", "origin_question": "In which turn did you see both \"the rectangular brown wooden furniture with a simple design\" and \"the stainless steel modern refrigerator with a smooth surface\" simultaneously? ", "option": [], "answer": "5", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the rectangular brown wooden furniture with a simple design\" and \"the stainless steel modern refrigerator with a smooth surface\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0011_01/01200.jpg", "scene0011_01/01240.jpg", "scene0011_01/01280.jpg", "scene0011_01/01320.jpg", "scene0011_01/01360.jpg" ], "turn_id": 7, "scan_id": "scene0011_01", "origin_question": "When did you first discover \"the stainless steel modern refrigerator with a smooth surface\"? (index of the turn)", "option": [], "answer": "5", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the stainless steel modern refrigerator with a smooth surface\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0011_01/01400.jpg", "scene0011_01/01440.jpg", "scene0011_01/01480.jpg", "scene0011_01/01520.jpg", "scene0011_01/01560.jpg" ], "turn_id": 8, "scan_id": "scene0011_01", "origin_question": "In which turn were you closest to \"the rectangular brown wooden furniture with a simple design\"? ", "option": [], "answer": "4", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the rectangular brown wooden furniture with a simple design\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0011_01/01600.jpg", "scene0011_01/01640.jpg", "scene0011_01/01680.jpg", "scene0011_01/01720.jpg", "scene0011_01/01760.jpg" ], "turn_id": 9, "scan_id": "scene0011_01", "origin_question": "In which turn were you closest to \"the long, rectangular wooden table with a deep brown surface\"? ", "option": [], "answer": "3", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the long, rectangular wooden table with a deep brown surface\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "scene0011_01/01800.jpg", "scene0011_01/01840.jpg", "scene0011_01/01880.jpg", "scene0011_01/01920.jpg", "scene0011_01/01960.jpg" ], "turn_id": 10, "scan_id": "scene0011_01", "origin_question": "Among these three objects(\"the long, rectangular wooden table with a deep brown surface\",\"the black wall-mounted flat-screen TV\",\"the sleek rectangular stainless steel sink in silver\" ), which one is the closest to you?", "option": [ "the long, rectangular wooden table with a deep brown surface", "the black wall-mounted flat-screen TV", "the sleek rectangular stainless steel sink in silver" ], "answer": "the long, rectangular wooden table with a deep brown surface", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the long, rectangular wooden table with a deep brown surface\",\"the black wall-mounted flat-screen TV\",\"the sleek rectangular stainless steel sink in silver\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the long, rectangular wooden table with a deep brown surface', 'the black wall-mounted flat-screen TV', 'the sleek rectangular stainless steel sink in silver'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "1mp3d_0044_region37/cc7ea01bcb0b4d358b8b34625e11d843_i1_4.jpg", "1mp3d_0044_region37/cc7ea01bcb0b4d358b8b34625e11d843_i2_4.jpg", "1mp3d_0044_region37/ed1015d3d3f84c79885ded306617eeb4_i1_5.jpg", "1mp3d_0044_region37/ed1015d3d3f84c79885ded306617eeb4_i2_5.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0044_region37", "origin_question": "Remember, how many spoon(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many spoon(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 2 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0044_region37/ed1015d3d3f84c79885ded306617eeb4_i1_0.jpg", "1mp3d_0044_region37/ed1015d3d3f84c79885ded306617eeb4_i2_0.jpg", "1mp3d_0044_region37/cc7ea01bcb0b4d358b8b34625e11d843_i1_1.jpg", "1mp3d_0044_region37/cc7ea01bcb0b4d358b8b34625e11d843_i2_1.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0044_region37", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 2 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0067_02/00000.jpg", "scene0067_02/00050.jpg", "scene0067_02/00080.jpg", "scene0067_02/00110.jpg" ], "turn_id": 1, "scan_id": "scene0067_02", "origin_question": "Remember, have you seen any bread(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any bread(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0067_02/00400.jpg", "scene0067_02/00430.jpg", "scene0067_02/00460.jpg", "scene0067_02/00490.jpg" ], "turn_id": 2, "scan_id": "scene0067_02", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the light\" from your position?", "option": [], "answer": "98.09", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the light\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0067_02/00520.jpg", "scene0067_02/00550.jpg", "scene0067_02/00590.jpg", "scene0067_02/00620.jpg" ], "turn_id": 3, "scan_id": "scene0067_02", "origin_question": "Which two objects are on the same side of you now? \"the rectangular glass mirror with a reflective surface\", \"the light\", \"the white ceramic toilet with a closed seat\". ", "option": [ "the rectangular glass mirror with a reflective surface, the light", "the rectangular glass mirror with a reflective surface, the white ceramic toilet with a closed seat", "the light, the white ceramic toilet with a closed seat" ], "answer": "the rectangular glass mirror with a reflective surface, the light", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the rectangular glass mirror with a reflective surface\", \"the light\", \"the white ceramic toilet with a closed seat\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular glass mirror with a reflective surface, the light', 'the rectangular glass mirror with a reflective surface, the white ceramic toilet with a closed seat', 'the light, the white ceramic toilet with a closed seat'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0067_02/00650.jpg", "scene0067_02/00680.jpg", "scene0067_02/00710.jpg", "scene0067_02/00740.jpg" ], "turn_id": 4, "scan_id": "scene0067_02", "origin_question": "Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0067_02/00940.jpg", "scene0067_02/00970.jpg", "scene0067_02/01000.jpg", "scene0067_02/01030.jpg" ], "turn_id": 5, "scan_id": "scene0067_02", "origin_question": "When did you first discover \"the rectangular glass mirror with a reflective surface\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the rectangular glass mirror with a reflective surface\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0609_01/00000.jpg", "scene0609_01/00040.jpg", "scene0609_01/00080.jpg", "scene0609_01/00120.jpg" ], "turn_id": 1, "scan_id": "scene0609_01", "origin_question": "Remember, how many chair(s) have you seen so far? ", "option": [], "answer": "3", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many chair(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0609_01/00160.jpg", "scene0609_01/00200.jpg", "scene0609_01/00240.jpg", "scene0609_01/00280.jpg" ], "turn_id": 2, "scan_id": "scene0609_01", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0609_01/00320.jpg", "scene0609_01/00360.jpg", "scene0609_01/00400.jpg", "scene0609_01/00440.jpg" ], "turn_id": 3, "scan_id": "scene0609_01", "origin_question": "Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "91.79", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0609_01/00480.jpg", "scene0609_01/00520.jpg", "scene0609_01/00560.jpg", "scene0609_01/00600.jpg" ], "turn_id": 4, "scan_id": "scene0609_01", "origin_question": "How far is your current position from where you were at the end of the turn 1? (in meters)", "option": [], "answer": "2.11", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 1? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0609_01/00640.jpg", "scene0609_01/00680.jpg", "scene0609_01/00720.jpg", "scene0609_01/00760.jpg" ], "turn_id": 5, "scan_id": "scene0609_01", "origin_question": "Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0609_01/00800.jpg", "scene0609_01/00840.jpg", "scene0609_01/00880.jpg", "scene0609_01/00920.jpg" ], "turn_id": 6, "scan_id": "scene0609_01", "origin_question": "", "option": [], "answer": "", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "1mp3d_0067_region8/3317348467784b19925a007204cb7f17_i1_5.jpg", "1mp3d_0067_region8/3317348467784b19925a007204cb7f17_i2_5.jpg", "1mp3d_0067_region8/4382fba86a6a4956b7007f079a990268_i1_0.jpg", "1mp3d_0067_region8/4382fba86a6a4956b7007f079a990268_i2_0.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0067_region8", "origin_question": "Remember, how many picture(s) have you seen so far? ", "option": [], "answer": "2", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many picture(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "1mp3d_0067_region8/0260ad22c1f64b88b7c7ffa843d86aa8_i1_0.jpg", "1mp3d_0067_region8/0260ad22c1f64b88b7c7ffa843d86aa8_i2_0.jpg", "1mp3d_0067_region8/caa5b0dce8fc45a8982f917d041cf441_i1_5.jpg", "1mp3d_0067_region8/caa5b0dce8fc45a8982f917d041cf441_i2_5.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0067_region8", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the rectangular sliding glass door with a blue frame\"; \"the brown deer antler chandelier with candle-like bulbs\"; \"the large rectangular dark brown window with a wooden frame\" ", "option": [ "the rectangular sliding glass door with a blue frame", "the brown deer antler chandelier with candle-like bulbs", "the large rectangular dark brown window with a wooden frame" ], "answer": "the large rectangular dark brown window with a wooden frame", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the rectangular sliding glass door with a blue frame\"; \"the brown deer antler chandelier with candle-like bulbs\"; \"the large rectangular dark brown window with a wooden frame\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular sliding glass door with a blue frame', 'the brown deer antler chandelier with candle-like bulbs', 'the large rectangular dark brown window with a wooden frame'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "1mp3d_0067_region8/e394a9eecb63432b804ebb2e96a563b5_i1_4.jpg", "1mp3d_0067_region8/e394a9eecb63432b804ebb2e96a563b5_i2_4.jpg", "1mp3d_0067_region8/24a5c4d4de2e47fba4cd35a2cf63bafe_i1_5.jpg", "1mp3d_0067_region8/24a5c4d4de2e47fba4cd35a2cf63bafe_i2_5.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0067_region8", "origin_question": "When did you first discover \"the sink\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the sink\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0067_region8/24a5c4d4de2e47fba4cd35a2cf63bafe_i1_0.jpg", "1mp3d_0067_region8/24a5c4d4de2e47fba4cd35a2cf63bafe_i2_0.jpg", "1mp3d_0067_region8/24a5c4d4de2e47fba4cd35a2cf63bafe_i1_4.jpg", "1mp3d_0067_region8/24a5c4d4de2e47fba4cd35a2cf63bafe_i2_4.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0067_region8", "origin_question": "Remember, have you seen any sack(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any sack(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "1mp3d_0067_region8/e394a9eecb63432b804ebb2e96a563b5_i1_0.jpg", "1mp3d_0067_region8/e394a9eecb63432b804ebb2e96a563b5_i2_0.jpg", "1mp3d_0067_region8/e394a9eecb63432b804ebb2e96a563b5_i1_5.jpg", "1mp3d_0067_region8/e394a9eecb63432b804ebb2e96a563b5_i2_5.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0067_region8", "origin_question": "Which direction is the sink in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the sink in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "1mp3d_0067_region8/caa5b0dce8fc45a8982f917d041cf441_i1_3.jpg", "1mp3d_0067_region8/caa5b0dce8fc45a8982f917d041cf441_i2_3.jpg", "1mp3d_0067_region8/caa5b0dce8fc45a8982f917d041cf441_i1_4.jpg", "1mp3d_0067_region8/caa5b0dce8fc45a8982f917d041cf441_i2_4.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0067_region8", "origin_question": "What will be the first-time appearance order of the following categories: beam, chair, stool?", "option": [ "chair, stool, beam", "stool, chair, beam", "chair, beam, stool", "stool, beam, chair" ], "answer": "chair, beam, stool", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: beam, chair, stool? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['chair, stool, beam', 'stool, chair, beam', 'chair, beam, stool', 'stool, beam, chair'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0067_region8/0260ad22c1f64b88b7c7ffa843d86aa8_i1_5.jpg", "1mp3d_0067_region8/0260ad22c1f64b88b7c7ffa843d86aa8_i2_5.jpg", "1mp3d_0067_region8/4382fba86a6a4956b7007f079a990268_i1_1.jpg", "1mp3d_0067_region8/4382fba86a6a4956b7007f079a990268_i2_1.jpg" ], "turn_id": 7, "scan_id": "1mp3d_0067_region8", "origin_question": "Compared to your position at the end of the 5 turn, are you now closer or farther away from \"the sink\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 5 turn, are you now closer or farther away from \"the sink\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0057_region18/79103bf93d81464786b23dc26991d23e_i1_5.jpg", "1mp3d_0057_region18/79103bf93d81464786b23dc26991d23e_i2_5.jpg", "1mp3d_0057_region18/1e27234cce4e499b853cc41866186663_i1_1.jpg", "1mp3d_0057_region18/1e27234cce4e499b853cc41866186663_i2_1.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0057_region18", "origin_question": "Remember, have you seen any door(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any door(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0057_region18/1e27234cce4e499b853cc41866186663_i2_2.jpg", "1mp3d_0057_region18/1e27234cce4e499b853cc41866186663_i2_4.jpg", "1mp3d_0057_region18/1e27234cce4e499b853cc41866186663_i1_0.jpg", "1mp3d_0057_region18/1e27234cce4e499b853cc41866186663_i2_0.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0057_region18", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "1mp3d_0057_region18/1e27234cce4e499b853cc41866186663_i1_3.jpg", "1mp3d_0057_region18/1e27234cce4e499b853cc41866186663_i2_3.jpg", "1mp3d_0057_region18/79103bf93d81464786b23dc26991d23e_i1_4.jpg", "1mp3d_0057_region18/79103bf93d81464786b23dc26991d23e_i2_4.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0057_region18", "origin_question": "Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "96.21", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "1mp3d_0057_region18/79103bf93d81464786b23dc26991d23e_i1_3.jpg", "1mp3d_0057_region18/79103bf93d81464786b23dc26991d23e_i2_3.jpg", "1mp3d_0057_region18/71834224b28f45aca00f1c19f2ba0668_i1_2.jpg", "1mp3d_0057_region18/71834224b28f45aca00f1c19f2ba0668_i2_2.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0057_region18", "origin_question": "By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular wooden door with a light brown hue\", Y: \"the fluffy, white cotton bath towel\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to X and farther away from Y", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular wooden door with a light brown hue\", Y: \"the fluffy, white cotton bath towel\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0057_region18/71834224b28f45aca00f1c19f2ba0668_i1_3.jpg", "1mp3d_0057_region18/71834224b28f45aca00f1c19f2ba0668_i2_3.jpg", "1mp3d_0057_region18/71834224b28f45aca00f1c19f2ba0668_i2_5.jpg", "1mp3d_0057_region18/71834224b28f45aca00f1c19f2ba0668_i1_4.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0057_region18", "origin_question": "Remember, have you seen any bathtub(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any bathtub(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0111_00/00000.jpg", "scene0111_00/00040.jpg", "scene0111_00/00080.jpg", "scene0111_00/00120.jpg", "scene0111_00/00160.jpg" ], "turn_id": 1, "scan_id": "scene0111_00", "origin_question": "Remember, have you seen any candlestick(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any candlestick(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0111_00/00200.jpg", "scene0111_00/00240.jpg", "scene0111_00/00280.jpg", "scene0111_00/00320.jpg", "scene0111_00/00360.jpg" ], "turn_id": 2, "scan_id": "scene0111_00", "origin_question": "In which turn were you farthest from \"the rectangular framed picture of bicycles\"? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the rectangular framed picture of bicycles\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0111_00/00400.jpg", "scene0111_00/00440.jpg", "scene0111_00/00480.jpg", "scene0111_00/00520.jpg", "scene0111_00/00560.jpg" ], "turn_id": 3, "scan_id": "scene0111_00", "origin_question": "Is the standard cylindrical white cup made of plastic or ceramic to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the standard cylindrical white cup made of plastic or ceramic to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0111_00/00600.jpg", "scene0111_00/00760.jpg", "scene0111_00/00800.jpg", "scene0111_00/00840.jpg", "scene0111_00/00880.jpg" ], "turn_id": 4, "scan_id": "scene0111_00", "origin_question": "Which two objects are on the same side of you now? \"the large rectangular flat-screen TV with matte finish\", \"the modern compact black microwave with a shiny surface\", \"the round red plastic bowl with a smooth surface\". ", "option": [ "the large rectangular flat-screen TV with matte finish, the modern compact black microwave with a shiny surface", "the large rectangular flat-screen TV with matte finish, the round red plastic bowl with a smooth surface", "the modern compact black microwave with a shiny surface, the round red plastic bowl with a smooth surface" ], "answer": "the modern compact black microwave with a shiny surface, the round red plastic bowl with a smooth surface", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the large rectangular flat-screen TV with matte finish\", \"the modern compact black microwave with a shiny surface\", \"the round red plastic bowl with a smooth surface\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the large rectangular flat-screen TV with matte finish, the modern compact black microwave with a shiny surface', 'the large rectangular flat-screen TV with matte finish, the round red plastic bowl with a smooth surface', 'the modern compact black microwave with a shiny surface, the round red plastic bowl with a smooth surface'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0111_00/00920.jpg", "scene0111_00/00960.jpg", "scene0111_00/01000.jpg", "scene0111_00/01040.jpg", "scene0111_00/01080.jpg" ], "turn_id": 5, "scan_id": "scene0111_00", "origin_question": "In which turn did you see both \"the toaster\" and \"the refrigerator\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the toaster\" and \"the refrigerator\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0111_00/01120.jpg", "scene0111_00/01160.jpg", "scene0111_00/01200.jpg", "scene0111_00/01240.jpg", "scene0111_00/01280.jpg" ], "turn_id": 6, "scan_id": "scene0111_00", "origin_question": "Which direction is the toaster in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the toaster in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0111_00/01320.jpg", "scene0111_00/01360.jpg", "scene0111_00/01400.jpg", "scene0111_00/01440.jpg", "scene0111_00/01480.jpg" ], "turn_id": 7, "scan_id": "scene0111_00", "origin_question": "Which two objects are on the same side of you now? \"the standard cylindrical white cup made of plastic or ceramic\", \"the rectangular framed picture of bicycles\", \"the round red plastic bowl with a smooth surface\". ", "option": [ "the standard cylindrical white cup made of plastic or ceramic, the rectangular framed picture of bicycles", "the standard cylindrical white cup made of plastic or ceramic, the round red plastic bowl with a smooth surface", "the rectangular framed picture of bicycles, the round red plastic bowl with a smooth surface" ], "answer": "the rectangular framed picture of bicycles, the round red plastic bowl with a smooth surface", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the standard cylindrical white cup made of plastic or ceramic\", \"the rectangular framed picture of bicycles\", \"the round red plastic bowl with a smooth surface\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the standard cylindrical white cup made of plastic or ceramic, the rectangular framed picture of bicycles', 'the standard cylindrical white cup made of plastic or ceramic, the round red plastic bowl with a smooth surface', 'the rectangular framed picture of bicycles, the round red plastic bowl with a smooth surface'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0111_00/01520.jpg", "scene0111_00/01560.jpg", "scene0111_00/01600.jpg", "scene0111_00/01640.jpg", "scene0111_00/01680.jpg" ], "turn_id": 8, "scan_id": "scene0111_00", "origin_question": "Which direction is the versatile pair of red-handled metal scissors in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the versatile pair of red-handled metal scissors in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0111_00/01720.jpg", "scene0111_00/01760.jpg", "scene0111_00/01800.jpg", "scene0111_00/01840.jpg", "scene0111_00/01880.jpg" ], "turn_id": 9, "scan_id": "scene0111_00", "origin_question": "Which two objects are on the same side of you now? \"the round red plastic bowl with a smooth surface\", \"the large rectangular flat-screen TV with matte finish\", \"the rectangular silver wire dish rack\". ", "option": [ "the round red plastic bowl with a smooth surface, the large rectangular flat-screen TV with matte finish", "the round red plastic bowl with a smooth surface, the rectangular silver wire dish rack", "the large rectangular flat-screen TV with matte finish, the rectangular silver wire dish rack" ], "answer": "the round red plastic bowl with a smooth surface, the large rectangular flat-screen TV with matte finish", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the round red plastic bowl with a smooth surface\", \"the large rectangular flat-screen TV with matte finish\", \"the rectangular silver wire dish rack\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the round red plastic bowl with a smooth surface, the large rectangular flat-screen TV with matte finish', 'the round red plastic bowl with a smooth surface, the rectangular silver wire dish rack', 'the large rectangular flat-screen TV with matte finish, the rectangular silver wire dish rack'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0111_00/01920.jpg", "scene0111_00/01960.jpg", "scene0111_00/02000.jpg", "scene0111_00/02040.jpg", "scene0111_00/02080.jpg" ], "turn_id": 10, "scan_id": "scene0111_00", "origin_question": "Is the rectangular wooden door with a smooth light-colored surface to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the rectangular wooden door with a smooth light-colored surface to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0078_region7/2296da968e464b8db6e6bdba81fbe274_i1_4.jpg", "1mp3d_0078_region7/2296da968e464b8db6e6bdba81fbe274_i2_4.jpg", "1mp3d_0078_region7/6683adca11334414925158c4d21105b1_i1_0.jpg", "1mp3d_0078_region7/6683adca11334414925158c4d21105b1_i2_0.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0078_region7", "origin_question": "", "option": [], "answer": "", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "1mp3d_0078_region7/6683adca11334414925158c4d21105b1_i1_4.jpg", "1mp3d_0078_region7/6683adca11334414925158c4d21105b1_i2_4.jpg", "1mp3d_0078_region7/36495e0d819442c4a0666597566b3fa2_i1_3.jpg", "1mp3d_0078_region7/36495e0d819442c4a0666597566b3fa2_i2_3.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0078_region7", "origin_question": "Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Right", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "1mp3d_0078_region7/6683adca11334414925158c4d21105b1_i1_2.jpg", "1mp3d_0078_region7/6683adca11334414925158c4d21105b1_i2_2.jpg", "1mp3d_0078_region7/2296da968e464b8db6e6bdba81fbe274_i1_2.jpg", "1mp3d_0078_region7/2296da968e464b8db6e6bdba81fbe274_i2_2.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0078_region7", "origin_question": "How far is your current position from where you were at the end of the turn 1? (in meters)", "option": [], "answer": "1.21", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 1? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "1mp3d_0078_region7/59bb2a23ea8547ee9a9d05cb436a621b_i1_2.jpg", "1mp3d_0078_region7/59bb2a23ea8547ee9a9d05cb436a621b_i2_2.jpg", "1mp3d_0078_region7/59bb2a23ea8547ee9a9d05cb436a621b_i1_1.jpg", "1mp3d_0078_region7/59bb2a23ea8547ee9a9d05cb436a621b_i2_1.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0078_region7", "origin_question": "Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance forward or backward from that position?", "option": [ "Forward", "Backward" ], "answer": "Forward", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance forward or backward from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Forward', 'Backward'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "1mp3d_0078_region7/2296da968e464b8db6e6bdba81fbe274_i1_3.jpg", "1mp3d_0078_region7/2296da968e464b8db6e6bdba81fbe274_i2_3.jpg", "1mp3d_0078_region7/59bb2a23ea8547ee9a9d05cb436a621b_i1_4.jpg", "1mp3d_0078_region7/59bb2a23ea8547ee9a9d05cb436a621b_i2_4.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0078_region7", "origin_question": "Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "122.63", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0414_00/00000.jpg", "scene0414_00/00030.jpg", "scene0414_00/00060.jpg", "scene0414_00/00090.jpg" ], "turn_id": 1, "scan_id": "scene0414_00", "origin_question": "Remember, have you seen any folder(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any folder(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0414_00/00120.jpg", "scene0414_00/00540.jpg", "scene0414_00/00570.jpg", "scene0414_00/00600.jpg" ], "turn_id": 2, "scan_id": "scene0414_00", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the bright red medium-sized backpack with black straps\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the bright red medium-sized backpack with black straps\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0414_00/00630.jpg", "scene0414_00/00660.jpg", "scene0414_00/00690.jpg", "scene0414_00/00720.jpg" ], "turn_id": 3, "scan_id": "scene0414_00", "origin_question": "In which turn did you see both \"the bright red medium-sized backpack with black straps\" and \"the blackboard\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the bright red medium-sized backpack with black straps\" and \"the blackboard\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0414_00/00750.jpg", "scene0414_00/00780.jpg", "scene0414_00/00810.jpg", "scene0414_00/00840.jpg" ], "turn_id": 4, "scan_id": "scene0414_00", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the bright red medium-sized backpack with black straps\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the bright red medium-sized backpack with black straps\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0414_00/00870.jpg", "scene0414_00/00900.jpg", "scene0414_00/00930.jpg", "scene0414_00/00960.jpg" ], "turn_id": 5, "scan_id": "scene0414_00", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the small rectangular light brown wooden cabinet\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the small rectangular light brown wooden cabinet\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0414_00/00990.jpg", "scene0414_00/01020.jpg", "scene0414_00/01050.jpg", "scene0414_00/01080.jpg" ], "turn_id": 6, "scan_id": "scene0414_00", "origin_question": "By the end of the 3 turn, were you getting closer to or farther away from these two objects: X: \"the small rectangular light brown wooden cabinet\", Y: \"the bin\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 3 turn, were you getting closer to or farther away from these two objects: X: \"the small rectangular light brown wooden cabinet\", Y: \"the bin\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0423_02/00000.jpg", "scene0423_02/00030.jpg", "scene0423_02/00060.jpg", "scene0423_02/00090.jpg" ], "turn_id": 1, "scan_id": "scene0423_02", "origin_question": "Remember, have you seen any chair(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any chair(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0423_02/00120.jpg", "scene0423_02/00150.jpg", "scene0423_02/00180.jpg", "scene0423_02/00210.jpg" ], "turn_id": 2, "scan_id": "scene0423_02", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0423_02/00240.jpg", "scene0423_02/00270.jpg", "scene0423_02/00300.jpg", "scene0423_02/00330.jpg" ], "turn_id": 3, "scan_id": "scene0423_02", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0423_02/00360.jpg", "scene0423_02/00390.jpg", "scene0423_02/00420.jpg", "scene0423_02/00450.jpg" ], "turn_id": 4, "scan_id": "scene0423_02", "origin_question": "Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "57.63", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0423_02/00480.jpg", "scene0423_02/00510.jpg", "scene0423_02/00540.jpg", "scene0423_02/00570.jpg" ], "turn_id": 5, "scan_id": "scene0423_02", "origin_question": "", "option": [], "answer": "", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0423_02/00600.jpg", "scene0423_02/00630.jpg", "scene0423_02/00660.jpg", "scene0423_02/00690.jpg" ], "turn_id": 6, "scan_id": "scene0423_02", "origin_question": "Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0423_02/00720.jpg", "scene0423_02/00750.jpg", "scene0423_02/00780.jpg", "scene0423_02/00810.jpg" ], "turn_id": 7, "scan_id": "scene0423_02", "origin_question": "Using your orientation at the end of turn 6 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "67.56", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 6 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0067_region1/2291e23cde8e4eaebe6b9fe9cbdeb888_i1_0.jpg", "1mp3d_0067_region1/2291e23cde8e4eaebe6b9fe9cbdeb888_i2_0.jpg", "1mp3d_0067_region1/8dc2bb01c486457b9a0348c413d62f67_i1_5.jpg", "1mp3d_0067_region1/8dc2bb01c486457b9a0348c413d62f67_i2_5.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0067_region1", "origin_question": "Remember, have you seen any window(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any window(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "1mp3d_0067_region1/8dc2bb01c486457b9a0348c413d62f67_i1_1.jpg", "1mp3d_0067_region1/8dc2bb01c486457b9a0348c413d62f67_i2_1.jpg", "1mp3d_0067_region1/8dc2bb01c486457b9a0348c413d62f67_i1_2.jpg", "1mp3d_0067_region1/8dc2bb01c486457b9a0348c413d62f67_i2_2.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0067_region1", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the rectangular woven couch with a high backrest\"; \"the light beige rectangular glass window with a wooden frame\"; \"the bed\" ", "option": [ "the rectangular woven couch with a high backrest", "the light beige rectangular glass window with a wooden frame", "the bed" ], "answer": "the rectangular woven couch with a high backrest", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the rectangular woven couch with a high backrest\"; \"the light beige rectangular glass window with a wooden frame\"; \"the bed\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular woven couch with a high backrest', 'the light beige rectangular glass window with a wooden frame', 'the bed'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "1mp3d_0067_region1/fa1974a668c94fa393eccece0be0ede3_i1_0.jpg", "1mp3d_0067_region1/fa1974a668c94fa393eccece0be0ede3_i2_0.jpg", "1mp3d_0067_region1/58f71a450eed4d878a9b9e6a67ad6f39_i1_0.jpg", "1mp3d_0067_region1/58f71a450eed4d878a9b9e6a67ad6f39_i2_0.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0067_region1", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "2.77", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "1mp3d_0067_region1/58f71a450eed4d878a9b9e6a67ad6f39_i1_1.jpg", "1mp3d_0067_region1/58f71a450eed4d878a9b9e6a67ad6f39_i2_1.jpg", "1mp3d_0067_region1/fa1974a668c94fa393eccece0be0ede3_i1_5.jpg", "1mp3d_0067_region1/fa1974a668c94fa393eccece0be0ede3_i2_5.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0067_region1", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "1mp3d_0067_region1/fa1974a668c94fa393eccece0be0ede3_i1_4.jpg", "1mp3d_0067_region1/fa1974a668c94fa393eccece0be0ede3_i2_4.jpg", "1mp3d_0067_region1/fa1974a668c94fa393eccece0be0ede3_i1_3.jpg", "1mp3d_0067_region1/fa1974a668c94fa393eccece0be0ede3_i2_3.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0067_region1", "origin_question": "At the end of which turn were both of \"the light beige rectangular glass window with a wooden frame\" and \"the rectangular woven couch with a high backrest\" on your right side?", "option": [], "answer": "3", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the light beige rectangular glass window with a wooden frame\" and \"the rectangular woven couch with a high backrest\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "1mp3d_0067_region1/1674e74d24cc478e832b252de024bc9b_i1_4.jpg", "1mp3d_0067_region1/1674e74d24cc478e832b252de024bc9b_i2_4.jpg", "1mp3d_0067_region1/1674e74d24cc478e832b252de024bc9b_i1_5.jpg", "1mp3d_0067_region1/1674e74d24cc478e832b252de024bc9b_i2_5.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0067_region1", "origin_question": "At the end of which turn were both of \"the light beige rectangular glass window with a wooden frame\" and \"the bed\" on your left side?", "option": [], "answer": "2", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the light beige rectangular glass window with a wooden frame\" and \"the bed\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0315_00/00000.jpg", "scene0315_00/00040.jpg", "scene0315_00/00080.jpg", "scene0315_00/00120.jpg" ], "turn_id": 1, "scan_id": "scene0315_00", "origin_question": "Remember, have you seen any knife(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any knife(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0315_00/00160.jpg", "scene0315_00/00200.jpg", "scene0315_00/00240.jpg", "scene0315_00/00280.jpg" ], "turn_id": 2, "scan_id": "scene0315_00", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the black rectangular wood coffee table with a smooth surface\"; \"the rectangular white velvet ottoman\"; \"the dark brown rectangular wooden cabinet\" ", "option": [ "the black rectangular wood coffee table with a smooth surface", "the rectangular white velvet ottoman", "the dark brown rectangular wooden cabinet" ], "answer": "the rectangular white velvet ottoman", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the black rectangular wood coffee table with a smooth surface\"; \"the rectangular white velvet ottoman\"; \"the dark brown rectangular wooden cabinet\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the black rectangular wood coffee table with a smooth surface', 'the rectangular white velvet ottoman', 'the dark brown rectangular wooden cabinet'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0315_00/00320.jpg", "scene0315_00/00360.jpg", "scene0315_00/00400.jpg", "scene0315_00/00440.jpg" ], "turn_id": 3, "scan_id": "scene0315_00", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the rectangular white velvet ottoman\"; \"the black rectangular wooden shelf with multiple layers\"; \"the long, wide white curtain made of lightweight fabric\" ", "option": [ "the rectangular white velvet ottoman", "the black rectangular wooden shelf with multiple layers", "the long, wide white curtain made of lightweight fabric" ], "answer": "the long, wide white curtain made of lightweight fabric", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the rectangular white velvet ottoman\"; \"the black rectangular wooden shelf with multiple layers\"; \"the long, wide white curtain made of lightweight fabric\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular white velvet ottoman', 'the black rectangular wooden shelf with multiple layers', 'the long, wide white curtain made of lightweight fabric'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0315_00/00480.jpg", "scene0315_00/00520.jpg", "scene0315_00/00560.jpg", "scene0315_00/00600.jpg" ], "turn_id": 4, "scan_id": "scene0315_00", "origin_question": "By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular brown wicker storage box with a lid\", Y: \"the rectangular white velvet ottoman\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "farther away from both X and Y", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 2 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular brown wicker storage box with a lid\", Y: \"the rectangular white velvet ottoman\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0315_00/00640.jpg", "scene0315_00/00680.jpg", "scene0315_00/00720.jpg", "scene0315_00/00760.jpg" ], "turn_id": 5, "scan_id": "scene0315_00", "origin_question": "Please recall \"the rectangular white velvet ottoman\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "3.03", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the rectangular white velvet ottoman\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0315_00/00800.jpg", "scene0315_00/00840.jpg", "scene0315_00/00880.jpg", "scene0315_00/00920.jpg" ], "turn_id": 6, "scan_id": "scene0315_00", "origin_question": "In which turn did you see both \"the rectangular brown wicker storage box with a lid\" and \"the rectangular white velvet ottoman\" simultaneously? ", "option": [], "answer": "6", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the rectangular brown wicker storage box with a lid\" and \"the rectangular white velvet ottoman\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0315_00/00960.jpg", "scene0315_00/01000.jpg", "scene0315_00/01040.jpg", "scene0315_00/01080.jpg" ], "turn_id": 7, "scan_id": "scene0315_00", "origin_question": "By the end of the 5 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular brown wicker storage box with a lid\", Y: \"the rectangular white velvet ottoman\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 5 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular brown wicker storage box with a lid\", Y: \"the rectangular white velvet ottoman\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0315_00/01120.jpg", "scene0315_00/01160.jpg", "scene0315_00/01200.jpg", "scene0315_00/01240.jpg" ], "turn_id": 8, "scan_id": "scene0315_00", "origin_question": "By the end of the 5 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular brown wicker storage box with a lid\", Y: \"the rectangular white velvet ottoman\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 5 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular brown wicker storage box with a lid\", Y: \"the rectangular white velvet ottoman\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0315_00/01280.jpg", "scene0315_00/01320.jpg", "scene0315_00/01360.jpg", "scene0315_00/01400.jpg" ], "turn_id": 9, "scan_id": "scene0315_00", "origin_question": "How far is your current position from where you were at the end of the turn 5? (in meters)", "option": [], "answer": "4.09", "user_message": "For the 9 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 5? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0451_03/00000.jpg", "scene0451_03/00040.jpg", "scene0451_03/00080.jpg", "scene0451_03/00120.jpg", "scene0451_03/00160.jpg" ], "turn_id": 1, "scan_id": "scene0451_03", "origin_question": "Remember, how many wardrobe(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many wardrobe(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0451_03/00200.jpg", "scene0451_03/00240.jpg", "scene0451_03/00280.jpg", "scene0451_03/00320.jpg", "scene0451_03/00360.jpg" ], "turn_id": 2, "scan_id": "scene0451_03", "origin_question": "When did you first discover \"the curtain\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the curtain\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0451_03/00400.jpg", "scene0451_03/00440.jpg", "scene0451_03/00480.jpg", "scene0451_03/00520.jpg", "scene0451_03/00560.jpg" ], "turn_id": 3, "scan_id": "scene0451_03", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the light brown rectangular wooden bench with a slanted backrest\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the light brown rectangular wooden bench with a slanted backrest\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0451_03/00600.jpg", "scene0451_03/00640.jpg", "scene0451_03/00680.jpg", "scene0451_03/00720.jpg", "scene0451_03/00760.jpg" ], "turn_id": 4, "scan_id": "scene0451_03", "origin_question": "By the end of the 3 turn, were you getting closer to or farther away from these two objects: X: \"the large rectangular glass window with a white frame\", Y: \"the light brown rectangular wooden bench with a slanted backrest\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "farther away from both X and Y", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 3 turn, were you getting closer to or farther away from these two objects: X: \"the large rectangular glass window with a white frame\", Y: \"the light brown rectangular wooden bench with a slanted backrest\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0451_03/00800.jpg", "scene0451_03/00840.jpg", "scene0451_03/00880.jpg", "scene0451_03/00920.jpg", "scene0451_03/00960.jpg" ], "turn_id": 5, "scan_id": "scene0451_03", "origin_question": "Is the light brown rectangular wooden bench with a slanted backrest to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the light brown rectangular wooden bench with a slanted backrest to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0451_03/01000.jpg", "scene0451_03/01040.jpg", "scene0451_03/01080.jpg", "scene0451_03/01120.jpg", "scene0451_03/01160.jpg" ], "turn_id": 6, "scan_id": "scene0451_03", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0451_03/01200.jpg", "scene0451_03/01240.jpg", "scene0451_03/01280.jpg", "scene0451_03/01320.jpg", "scene0451_03/01360.jpg" ], "turn_id": 7, "scan_id": "scene0451_03", "origin_question": "", "option": [], "answer": "", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0451_03/01400.jpg", "scene0451_03/01440.jpg", "scene0451_03/01480.jpg", "scene0451_03/01520.jpg", "scene0451_03/01560.jpg" ], "turn_id": 8, "scan_id": "scene0451_03", "origin_question": "Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "141.48", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0317_00/00000.jpg", "scene0317_00/00030.jpg", "scene0317_00/00060.jpg", "scene0317_00/00090.jpg" ], "turn_id": 1, "scan_id": "scene0317_00", "origin_question": "Remember, have you seen any bottle(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any bottle(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0317_00/00120.jpg", "scene0317_00/00150.jpg", "scene0317_00/00180.jpg", "scene0317_00/00210.jpg" ], "turn_id": 2, "scan_id": "scene0317_00", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular vibrant orange plush pillow\", Y: \"the vibrant orange rectangular stool with rounded edges\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular vibrant orange plush pillow\", Y: \"the vibrant orange rectangular stool with rounded edges\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0317_00/00240.jpg", "scene0317_00/00270.jpg", "scene0317_00/00370.jpg", "scene0317_00/00400.jpg" ], "turn_id": 3, "scan_id": "scene0317_00", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the rectangular vibrant orange plush pillow\" from your position?", "option": [], "answer": "107.73", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the rectangular vibrant orange plush pillow\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0317_00/00430.jpg", "scene0317_00/00460.jpg", "scene0317_00/00490.jpg", "scene0317_00/00520.jpg" ], "turn_id": 4, "scan_id": "scene0317_00", "origin_question": "In which turn did you see both \"the rectangular vibrant orange plush pillow\" and \"the black rectangular flat-panel monitor made of metal\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the rectangular vibrant orange plush pillow\" and \"the black rectangular flat-panel monitor made of metal\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0317_00/00550.jpg", "scene0317_00/00580.jpg", "scene0317_00/00610.jpg", "scene0317_00/00640.jpg" ], "turn_id": 5, "scan_id": "scene0317_00", "origin_question": "Which direction is the coiled gray wire with a black plug in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the coiled gray wire with a black plug in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0317_00/00670.jpg", "scene0317_00/00700.jpg", "scene0317_00/00730.jpg", "scene0317_00/00760.jpg" ], "turn_id": 6, "scan_id": "scene0317_00", "origin_question": "Using your orientation at the end of turn 5 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "124.09", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 5 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0317_00/00790.jpg", "scene0317_00/00820.jpg", "scene0317_00/00850.jpg", "scene0317_00/00880.jpg" ], "turn_id": 7, "scan_id": "scene0317_00", "origin_question": "Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the coiled gray wire with a black plug\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the coiled gray wire with a black plug\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0359_00/00000.jpg", "scene0359_00/00030.jpg", "scene0359_00/00060.jpg", "scene0359_00/00090.jpg" ], "turn_id": 1, "scan_id": "scene0359_00", "origin_question": "Remember, have you seen any lamp(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any lamp(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0359_00/00120.jpg", "scene0359_00/00150.jpg", "scene0359_00/00180.jpg", "scene0359_00/00210.jpg" ], "turn_id": 2, "scan_id": "scene0359_00", "origin_question": "In which turn did you see both \"the bed\" and \"the vertical light-colored fabric curtain with subtle texture\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the bed\" and \"the vertical light-colored fabric curtain with subtle texture\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0359_00/00240.jpg", "scene0359_00/00270.jpg", "scene0359_00/00300.jpg", "scene0359_00/00330.jpg" ], "turn_id": 3, "scan_id": "scene0359_00", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "122.15", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0359_00/00360.jpg", "scene0359_00/00390.jpg", "scene0359_00/00420.jpg", "scene0359_00/00450.jpg" ], "turn_id": 4, "scan_id": "scene0359_00", "origin_question": "Which two objects are on the same side of you now? \"the bed\", \"the rectangular wooden shelf with two layers\", \"the cylindrical bottle with a green nozzle\". ", "option": [ "the bed, the rectangular wooden shelf with two layers", "the bed, the cylindrical bottle with a green nozzle", "the rectangular wooden shelf with two layers, the cylindrical bottle with a green nozzle" ], "answer": "the rectangular wooden shelf with two layers, the cylindrical bottle with a green nozzle", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the bed\", \"the rectangular wooden shelf with two layers\", \"the cylindrical bottle with a green nozzle\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the bed, the rectangular wooden shelf with two layers', 'the bed, the cylindrical bottle with a green nozzle', 'the rectangular wooden shelf with two layers, the cylindrical bottle with a green nozzle'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0359_00/00480.jpg", "scene0359_00/00510.jpg", "scene0359_00/00540.jpg", "scene0359_00/00570.jpg" ], "turn_id": 5, "scan_id": "scene0359_00", "origin_question": "At the end of which turn were both of \"the horizontal white metal radiator with ventilation slots\" and \"the black ergonomic chair with a striped plastic backrest\" on your left side?", "option": [], "answer": "5", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the horizontal white metal radiator with ventilation slots\" and \"the black ergonomic chair with a striped plastic backrest\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0067_region24/5816276565a94b17ba8c667db5829250_i1_4.jpg", "1mp3d_0067_region24/5816276565a94b17ba8c667db5829250_i2_4.jpg", "1mp3d_0067_region24/5816276565a94b17ba8c667db5829250_i1_3.jpg", "1mp3d_0067_region24/5816276565a94b17ba8c667db5829250_i2_3.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0067_region24", "origin_question": "Remember, have you seen any window(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any window(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "1mp3d_0067_region24/0ffa6fd0a65a4462b2023f28b14741e5_i1_1.jpg", "1mp3d_0067_region24/0ffa6fd0a65a4462b2023f28b14741e5_i2_1.jpg", "1mp3d_0067_region24/0ffa6fd0a65a4462b2023f28b14741e5_i1_2.jpg", "1mp3d_0067_region24/0ffa6fd0a65a4462b2023f28b14741e5_i2_2.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0067_region24", "origin_question": "At the end of which turn were both of \"the window\" and \"the large, semi-circular decoration in deep blue\" on your left side?", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the window\" and \"the large, semi-circular decoration in deep blue\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "1mp3d_0067_region24/65818fac73834781a2635832cb6a921b_i1_1.jpg", "1mp3d_0067_region24/65818fac73834781a2635832cb6a921b_i2_1.jpg", "1mp3d_0067_region24/65818fac73834781a2635832cb6a921b_i1_0.jpg", "1mp3d_0067_region24/65818fac73834781a2635832cb6a921b_i2_0.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0067_region24", "origin_question": "When did you first discover \"the large, semi-circular decoration in deep blue\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the large, semi-circular decoration in deep blue\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "1mp3d_0067_region24/1f5d75e4b97147febba7d696e2efbc08_i1_1.jpg", "1mp3d_0067_region24/1f5d75e4b97147febba7d696e2efbc08_i2_1.jpg", "1mp3d_0067_region24/8249032cc7ce45588b8a93ad0bbbfc2f_i1_2.jpg", "1mp3d_0067_region24/8249032cc7ce45588b8a93ad0bbbfc2f_i2_2.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0067_region24", "origin_question": "How far is your current position from where you were at the end of the turn 3? (in meters)", "option": [], "answer": "4.02", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 3? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0067_region24/3a9a0cde4ef64fe59d63a99a93bd149d_i1_2.jpg", "1mp3d_0067_region24/3a9a0cde4ef64fe59d63a99a93bd149d_i2_2.jpg", "1mp3d_0067_region24/6e080f6fe8e64a0693b2495914f18dca_i1_1.jpg", "1mp3d_0067_region24/6e080f6fe8e64a0693b2495914f18dca_i2_1.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0067_region24", "origin_question": "Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the large, semi-circular decoration in deep blue\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the large, semi-circular decoration in deep blue\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0067_region24/6e080f6fe8e64a0693b2495914f18dca_i1_2.jpg", "1mp3d_0067_region24/6e080f6fe8e64a0693b2495914f18dca_i2_2.jpg", "1mp3d_0067_region24/3a9a0cde4ef64fe59d63a99a93bd149d_i1_5.jpg", "1mp3d_0067_region24/3a9a0cde4ef64fe59d63a99a93bd149d_i2_5.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0067_region24", "origin_question": "", "option": [], "answer": "", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0067_region24/8249032cc7ce45588b8a93ad0bbbfc2f_i1_4.jpg", "1mp3d_0067_region24/8249032cc7ce45588b8a93ad0bbbfc2f_i2_4.jpg", "1mp3d_0067_region24/8249032cc7ce45588b8a93ad0bbbfc2f_i1_5.jpg", "1mp3d_0067_region24/8249032cc7ce45588b8a93ad0bbbfc2f_i2_5.jpg" ], "turn_id": 7, "scan_id": "1mp3d_0067_region24", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the large, semi-circular decoration in deep blue\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the large, semi-circular decoration in deep blue\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "1mp3d_0067_region24/1f5d75e4b97147febba7d696e2efbc08_i1_3.jpg", "1mp3d_0067_region24/1f5d75e4b97147febba7d696e2efbc08_i2_3.jpg", "1mp3d_0067_region24/067ec17ab4314effbaf5e67e5acfacad_i1_5.jpg", "1mp3d_0067_region24/067ec17ab4314effbaf5e67e5acfacad_i2_5.jpg" ], "turn_id": 8, "scan_id": "1mp3d_0067_region24", "origin_question": "Assuming the direction you are facing at the end of the turn 5 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Right", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 5 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "1mp3d_0067_region24/067ec17ab4314effbaf5e67e5acfacad_i1_4.jpg", "1mp3d_0067_region24/067ec17ab4314effbaf5e67e5acfacad_i2_4.jpg", "1mp3d_0067_region24/067ec17ab4314effbaf5e67e5acfacad_i1_3.jpg", "1mp3d_0067_region24/067ec17ab4314effbaf5e67e5acfacad_i2_3.jpg" ], "turn_id": 9, "scan_id": "1mp3d_0067_region24", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 9 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0435_02/00000.jpg", "scene0435_02/00040.jpg", "scene0435_02/00080.jpg", "scene0435_02/00120.jpg", "scene0435_02/00160.jpg" ], "turn_id": 1, "scan_id": "scene0435_02", "origin_question": "Remember, have you seen any humidifier(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any humidifier(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0435_02/00200.jpg", "scene0435_02/00240.jpg", "scene0435_02/00280.jpg", "scene0435_02/00320.jpg", "scene0435_02/00360.jpg" ], "turn_id": 2, "scan_id": "scene0435_02", "origin_question": "At the end of which turn were both of \"the dark rectangular wooden desk for office use\" and \"the bin\" on your left side?", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the dark rectangular wooden desk for office use\" and \"the bin\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0435_02/00400.jpg", "scene0435_02/00440.jpg", "scene0435_02/00480.jpg", "scene0435_02/00520.jpg", "scene0435_02/00560.jpg" ], "turn_id": 3, "scan_id": "scene0435_02", "origin_question": "Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "106.33", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0435_02/00600.jpg", "scene0435_02/00640.jpg", "scene0435_02/00680.jpg", "scene0435_02/00720.jpg", "scene0435_02/00760.jpg" ], "turn_id": 4, "scan_id": "scene0435_02", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the black refrigerator made of metal and plastic\" from your position?", "option": [], "answer": "129.71", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the black refrigerator made of metal and plastic\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0435_02/00800.jpg", "scene0435_02/00840.jpg", "scene0435_02/00880.jpg", "scene0435_02/00920.jpg", "scene0435_02/00960.jpg" ], "turn_id": 5, "scan_id": "scene0435_02", "origin_question": "How far is your current position from where you were at the end of the turn 4? (in meters)", "option": [], "answer": "2.9", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 4? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0435_02/01000.jpg", "scene0435_02/01040.jpg", "scene0435_02/01080.jpg", "scene0435_02/01120.jpg", "scene0435_02/01160.jpg" ], "turn_id": 6, "scan_id": "scene0435_02", "origin_question": "Compared to your position at the end of the 5 turn, are you now closer or farther away from \"the sturdy black nylon backpack with padded straps\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 5 turn, are you now closer or farther away from \"the sturdy black nylon backpack with padded straps\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0435_02/01200.jpg", "scene0435_02/01240.jpg", "scene0435_02/01280.jpg", "scene0435_02/01320.jpg", "scene0435_02/01360.jpg" ], "turn_id": 7, "scan_id": "scene0435_02", "origin_question": "Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the black refrigerator made of metal and plastic\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the black refrigerator made of metal and plastic\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0435_02/01400.jpg", "scene0435_02/01440.jpg", "scene0435_02/01480.jpg", "scene0435_02/01520.jpg", "scene0435_02/01560.jpg" ], "turn_id": 8, "scan_id": "scene0435_02", "origin_question": "Is the black refrigerator made of metal and plastic to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the black refrigerator made of metal and plastic to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0030_region18/0891f9ad46d741a4844ff53e8623aec3_i1_4.jpg", "1mp3d_0030_region18/0891f9ad46d741a4844ff53e8623aec3_i2_4.jpg", "1mp3d_0030_region18/267636f4842a422ba13275cb995ca9f5_i1_2.jpg", "1mp3d_0030_region18/267636f4842a422ba13275cb995ca9f5_i2_2.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0030_region18", "origin_question": "Remember, have you seen any piano(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any piano(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 2 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "1mp3d_0030_region18/0891f9ad46d741a4844ff53e8623aec3_i1_5.jpg", "1mp3d_0030_region18/0891f9ad46d741a4844ff53e8623aec3_i2_5.jpg", "1mp3d_0030_region18/0891f9ad46d741a4844ff53e8623aec3_i1_0.jpg", "1mp3d_0030_region18/0891f9ad46d741a4844ff53e8623aec3_i2_0.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0030_region18", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the rectangular light brown bamboo mat\"; \"the tall, slender plant with curved green leaves\"; \"the wooden bench with red planks and a metal frame\" ", "option": [ "the rectangular light brown bamboo mat", "the tall, slender plant with curved green leaves", "the wooden bench with red planks and a metal frame" ], "answer": "the wooden bench with red planks and a metal frame", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the rectangular light brown bamboo mat\"; \"the tall, slender plant with curved green leaves\"; \"the wooden bench with red planks and a metal frame\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular light brown bamboo mat', 'the tall, slender plant with curved green leaves', 'the wooden bench with red planks and a metal frame'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 2 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0040_region29/ad048e19d6cc41d2b559988b26aea08a_i1_1.jpg", "1mp3d_0040_region29/ad048e19d6cc41d2b559988b26aea08a_i2_1.jpg", "1mp3d_0040_region29/86d522ead0d0483f94fb6fa567ddc997_i1_1.jpg", "1mp3d_0040_region29/86d522ead0d0483f94fb6fa567ddc997_i2_1.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0040_region29", "origin_question": "Remember, have you seen any stairs(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any stairs(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "1mp3d_0040_region29/c948e0adde424b22af74a18d3cf14ee6_i1_2.jpg", "1mp3d_0040_region29/c948e0adde424b22af74a18d3cf14ee6_i2_2.jpg", "1mp3d_0040_region29/c948e0adde424b22af74a18d3cf14ee6_i1_1.jpg", "1mp3d_0040_region29/c948e0adde424b22af74a18d3cf14ee6_i2_1.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0040_region29", "origin_question": "Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance forward or backward from that position?", "option": [ "Forward", "Backward" ], "answer": "Backward", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance forward or backward from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Forward', 'Backward'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "1mp3d_0040_region29/86d522ead0d0483f94fb6fa567ddc997_i1_2.jpg", "1mp3d_0040_region29/86d522ead0d0483f94fb6fa567ddc997_i2_2.jpg", "1mp3d_0040_region29/ad048e19d6cc41d2b559988b26aea08a_i1_2.jpg", "1mp3d_0040_region29/ad048e19d6cc41d2b559988b26aea08a_i2_2.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0040_region29", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0040_region29/42aed91e9ec84f9ea89f86c49db984e5_i2_3.jpg", "1mp3d_0040_region29/5103b137c4464b6ab143fbe4b7833e2f_i1_3.jpg", "1mp3d_0040_region29/5103b137c4464b6ab143fbe4b7833e2f_i2_3.jpg", "1mp3d_0040_region29/5103b137c4464b6ab143fbe4b7833e2f_i1_2.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0040_region29", "origin_question": "", "option": [], "answer": "", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "1mp3d_0040_region29/5103b137c4464b6ab143fbe4b7833e2f_i2_2.jpg", "1mp3d_0040_region29/42aed91e9ec84f9ea89f86c49db984e5_i1_2.jpg", "1mp3d_0040_region29/42aed91e9ec84f9ea89f86c49db984e5_i2_2.jpg", "1mp3d_0040_region29/230d9b641a9d4b129304032061a90427_i1_0.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0040_region29", "origin_question": "Assuming the direction you are facing at the end of the turn 3 is forward, did you move a certain distance forward or backward from that position?", "option": [ "Forward", "Backward" ], "answer": "Forward", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 3 is forward, did you move a certain distance forward or backward from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Forward', 'Backward'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "1mp3d_0040_region29/230d9b641a9d4b129304032061a90427_i2_0.jpg", "1mp3d_0040_region29/230d9b641a9d4b129304032061a90427_i1_1.jpg", "1mp3d_0040_region29/230d9b641a9d4b129304032061a90427_i2_1.jpg", "1mp3d_0040_region29/8789cbb1b52b44038196d39d43465947_i1_0.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0040_region29", "origin_question": "How far is your current position from where you were at the end of the turn 3? (in meters)", "option": [], "answer": "2.25", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 3? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "1mp3d_0040_region29/8789cbb1b52b44038196d39d43465947_i2_0.jpg", "1mp3d_0040_region29/21f3a12071494948a2b79c77a32a266d_i1_0.jpg", "1mp3d_0040_region29/21f3a12071494948a2b79c77a32a266d_i2_0.jpg", "1mp3d_0040_region29/21f3a12071494948a2b79c77a32a266d_i1_5.jpg" ], "turn_id": 7, "scan_id": "1mp3d_0040_region29", "origin_question": "How far is your current position from where you were at the end of the turn 1? (in meters)", "option": [], "answer": "2.3", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 1? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "1mp3d_0040_region29/21f3a12071494948a2b79c77a32a266d_i2_5.jpg", "1mp3d_0040_region29/8789cbb1b52b44038196d39d43465947_i1_1.jpg", "1mp3d_0040_region29/8789cbb1b52b44038196d39d43465947_i2_1.jpg", "1mp3d_0040_region29/084c9684a25a4c4f8cdea7afa21d6851_i1_0.jpg" ], "turn_id": 8, "scan_id": "1mp3d_0040_region29", "origin_question": "Using your orientation at the end of turn 5 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 5 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0378_02/00000.jpg", "scene0378_02/00040.jpg", "scene0378_02/00080.jpg", "scene0378_02/00120.jpg", "scene0378_02/00160.jpg" ], "turn_id": 1, "scan_id": "scene0378_02", "origin_question": "Remember, how many kitchen island(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many kitchen island(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0378_02/00200.jpg", "scene0378_02/00240.jpg", "scene0378_02/00280.jpg", "scene0378_02/00320.jpg", "scene0378_02/00360.jpg" ], "turn_id": 2, "scan_id": "scene0378_02", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "scene0378_02/00400.jpg", "scene0378_02/00440.jpg", "scene0378_02/00480.jpg", "scene0378_02/00520.jpg", "scene0378_02/00560.jpg" ], "turn_id": 3, "scan_id": "scene0378_02", "origin_question": "Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Left", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0378_02/00600.jpg", "scene0378_02/00640.jpg", "scene0378_02/00680.jpg", "scene0378_02/00720.jpg", "scene0378_02/00760.jpg" ], "turn_id": 4, "scan_id": "scene0378_02", "origin_question": "At the end of which turn were both of \"the window\" and \"the rectangular white picture with abstract designs\" on your left side?", "option": [], "answer": "4", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the window\" and \"the rectangular white picture with abstract designs\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0378_02/00800.jpg", "scene0378_02/00840.jpg", "scene0378_02/00880.jpg", "scene0378_02/00920.jpg", "scene0378_02/00960.jpg" ], "turn_id": 5, "scan_id": "scene0378_02", "origin_question": "Which direction is the window in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the window in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0378_02/01000.jpg", "scene0378_02/01040.jpg", "scene0378_02/01080.jpg", "scene0378_02/01120.jpg", "scene0378_02/01160.jpg" ], "turn_id": 6, "scan_id": "scene0378_02", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the pair of blue shoes with white stripes\"; \"the light beige shelf with multiple compartments\"; \"the casual deep blue or black synthetic jacket\" ", "option": [ "the pair of blue shoes with white stripes", "the light beige shelf with multiple compartments", "the casual deep blue or black synthetic jacket" ], "answer": "the light beige shelf with multiple compartments", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the pair of blue shoes with white stripes\"; \"the light beige shelf with multiple compartments\"; \"the casual deep blue or black synthetic jacket\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the pair of blue shoes with white stripes', 'the light beige shelf with multiple compartments', 'the casual deep blue or black synthetic jacket'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0378_02/01200.jpg", "scene0378_02/01240.jpg", "scene0378_02/01280.jpg", "scene0378_02/01320.jpg", "scene0378_02/01360.jpg" ], "turn_id": 7, "scan_id": "scene0378_02", "origin_question": "Please recall \"the modern multifunctional black printer with a touchscreen\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.51", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the modern multifunctional black printer with a touchscreen\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0378_02/01400.jpg", "scene0378_02/01440.jpg", "scene0378_02/01480.jpg", "scene0378_02/01520.jpg", "scene0378_02/01560.jpg" ], "turn_id": 8, "scan_id": "scene0378_02", "origin_question": "Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "32.18", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0378_02/01600.jpg", "scene0378_02/01640.jpg", "scene0378_02/01680.jpg", "scene0378_02/01720.jpg", "scene0378_02/01760.jpg" ], "turn_id": 9, "scan_id": "scene0378_02", "origin_question": "In which turn did you see both \"the silver flat-screen monitor with a black border\" and \"the rectangular smooth mustard-colored door\" simultaneously? ", "option": [], "answer": "8", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the silver flat-screen monitor with a black border\" and \"the rectangular smooth mustard-colored door\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0187_00/00000.jpg", "scene0187_00/00040.jpg", "scene0187_00/00080.jpg", "scene0187_00/00120.jpg", "scene0187_00/00160.jpg" ], "turn_id": 1, "scan_id": "scene0187_00", "origin_question": "Remember, have you seen any stopcock(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any stopcock(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0187_00/00200.jpg", "scene0187_00/00240.jpg", "scene0187_00/00280.jpg", "scene0187_00/00320.jpg", "scene0187_00/00360.jpg" ], "turn_id": 2, "scan_id": "scene0187_00", "origin_question": "Remember, how many salt(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many salt(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0187_00/00400.jpg", "scene0187_00/00440.jpg", "scene0187_00/00480.jpg", "scene0187_00/00520.jpg", "scene0187_00/00560.jpg" ], "turn_id": 3, "scan_id": "scene0187_00", "origin_question": "In which turn were you closest to \"the medium-sized rectangular dark backpack made of nylon\"? ", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the medium-sized rectangular dark backpack made of nylon\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "scene0187_00/00600.jpg", "scene0187_00/00640.jpg", "scene0187_00/00680.jpg", "scene0187_00/00720.jpg", "scene0187_00/00760.jpg" ], "turn_id": 4, "scan_id": "scene0187_00", "origin_question": "Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Left", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0187_00/00800.jpg", "scene0187_00/00840.jpg", "scene0187_00/00880.jpg", "scene0187_00/00920.jpg", "scene0187_00/00960.jpg" ], "turn_id": 5, "scan_id": "scene0187_00", "origin_question": "Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0187_00/01000.jpg", "scene0187_00/01040.jpg", "scene0187_00/01080.jpg", "scene0187_00/01120.jpg", "scene0187_00/01160.jpg" ], "turn_id": 6, "scan_id": "scene0187_00", "origin_question": "Using your orientation at the end of turn 4 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "125.4", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 4 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0187_00/01200.jpg", "scene0187_00/01240.jpg", "scene0187_00/01280.jpg", "scene0187_00/01320.jpg", "scene0187_00/01360.jpg" ], "turn_id": 7, "scan_id": "scene0187_00", "origin_question": "Compared to your position at the end of the 6 turn, are you now closer or farther away from \"the medium-sized rectangular dark backpack made of nylon\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 6 turn, are you now closer or farther away from \"the medium-sized rectangular dark backpack made of nylon\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "scene0187_00/01400.jpg", "scene0187_00/01440.jpg", "scene0187_00/01480.jpg", "scene0187_00/01520.jpg", "scene0187_00/01560.jpg" ], "turn_id": 8, "scan_id": "scene0187_00", "origin_question": "Assuming the direction you are facing at the end of the turn 5 is forward, did you move a certain distance forward or backward from that position?", "option": [ "Forward", "Backward" ], "answer": "Backward", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 5 is forward, did you move a certain distance forward or backward from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Forward', 'Backward'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0187_00/01600.jpg", "scene0187_00/01640.jpg", "scene0187_00/01680.jpg", "scene0187_00/01720.jpg", "scene0187_00/01760.jpg" ], "turn_id": 9, "scan_id": "scene0187_00", "origin_question": "Remember, how many crib(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many crib(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0187_00/01800.jpg", "scene0187_00/01840.jpg", "scene0187_00/01880.jpg", "scene0187_00/01920.jpg", "scene0187_00/01960.jpg" ], "turn_id": 10, "scan_id": "scene0187_00", "origin_question": "Remember, how many soap dish(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many soap dish(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0134_00/00000.jpg", "scene0134_00/00030.jpg", "scene0134_00/00060.jpg", "scene0134_00/00090.jpg" ], "turn_id": 1, "scan_id": "scene0134_00", "origin_question": "Remember, how many chair(s) have you seen so far? ", "option": [], "answer": "3", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many chair(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0134_00/00120.jpg", "scene0134_00/00150.jpg", "scene0134_00/00180.jpg", "scene0134_00/00210.jpg" ], "turn_id": 2, "scan_id": "scene0134_00", "origin_question": "In which turn were you farthest from \"the medium-sized framed picture with abstract natural elements\"? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the medium-sized framed picture with abstract natural elements\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0134_00/00240.jpg", "scene0134_00/00270.jpg", "scene0134_00/00300.jpg", "scene0134_00/00330.jpg" ], "turn_id": 3, "scan_id": "scene0134_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0134_00/00360.jpg", "scene0134_00/00390.jpg", "scene0134_00/00420.jpg", "scene0134_00/00450.jpg" ], "turn_id": 4, "scan_id": "scene0134_00", "origin_question": "How far is your current position from where you were at the end of the turn 1? (in meters)", "option": [], "answer": "1.62", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 1? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0134_00/00480.jpg", "scene0134_00/00510.jpg", "scene0134_00/00540.jpg", "scene0134_00/00570.jpg" ], "turn_id": 5, "scan_id": "scene0134_00", "origin_question": "Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0134_00/00600.jpg", "scene0134_00/00630.jpg", "scene0134_00/00660.jpg", "scene0134_00/00690.jpg" ], "turn_id": 6, "scan_id": "scene0134_00", "origin_question": "Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0134_00/00720.jpg", "scene0134_00/00750.jpg", "scene0134_00/00780.jpg", "scene0134_00/00810.jpg" ], "turn_id": 7, "scan_id": "scene0134_00", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the medium-sized framed picture with abstract natural elements\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the medium-sized framed picture with abstract natural elements\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0134_00/00840.jpg", "scene0134_00/00870.jpg", "scene0134_00/00900.jpg", "scene0134_00/00930.jpg" ], "turn_id": 8, "scan_id": "scene0134_00", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the medium-sized framed picture with abstract natural elements\", Y: \"the window\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to Y and farther away from X", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the medium-sized framed picture with abstract natural elements\", Y: \"the window\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0144_01/00000.jpg", "scene0144_01/00040.jpg", "scene0144_01/00080.jpg", "scene0144_01/00120.jpg" ], "turn_id": 1, "scan_id": "scene0144_01", "origin_question": "Remember, have you seen any bowl(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any bowl(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0144_01/00160.jpg", "scene0144_01/00200.jpg", "scene0144_01/00240.jpg", "scene0144_01/00280.jpg" ], "turn_id": 2, "scan_id": "scene0144_01", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0144_01/00320.jpg", "scene0144_01/00360.jpg", "scene0144_01/00400.jpg", "scene0144_01/00440.jpg" ], "turn_id": 3, "scan_id": "scene0144_01", "origin_question": "In which turn were you closest to \"the sleek white bed with green and white sheets\"? ", "option": [], "answer": "2", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the sleek white bed with green and white sheets\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0144_01/00480.jpg", "scene0144_01/00520.jpg", "scene0144_01/00560.jpg", "scene0144_01/00600.jpg" ], "turn_id": 4, "scan_id": "scene0144_01", "origin_question": "Which two objects are on the same side of you now? \"the stand\", \"the window\", \"the bottle\". ", "option": [ "the stand, the window", "the stand, the bottle", "the window, the bottle" ], "answer": "the stand, the window", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the stand\", \"the window\", \"the bottle\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the stand, the window', 'the stand, the bottle', 'the window, the bottle'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0144_01/00640.jpg", "scene0144_01/00680.jpg", "scene0144_01/00720.jpg", "scene0144_01/00760.jpg" ], "turn_id": 5, "scan_id": "scene0144_01", "origin_question": "In which turn did you see both \"the window\" and \"the standard brown wooden desk with a smooth surface\" simultaneously? ", "option": [], "answer": "4", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the window\" and \"the standard brown wooden desk with a smooth surface\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0144_01/00800.jpg", "scene0144_01/00840.jpg", "scene0144_01/00880.jpg", "scene0144_01/00920.jpg" ], "turn_id": 6, "scan_id": "scene0144_01", "origin_question": "Which two objects are on the same side of you now? \"the black multi-functional printer made of metal\", \"the stand\", \"the bottle\". ", "option": [ "the black multi-functional printer made of metal, the stand", "the black multi-functional printer made of metal, the bottle", "the stand, the bottle" ], "answer": "the black multi-functional printer made of metal, the stand", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the black multi-functional printer made of metal\", \"the stand\", \"the bottle\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the black multi-functional printer made of metal, the stand', 'the black multi-functional printer made of metal, the bottle', 'the stand, the bottle'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0144_01/00960.jpg", "scene0144_01/01000.jpg", "scene0144_01/01040.jpg", "scene0144_01/01080.jpg" ], "turn_id": 7, "scan_id": "scene0144_01", "origin_question": "What will be the first-time appearance order of the following categories: chair, pillow, printer?", "option": [ "chair, printer, pillow", "printer, chair, pillow", "printer, pillow, chair", "pillow, chair, printer" ], "answer": "printer, chair, pillow", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: chair, pillow, printer? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['chair, printer, pillow', 'printer, chair, pillow', 'printer, pillow, chair', 'pillow, chair, printer'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0144_01/01120.jpg", "scene0144_01/01160.jpg", "scene0144_01/01200.jpg", "scene0144_01/01240.jpg" ], "turn_id": 8, "scan_id": "scene0144_01", "origin_question": "Which direction is the standard brown wooden desk with a smooth surface in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the standard brown wooden desk with a smooth surface in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0356_01/00000.jpg", "scene0356_01/00040.jpg", "scene0356_01/00080.jpg", "scene0356_01/00120.jpg" ], "turn_id": 1, "scan_id": "scene0356_01", "origin_question": "Remember, have you seen any wardrobe(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any wardrobe(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0356_01/00160.jpg", "scene0356_01/00200.jpg", "scene0356_01/00240.jpg", "scene0356_01/00280.jpg" ], "turn_id": 2, "scan_id": "scene0356_01", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0356_01/00320.jpg", "scene0356_01/00360.jpg", "scene0356_01/00400.jpg", "scene0356_01/00440.jpg" ], "turn_id": 3, "scan_id": "scene0356_01", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the shallow rectangular brown wooden shelf\" from your position?", "option": [], "answer": "158.55", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the shallow rectangular brown wooden shelf\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0356_01/00480.jpg", "scene0356_01/00520.jpg", "scene0356_01/00560.jpg", "scene0356_01/00600.jpg" ], "turn_id": 4, "scan_id": "scene0356_01", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the black chair with a mesh backrest\" from your position?", "option": [], "answer": "101.39", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the black chair with a mesh backrest\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0356_01/00640.jpg", "scene0356_01/00680.jpg", "scene0356_01/00720.jpg", "scene0356_01/00760.jpg" ], "turn_id": 5, "scan_id": "scene0356_01", "origin_question": "Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0356_01/00800.jpg", "scene0356_01/00840.jpg", "scene0356_01/00880.jpg", "scene0356_01/00920.jpg" ], "turn_id": 6, "scan_id": "scene0356_01", "origin_question": "Which two objects are on the same side of you now? \"the rectangular light-colored wooden bed with a blue mattress\", \"the light-brown rectangular wooden desk\", \"the light brown wooden dresser with three drawers\". ", "option": [ "the rectangular light-colored wooden bed with a blue mattress, the light-brown rectangular wooden desk", "the rectangular light-colored wooden bed with a blue mattress, the light brown wooden dresser with three drawers", "the light-brown rectangular wooden desk, the light brown wooden dresser with three drawers" ], "answer": "the rectangular light-colored wooden bed with a blue mattress, the light-brown rectangular wooden desk", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the rectangular light-colored wooden bed with a blue mattress\", \"the light-brown rectangular wooden desk\", \"the light brown wooden dresser with three drawers\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular light-colored wooden bed with a blue mattress, the light-brown rectangular wooden desk', 'the rectangular light-colored wooden bed with a blue mattress, the light brown wooden dresser with three drawers', 'the light-brown rectangular wooden desk, the light brown wooden dresser with three drawers'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0356_01/00960.jpg", "scene0356_01/01000.jpg", "scene0356_01/01040.jpg", "scene0356_01/01080.jpg" ], "turn_id": 7, "scan_id": "scene0356_01", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the light-brown rectangular wooden desk\" from your position?", "option": [], "answer": "141.74", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the light-brown rectangular wooden desk\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0356_01/01120.jpg", "scene0356_01/01160.jpg", "scene0356_01/01200.jpg", "scene0356_01/01240.jpg" ], "turn_id": 8, "scan_id": "scene0356_01", "origin_question": "At the end of which turn were both of \"the black chair with a mesh backrest\" and \"the light brown wooden dresser with three drawers\" on your right side?", "option": [], "answer": "7", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the black chair with a mesh backrest\" and \"the light brown wooden dresser with three drawers\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0612_00/00000.jpg", "scene0612_00/00040.jpg", "scene0612_00/00080.jpg", "scene0612_00/00120.jpg" ], "turn_id": 1, "scan_id": "scene0612_00", "origin_question": "Remember, have you seen any window(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any window(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0612_00/00160.jpg", "scene0612_00/00200.jpg", "scene0612_00/00240.jpg", "scene0612_00/00280.jpg" ], "turn_id": 2, "scan_id": "scene0612_00", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the soap dispenser\"; \"the round, white porcelain sink with a chrome faucet\"; \"the small rectangular black plastic crate\" ", "option": [ "the soap dispenser", "the round, white porcelain sink with a chrome faucet", "the small rectangular black plastic crate" ], "answer": "the round, white porcelain sink with a chrome faucet", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the soap dispenser\"; \"the round, white porcelain sink with a chrome faucet\"; \"the small rectangular black plastic crate\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the soap dispenser', 'the round, white porcelain sink with a chrome faucet', 'the small rectangular black plastic crate'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0612_00/00320.jpg", "scene0612_00/00360.jpg", "scene0612_00/00400.jpg", "scene0612_00/00440.jpg" ], "turn_id": 3, "scan_id": "scene0612_00", "origin_question": "Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "144.01", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0612_00/00480.jpg", "scene0612_00/00520.jpg", "scene0612_00/00560.jpg", "scene0612_00/00600.jpg" ], "turn_id": 4, "scan_id": "scene0612_00", "origin_question": "Which direction is the smooth beige ceramic counter with a circular sink in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the smooth beige ceramic counter with a circular sink in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0612_00/00640.jpg", "scene0612_00/00680.jpg", "scene0612_00/00720.jpg", "scene0612_00/00760.jpg" ], "turn_id": 5, "scan_id": "scene0612_00", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the soap dispenser\" from your position?", "option": [], "answer": "73.28", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the soap dispenser\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0612_00/00800.jpg", "scene0612_00/00840.jpg", "scene0612_00/00880.jpg", "scene0612_00/00920.jpg" ], "turn_id": 6, "scan_id": "scene0612_00", "origin_question": "Which two objects are on the same side of you now? \"the large white rectangular plastic curtain\", \"the smooth beige ceramic counter with a circular sink\", \"the rectangular white plastic switch\". ", "option": [ "the large white rectangular plastic curtain, the smooth beige ceramic counter with a circular sink", "the large white rectangular plastic curtain, the rectangular white plastic switch", "the smooth beige ceramic counter with a circular sink, the rectangular white plastic switch" ], "answer": "the smooth beige ceramic counter with a circular sink, the rectangular white plastic switch", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the large white rectangular plastic curtain\", \"the smooth beige ceramic counter with a circular sink\", \"the rectangular white plastic switch\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the large white rectangular plastic curtain, the smooth beige ceramic counter with a circular sink', 'the large white rectangular plastic curtain, the rectangular white plastic switch', 'the smooth beige ceramic counter with a circular sink, the rectangular white plastic switch'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0136_00/00000.jpg", "scene0136_00/00040.jpg", "scene0136_00/00080.jpg", "scene0136_00/00120.jpg" ], "turn_id": 1, "scan_id": "scene0136_00", "origin_question": "Remember, have you seen any table(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any table(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0136_00/00160.jpg", "scene0136_00/00200.jpg", "scene0136_00/00240.jpg", "scene0136_00/00280.jpg" ], "turn_id": 2, "scan_id": "scene0136_00", "origin_question": "In which turn were you farthest from \"the small rectangular black metal table\"? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the small rectangular black metal table\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0136_00/00320.jpg", "scene0136_00/00360.jpg", "scene0136_00/00400.jpg", "scene0136_00/00440.jpg" ], "turn_id": 3, "scan_id": "scene0136_00", "origin_question": "Please recall \"the sturdy grey rectangular metal cabinet with two doors\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.54", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the sturdy grey rectangular metal cabinet with two doors\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0136_00/00480.jpg", "scene0136_00/00520.jpg", "scene0136_00/00560.jpg", "scene0136_00/00600.jpg" ], "turn_id": 4, "scan_id": "scene0136_00", "origin_question": "Is the white cylindrical plastic bucket with a black label to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the white cylindrical plastic bucket with a black label to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0136_00/00640.jpg", "scene0136_00/00680.jpg", "scene0136_00/00720.jpg", "scene0136_00/00760.jpg" ], "turn_id": 5, "scan_id": "scene0136_00", "origin_question": "Which direction is the white cylindrical plastic bucket with a black label in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the white cylindrical plastic bucket with a black label in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0136_00/00800.jpg", "scene0136_00/00840.jpg", "scene0136_00/00880.jpg", "scene0136_00/00920.jpg" ], "turn_id": 6, "scan_id": "scene0136_00", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the paper cutter\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the paper cutter\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0277_02/00000.jpg", "scene0277_02/00030.jpg", "scene0277_02/00060.jpg", "scene0277_02/00090.jpg" ], "turn_id": 1, "scan_id": "scene0277_02", "origin_question": "Remember, have you seen any bed(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any bed(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0277_02/00120.jpg", "scene0277_02/00150.jpg", "scene0277_02/00180.jpg", "scene0277_02/00210.jpg" ], "turn_id": 2, "scan_id": "scene0277_02", "origin_question": "In which turn did you see both \"the rectangular bed with a blue mattress\" and \"the tall, light brown wooden cabinet\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the rectangular bed with a blue mattress\" and \"the tall, light brown wooden cabinet\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0277_02/00240.jpg", "scene0277_02/00270.jpg", "scene0277_02/00300.jpg", "scene0277_02/00330.jpg" ], "turn_id": 3, "scan_id": "scene0277_02", "origin_question": "At the end of which turn were both of \"the small cylindrical, white plastic bin\" and \"the desk\" on your left side?", "option": [], "answer": "2", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the small cylindrical, white plastic bin\" and \"the desk\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0277_02/00360.jpg", "scene0277_02/00390.jpg", "scene0277_02/00420.jpg", "scene0277_02/00450.jpg" ], "turn_id": 4, "scan_id": "scene0277_02", "origin_question": "Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "66.42", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0277_02/00480.jpg", "scene0277_02/00510.jpg", "scene0277_02/00540.jpg", "scene0277_02/00570.jpg" ], "turn_id": 5, "scan_id": "scene0277_02", "origin_question": "Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the rectangular bed with a blue mattress\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the rectangular bed with a blue mattress\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0277_02/00600.jpg", "scene0277_02/00630.jpg", "scene0277_02/00660.jpg", "scene0277_02/00690.jpg" ], "turn_id": 6, "scan_id": "scene0277_02", "origin_question": "Please recall \"the black chair with a curved backrest\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.83", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the black chair with a curved backrest\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0277_02/00720.jpg", "scene0277_02/00750.jpg", "scene0277_02/00780.jpg", "scene0277_02/00830.jpg" ], "turn_id": 7, "scan_id": "scene0277_02", "origin_question": "Using your orientation at the end of turn 6 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "128.8", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 6 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0029_region6/d8cae4c7073346e392d3996254bf741c_i1_5.jpg", "1mp3d_0029_region6/d8cae4c7073346e392d3996254bf741c_i2_5.jpg", "1mp3d_0029_region6/e29484c59af4429981a6a5d5afe4d5a0_i1_2.jpg", "1mp3d_0029_region6/e29484c59af4429981a6a5d5afe4d5a0_i2_2.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0029_region6", "origin_question": "Remember, have you seen any door(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any door(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 3 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0029_region6/5c39473b25b74307858764d1a2045b9e_i1_2.jpg", "1mp3d_0029_region6/5c39473b25b74307858764d1a2045b9e_i2_2.jpg", "1mp3d_0029_region6/5c39473b25b74307858764d1a2045b9e_i2_1.jpg", "1mp3d_0029_region6/e29484c59af4429981a6a5d5afe4d5a0_i1_0.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0029_region6", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 3 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0029_region6/e29484c59af4429981a6a5d5afe4d5a0_i2_0.jpg", "1mp3d_0029_region6/d8cae4c7073346e392d3996254bf741c_i1_2.jpg", "1mp3d_0029_region6/d8cae4c7073346e392d3996254bf741c_i2_2.jpg", "1mp3d_0029_region6/af5f1942b4374015823bd32e31b31c64_i1_4.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0029_region6", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the door\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the door\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 3 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0580_01/00000.jpg", "scene0580_01/00040.jpg", "scene0580_01/00080.jpg", "scene0580_01/00120.jpg", "scene0580_01/00160.jpg" ], "turn_id": 1, "scan_id": "scene0580_01", "origin_question": "Remember, have you seen any lamp(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any lamp(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0580_01/00200.jpg", "scene0580_01/00240.jpg", "scene0580_01/00280.jpg", "scene0580_01/00320.jpg", "scene0580_01/00360.jpg" ], "turn_id": 2, "scan_id": "scene0580_01", "origin_question": "When did you first discover \"the shoe\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the shoe\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0580_01/00400.jpg", "scene0580_01/00440.jpg", "scene0580_01/00480.jpg", "scene0580_01/00520.jpg", "scene0580_01/00560.jpg" ], "turn_id": 3, "scan_id": "scene0580_01", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0580_01/00600.jpg", "scene0580_01/00640.jpg", "scene0580_01/00680.jpg", "scene0580_01/00720.jpg", "scene0580_01/00760.jpg" ], "turn_id": 4, "scan_id": "scene0580_01", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the shoe\" from your position?", "option": [], "answer": "54.43", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the shoe\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0580_01/00800.jpg", "scene0580_01/00840.jpg", "scene0580_01/00880.jpg", "scene0580_01/00920.jpg", "scene0580_01/00960.jpg" ], "turn_id": 5, "scan_id": "scene0580_01", "origin_question": "", "option": [], "answer": "", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0580_01/01000.jpg", "scene0580_01/01040.jpg", "scene0580_01/01080.jpg", "scene0580_01/01120.jpg", "scene0580_01/01160.jpg" ], "turn_id": 6, "scan_id": "scene0580_01", "origin_question": "", "option": [], "answer": "", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0580_01/01200.jpg", "scene0580_01/01240.jpg", "scene0580_01/01280.jpg", "scene0580_01/01320.jpg", "scene0580_01/01360.jpg" ], "turn_id": 7, "scan_id": "scene0580_01", "origin_question": "Remember, how many pillow(s) have you seen so far? ", "option": [], "answer": "2", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many pillow(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0580_01/01400.jpg", "scene0580_01/01440.jpg", "scene0580_01/01480.jpg", "scene0580_01/01520.jpg", "scene0580_01/01560.jpg" ], "turn_id": 8, "scan_id": "scene0580_01", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the red vase-shaped ceramic lamp with a traditional shade\" from your position?", "option": [], "answer": "114.07", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the red vase-shaped ceramic lamp with a traditional shade\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0580_01/01600.jpg", "scene0580_01/01640.jpg", "scene0580_01/01680.jpg", "scene0580_01/01720.jpg", "scene0580_01/01760.jpg" ], "turn_id": 9, "scan_id": "scene0580_01", "origin_question": "Please recall \"the shoe\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.89", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the shoe\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0580_01/01800.jpg", "scene0580_01/01840.jpg", "scene0580_01/01880.jpg", "scene0580_01/01920.jpg", "scene0580_01/01960.jpg" ], "turn_id": 10, "scan_id": "scene0580_01", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the red vase-shaped ceramic lamp with a traditional shade\" from your position?", "option": [], "answer": "96.83", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the red vase-shaped ceramic lamp with a traditional shade\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0203_01/00000.jpg", "scene0203_01/00040.jpg", "scene0203_01/00080.jpg", "scene0203_01/00120.jpg" ], "turn_id": 1, "scan_id": "scene0203_01", "origin_question": "Remember, how many radiator(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many radiator(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0203_01/00160.jpg", "scene0203_01/00200.jpg", "scene0203_01/00240.jpg", "scene0203_01/00280.jpg" ], "turn_id": 2, "scan_id": "scene0203_01", "origin_question": "At the end of which turn were both of \"the chair\" and \"the rounded-back stool with a black cushion and legs\" on your left side?", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the chair\" and \"the rounded-back stool with a black cushion and legs\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0203_01/00320.jpg", "scene0203_01/00360.jpg", "scene0203_01/00400.jpg", "scene0203_01/00440.jpg" ], "turn_id": 3, "scan_id": "scene0203_01", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the chair\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the chair\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "scene0203_01/00480.jpg", "scene0203_01/00520.jpg", "scene0203_01/00560.jpg", "scene0203_01/00600.jpg" ], "turn_id": 4, "scan_id": "scene0203_01", "origin_question": "Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Right", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0203_01/00670.jpg", "scene0203_01/00710.jpg", "scene0203_01/00750.jpg", "scene0203_01/00790.jpg" ], "turn_id": 5, "scan_id": "scene0203_01", "origin_question": "At the end of which turn were both of \"the rounded-back stool with a black cushion and legs\" and \"the small, elegant black and silver pen\" on your right side?", "option": [], "answer": "5", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the rounded-back stool with a black cushion and legs\" and \"the small, elegant black and silver pen\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0203_01/00830.jpg", "scene0203_01/00870.jpg", "scene0203_01/00910.jpg", "scene0203_01/00950.jpg" ], "turn_id": 6, "scan_id": "scene0203_01", "origin_question": "", "option": [], "answer": "", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0203_01/00990.jpg", "scene0203_01/01030.jpg", "scene0203_01/01070.jpg", "scene0203_01/01110.jpg" ], "turn_id": 7, "scan_id": "scene0203_01", "origin_question": "In which turn were you closest to \"the rectangular hardcover book with a light blue cover\"? ", "option": [], "answer": "2", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the rectangular hardcover book with a light blue cover\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0203_01/01150.jpg", "scene0203_01/01190.jpg", "scene0203_01/01230.jpg", "scene0203_01/01270.jpg" ], "turn_id": 8, "scan_id": "scene0203_01", "origin_question": "Which two objects are on the same side of you now? \"the rounded-back stool with a black cushion and legs\", \"the rectangular hardcover book with a light blue cover\", \"the standard-sized semi-closed white notebook with black pages\". ", "option": [ "the rounded-back stool with a black cushion and legs, the rectangular hardcover book with a light blue cover", "the rounded-back stool with a black cushion and legs, the standard-sized semi-closed white notebook with black pages", "the rectangular hardcover book with a light blue cover, the standard-sized semi-closed white notebook with black pages" ], "answer": "the rectangular hardcover book with a light blue cover, the standard-sized semi-closed white notebook with black pages", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the rounded-back stool with a black cushion and legs\", \"the rectangular hardcover book with a light blue cover\", \"the standard-sized semi-closed white notebook with black pages\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rounded-back stool with a black cushion and legs, the rectangular hardcover book with a light blue cover', 'the rounded-back stool with a black cushion and legs, the standard-sized semi-closed white notebook with black pages', 'the rectangular hardcover book with a light blue cover, the standard-sized semi-closed white notebook with black pages'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0203_01/01310.jpg", "scene0203_01/01350.jpg", "scene0203_01/01390.jpg", "scene0203_01/01430.jpg" ], "turn_id": 9, "scan_id": "scene0203_01", "origin_question": "Which direction is the rounded-back stool with a black cushion and legs in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 9 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the rounded-back stool with a black cushion and legs in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0553_02/00000.jpg", "scene0553_02/00040.jpg", "scene0553_02/00080.jpg", "scene0553_02/00120.jpg" ], "turn_id": 1, "scan_id": "scene0553_02", "origin_question": "Remember, have you seen any stool(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any stool(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0553_02/00160.jpg", "scene0553_02/00200.jpg", "scene0553_02/00240.jpg", "scene0553_02/00280.jpg" ], "turn_id": 2, "scan_id": "scene0553_02", "origin_question": "Please recall \"the cylindrical roll of white tissue paper\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.22", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the cylindrical roll of white tissue paper\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0553_02/00320.jpg", "scene0553_02/00360.jpg", "scene0553_02/00400.jpg", "scene0553_02/00440.jpg" ], "turn_id": 3, "scan_id": "scene0553_02", "origin_question": "In which turn did you see both \"the white round ceramic toilet with a smooth surface\" and \"the toilet paper\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the white round ceramic toilet with a smooth surface\" and \"the toilet paper\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0553_02/00480.jpg", "scene0553_02/00520.jpg", "scene0553_02/00560.jpg", "scene0553_02/00600.jpg" ], "turn_id": 4, "scan_id": "scene0553_02", "origin_question": "Please recall \"the cylindrical roll of white tissue paper\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.03", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the cylindrical roll of white tissue paper\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0553_02/00640.jpg", "scene0553_02/00680.jpg", "scene0553_02/00720.jpg", "scene0553_02/00760.jpg" ], "turn_id": 5, "scan_id": "scene0553_02", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the medium-sized black-gray backpack made of sturdy nylon\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the medium-sized black-gray backpack made of sturdy nylon\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0553_02/00800.jpg", "scene0553_02/00840.jpg", "scene0553_02/00880.jpg", "scene0553_02/00920.jpg" ], "turn_id": 6, "scan_id": "scene0553_02", "origin_question": "In which turn did you see both \"the white round ceramic toilet with a smooth surface\" and \"the cylindrical roll of white tissue paper\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the white round ceramic toilet with a smooth surface\" and \"the cylindrical roll of white tissue paper\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0553_02/00960.jpg", "scene0553_02/01000.jpg", "scene0553_02/01040.jpg", "scene0553_02/01080.jpg" ], "turn_id": 7, "scan_id": "scene0553_02", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the medium-sized black-gray backpack made of sturdy nylon\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the medium-sized black-gray backpack made of sturdy nylon\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0553_02/01120.jpg", "scene0553_02/01160.jpg", "scene0553_02/01200.jpg", "scene0553_02/01240.jpg" ], "turn_id": 8, "scan_id": "scene0553_02", "origin_question": "Which direction is the medium-sized black-gray backpack made of sturdy nylon in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the medium-sized black-gray backpack made of sturdy nylon in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0553_02/01280.jpg", "scene0553_02/01340.jpg", "scene0553_02/01380.jpg", "scene0553_02/01420.jpg" ], "turn_id": 9, "scan_id": "scene0553_02", "origin_question": "Which direction is the small rectangular stainless steel bin in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 9 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the small rectangular stainless steel bin in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0040_region3/c4376e6a0d66466ca75505d35e17bc9b_i1_2.jpg", "1mp3d_0040_region3/c4376e6a0d66466ca75505d35e17bc9b_i2_2.jpg", "1mp3d_0040_region3/c4376e6a0d66466ca75505d35e17bc9b_i1_1.jpg", "1mp3d_0040_region3/c4376e6a0d66466ca75505d35e17bc9b_i2_1.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0040_region3", "origin_question": "Remember, have you seen any tv(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any tv(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0040_region3/51ddde8be586471b8cbd18f793fbf2cb_i1_1.jpg", "1mp3d_0040_region3/51ddde8be586471b8cbd18f793fbf2cb_i2_1.jpg", "1mp3d_0040_region3/fd8580274f564f31b5b5f042b2b85aa6_i1_2.jpg", "1mp3d_0040_region3/fd8580274f564f31b5b5f042b2b85aa6_i2_2.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0040_region3", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "1mp3d_0040_region3/fd8580274f564f31b5b5f042b2b85aa6_i1_3.jpg", "1mp3d_0040_region3/fd8580274f564f31b5b5f042b2b85aa6_i2_3.jpg", "1mp3d_0040_region3/51ddde8be586471b8cbd18f793fbf2cb_i1_3.jpg", "1mp3d_0040_region3/51ddde8be586471b8cbd18f793fbf2cb_i2_3.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0040_region3", "origin_question": "How far is your current position from where you were at the end of the turn 1? (in meters)", "option": [], "answer": "1.62", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 1? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "1mp3d_0040_region3/51ddde8be586471b8cbd18f793fbf2cb_i1_2.jpg", "1mp3d_0040_region3/51ddde8be586471b8cbd18f793fbf2cb_i2_2.jpg", "1mp3d_0040_region3/c22b939107ee4035bff16a85337942c2_i1_1.jpg", "1mp3d_0040_region3/c22b939107ee4035bff16a85337942c2_i2_1.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0040_region3", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "68.19", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0645_01/00000.jpg", "scene0645_01/00040.jpg", "scene0645_01/00080.jpg", "scene0645_01/00120.jpg", "scene0645_01/00160.jpg" ], "turn_id": 1, "scan_id": "scene0645_01", "origin_question": "Remember, how many bed(s) have you seen so far? ", "option": [], "answer": "2", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many bed(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0645_01/00200.jpg", "scene0645_01/00240.jpg", "scene0645_01/00280.jpg", "scene0645_01/00320.jpg", "scene0645_01/00360.jpg" ], "turn_id": 2, "scan_id": "scene0645_01", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the rectangular black plastic album\"; \"the bottle\"; \"the cylindrical white lampshade on a round black lamp\" ", "option": [ "the rectangular black plastic album", "the bottle", "the cylindrical white lampshade on a round black lamp" ], "answer": "the rectangular black plastic album", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the rectangular black plastic album\"; \"the bottle\"; \"the cylindrical white lampshade on a round black lamp\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular black plastic album', 'the bottle', 'the cylindrical white lampshade on a round black lamp'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0645_01/00400.jpg", "scene0645_01/00440.jpg", "scene0645_01/00480.jpg", "scene0645_01/00520.jpg", "scene0645_01/00560.jpg" ], "turn_id": 3, "scan_id": "scene0645_01", "origin_question": "Which direction is the bottle in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-left", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the bottle in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0645_01/00600.jpg", "scene0645_01/00640.jpg", "scene0645_01/00680.jpg", "scene0645_01/00720.jpg", "scene0645_01/00760.jpg" ], "turn_id": 4, "scan_id": "scene0645_01", "origin_question": "In which turn did you see both \"the bottle\" and \"the curtain\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the bottle\" and \"the curtain\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0645_01/00800.jpg", "scene0645_01/00840.jpg", "scene0645_01/00880.jpg", "scene0645_01/00920.jpg", "scene0645_01/00960.jpg" ], "turn_id": 5, "scan_id": "scene0645_01", "origin_question": "At the end of which turn were both of \"the slender plastic remote control in dark color\" and \"the window\" on your right side?", "option": [], "answer": "1", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the slender plastic remote control in dark color\" and \"the window\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0645_01/01000.jpg", "scene0645_01/01040.jpg", "scene0645_01/01080.jpg", "scene0645_01/01120.jpg", "scene0645_01/01160.jpg" ], "turn_id": 6, "scan_id": "scene0645_01", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the bottle\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the bottle\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0645_01/01200.jpg", "scene0645_01/01240.jpg", "scene0645_01/01280.jpg", "scene0645_01/01320.jpg", "scene0645_01/01360.jpg" ], "turn_id": 7, "scan_id": "scene0645_01", "origin_question": "Which direction is the curtain in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the curtain in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0645_01/01400.jpg", "scene0645_01/01440.jpg", "scene0645_01/01480.jpg", "scene0645_01/01520.jpg", "scene0645_01/01560.jpg" ], "turn_id": 8, "scan_id": "scene0645_01", "origin_question": "Using your orientation at the end of turn 6 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 6 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0645_01/01600.jpg", "scene0645_01/01640.jpg", "scene0645_01/01680.jpg", "scene0645_01/01720.jpg", "scene0645_01/01760.jpg" ], "turn_id": 9, "scan_id": "scene0645_01", "origin_question": "", "option": [], "answer": "", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0645_01/01800.jpg", "scene0645_01/01840.jpg", "scene0645_01/01880.jpg", "scene0645_01/01920.jpg", "scene0645_01/01960.jpg" ], "turn_id": 10, "scan_id": "scene0645_01", "origin_question": "Which two objects are on the same side of you now? \"the slender plastic remote control in dark color\", \"the moderate-sized wooden stand in deep brown\", \"the brown, three-seater curved couch made of plush material\". ", "option": [ "the slender plastic remote control in dark color, the moderate-sized wooden stand in deep brown", "the slender plastic remote control in dark color, the brown, three-seater curved couch made of plush material", "the moderate-sized wooden stand in deep brown, the brown, three-seater curved couch made of plush material" ], "answer": "the slender plastic remote control in dark color, the moderate-sized wooden stand in deep brown", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the slender plastic remote control in dark color\", \"the moderate-sized wooden stand in deep brown\", \"the brown, three-seater curved couch made of plush material\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the slender plastic remote control in dark color, the moderate-sized wooden stand in deep brown', 'the slender plastic remote control in dark color, the brown, three-seater curved couch made of plush material', 'the moderate-sized wooden stand in deep brown, the brown, three-seater curved couch made of plush material'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0621_00/00000.jpg", "scene0621_00/00040.jpg", "scene0621_00/00080.jpg", "scene0621_00/00120.jpg", "scene0621_00/00160.jpg" ], "turn_id": 1, "scan_id": "scene0621_00", "origin_question": "Remember, how many table(s) have you seen so far? ", "option": [], "answer": "2", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many table(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "scene0621_00/00200.jpg", "scene0621_00/00240.jpg", "scene0621_00/00280.jpg", "scene0621_00/00320.jpg", "scene0621_00/00360.jpg" ], "turn_id": 2, "scan_id": "scene0621_00", "origin_question": "Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Left", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0621_00/00400.jpg", "scene0621_00/00440.jpg", "scene0621_00/00480.jpg", "scene0621_00/00520.jpg", "scene0621_00/00560.jpg" ], "turn_id": 3, "scan_id": "scene0621_00", "origin_question": "When did you first discover \"the rectangular light-brown wooden door\"? (index of the turn)", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the rectangular light-brown wooden door\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0621_00/00600.jpg", "scene0621_00/00640.jpg", "scene0621_00/00680.jpg", "scene0621_00/00720.jpg", "scene0621_00/00760.jpg" ], "turn_id": 4, "scan_id": "scene0621_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0621_00/00800.jpg", "scene0621_00/00840.jpg", "scene0621_00/00880.jpg", "scene0621_00/00920.jpg", "scene0621_00/00960.jpg" ], "turn_id": 5, "scan_id": "scene0621_00", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "55.77", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0621_00/01000.jpg", "scene0621_00/01040.jpg", "scene0621_00/01080.jpg", "scene0621_00/01120.jpg", "scene0621_00/01160.jpg" ], "turn_id": 6, "scan_id": "scene0621_00", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the box\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the box\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0621_00/01200.jpg", "scene0621_00/01240.jpg", "scene0621_00/01280.jpg", "scene0621_00/01320.jpg", "scene0621_00/01360.jpg" ], "turn_id": 7, "scan_id": "scene0621_00", "origin_question": "Using your orientation at the end of turn 4 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "154.07", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 4 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0621_00/01400.jpg", "scene0621_00/01440.jpg", "scene0621_00/01480.jpg", "scene0621_00/01520.jpg", "scene0621_00/01560.jpg" ], "turn_id": 8, "scan_id": "scene0621_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0621_00/01600.jpg", "scene0621_00/01640.jpg", "scene0621_00/01680.jpg", "scene0621_00/01720.jpg", "scene0621_00/01760.jpg" ], "turn_id": 9, "scan_id": "scene0621_00", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the box\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the box\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0621_00/01800.jpg", "scene0621_00/01840.jpg", "scene0621_00/01880.jpg", "scene0621_00/01920.jpg", "scene0621_00/01960.jpg" ], "turn_id": 10, "scan_id": "scene0621_00", "origin_question": "Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "143.8", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0566_00/00000.jpg", "scene0566_00/00040.jpg", "scene0566_00/00080.jpg", "scene0566_00/00120.jpg", "scene0566_00/00160.jpg" ], "turn_id": 1, "scan_id": "scene0566_00", "origin_question": "Remember, have you seen any blackboard(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any blackboard(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0566_00/00200.jpg", "scene0566_00/00240.jpg", "scene0566_00/00280.jpg", "scene0566_00/00320.jpg", "scene0566_00/00360.jpg" ], "turn_id": 2, "scan_id": "scene0566_00", "origin_question": "When did you first discover \"the standard brown wooden door with a simple design\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the standard brown wooden door with a simple design\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0566_00/00400.jpg", "scene0566_00/00440.jpg", "scene0566_00/00480.jpg", "scene0566_00/00520.jpg", "scene0566_00/00560.jpg" ], "turn_id": 3, "scan_id": "scene0566_00", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0566_00/00600.jpg", "scene0566_00/00640.jpg", "scene0566_00/00680.jpg", "scene0566_00/00770.jpg", "scene0566_00/00810.jpg" ], "turn_id": 4, "scan_id": "scene0566_00", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the rectangular white table made of sturdy material\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the rectangular white table made of sturdy material\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0566_00/00850.jpg", "scene0566_00/00890.jpg", "scene0566_00/00930.jpg", "scene0566_00/00970.jpg", "scene0566_00/01010.jpg" ], "turn_id": 5, "scan_id": "scene0566_00", "origin_question": "When did you first discover \"the rectangular white table made of sturdy material\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the rectangular white table made of sturdy material\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0566_00/01050.jpg", "scene0566_00/01090.jpg", "scene0566_00/01130.jpg", "scene0566_00/01170.jpg", "scene0566_00/01210.jpg" ], "turn_id": 6, "scan_id": "scene0566_00", "origin_question": "How far is your current position from where you were at the end of the turn 1? (in meters)", "option": [], "answer": "2.47", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 1? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0566_00/01250.jpg", "scene0566_00/01290.jpg", "scene0566_00/01330.jpg", "scene0566_00/01370.jpg", "scene0566_00/01410.jpg" ], "turn_id": 7, "scan_id": "scene0566_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0566_00/01450.jpg", "scene0566_00/01490.jpg", "scene0566_00/01530.jpg", "scene0566_00/01570.jpg", "scene0566_00/01610.jpg" ], "turn_id": 8, "scan_id": "scene0566_00", "origin_question": "Using your orientation at the end of turn 4 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "39.49", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 4 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0599_00/00000.jpg", "scene0599_00/00040.jpg", "scene0599_00/00080.jpg", "scene0599_00/00120.jpg", "scene0599_00/00160.jpg" ], "turn_id": 1, "scan_id": "scene0599_00", "origin_question": "Remember, have you seen any window(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any window(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0599_00/00200.jpg", "scene0599_00/00240.jpg", "scene0599_00/00280.jpg", "scene0599_00/00320.jpg", "scene0599_00/00360.jpg" ], "turn_id": 2, "scan_id": "scene0599_00", "origin_question": "In which turn did you see both \"the medium-sized flat-screen TV made of plastic\" and \"the rectangular glass window with a white border\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the medium-sized flat-screen TV made of plastic\" and \"the rectangular glass window with a white border\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0599_00/00400.jpg", "scene0599_00/00440.jpg", "scene0599_00/00480.jpg", "scene0599_00/00520.jpg", "scene0599_00/00560.jpg" ], "turn_id": 3, "scan_id": "scene0599_00", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the telephone\"; \"the standard-sized rectangular white document\"; \"the long, rectangular light brown wooden table\" ", "option": [ "the telephone", "the standard-sized rectangular white document", "the long, rectangular light brown wooden table" ], "answer": "the telephone", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the telephone\"; \"the standard-sized rectangular white document\"; \"the long, rectangular light brown wooden table\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the telephone', 'the standard-sized rectangular white document', 'the long, rectangular light brown wooden table'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0599_00/00600.jpg", "scene0599_00/00640.jpg", "scene0599_00/00680.jpg", "scene0599_00/00720.jpg", "scene0599_00/00760.jpg" ], "turn_id": 4, "scan_id": "scene0599_00", "origin_question": "When did you first discover \"the rectangular blackboard with a wooden frame\"? (index of the turn)", "option": [], "answer": "4", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the rectangular blackboard with a wooden frame\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0599_00/00800.jpg", "scene0599_00/00840.jpg", "scene0599_00/00880.jpg", "scene0599_00/00920.jpg", "scene0599_00/00960.jpg" ], "turn_id": 5, "scan_id": "scene0599_00", "origin_question": "In which turn did you see both \"the medium-sized purple and black backpack\" and \"the rectangular blackboard with a wooden frame\" simultaneously? ", "option": [], "answer": "4", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the medium-sized purple and black backpack\" and \"the rectangular blackboard with a wooden frame\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "scene0599_00/01000.jpg", "scene0599_00/01040.jpg", "scene0599_00/01080.jpg", "scene0599_00/01120.jpg", "scene0599_00/01160.jpg" ], "turn_id": 6, "scan_id": "scene0599_00", "origin_question": "Assuming the direction you are facing at the end of the turn 5 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Left", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 5 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0599_00/01200.jpg", "scene0599_00/01240.jpg", "scene0599_00/01280.jpg", "scene0599_00/01320.jpg", "scene0599_00/01360.jpg" ], "turn_id": 7, "scan_id": "scene0599_00", "origin_question": "Please recall \"the medium-sized purple and black backpack\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.53", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the medium-sized purple and black backpack\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "scene0599_00/01400.jpg", "scene0599_00/01440.jpg", "scene0599_00/01480.jpg", "scene0599_00/01520.jpg", "scene0599_00/01560.jpg" ], "turn_id": 8, "scan_id": "scene0599_00", "origin_question": "Among these three objects(\"the medium-sized purple and black backpack\",\"the medium-sized rectangular gray plastic bin\",\"the rectangular gray dispenser made of plastic\" ), which one is the farthest from you?", "option": [ "the medium-sized purple and black backpack", "the medium-sized rectangular gray plastic bin", "the rectangular gray dispenser made of plastic" ], "answer": "the medium-sized purple and black backpack", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the medium-sized purple and black backpack\",\"the medium-sized rectangular gray plastic bin\",\"the rectangular gray dispenser made of plastic\" ), which one is the farthest from you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the medium-sized purple and black backpack', 'the medium-sized rectangular gray plastic bin', 'the rectangular gray dispenser made of plastic'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0599_00/01600.jpg", "scene0599_00/01640.jpg", "scene0599_00/01680.jpg", "scene0599_00/01720.jpg", "scene0599_00/01760.jpg" ], "turn_id": 9, "scan_id": "scene0599_00", "origin_question": "What will be the first-time appearance order of the following categories: bin, table, telephone?", "option": [ "bin, telephone, table", "telephone, table, bin", "bin, table, telephone", "table, telephone, bin" ], "answer": "table, telephone, bin", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bin, table, telephone? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['bin, telephone, table', 'telephone, table, bin', 'bin, table, telephone', 'table, telephone, bin'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0684_00/00000.jpg", "scene0684_00/00030.jpg", "scene0684_00/00060.jpg", "scene0684_00/00090.jpg" ], "turn_id": 1, "scan_id": "scene0684_00", "origin_question": "Remember, have you seen any door(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any door(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0684_00/00120.jpg", "scene0684_00/00150.jpg", "scene0684_00/00180.jpg", "scene0684_00/00210.jpg" ], "turn_id": 2, "scan_id": "scene0684_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0684_00/00240.jpg", "scene0684_00/00270.jpg", "scene0684_00/00300.jpg", "scene0684_00/00330.jpg" ], "turn_id": 3, "scan_id": "scene0684_00", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the black office chair with a rounded backrest\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the black office chair with a rounded backrest\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0684_00/00360.jpg", "scene0684_00/00390.jpg", "scene0684_00/00420.jpg", "scene0684_00/00450.jpg" ], "turn_id": 4, "scan_id": "scene0684_00", "origin_question": "Which direction is the modern rectangular telephone with a dark casing in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the modern rectangular telephone with a dark casing in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0684_00/00480.jpg", "scene0684_00/00510.jpg", "scene0684_00/00540.jpg", "scene0684_00/00570.jpg" ], "turn_id": 5, "scan_id": "scene0684_00", "origin_question": "When did you first discover \"the medium-sized brown paper bag with a rectangular shape\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the medium-sized brown paper bag with a rectangular shape\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0684_00/00600.jpg", "scene0684_00/00630.jpg", "scene0684_00/00660.jpg", "scene0684_00/00690.jpg" ], "turn_id": 6, "scan_id": "scene0684_00", "origin_question": "Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "117.71", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0684_00/00720.jpg", "scene0684_00/00750.jpg", "scene0684_00/00780.jpg", "scene0684_00/00810.jpg" ], "turn_id": 7, "scan_id": "scene0684_00", "origin_question": "By the end of the 4 turn, were you getting closer to or farther away from these two objects: X: \"the black office chair with a rounded backrest\", Y: \"the white cylindrical ceramic cup with a glossy surface\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "farther away from both X and Y", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 4 turn, were you getting closer to or farther away from these two objects: X: \"the black office chair with a rounded backrest\", Y: \"the white cylindrical ceramic cup with a glossy surface\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0067_region12/48468989fb7f4fcbba1f7be7877bcc8e_i1_5.jpg", "1mp3d_0067_region12/48468989fb7f4fcbba1f7be7877bcc8e_i2_5.jpg", "1mp3d_0067_region12/9ab3d35f9bc44163bcbafaf0e2d97ce7_i1_0.jpg", "1mp3d_0067_region12/9ab3d35f9bc44163bcbafaf0e2d97ce7_i2_0.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0067_region12", "origin_question": "Remember, have you seen any knife(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any knife(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "1mp3d_0067_region12/3d53789d2b2b4eeaadba85394cb14fdb_i1_0.jpg", "1mp3d_0067_region12/3d53789d2b2b4eeaadba85394cb14fdb_i2_0.jpg", "1mp3d_0067_region12/dbd37b4b3597425fae7fa1b818d5e29e_i1_0.jpg", "1mp3d_0067_region12/dbd37b4b3597425fae7fa1b818d5e29e_i2_0.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0067_region12", "origin_question": "In which turn were you closest to \"the modern fan with curved deep brown wooden blades\"? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the modern fan with curved deep brown wooden blades\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "1mp3d_0067_region12/d2bf18703fd645bcb077523c3c200c30_i1_4.jpg", "1mp3d_0067_region12/d2bf18703fd645bcb077523c3c200c30_i2_4.jpg", "1mp3d_0067_region12/6f8be003180e485aa6312a534469785b_i1_0.jpg", "1mp3d_0067_region12/6f8be003180e485aa6312a534469785b_i2_0.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0067_region12", "origin_question": "Which direction is the rectangular picture with a lavish golden frame in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the rectangular picture with a lavish golden frame in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "1mp3d_0067_region12/6f8be003180e485aa6312a534469785b_i1_1.jpg", "1mp3d_0067_region12/6f8be003180e485aa6312a534469785b_i2_1.jpg", "1mp3d_0067_region12/d2bf18703fd645bcb077523c3c200c30_i1_5.jpg", "1mp3d_0067_region12/d2bf18703fd645bcb077523c3c200c30_i2_5.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0067_region12", "origin_question": "What will be the first-time appearance order of the following categories: clock, fan, fireplace?", "option": [ "clock, fan, fireplace", "fan, clock, fireplace", "clock, fireplace, fan", "fireplace, clock, fan" ], "answer": "fireplace, clock, fan", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: clock, fan, fireplace? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['clock, fan, fireplace', 'fan, clock, fireplace', 'clock, fireplace, fan', 'fireplace, clock, fan'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "1mp3d_0067_region12/d2bf18703fd645bcb077523c3c200c30_i1_0.jpg", "1mp3d_0067_region12/d2bf18703fd645bcb077523c3c200c30_i2_0.jpg", "1mp3d_0067_region12/dbd37b4b3597425fae7fa1b818d5e29e_i1_2.jpg", "1mp3d_0067_region12/dbd37b4b3597425fae7fa1b818d5e29e_i2_2.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0067_region12", "origin_question": "Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "110.6", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "1mp3d_0067_region12/dbd37b4b3597425fae7fa1b818d5e29e_i1_1.jpg", "1mp3d_0067_region12/dbd37b4b3597425fae7fa1b818d5e29e_i2_1.jpg", "1mp3d_0067_region12/3d53789d2b2b4eeaadba85394cb14fdb_i1_1.jpg", "1mp3d_0067_region12/3d53789d2b2b4eeaadba85394cb14fdb_i2_1.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0067_region12", "origin_question": "What will be the first-time appearance order of the following categories: fan, fireplace, ottoman?", "option": [ "fan, fireplace, ottoman", "ottoman, fireplace, fan", "fireplace, ottoman, fan", "fan, ottoman, fireplace" ], "answer": "fireplace, ottoman, fan", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: fan, fireplace, ottoman? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['fan, fireplace, ottoman', 'ottoman, fireplace, fan', 'fireplace, ottoman, fan', 'fan, ottoman, fireplace'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "1mp3d_0067_region12/3d53789d2b2b4eeaadba85394cb14fdb_i1_2.jpg", "1mp3d_0067_region12/3d53789d2b2b4eeaadba85394cb14fdb_i2_2.jpg", "1mp3d_0067_region12/9ab3d35f9bc44163bcbafaf0e2d97ce7_i1_2.jpg", "1mp3d_0067_region12/9ab3d35f9bc44163bcbafaf0e2d97ce7_i2_2.jpg" ], "turn_id": 7, "scan_id": "1mp3d_0067_region12", "origin_question": "Is the large circular wooden clock with a deep brown color to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the large circular wooden clock with a deep brown color to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0067_region12/48468989fb7f4fcbba1f7be7877bcc8e_i1_1.jpg", "1mp3d_0067_region12/48468989fb7f4fcbba1f7be7877bcc8e_i2_1.jpg", "1mp3d_0067_region12/48468989fb7f4fcbba1f7be7877bcc8e_i1_2.jpg", "1mp3d_0067_region12/48468989fb7f4fcbba1f7be7877bcc8e_i2_2.jpg" ], "turn_id": 8, "scan_id": "1mp3d_0067_region12", "origin_question": "", "option": [], "answer": "", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "1mp3d_0067_region12/55d758f8ca7243b5be0fc24557452329_i1_1.jpg", "1mp3d_0067_region12/55d758f8ca7243b5be0fc24557452329_i2_1.jpg", "1mp3d_0067_region12/55d758f8ca7243b5be0fc24557452329_i1_0.jpg", "1mp3d_0067_region12/55d758f8ca7243b5be0fc24557452329_i2_0.jpg" ], "turn_id": 9, "scan_id": "1mp3d_0067_region12", "origin_question": "Using your orientation at the end of turn 6 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 9 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 6 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0496_00/00000.jpg", "scene0496_00/00040.jpg", "scene0496_00/00240.jpg", "scene0496_00/00280.jpg" ], "turn_id": 1, "scan_id": "scene0496_00", "origin_question": "Remember, have you seen any blackboard(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any blackboard(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "scene0496_00/00320.jpg", "scene0496_00/00360.jpg", "scene0496_00/00400.jpg", "scene0496_00/00440.jpg" ], "turn_id": 2, "scan_id": "scene0496_00", "origin_question": "Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Left", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0496_00/00480.jpg", "scene0496_00/00520.jpg", "scene0496_00/00560.jpg", "scene0496_00/00600.jpg" ], "turn_id": 3, "scan_id": "scene0496_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0496_00/00640.jpg", "scene0496_00/00680.jpg", "scene0496_00/00720.jpg", "scene0496_00/00760.jpg" ], "turn_id": 4, "scan_id": "scene0496_00", "origin_question": "In which turn were you closest to \"the rectangular deep gray blackboard with a wooden frame\"? ", "option": [], "answer": "2", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the rectangular deep gray blackboard with a wooden frame\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0496_00/00800.jpg", "scene0496_00/00840.jpg", "scene0496_00/00880.jpg", "scene0496_00/00920.jpg" ], "turn_id": 5, "scan_id": "scene0496_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0496_00/00960.jpg", "scene0496_00/01000.jpg", "scene0496_00/01040.jpg", "scene0496_00/01080.jpg" ], "turn_id": 6, "scan_id": "scene0496_00", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0496_00/01120.jpg", "scene0496_00/01160.jpg", "scene0496_00/01200.jpg", "scene0496_00/01240.jpg" ], "turn_id": 7, "scan_id": "scene0496_00", "origin_question": "Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 4 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0040_region20/a0e69514034f4b3f81f1addeef642e66_i1_1.jpg", "1mp3d_0040_region20/a0e69514034f4b3f81f1addeef642e66_i2_1.jpg", "1mp3d_0040_region20/85f96a391fcc4e23a5c3b3830a5955e7_i1_3.jpg", "1mp3d_0040_region20/85f96a391fcc4e23a5c3b3830a5955e7_i2_3.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0040_region20", "origin_question": "Remember, have you seen any mirror(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any mirror(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "1mp3d_0040_region20/85f96a391fcc4e23a5c3b3830a5955e7_i1_2.jpg", "1mp3d_0040_region20/85f96a391fcc4e23a5c3b3830a5955e7_i2_2.jpg", "1mp3d_0040_region20/a0e69514034f4b3f81f1addeef642e66_i1_4.jpg", "1mp3d_0040_region20/a0e69514034f4b3f81f1addeef642e66_i2_4.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0040_region20", "origin_question": "Which two objects are on the same side of you now? \"the white textured cotton towel, medium-sized and folded\", \"the rectangular picture with a deep brown frame\", \"the rectangular, creamy white fuzzy mat\". ", "option": [ "the white textured cotton towel, medium-sized and folded, the rectangular picture with a deep brown frame", "the white textured cotton towel, medium-sized and folded, the rectangular, creamy white fuzzy mat", "the rectangular picture with a deep brown frame, the rectangular, creamy white fuzzy mat" ], "answer": "the rectangular picture with a deep brown frame, the rectangular, creamy white fuzzy mat", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the white textured cotton towel, medium-sized and folded\", \"the rectangular picture with a deep brown frame\", \"the rectangular, creamy white fuzzy mat\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the white textured cotton towel, medium-sized and folded, the rectangular picture with a deep brown frame', 'the white textured cotton towel, medium-sized and folded, the rectangular, creamy white fuzzy mat', 'the rectangular picture with a deep brown frame, the rectangular, creamy white fuzzy mat'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "1mp3d_0040_region20/bf1d077ef3214d8ca1e74b5a2e378080_i1_3.jpg", "1mp3d_0040_region20/bf1d077ef3214d8ca1e74b5a2e378080_i2_3.jpg", "1mp3d_0040_region20/ac28cb6fd2884edd8cdf715e58be5bc7_i1_2.jpg", "1mp3d_0040_region20/ac28cb6fd2884edd8cdf715e58be5bc7_i2_2.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0040_region20", "origin_question": "In which turn did you see both \"the white textured cotton towel, medium-sized and folded\" and \"the rectangular mirror with a dark brown ornate frame\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the white textured cotton towel, medium-sized and folded\" and \"the rectangular mirror with a dark brown ornate frame\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "1mp3d_0040_region20/ac28cb6fd2884edd8cdf715e58be5bc7_i1_1.jpg", "1mp3d_0040_region20/ac28cb6fd2884edd8cdf715e58be5bc7_i2_1.jpg", "1mp3d_0040_region20/bf1d077ef3214d8ca1e74b5a2e378080_i1_1.jpg", "1mp3d_0040_region20/bf1d077ef3214d8ca1e74b5a2e378080_i2_1.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0040_region20", "origin_question": "Is the rectangular light brown wooden window frame to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the rectangular light brown wooden window frame to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0040_region20/bf1d077ef3214d8ca1e74b5a2e378080_i1_2.jpg", "1mp3d_0040_region20/bf1d077ef3214d8ca1e74b5a2e378080_i2_2.jpg", "1mp3d_0040_region20/a0e69514034f4b3f81f1addeef642e66_i1_2.jpg", "1mp3d_0040_region20/a0e69514034f4b3f81f1addeef642e66_i2_2.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0040_region20", "origin_question": "Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the white textured cotton towel, medium-sized and folded\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the white textured cotton towel, medium-sized and folded\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "1mp3d_0040_region20/a0e69514034f4b3f81f1addeef642e66_i1_3.jpg", "1mp3d_0040_region20/a0e69514034f4b3f81f1addeef642e66_i2_3.jpg", "1mp3d_0040_region20/85f96a391fcc4e23a5c3b3830a5955e7_i1_1.jpg", "1mp3d_0040_region20/85f96a391fcc4e23a5c3b3830a5955e7_i2_1.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0040_region20", "origin_question": "How far is your current position from where you were at the end of the turn 3? (in meters)", "option": [], "answer": "2.24", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 3? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0044_region24/85bd2dfb84fc44a8a34eef408ba1b259_i1_2.jpg", "1mp3d_0044_region24/85bd2dfb84fc44a8a34eef408ba1b259_i2_2.jpg", "1mp3d_0044_region24/53f23896bfe34da397d986239f65b5d3_i1_4.jpg", "1mp3d_0044_region24/53f23896bfe34da397d986239f65b5d3_i2_4.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0044_region24", "origin_question": "Remember, have you seen any door(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any door(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "1mp3d_0044_region24/85bd2dfb84fc44a8a34eef408ba1b259_i1_5.jpg", "1mp3d_0044_region24/85bd2dfb84fc44a8a34eef408ba1b259_i2_5.jpg", "1mp3d_0044_region24/53f23896bfe34da397d986239f65b5d3_i1_3.jpg", "1mp3d_0044_region24/53f23896bfe34da397d986239f65b5d3_i2_3.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0044_region24", "origin_question": "In which turn did you see both \"the large black chair with a mesh back\" and \"the classic semicircular light beige arch\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the large black chair with a mesh back\" and \"the classic semicircular light beige arch\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "1mp3d_0044_region24/85bd2dfb84fc44a8a34eef408ba1b259_i1_4.jpg", "1mp3d_0044_region24/85bd2dfb84fc44a8a34eef408ba1b259_i2_4.jpg", "1mp3d_0044_region24/53f23896bfe34da397d986239f65b5d3_i1_1.jpg", "1mp3d_0044_region24/53f23896bfe34da397d986239f65b5d3_i2_1.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0044_region24", "origin_question": "In which turn did you see both \"the classic wooden door with a warm brown color\" and \"the black plastic printer on a black marble countertop\" simultaneously? ", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the classic wooden door with a warm brown color\" and \"the black plastic printer on a black marble countertop\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "1mp3d_0044_region24/53f23896bfe34da397d986239f65b5d3_i1_0.jpg", "1mp3d_0044_region24/53f23896bfe34da397d986239f65b5d3_i2_0.jpg", "1mp3d_0044_region24/75715b08ec6844b69b7444db18949855_i1_4.jpg", "1mp3d_0044_region24/75715b08ec6844b69b7444db18949855_i2_4.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0044_region24", "origin_question": "When did you first discover \"the classic wooden door with a warm brown color\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the classic wooden door with a warm brown color\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "1mp3d_0044_region24/6c76d43a72e14b71956c5444d32e129d_i1_0.jpg", "1mp3d_0044_region24/6c76d43a72e14b71956c5444d32e129d_i2_0.jpg", "1mp3d_0044_region24/75715b08ec6844b69b7444db18949855_i1_0.jpg", "1mp3d_0044_region24/75715b08ec6844b69b7444db18949855_i2_0.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0044_region24", "origin_question": "How far is your current position from where you were at the end of the turn 3? (in meters)", "option": [], "answer": "1.87", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 3? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0387_02/00000.jpg", "scene0387_02/00030.jpg", "scene0387_02/00060.jpg", "scene0387_02/00090.jpg" ], "turn_id": 1, "scan_id": "scene0387_02", "origin_question": "Remember, have you seen any wine(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any wine(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0387_02/00120.jpg", "scene0387_02/00150.jpg", "scene0387_02/00180.jpg", "scene0387_02/00210.jpg" ], "turn_id": 2, "scan_id": "scene0387_02", "origin_question": "Remember, how many vase(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many vase(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0387_02/00240.jpg", "scene0387_02/00270.jpg", "scene0387_02/00300.jpg", "scene0387_02/00330.jpg" ], "turn_id": 3, "scan_id": "scene0387_02", "origin_question": "In which turn were you closest to \"the rectangular plastic bin with a blue lid\"? ", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the rectangular plastic bin with a blue lid\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0387_02/00360.jpg", "scene0387_02/00390.jpg", "scene0387_02/00420.jpg", "scene0387_02/00450.jpg" ], "turn_id": 4, "scan_id": "scene0387_02", "origin_question": "When did you first discover \"the window\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the window\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0387_02/00480.jpg", "scene0387_02/00510.jpg", "scene0387_02/00540.jpg", "scene0387_02/00570.jpg" ], "turn_id": 5, "scan_id": "scene0387_02", "origin_question": "Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "79.04", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "scene0387_02/00600.jpg", "scene0387_02/00630.jpg", "scene0387_02/00660.jpg", "scene0387_02/00690.jpg" ], "turn_id": 6, "scan_id": "scene0387_02", "origin_question": "Assuming the direction you are facing at the end of the turn 3 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Right", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 3 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0387_02/00720.jpg", "scene0387_02/00750.jpg", "scene0387_02/00780.jpg", "scene0387_02/00810.jpg" ], "turn_id": 7, "scan_id": "scene0387_02", "origin_question": "When did you first discover \"the rectangular glass door with a white frame\"? (index of the turn)", "option": [], "answer": "4", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the rectangular glass door with a white frame\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0387_02/00840.jpg", "scene0387_02/00870.jpg", "scene0387_02/00900.jpg", "scene0387_02/00930.jpg" ], "turn_id": 8, "scan_id": "scene0387_02", "origin_question": "In which turn did you see both \"the window\" and \"the rectangular plastic bin with a blue lid\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the window\" and \"the rectangular plastic bin with a blue lid\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0074_region11/7dbbeb8448f94a0ca912bb73561fb1f7_i1_0.jpg", "1mp3d_0074_region11/7dbbeb8448f94a0ca912bb73561fb1f7_i2_0.jpg", "1mp3d_0074_region11/7dbbeb8448f94a0ca912bb73561fb1f7_i1_5.jpg", "1mp3d_0074_region11/7dbbeb8448f94a0ca912bb73561fb1f7_i2_5.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0074_region11", "origin_question": "Remember, have you seen any mouse(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any mouse(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "1mp3d_0074_region11/70d867fa66c445ab847cb6f12f219ff4_i1_5.jpg", "1mp3d_0074_region11/70d867fa66c445ab847cb6f12f219ff4_i2_5.jpg", "1mp3d_0074_region11/7dbbeb8448f94a0ca912bb73561fb1f7_i1_1.jpg", "1mp3d_0074_region11/7dbbeb8448f94a0ca912bb73561fb1f7_i2_1.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0074_region11", "origin_question": "In which turn were you farthest from \"the rectangular white door with a glass panel\"? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the rectangular white door with a glass panel\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "1mp3d_0074_region11/7dbbeb8448f94a0ca912bb73561fb1f7_i2_2.jpg", "1mp3d_0074_region11/7dbbeb8448f94a0ca912bb73561fb1f7_i1_3.jpg", "1mp3d_0074_region11/7dbbeb8448f94a0ca912bb73561fb1f7_i2_3.jpg", "1mp3d_0074_region11/7dbbeb8448f94a0ca912bb73561fb1f7_i1_4.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0074_region11", "origin_question": "What will be the first-time appearance order of the following categories: door, paper, plate?", "option": [ "paper, door, plate", "paper, plate, door", "plate, paper, door", "plate, door, paper" ], "answer": "plate, door, paper", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: door, paper, plate? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['paper, door, plate', 'paper, plate, door', 'plate, paper, door', 'plate, door, paper'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "1mp3d_0074_region11/7dbbeb8448f94a0ca912bb73561fb1f7_i2_4.jpg", "1mp3d_0074_region11/70d867fa66c445ab847cb6f12f219ff4_i1_3.jpg", "1mp3d_0074_region11/70d867fa66c445ab847cb6f12f219ff4_i2_3.jpg", "1mp3d_0074_region11/70d867fa66c445ab847cb6f12f219ff4_i1_4.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0074_region11", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular window with a white frame made of glass\", Y: \"the medium-sized black rectangular box made of sturdy material\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the rectangular window with a white frame made of glass\", Y: \"the medium-sized black rectangular box made of sturdy material\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "1mp3d_0074_region11/70d867fa66c445ab847cb6f12f219ff4_i2_4.jpg", "1mp3d_0074_region11/70d867fa66c445ab847cb6f12f219ff4_i1_2.jpg", "1mp3d_0074_region11/70d867fa66c445ab847cb6f12f219ff4_i2_2.jpg", "1mp3d_0074_region11/70d867fa66c445ab847cb6f12f219ff4_i1_1.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0074_region11", "origin_question": "What will be the first-time appearance order of the following categories: door, keyboard, plate?", "option": [ "keyboard, door, plate", "keyboard, plate, door", "plate, door, keyboard", "door, plate, keyboard" ], "answer": "plate, door, keyboard", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: door, keyboard, plate? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['keyboard, door, plate', 'keyboard, plate, door', 'plate, door, keyboard', 'door, plate, keyboard'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0063_00/00000.jpg", "scene0063_00/00030.jpg", "scene0063_00/00060.jpg", "scene0063_00/00090.jpg" ], "turn_id": 1, "scan_id": "scene0063_00", "origin_question": "Remember, have you seen any garage door(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any garage door(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0063_00/00120.jpg", "scene0063_00/00150.jpg", "scene0063_00/00180.jpg", "scene0063_00/00210.jpg" ], "turn_id": 2, "scan_id": "scene0063_00", "origin_question": "In which turn did you see both \"the stainless steel refrigerator with a smooth surface\" and \"the stylish rectangular black granite counter\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the stainless steel refrigerator with a smooth surface\" and \"the stylish rectangular black granite counter\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0063_00/00240.jpg", "scene0063_00/00270.jpg", "scene0063_00/00300.jpg", "scene0063_00/00330.jpg" ], "turn_id": 3, "scan_id": "scene0063_00", "origin_question": "Which two objects are on the same side of you now? \"the small square wooden table with a metal base\", \"the dark rectangular cabinet with speckles\", \"the sleek black and silver stainless steel dishwasher\". ", "option": [ "the small square wooden table with a metal base, the dark rectangular cabinet with speckles", "the small square wooden table with a metal base, the sleek black and silver stainless steel dishwasher", "the dark rectangular cabinet with speckles, the sleek black and silver stainless steel dishwasher" ], "answer": "the dark rectangular cabinet with speckles, the sleek black and silver stainless steel dishwasher", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the small square wooden table with a metal base\", \"the dark rectangular cabinet with speckles\", \"the sleek black and silver stainless steel dishwasher\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the small square wooden table with a metal base, the dark rectangular cabinet with speckles', 'the small square wooden table with a metal base, the sleek black and silver stainless steel dishwasher', 'the dark rectangular cabinet with speckles, the sleek black and silver stainless steel dishwasher'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0063_00/00360.jpg", "scene0063_00/00390.jpg", "scene0063_00/00420.jpg", "scene0063_00/00450.jpg" ], "turn_id": 4, "scan_id": "scene0063_00", "origin_question": "Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the stainless steel refrigerator with a smooth surface\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the stainless steel refrigerator with a smooth surface\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 4 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0696_02/00000.jpg", "scene0696_02/00040.jpg", "scene0696_02/00080.jpg", "scene0696_02/00120.jpg" ], "turn_id": 1, "scan_id": "scene0696_02", "origin_question": "Remember, how many chair(s) have you seen so far? ", "option": [], "answer": "3", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many chair(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0696_02/00160.jpg", "scene0696_02/00200.jpg", "scene0696_02/00240.jpg", "scene0696_02/00280.jpg" ], "turn_id": 2, "scan_id": "scene0696_02", "origin_question": "Which direction is the medium blue rectangular tote bag made of sturdy material in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the medium blue rectangular tote bag made of sturdy material in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0696_02/00320.jpg", "scene0696_02/00360.jpg", "scene0696_02/00400.jpg", "scene0696_02/00440.jpg" ], "turn_id": 3, "scan_id": "scene0696_02", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the medium blue rectangular tote bag made of sturdy material\" from your position?", "option": [], "answer": "144.8", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the medium blue rectangular tote bag made of sturdy material\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0696_02/00480.jpg", "scene0696_02/00520.jpg", "scene0696_02/00560.jpg", "scene0696_02/00600.jpg" ], "turn_id": 4, "scan_id": "scene0696_02", "origin_question": "Is the standard-sized mattress with a patterned cover to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the standard-sized mattress with a patterned cover to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0696_02/00640.jpg", "scene0696_02/00680.jpg", "scene0696_02/00720.jpg", "scene0696_02/00760.jpg" ], "turn_id": 5, "scan_id": "scene0696_02", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the rectangular green ping-pong table made of wood\"; \"the standard-sized mattress with a patterned cover\"; \"the white rectangular wooden door with a glass section\" ", "option": [ "the rectangular green ping-pong table made of wood", "the standard-sized mattress with a patterned cover", "the white rectangular wooden door with a glass section" ], "answer": "the white rectangular wooden door with a glass section", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the rectangular green ping-pong table made of wood\"; \"the standard-sized mattress with a patterned cover\"; \"the white rectangular wooden door with a glass section\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular green ping-pong table made of wood', 'the standard-sized mattress with a patterned cover', 'the white rectangular wooden door with a glass section'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0696_02/00800.jpg", "scene0696_02/00840.jpg", "scene0696_02/00880.jpg", "scene0696_02/00920.jpg" ], "turn_id": 6, "scan_id": "scene0696_02", "origin_question": "", "option": [], "answer": "", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0696_02/00960.jpg", "scene0696_02/01000.jpg", "scene0696_02/01040.jpg", "scene0696_02/01080.jpg" ], "turn_id": 7, "scan_id": "scene0696_02", "origin_question": "Which two objects are on the same side of you now? \"the rectangular green ping-pong table made of wood\", \"the large round brown wicker basket with a handle\", \"the white rectangular wooden door with a glass section\". ", "option": [ "the rectangular green ping-pong table made of wood, the large round brown wicker basket with a handle", "the rectangular green ping-pong table made of wood, the white rectangular wooden door with a glass section", "the large round brown wicker basket with a handle, the white rectangular wooden door with a glass section" ], "answer": "the rectangular green ping-pong table made of wood, the large round brown wicker basket with a handle", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the rectangular green ping-pong table made of wood\", \"the large round brown wicker basket with a handle\", \"the white rectangular wooden door with a glass section\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular green ping-pong table made of wood, the large round brown wicker basket with a handle', 'the rectangular green ping-pong table made of wood, the white rectangular wooden door with a glass section', 'the large round brown wicker basket with a handle, the white rectangular wooden door with a glass section'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0696_02/01120.jpg", "scene0696_02/01160.jpg", "scene0696_02/01200.jpg", "scene0696_02/01240.jpg" ], "turn_id": 8, "scan_id": "scene0696_02", "origin_question": "Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the white rectangular metal radiator\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the white rectangular metal radiator\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0696_02/01280.jpg", "scene0696_02/01320.jpg", "scene0696_02/01360.jpg", "scene0696_02/01400.jpg" ], "turn_id": 9, "scan_id": "scene0696_02", "origin_question": "How far is your current position from where you were at the end of the turn 3? (in meters)", "option": [], "answer": "2.27", "user_message": "For the 9 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 3? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0458_00/00000.jpg", "scene0458_00/00040.jpg", "scene0458_00/00080.jpg", "scene0458_00/00120.jpg", "scene0458_00/00160.jpg" ], "turn_id": 1, "scan_id": "scene0458_00", "origin_question": "Remember, have you seen any range hood(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any range hood(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0458_00/00200.jpg", "scene0458_00/00240.jpg", "scene0458_00/00280.jpg", "scene0458_00/00320.jpg", "scene0458_00/00360.jpg" ], "turn_id": 2, "scan_id": "scene0458_00", "origin_question": "Which direction is the bathroom vanity with a marble countertop and wooden cabinet in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the bathroom vanity with a marble countertop and wooden cabinet in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0458_00/00400.jpg", "scene0458_00/00440.jpg", "scene0458_00/00480.jpg", "scene0458_00/00520.jpg", "scene0458_00/00560.jpg" ], "turn_id": 3, "scan_id": "scene0458_00", "origin_question": "What will be the first-time appearance order of the following categories: dustpan, sink, towel?", "option": [ "towel, sink, dustpan", "sink, towel, dustpan", "sink, dustpan, towel", "towel, dustpan, sink" ], "answer": "sink, dustpan, towel", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: dustpan, sink, towel? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['towel, sink, dustpan', 'sink, towel, dustpan', 'sink, dustpan, towel', 'towel, dustpan, sink'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0458_00/00600.jpg", "scene0458_00/00640.jpg", "scene0458_00/00680.jpg", "scene0458_00/00720.jpg", "scene0458_00/00760.jpg" ], "turn_id": 4, "scan_id": "scene0458_00", "origin_question": "Remember, how many mailbox(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many mailbox(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0458_00/00800.jpg", "scene0458_00/00840.jpg", "scene0458_00/00880.jpg", "scene0458_00/00920.jpg", "scene0458_00/00960.jpg" ], "turn_id": 5, "scan_id": "scene0458_00", "origin_question": "In which turn did you see both \"the black and silver three-step metal ladder\" and \"the small rectangular grey plastic dustpan\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the black and silver three-step metal ladder\" and \"the small rectangular grey plastic dustpan\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0458_00/01000.jpg", "scene0458_00/01040.jpg", "scene0458_00/01080.jpg", "scene0458_00/01120.jpg", "scene0458_00/01160.jpg" ], "turn_id": 6, "scan_id": "scene0458_00", "origin_question": "Which direction is the white oval ceramic sink with a silver faucet in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the white oval ceramic sink with a silver faucet in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0458_00/01200.jpg", "scene0458_00/01240.jpg", "scene0458_00/01280.jpg", "scene0458_00/01320.jpg", "scene0458_00/01360.jpg" ], "turn_id": 7, "scan_id": "scene0458_00", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the rectangular wooden window with closed horizontal blinds\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the rectangular wooden window with closed horizontal blinds\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0458_00/01400.jpg", "scene0458_00/01440.jpg", "scene0458_00/01480.jpg", "scene0458_00/01520.jpg", "scene0458_00/01560.jpg" ], "turn_id": 8, "scan_id": "scene0458_00", "origin_question": "Please recall \"the white oval ceramic sink with a silver faucet\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.68", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the white oval ceramic sink with a silver faucet\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0458_00/01600.jpg", "scene0458_00/01640.jpg", "scene0458_00/01680.jpg", "scene0458_00/01720.jpg", "scene0458_00/01760.jpg" ], "turn_id": 9, "scan_id": "scene0458_00", "origin_question": "Which direction is the rectangular wooden window with closed horizontal blinds in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the rectangular wooden window with closed horizontal blinds in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0425_00/00000.jpg", "scene0425_00/00030.jpg", "scene0425_00/00060.jpg", "scene0425_00/00090.jpg" ], "turn_id": 1, "scan_id": "scene0425_00", "origin_question": "Remember, have you seen any tissue box(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any tissue box(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0425_00/00120.jpg", "scene0425_00/00150.jpg", "scene0425_00/00180.jpg", "scene0425_00/00210.jpg" ], "turn_id": 2, "scan_id": "scene0425_00", "origin_question": "When did you first discover \"the bed\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the bed\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "scene0425_00/00240.jpg", "scene0425_00/00270.jpg", "scene0425_00/00300.jpg", "scene0425_00/00330.jpg" ], "turn_id": 3, "scan_id": "scene0425_00", "origin_question": "Among these three objects(\"the bed\",\"the red and black durable medium to large backpack\",\"the clear plastic pitcher with a white lid\" ), which one is the farthest from you?", "option": [ "the bed", "the red and black durable medium to large backpack", "the clear plastic pitcher with a white lid" ], "answer": "the clear plastic pitcher with a white lid", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects(\"the bed\",\"the red and black durable medium to large backpack\",\"the clear plastic pitcher with a white lid\" ), which one is the farthest from you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the bed', 'the red and black durable medium to large backpack', 'the clear plastic pitcher with a white lid'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0425_00/00360.jpg", "scene0425_00/00390.jpg", "scene0425_00/00420.jpg", "scene0425_00/00450.jpg" ], "turn_id": 4, "scan_id": "scene0425_00", "origin_question": "Please recall \"the rectangular wooden desk with a brownish hue\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.3", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the rectangular wooden desk with a brownish hue\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0425_00/00480.jpg", "scene0425_00/00510.jpg", "scene0425_00/00540.jpg", "scene0425_00/00570.jpg" ], "turn_id": 5, "scan_id": "scene0425_00", "origin_question": "Is the bed to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the bed to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0425_00/00600.jpg", "scene0425_00/00630.jpg", "scene0425_00/00660.jpg", "scene0425_00/00690.jpg" ], "turn_id": 6, "scan_id": "scene0425_00", "origin_question": "Remember, have you seen any bin(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any bin(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0425_00/00720.jpg", "scene0425_00/00750.jpg", "scene0425_00/00780.jpg", "scene0425_00/00810.jpg" ], "turn_id": 7, "scan_id": "scene0425_00", "origin_question": "Please recall \"the white rectangular paper with red and black text\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.0", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the white rectangular paper with red and black text\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0555_00/00000.jpg", "scene0555_00/00040.jpg", "scene0555_00/00080.jpg", "scene0555_00/00120.jpg", "scene0555_00/00160.jpg" ], "turn_id": 1, "scan_id": "scene0555_00", "origin_question": "Remember, how many chair(s) have you seen so far? ", "option": [], "answer": "3", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many chair(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0555_00/00200.jpg", "scene0555_00/00240.jpg", "scene0555_00/00280.jpg", "scene0555_00/00320.jpg", "scene0555_00/00360.jpg" ], "turn_id": 2, "scan_id": "scene0555_00", "origin_question": "When did you first discover \"the flat panel computer monitor with a black frame\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the flat panel computer monitor with a black frame\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0555_00/00400.jpg", "scene0555_00/00440.jpg", "scene0555_00/00480.jpg", "scene0555_00/00520.jpg", "scene0555_00/00560.jpg" ], "turn_id": 3, "scan_id": "scene0555_00", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "146.47", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0555_00/00600.jpg", "scene0555_00/00640.jpg", "scene0555_00/00680.jpg", "scene0555_00/00720.jpg", "scene0555_00/00760.jpg" ], "turn_id": 4, "scan_id": "scene0555_00", "origin_question": "How far is your current position from where you were at the end of the turn 3? (in meters)", "option": [], "answer": "1.47", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 3? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0555_00/00800.jpg", "scene0555_00/00840.jpg", "scene0555_00/00880.jpg", "scene0555_00/00920.jpg", "scene0555_00/00960.jpg" ], "turn_id": 5, "scan_id": "scene0555_00", "origin_question": "In which turn were you closest to \"the medium-sized rectangular white table made of smooth wood\"? ", "option": [], "answer": "5", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the medium-sized rectangular white table made of smooth wood\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0555_00/01000.jpg", "scene0555_00/01040.jpg", "scene0555_00/01080.jpg", "scene0555_00/01120.jpg", "scene0555_00/01160.jpg" ], "turn_id": 6, "scan_id": "scene0555_00", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0555_00/01200.jpg", "scene0555_00/01240.jpg", "scene0555_00/01280.jpg", "scene0555_00/01320.jpg", "scene0555_00/01360.jpg" ], "turn_id": 7, "scan_id": "scene0555_00", "origin_question": "In which turn were you closest to \"the rectangular white counter made of wood and metal\"? ", "option": [], "answer": "7", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the rectangular white counter made of wood and metal\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0555_00/01400.jpg", "scene0555_00/01440.jpg", "scene0555_00/01480.jpg", "scene0555_00/01520.jpg", "scene0555_00/01560.jpg" ], "turn_id": 8, "scan_id": "scene0555_00", "origin_question": "Remember, have you seen any drawer(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any drawer(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0555_00/01600.jpg", "scene0555_00/01640.jpg", "scene0555_00/01680.jpg", "scene0555_00/01720.jpg", "scene0555_00/01760.jpg" ], "turn_id": 9, "scan_id": "scene0555_00", "origin_question": "Is the rectangular white counter made of wood and metal to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the rectangular white counter made of wood and metal to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0509_02/00000.jpg", "scene0509_02/00030.jpg", "scene0509_02/00060.jpg", "scene0509_02/00090.jpg" ], "turn_id": 1, "scan_id": "scene0509_02", "origin_question": "Remember, have you seen any toilet(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any toilet(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0509_02/00120.jpg", "scene0509_02/00150.jpg", "scene0509_02/00180.jpg", "scene0509_02/00210.jpg" ], "turn_id": 2, "scan_id": "scene0509_02", "origin_question": "When did you first discover \"the rectangular black wooden cabinet with a smooth surface\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the rectangular black wooden cabinet with a smooth surface\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0509_02/00240.jpg", "scene0509_02/00270.jpg", "scene0509_02/00300.jpg", "scene0509_02/00330.jpg" ], "turn_id": 3, "scan_id": "scene0509_02", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the standard-sized roll of white toilet paper\" from your position?", "option": [], "answer": "178.11", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the standard-sized roll of white toilet paper\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0509_02/00360.jpg", "scene0509_02/00390.jpg", "scene0509_02/00420.jpg", "scene0509_02/00450.jpg" ], "turn_id": 4, "scan_id": "scene0509_02", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0509_02/00480.jpg", "scene0509_02/00510.jpg", "scene0509_02/00540.jpg", "scene0509_02/00570.jpg" ], "turn_id": 5, "scan_id": "scene0509_02", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the sink\" from your position?", "option": [], "answer": "164.67", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the sink\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0509_02/00600.jpg", "scene0509_02/00630.jpg", "scene0509_02/00660.jpg", "scene0509_02/00690.jpg" ], "turn_id": 6, "scan_id": "scene0509_02", "origin_question": "At the end of which turn were both of \"the standard-sized roll of white toilet paper\" and \"the rectangular black wooden cabinet with a smooth surface\" on your right side?", "option": [], "answer": "6", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the standard-sized roll of white toilet paper\" and \"the rectangular black wooden cabinet with a smooth surface\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0509_02/00720.jpg", "scene0509_02/00750.jpg", "scene0509_02/00780.jpg", "scene0509_02/00810.jpg" ], "turn_id": 7, "scan_id": "scene0509_02", "origin_question": "By the end of the 3 turn, were you getting closer to or farther away from these two objects: X: \"the standard-sized roll of white toilet paper\", Y: \"the sink\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "farther away from both X and Y", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 3 turn, were you getting closer to or farther away from these two objects: X: \"the standard-sized roll of white toilet paper\", Y: \"the sink\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0509_02/00840.jpg", "scene0509_02/00870.jpg", "scene0509_02/00900.jpg", "scene0509_02/00930.jpg" ], "turn_id": 8, "scan_id": "scene0509_02", "origin_question": "Which two objects are on the same side of you now? \"the white toilet with a round bowl and lid\", \"the rail\", \"the rectangular black wooden cabinet with a smooth surface\". ", "option": [ "the white toilet with a round bowl and lid, the rail", "the white toilet with a round bowl and lid, the rectangular black wooden cabinet with a smooth surface", "the rail, the rectangular black wooden cabinet with a smooth surface" ], "answer": "the white toilet with a round bowl and lid, the rail", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the white toilet with a round bowl and lid\", \"the rail\", \"the rectangular black wooden cabinet with a smooth surface\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the white toilet with a round bowl and lid, the rail', 'the white toilet with a round bowl and lid, the rectangular black wooden cabinet with a smooth surface', 'the rail, the rectangular black wooden cabinet with a smooth surface'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0029_region24/1838a8fb8e9f48cebcd079d5373cba02_i1_5.jpg", "1mp3d_0029_region24/1838a8fb8e9f48cebcd079d5373cba02_i2_5.jpg", "1mp3d_0029_region24/d2f529abd1404526bb735dc0b5588df3_i1_5.jpg", "1mp3d_0029_region24/d2f529abd1404526bb735dc0b5588df3_i2_5.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0029_region24", "origin_question": "Remember, have you seen any door(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any door(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0029_region24/d2f529abd1404526bb735dc0b5588df3_i1_4.jpg", "1mp3d_0029_region24/d2f529abd1404526bb735dc0b5588df3_i2_4.jpg", "1mp3d_0029_region24/1838a8fb8e9f48cebcd079d5373cba02_i1_0.jpg", "1mp3d_0029_region24/1838a8fb8e9f48cebcd079d5373cba02_i2_0.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0029_region24", "origin_question": "Remember, have you seen any toothbrush(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any toothbrush(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "1mp3d_0029_region24/dfe1f8e8b58d4da49f0c73516e7a4b67_i1_0.jpg", "1mp3d_0029_region24/dfe1f8e8b58d4da49f0c73516e7a4b67_i2_0.jpg", "1mp3d_0029_region24/3400e984cd9642008cdb2cbbff0989a9_i1_1.jpg", "1mp3d_0029_region24/3400e984cd9642008cdb2cbbff0989a9_i2_1.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0029_region24", "origin_question": "Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Left", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 2 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "1mp3d_0029_region24/b47475370e824b9e923a4a96bc98e1e3_i1_2.jpg", "1mp3d_0029_region24/b47475370e824b9e923a4a96bc98e1e3_i2_2.jpg", "1mp3d_0029_region24/64b5e1a6f2c64d779bd3a7286b1bf7d3_i1_1.jpg", "1mp3d_0029_region24/64b5e1a6f2c64d779bd3a7286b1bf7d3_i2_1.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0029_region24", "origin_question": "Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "1mp3d_0029_region24/3d21c0cbc98b4c3da6585fabef9c68c3_i1_3.jpg", "1mp3d_0029_region24/3d21c0cbc98b4c3da6585fabef9c68c3_i2_3.jpg", "1mp3d_0029_region24/64b5e1a6f2c64d779bd3a7286b1bf7d3_i1_2.jpg", "1mp3d_0029_region24/64b5e1a6f2c64d779bd3a7286b1bf7d3_i2_2.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0029_region24", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "3.09", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "1mp3d_0029_region24/b47475370e824b9e923a4a96bc98e1e3_i1_4.jpg", "1mp3d_0029_region24/b47475370e824b9e923a4a96bc98e1e3_i2_4.jpg", "1mp3d_0029_region24/b47475370e824b9e923a4a96bc98e1e3_i1_5.jpg", "1mp3d_0029_region24/b47475370e824b9e923a4a96bc98e1e3_i2_5.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0029_region24", "origin_question": "What will be the first-time appearance order of the following categories: door, plant, switch?", "option": [ "plant, door, switch", "switch, plant, door", "switch, door, plant", "door, plant, switch" ], "answer": "door, plant, switch", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: door, plant, switch? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['plant, door, switch', 'switch, plant, door', 'switch, door, plant', 'door, plant, switch'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "1mp3d_0029_region24/3400e984cd9642008cdb2cbbff0989a9_i1_4.jpg", "1mp3d_0029_region24/3400e984cd9642008cdb2cbbff0989a9_i2_4.jpg", "1mp3d_0029_region24/3400e984cd9642008cdb2cbbff0989a9_i1_5.jpg", "1mp3d_0029_region24/3400e984cd9642008cdb2cbbff0989a9_i2_5.jpg" ], "turn_id": 7, "scan_id": "1mp3d_0029_region24", "origin_question": "Which direction is the vent in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the vent in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "1mp3d_0029_region24/dfe1f8e8b58d4da49f0c73516e7a4b67_i1_4.jpg", "1mp3d_0029_region24/dfe1f8e8b58d4da49f0c73516e7a4b67_i2_4.jpg", "1mp3d_0029_region24/dd83fb40a2e14ac99de9fe9bcfaf44df_i1_2.jpg", "1mp3d_0029_region24/dd83fb40a2e14ac99de9fe9bcfaf44df_i2_2.jpg" ], "turn_id": 8, "scan_id": "1mp3d_0029_region24", "origin_question": "Assuming the direction you are facing at the end of the turn 7 is forward, did you move a certain distance forward or backward from that position?", "option": [ "Forward", "Backward" ], "answer": "Backward", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 7 is forward, did you move a certain distance forward or backward from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Forward', 'Backward'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0606_00/00000.jpg", "scene0606_00/00040.jpg", "scene0606_00/00080.jpg", "scene0606_00/00120.jpg", "scene0606_00/00160.jpg" ], "turn_id": 1, "scan_id": "scene0606_00", "origin_question": "Remember, have you seen any chair(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any chair(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0606_00/00200.jpg", "scene0606_00/00240.jpg", "scene0606_00/00280.jpg", "scene0606_00/00320.jpg", "scene0606_00/00360.jpg" ], "turn_id": 2, "scan_id": "scene0606_00", "origin_question": "At the end of which turn were both of \"the large rectangular white plastic bag with orange straps\" and \"the painting with a dark wooden frame\" on your right side?", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the large rectangular white plastic bag with orange straps\" and \"the painting with a dark wooden frame\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0606_00/00400.jpg", "scene0606_00/00440.jpg", "scene0606_00/00480.jpg", "scene0606_00/00520.jpg", "scene0606_00/00560.jpg" ], "turn_id": 3, "scan_id": "scene0606_00", "origin_question": "Please recall \"the large rectangular white plastic bag with orange straps\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "3.14", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the large rectangular white plastic bag with orange straps\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0606_00/00600.jpg", "scene0606_00/00640.jpg", "scene0606_00/00680.jpg", "scene0606_00/00720.jpg", "scene0606_00/00760.jpg" ], "turn_id": 4, "scan_id": "scene0606_00", "origin_question": "Which direction is the large blue plastic case with compartments in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the large blue plastic case with compartments in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0606_00/00800.jpg", "scene0606_00/00840.jpg", "scene0606_00/00880.jpg", "scene0606_00/00920.jpg", "scene0606_00/00960.jpg" ], "turn_id": 5, "scan_id": "scene0606_00", "origin_question": "Which direction is the rectangular countertop oven with a metallic sheen in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the rectangular countertop oven with a metallic sheen in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0606_00/01000.jpg", "scene0606_00/01040.jpg", "scene0606_00/01080.jpg", "scene0606_00/01120.jpg", "scene0606_00/01160.jpg" ], "turn_id": 6, "scan_id": "scene0606_00", "origin_question": "Which two objects are on the same side of you now? \"the clear plastic cylindrical bottle with a white lid\", \"the neutral textured stone or ceramic vase\", \"the medium rectangular beige pillow made of soft material\". ", "option": [ "the clear plastic cylindrical bottle with a white lid, the neutral textured stone or ceramic vase", "the clear plastic cylindrical bottle with a white lid, the medium rectangular beige pillow made of soft material", "the neutral textured stone or ceramic vase, the medium rectangular beige pillow made of soft material" ], "answer": "the clear plastic cylindrical bottle with a white lid, the neutral textured stone or ceramic vase", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the clear plastic cylindrical bottle with a white lid\", \"the neutral textured stone or ceramic vase\", \"the medium rectangular beige pillow made of soft material\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the clear plastic cylindrical bottle with a white lid, the neutral textured stone or ceramic vase', 'the clear plastic cylindrical bottle with a white lid, the medium rectangular beige pillow made of soft material', 'the neutral textured stone or ceramic vase, the medium rectangular beige pillow made of soft material'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "scene0606_00/01200.jpg", "scene0606_00/01240.jpg", "scene0606_00/01280.jpg", "scene0606_00/01320.jpg", "scene0606_00/01360.jpg" ], "turn_id": 7, "scan_id": "scene0606_00", "origin_question": "Among these three objects(\"the painting with a dark wooden frame\",\"the clear plastic cylindrical bottle with a white lid\",\"the medium rectangular beige pillow made of soft material\" ), which one is the farthest from you?", "option": [ "the painting with a dark wooden frame", "the clear plastic cylindrical bottle with a white lid", "the medium rectangular beige pillow made of soft material" ], "answer": "the painting with a dark wooden frame", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the painting with a dark wooden frame\",\"the clear plastic cylindrical bottle with a white lid\",\"the medium rectangular beige pillow made of soft material\" ), which one is the farthest from you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the painting with a dark wooden frame', 'the clear plastic cylindrical bottle with a white lid', 'the medium rectangular beige pillow made of soft material'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0606_00/01400.jpg", "scene0606_00/01440.jpg", "scene0606_00/01480.jpg", "scene0606_00/01520.jpg", "scene0606_00/01560.jpg" ], "turn_id": 8, "scan_id": "scene0606_00", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the light-colored rectangular kitchen counter\" from your position?", "option": [], "answer": "100.07", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the light-colored rectangular kitchen counter\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0606_00/01600.jpg", "scene0606_00/01640.jpg", "scene0606_00/01680.jpg", "scene0606_00/01720.jpg", "scene0606_00/01760.jpg" ], "turn_id": 9, "scan_id": "scene0606_00", "origin_question": "Which direction is the rectangular glass window in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the rectangular glass window in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0606_00/01800.jpg", "scene0606_00/01840.jpg", "scene0606_00/01880.jpg", "scene0606_00/01920.jpg", "scene0606_00/01960.jpg" ], "turn_id": 10, "scan_id": "scene0606_00", "origin_question": "In which turn were you closest to \"the large rectangular white plastic bag with orange straps\"? ", "option": [], "answer": "10", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the large rectangular white plastic bag with orange straps\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0047_region11/62ee370ab6294deab2801a0d8b9d7b4e_i1_0.jpg", "1mp3d_0047_region11/62ee370ab6294deab2801a0d8b9d7b4e_i2_0.jpg", "1mp3d_0047_region11/51249b2653e1438cb8d455c96fdabf2c_i1_3.jpg", "1mp3d_0047_region11/51249b2653e1438cb8d455c96fdabf2c_i2_3.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0047_region11", "origin_question": "Remember, have you seen any fire extinguisher(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any fire extinguisher(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "1mp3d_0047_region11/6a5844722dd34235a4fc3270381de880_i1_0.jpg", "1mp3d_0047_region11/6a5844722dd34235a4fc3270381de880_i2_0.jpg", "1mp3d_0047_region11/6a5844722dd34235a4fc3270381de880_i1_5.jpg", "1mp3d_0047_region11/6a5844722dd34235a4fc3270381de880_i2_5.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0047_region11", "origin_question": "When did you first discover \"the stainless steel gas stove with a black top\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the stainless steel gas stove with a black top\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "1mp3d_0047_region11/b24fe936bb9b4b86be143c0cb4dfff9a_i1_0.jpg", "1mp3d_0047_region11/b24fe936bb9b4b86be143c0cb4dfff9a_i2_0.jpg", "1mp3d_0047_region11/6a5844722dd34235a4fc3270381de880_i1_1.jpg", "1mp3d_0047_region11/6a5844722dd34235a4fc3270381de880_i2_1.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0047_region11", "origin_question": "Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "96.33", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "1mp3d_0047_region11/51249b2653e1438cb8d455c96fdabf2c_i1_4.jpg", "1mp3d_0047_region11/51249b2653e1438cb8d455c96fdabf2c_i2_4.jpg", "1mp3d_0047_region11/62ee370ab6294deab2801a0d8b9d7b4e_i1_4.jpg", "1mp3d_0047_region11/62ee370ab6294deab2801a0d8b9d7b4e_i2_4.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0047_region11", "origin_question": "Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "143.92", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "1mp3d_0047_region11/62ee370ab6294deab2801a0d8b9d7b4e_i1_3.jpg", "1mp3d_0047_region11/62ee370ab6294deab2801a0d8b9d7b4e_i2_3.jpg", "1mp3d_0047_region11/d5a2d6c9ec8a47d0a7f17282c62dc5d7_i1_2.jpg", "1mp3d_0047_region11/d5a2d6c9ec8a47d0a7f17282c62dc5d7_i2_2.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0047_region11", "origin_question": "Is the large rectangular stainless steel sink with a shiny surface to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the large rectangular stainless steel sink with a shiny surface to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0040_region33/7fe65eed53f343edbdc30fe083d40c03_i1_0.jpg", "1mp3d_0040_region33/7fe65eed53f343edbdc30fe083d40c03_i2_0.jpg", "1mp3d_0040_region33/b5a0556b734741f5be20682706ab2ceb_i1_0.jpg", "1mp3d_0040_region33/b5a0556b734741f5be20682706ab2ceb_i2_0.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0040_region33", "origin_question": "Remember, have you seen any door(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any door(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "1mp3d_0040_region33/7fe65eed53f343edbdc30fe083d40c03_i1_1.jpg", "1mp3d_0040_region33/7fe65eed53f343edbdc30fe083d40c03_i2_1.jpg", "1mp3d_0040_region33/bad3bec20c0348229818447d00959338_i1_0.jpg", "1mp3d_0040_region33/bad3bec20c0348229818447d00959338_i2_0.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0040_region33", "origin_question": "In which turn were you farthest from \"the rectangular glass table with a black metal frame\"? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the rectangular glass table with a black metal frame\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "1mp3d_0040_region33/605736f22ba342a19dd128b758abe427_i1_0.jpg", "1mp3d_0040_region33/605736f22ba342a19dd128b758abe427_i2_0.jpg", "1mp3d_0040_region33/de9645c3960e474cb2b641afa18ff585_i1_1.jpg", "1mp3d_0040_region33/de9645c3960e474cb2b641afa18ff585_i2_1.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0040_region33", "origin_question": "What will be the first-time appearance order of the following categories: fan, pillar, table?", "option": [ "pillar, fan, table", "table, pillar, fan", "fan, pillar, table", "fan, table, pillar" ], "answer": "pillar, fan, table", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: fan, pillar, table? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['pillar, fan, table', 'table, pillar, fan', 'fan, pillar, table', 'fan, table, pillar'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "1mp3d_0040_region33/605736f22ba342a19dd128b758abe427_i1_2.jpg", "1mp3d_0040_region33/605736f22ba342a19dd128b758abe427_i2_2.jpg", "1mp3d_0040_region33/6cd3f0b6fb694de88dccb49a659b30f4_i1_2.jpg", "1mp3d_0040_region33/6cd3f0b6fb694de88dccb49a659b30f4_i2_2.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0040_region33", "origin_question": "At the end of which turn were both of \"the rectangular wall-mounted flat-screen TV with a black frame\" and \"the medium-sized fan with metallic blades\" on your right side?", "option": [], "answer": "2", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the rectangular wall-mounted flat-screen TV with a black frame\" and \"the medium-sized fan with metallic blades\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "1mp3d_0040_region33/6cd3f0b6fb694de88dccb49a659b30f4_i1_3.jpg", "1mp3d_0040_region33/6cd3f0b6fb694de88dccb49a659b30f4_i2_3.jpg", "1mp3d_0040_region33/80984eb5f1b446e7941cf10d542554e2_i1_3.jpg", "1mp3d_0040_region33/80984eb5f1b446e7941cf10d542554e2_i2_3.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0040_region33", "origin_question": "Please recall \"the compact rectangular metal cabinet with silver doors\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "5.28", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the compact rectangular metal cabinet with silver doors\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "1mp3d_0040_region33/80984eb5f1b446e7941cf10d542554e2_i1_2.jpg", "1mp3d_0040_region33/80984eb5f1b446e7941cf10d542554e2_i2_2.jpg", "1mp3d_0040_region33/2179e9aa345b47e6b4ae05588c79f2c7_i1_3.jpg", "1mp3d_0040_region33/2179e9aa345b47e6b4ae05588c79f2c7_i2_3.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0040_region33", "origin_question": "Remember, how many shampoo(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many shampoo(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "1mp3d_0040_region33/2179e9aa345b47e6b4ae05588c79f2c7_i1_2.jpg", "1mp3d_0040_region33/2179e9aa345b47e6b4ae05588c79f2c7_i2_2.jpg", "1mp3d_0040_region33/7f69997a170f44669146d8055599b59c_i1_2.jpg", "1mp3d_0040_region33/7f69997a170f44669146d8055599b59c_i2_2.jpg" ], "turn_id": 7, "scan_id": "1mp3d_0040_region33", "origin_question": "Is the rectangular wall-mounted flat-screen TV with a black frame to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the rectangular wall-mounted flat-screen TV with a black frame to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "1mp3d_0040_region33/7f69997a170f44669146d8055599b59c_i1_3.jpg", "1mp3d_0040_region33/7f69997a170f44669146d8055599b59c_i2_3.jpg", "1mp3d_0040_region33/9cdfad544440432889ef083904548041_i1_4.jpg", "1mp3d_0040_region33/9cdfad544440432889ef083904548041_i2_4.jpg" ], "turn_id": 8, "scan_id": "1mp3d_0040_region33", "origin_question": "What will be the first-time appearance order of the following categories: couch, fan, pillar?", "option": [ "fan, couch, pillar", "fan, pillar, couch", "pillar, couch, fan", "pillar, fan, couch" ], "answer": "pillar, couch, fan", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: couch, fan, pillar? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['fan, couch, pillar', 'fan, pillar, couch', 'pillar, couch, fan', 'pillar, fan, couch'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "1mp3d_0040_region33/19f9a14a488e44c1af958903be20c6ec_i1_4.jpg", "1mp3d_0040_region33/19f9a14a488e44c1af958903be20c6ec_i2_4.jpg", "1mp3d_0040_region33/eb2a69f69ce44fb69d68de6055a0e295_i1_4.jpg", "1mp3d_0040_region33/eb2a69f69ce44fb69d68de6055a0e295_i2_4.jpg" ], "turn_id": 9, "scan_id": "1mp3d_0040_region33", "origin_question": "Please recall \"the rectangular wall-mounted flat-screen TV with a black frame\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.28", "user_message": "For the 9 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the rectangular wall-mounted flat-screen TV with a black frame\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0046_00/00000.jpg", "scene0046_00/00040.jpg", "scene0046_00/00080.jpg", "scene0046_00/00120.jpg", "scene0046_00/00160.jpg" ], "turn_id": 1, "scan_id": "scene0046_00", "origin_question": "Remember, how many backpack(s) have you seen so far? ", "option": [], "answer": "3", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many backpack(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0046_00/00200.jpg", "scene0046_00/00240.jpg", "scene0046_00/00280.jpg", "scene0046_00/00320.jpg", "scene0046_00/00360.jpg" ], "turn_id": 2, "scan_id": "scene0046_00", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the soft, white standard pillow made of plush material\"; \"the tall rectangular light brown wooden wardrobe\"; \"the small light brown wooden stand\" ", "option": [ "the soft, white standard pillow made of plush material", "the tall rectangular light brown wooden wardrobe", "the small light brown wooden stand" ], "answer": "the soft, white standard pillow made of plush material", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the soft, white standard pillow made of plush material\"; \"the tall rectangular light brown wooden wardrobe\"; \"the small light brown wooden stand\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the soft, white standard pillow made of plush material', 'the tall rectangular light brown wooden wardrobe', 'the small light brown wooden stand'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0046_00/00400.jpg", "scene0046_00/00440.jpg", "scene0046_00/00480.jpg", "scene0046_00/00520.jpg", "scene0046_00/00560.jpg" ], "turn_id": 3, "scan_id": "scene0046_00", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the soft, white standard pillow made of plush material\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the soft, white standard pillow made of plush material\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0046_00/00600.jpg", "scene0046_00/00640.jpg", "scene0046_00/00680.jpg", "scene0046_00/00720.jpg", "scene0046_00/00760.jpg" ], "turn_id": 4, "scan_id": "scene0046_00", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the curtain\"; \"the black rectangular telephone with rounded corners\"; \"the bed\" ", "option": [ "the curtain", "the black rectangular telephone with rounded corners", "the bed" ], "answer": "the black rectangular telephone with rounded corners", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the curtain\"; \"the black rectangular telephone with rounded corners\"; \"the bed\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the curtain', 'the black rectangular telephone with rounded corners', 'the bed'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0046_00/00800.jpg", "scene0046_00/00840.jpg", "scene0046_00/00880.jpg", "scene0046_00/00920.jpg", "scene0046_00/00960.jpg" ], "turn_id": 5, "scan_id": "scene0046_00", "origin_question": "What will be the first-time appearance order of the following categories: bed, mouse, pillow?", "option": [ "bed, mouse, pillow", "bed, pillow, mouse", "mouse, pillow, bed", "mouse, bed, pillow" ], "answer": "mouse, bed, pillow", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bed, mouse, pillow? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['bed, mouse, pillow', 'bed, pillow, mouse', 'mouse, pillow, bed', 'mouse, bed, pillow'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0046_00/01000.jpg", "scene0046_00/01040.jpg", "scene0046_00/01080.jpg", "scene0046_00/01120.jpg", "scene0046_00/01160.jpg" ], "turn_id": 6, "scan_id": "scene0046_00", "origin_question": "By the end of the 4 turn, were you getting closer to or farther away from these two objects: X: \"the black rectangular remote control with curved edges\", Y: \"the modern flat-screen TV with a sleek black frame\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "farther away from both X and Y", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: By the end of the 4 turn, were you getting closer to or farther away from these two objects: X: \"the black rectangular remote control with curved edges\", Y: \"the modern flat-screen TV with a sleek black frame\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0046_00/01200.jpg", "scene0046_00/01240.jpg", "scene0046_00/01280.jpg", "scene0046_00/01320.jpg", "scene0046_00/01360.jpg" ], "turn_id": 7, "scan_id": "scene0046_00", "origin_question": "In which turn were you closest to \"the shelf\"? ", "option": [], "answer": "4", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the shelf\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0046_00/01400.jpg", "scene0046_00/01440.jpg", "scene0046_00/01480.jpg", "scene0046_00/01520.jpg", "scene0046_00/01560.jpg" ], "turn_id": 8, "scan_id": "scene0046_00", "origin_question": "Using your orientation at the end of turn 6 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "60.73", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 6 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0046_00/01600.jpg", "scene0046_00/01640.jpg", "scene0046_00/01680.jpg", "scene0046_00/01720.jpg", "scene0046_00/01760.jpg" ], "turn_id": 9, "scan_id": "scene0046_00", "origin_question": "What will be the first-time appearance order of the following categories: bathtub, case, remote control?", "option": [ "case, remote control, bathtub", "bathtub, remote control, case", "bathtub, case, remote control", "remote control, case, bathtub" ], "answer": "case, remote control, bathtub", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bathtub, case, remote control? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['case, remote control, bathtub', 'bathtub, remote control, case', 'bathtub, case, remote control', 'remote control, case, bathtub'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0046_00/01800.jpg", "scene0046_00/01840.jpg", "scene0046_00/01880.jpg", "scene0046_00/01920.jpg", "scene0046_00/01960.jpg" ], "turn_id": 10, "scan_id": "scene0046_00", "origin_question": "Please recall \"the medium-sized black plastic mouse\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "3.39", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the medium-sized black plastic mouse\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0469_00/00000.jpg", "scene0469_00/00030.jpg", "scene0469_00/00060.jpg", "scene0469_00/00090.jpg" ], "turn_id": 1, "scan_id": "scene0469_00", "origin_question": "Remember, have you seen any broom(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any broom(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0469_00/00120.jpg", "scene0469_00/00150.jpg", "scene0469_00/00180.jpg", "scene0469_00/00210.jpg" ], "turn_id": 2, "scan_id": "scene0469_00", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the rectangular, standard blackboard with green writing\"; \"the rectangular, light beige wooden desk\"; \"the bag\" ", "option": [ "the rectangular, standard blackboard with green writing", "the rectangular, light beige wooden desk", "the bag" ], "answer": "the bag", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the rectangular, standard blackboard with green writing\"; \"the rectangular, light beige wooden desk\"; \"the bag\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular, standard blackboard with green writing', 'the rectangular, light beige wooden desk', 'the bag'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0469_00/00240.jpg", "scene0469_00/00270.jpg", "scene0469_00/00300.jpg", "scene0469_00/00330.jpg" ], "turn_id": 3, "scan_id": "scene0469_00", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the bag\", Y: \"the rectangular white sheet of glossy paper\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the bag\", Y: \"the rectangular white sheet of glossy paper\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0469_00/00360.jpg", "scene0469_00/00390.jpg", "scene0469_00/00420.jpg", "scene0469_00/00450.jpg" ], "turn_id": 4, "scan_id": "scene0469_00", "origin_question": "By the end of the 3 turn, were you getting closer to or farther away from these two objects: X: \"the standard-sized hardcover book with a white cover\", Y: \"the large white and green cylindrical pen\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "farther away from both X and Y", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 3 turn, were you getting closer to or farther away from these two objects: X: \"the standard-sized hardcover book with a white cover\", Y: \"the large white and green cylindrical pen\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0469_00/00480.jpg", "scene0469_00/00510.jpg", "scene0469_00/00540.jpg", "scene0469_00/00570.jpg" ], "turn_id": 5, "scan_id": "scene0469_00", "origin_question": "Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "149.63", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0469_00/00600.jpg", "scene0469_00/00630.jpg", "scene0469_00/00660.jpg", "scene0469_00/00690.jpg" ], "turn_id": 6, "scan_id": "scene0469_00", "origin_question": "By the end of the 5 turn, were you getting closer to or farther away from these two objects: X: \"the standard-sized hardcover book with a white cover\", Y: \"the rectangular white sheet of glossy paper\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "farther away from both X and Y", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 5 turn, were you getting closer to or farther away from these two objects: X: \"the standard-sized hardcover book with a white cover\", Y: \"the rectangular white sheet of glossy paper\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0469_00/00720.jpg", "scene0469_00/00750.jpg", "scene0469_00/00780.jpg", "scene0469_00/00810.jpg" ], "turn_id": 7, "scan_id": "scene0469_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0466_01/00000.jpg", "scene0466_01/00030.jpg", "scene0466_01/00060.jpg", "scene0466_01/00090.jpg" ], "turn_id": 1, "scan_id": "scene0466_01", "origin_question": "Remember, how many scale(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many scale(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "scene0466_01/00120.jpg", "scene0466_01/00150.jpg", "scene0466_01/00180.jpg", "scene0466_01/00210.jpg" ], "turn_id": 2, "scan_id": "scene0466_01", "origin_question": "Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance forward or backward from that position?", "option": [ "Forward", "Backward" ], "answer": "Forward", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance forward or backward from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Forward', 'Backward'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0466_01/00240.jpg", "scene0466_01/00270.jpg", "scene0466_01/00300.jpg", "scene0466_01/00330.jpg" ], "turn_id": 3, "scan_id": "scene0466_01", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "1.81", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0466_01/00360.jpg", "scene0466_01/00390.jpg", "scene0466_01/00420.jpg", "scene0466_01/00450.jpg" ], "turn_id": 4, "scan_id": "scene0466_01", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the small white rectangular plastic water cooler\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the small white rectangular plastic water cooler\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0466_01/00480.jpg", "scene0466_01/00510.jpg", "scene0466_01/00540.jpg", "scene0466_01/00570.jpg" ], "turn_id": 5, "scan_id": "scene0466_01", "origin_question": "How far is your current position from where you were at the end of the turn 3? (in meters)", "option": [], "answer": "2.95", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 3? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0004_region16/5bc66b50e1374ca7a49ca01316891481_i1_2.jpg", "1mp3d_0004_region16/5bc66b50e1374ca7a49ca01316891481_i2_2.jpg", "1mp3d_0004_region16/5bc66b50e1374ca7a49ca01316891481_i1_1.jpg", "1mp3d_0004_region16/5bc66b50e1374ca7a49ca01316891481_i2_1.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0004_region16", "origin_question": "Remember, have you seen any coat hanger(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any coat hanger(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0004_region16/f50dc2ed4dd34dd1942c3c2600c2da50_i1_4.jpg", "1mp3d_0004_region16/f50dc2ed4dd34dd1942c3c2600c2da50_i2_4.jpg", "1mp3d_0004_region16/f50dc2ed4dd34dd1942c3c2600c2da50_i1_0.jpg", "1mp3d_0004_region16/f50dc2ed4dd34dd1942c3c2600c2da50_i2_0.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0004_region16", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "1mp3d_0004_region16/f50dc2ed4dd34dd1942c3c2600c2da50_i1_5.jpg", "1mp3d_0004_region16/f50dc2ed4dd34dd1942c3c2600c2da50_i2_5.jpg", "1mp3d_0004_region16/1962d6bb7af4424c9902b0246ba766b8_i1_5.jpg", "1mp3d_0004_region16/1962d6bb7af4424c9902b0246ba766b8_i2_5.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0004_region16", "origin_question": "Please recall \"then ornate, tall, arched mirror with intricate patterns\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.95", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"then ornate, tall, arched mirror with intricate patterns\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "1mp3d_0004_region16/700e21e51cd343699d77f9bfe065cb43_i1_0.jpg", "1mp3d_0004_region16/700e21e51cd343699d77f9bfe065cb43_i2_0.jpg", "1mp3d_0004_region16/7f685a7931294290a830dcf281863746_i1_2.jpg", "1mp3d_0004_region16/7f685a7931294290a830dcf281863746_i2_2.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0004_region16", "origin_question": "Is then ornate, tall, arched mirror with intricate patterns to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Is then ornate, tall, arched mirror with intricate patterns to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "1mp3d_0004_region16/700e21e51cd343699d77f9bfe065cb43_i1_1.jpg", "1mp3d_0004_region16/700e21e51cd343699d77f9bfe065cb43_i2_1.jpg", "1mp3d_0004_region16/700e21e51cd343699d77f9bfe065cb43_i1_5.jpg", "1mp3d_0004_region16/700e21e51cd343699d77f9bfe065cb43_i2_5.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0004_region16", "origin_question": "When did you first discover \"then ornate, tall, arched mirror with intricate patterns\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"then ornate, tall, arched mirror with intricate patterns\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0557_00/00000.jpg", "scene0557_00/00040.jpg", "scene0557_00/00080.jpg", "scene0557_00/00120.jpg" ], "turn_id": 1, "scan_id": "scene0557_00", "origin_question": "Remember, have you seen any curtain(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any curtain(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0557_00/00160.jpg", "scene0557_00/00200.jpg", "scene0557_00/00240.jpg", "scene0557_00/00280.jpg" ], "turn_id": 2, "scan_id": "scene0557_00", "origin_question": "Which two objects are on the same side of you now? \"the ottoman\", \"the rectangular red pillow with white edges\", \"the rectangular brown cardboard box\". ", "option": [ "the ottoman, the rectangular red pillow with white edges", "the ottoman, the rectangular brown cardboard box", "the rectangular red pillow with white edges, the rectangular brown cardboard box" ], "answer": "the ottoman, the rectangular red pillow with white edges", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the ottoman\", \"the rectangular red pillow with white edges\", \"the rectangular brown cardboard box\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the ottoman, the rectangular red pillow with white edges', 'the ottoman, the rectangular brown cardboard box', 'the rectangular red pillow with white edges, the rectangular brown cardboard box'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0557_00/00320.jpg", "scene0557_00/00360.jpg", "scene0557_00/00400.jpg", "scene0557_00/00440.jpg" ], "turn_id": 3, "scan_id": "scene0557_00", "origin_question": "Is the ottoman to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the ottoman to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0557_00/00480.jpg", "scene0557_00/00520.jpg", "scene0557_00/00560.jpg", "scene0557_00/00600.jpg" ], "turn_id": 4, "scan_id": "scene0557_00", "origin_question": "Remember, how many stand(s) have you seen so far? ", "option": [], "answer": "2", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many stand(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0557_00/00640.jpg", "scene0557_00/00680.jpg", "scene0557_00/00730.jpg", "scene0557_00/00770.jpg" ], "turn_id": 5, "scan_id": "scene0557_00", "origin_question": "Please recall \"the rectangular brown cardboard box\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.98", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the rectangular brown cardboard box\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0557_00/00810.jpg", "scene0557_00/00850.jpg", "scene0557_00/00890.jpg", "scene0557_00/00930.jpg" ], "turn_id": 6, "scan_id": "scene0557_00", "origin_question": "Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the blanket\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the blanket\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0557_00/00970.jpg", "scene0557_00/01010.jpg", "scene0557_00/01050.jpg", "scene0557_00/01090.jpg" ], "turn_id": 7, "scan_id": "scene0557_00", "origin_question": "In which turn did you see both \"the ottoman\" and \"the rectangular glass mirror with a wooden frame\" simultaneously? ", "option": [], "answer": "7", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the ottoman\" and \"the rectangular glass mirror with a wooden frame\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0557_00/01130.jpg", "scene0557_00/01170.jpg", "scene0557_00/01210.jpg", "scene0557_00/01250.jpg" ], "turn_id": 8, "scan_id": "scene0557_00", "origin_question": "When did you first discover \"the small rectangular wooden shelf in light color\"? (index of the turn)", "option": [], "answer": "4", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the small rectangular wooden shelf in light color\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0149_00/00000.jpg", "scene0149_00/00040.jpg", "scene0149_00/00080.jpg", "scene0149_00/00120.jpg" ], "turn_id": 1, "scan_id": "scene0149_00", "origin_question": "Remember, have you seen any refrigerator(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any refrigerator(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0149_00/00160.jpg", "scene0149_00/00200.jpg", "scene0149_00/00240.jpg", "scene0149_00/00280.jpg" ], "turn_id": 2, "scan_id": "scene0149_00", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the pitcher\"; \"the medium-sized upright white refrigerator\"; \"the small rectangular white gas stove\" ", "option": [ "the pitcher", "the medium-sized upright white refrigerator", "the small rectangular white gas stove" ], "answer": "the pitcher", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the pitcher\"; \"the medium-sized upright white refrigerator\"; \"the small rectangular white gas stove\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the pitcher', 'the medium-sized upright white refrigerator', 'the small rectangular white gas stove'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0149_00/00320.jpg", "scene0149_00/00360.jpg", "scene0149_00/00400.jpg", "scene0149_00/00440.jpg" ], "turn_id": 3, "scan_id": "scene0149_00", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0149_00/00480.jpg", "scene0149_00/00520.jpg", "scene0149_00/00560.jpg", "scene0149_00/00600.jpg" ], "turn_id": 4, "scan_id": "scene0149_00", "origin_question": "Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the round white frying pan with a black handle\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the round white frying pan with a black handle\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0149_00/00640.jpg", "scene0149_00/00680.jpg", "scene0149_00/00720.jpg", "scene0149_00/00760.jpg" ], "turn_id": 5, "scan_id": "scene0149_00", "origin_question": "Which two objects are on the same side of you now? \"the white rectangular microwave with a glass door\", \"the small rectangular window with a white frame\", \"the bag\". ", "option": [ "the white rectangular microwave with a glass door, the small rectangular window with a white frame", "the white rectangular microwave with a glass door, the bag", "the small rectangular window with a white frame, the bag" ], "answer": "the small rectangular window with a white frame, the bag", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the white rectangular microwave with a glass door\", \"the small rectangular window with a white frame\", \"the bag\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the white rectangular microwave with a glass door, the small rectangular window with a white frame', 'the white rectangular microwave with a glass door, the bag', 'the small rectangular window with a white frame, the bag'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0149_00/00800.jpg", "scene0149_00/00840.jpg", "scene0149_00/00880.jpg", "scene0149_00/00920.jpg" ], "turn_id": 6, "scan_id": "scene0149_00", "origin_question": "Which two objects are on the same side of you now? \"the round white frying pan with a black handle\", \"the cylindrical light brown tissue roll\", \"the bag\". ", "option": [ "the round white frying pan with a black handle, the cylindrical light brown tissue roll", "the round white frying pan with a black handle, the bag", "the cylindrical light brown tissue roll, the bag" ], "answer": "the round white frying pan with a black handle, the cylindrical light brown tissue roll", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the round white frying pan with a black handle\", \"the cylindrical light brown tissue roll\", \"the bag\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the round white frying pan with a black handle, the cylindrical light brown tissue roll', 'the round white frying pan with a black handle, the bag', 'the cylindrical light brown tissue roll, the bag'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0149_00/00960.jpg", "scene0149_00/01000.jpg", "scene0149_00/01040.jpg", "scene0149_00/01100.jpg" ], "turn_id": 7, "scan_id": "scene0149_00", "origin_question": "Which two objects are on the same side of you now? \"the small rectangular white gas stove\", \"the small rectangular window with a white frame\", \"the bin\". ", "option": [ "the small rectangular white gas stove, the small rectangular window with a white frame", "the small rectangular white gas stove, the bin", "the small rectangular window with a white frame, the bin" ], "answer": "the small rectangular window with a white frame, the bin", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the small rectangular white gas stove\", \"the small rectangular window with a white frame\", \"the bin\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the small rectangular white gas stove, the small rectangular window with a white frame', 'the small rectangular white gas stove, the bin', 'the small rectangular window with a white frame, the bin'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0149_00/01140.jpg", "scene0149_00/01180.jpg", "scene0149_00/01220.jpg", "scene0149_00/01260.jpg" ], "turn_id": 8, "scan_id": "scene0149_00", "origin_question": "Please recall \"the black wooden chair with a white cushion\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "2.4", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the black wooden chair with a white cushion\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0218_00/00000.jpg", "scene0218_00/00040.jpg", "scene0218_00/00080.jpg", "scene0218_00/00120.jpg" ], "turn_id": 1, "scan_id": "scene0218_00", "origin_question": "Remember, have you seen any desk(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any desk(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0218_00/00160.jpg", "scene0218_00/00200.jpg", "scene0218_00/00240.jpg", "scene0218_00/00280.jpg" ], "turn_id": 2, "scan_id": "scene0218_00", "origin_question": "In which turn did you see both \"the desk\" and \"the small round black and white plastic stool\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the desk\" and \"the small round black and white plastic stool\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0218_00/00320.jpg", "scene0218_00/00360.jpg", "scene0218_00/00400.jpg", "scene0218_00/00440.jpg" ], "turn_id": 3, "scan_id": "scene0218_00", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "113.47", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0218_00/00480.jpg", "scene0218_00/00520.jpg", "scene0218_00/00560.jpg", "scene0218_00/00600.jpg" ], "turn_id": 4, "scan_id": "scene0218_00", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the small round black and white plastic stool\" from your position?", "option": [], "answer": "79.87", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the small round black and white plastic stool\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0218_00/00640.jpg", "scene0218_00/00680.jpg", "scene0218_00/00720.jpg", "scene0218_00/00760.jpg" ], "turn_id": 5, "scan_id": "scene0218_00", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0218_00/00800.jpg", "scene0218_00/00840.jpg", "scene0218_00/00880.jpg", "scene0218_00/00920.jpg" ], "turn_id": 6, "scan_id": "scene0218_00", "origin_question": "Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the small round black and white plastic stool\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the small round black and white plastic stool\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "scene0218_00/00960.jpg", "scene0218_00/01000.jpg", "scene0218_00/01040.jpg", "scene0218_00/01080.jpg" ], "turn_id": 7, "scan_id": "scene0218_00", "origin_question": "Assuming the direction you are facing at the end of the turn 6 is forward, did you move a certain distance left or right from that position?", "option": [ "Left", "Right" ], "answer": "Left", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 6 is forward, did you move a certain distance left or right from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Left', 'Right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0218_00/01120.jpg", "scene0218_00/01160.jpg", "scene0218_00/01200.jpg", "scene0218_00/01240.jpg" ], "turn_id": 8, "scan_id": "scene0218_00", "origin_question": "What will be the first-time appearance order of the following categories: couch, desk, stool?", "option": [ "desk, stool, couch", "stool, couch, desk", "couch, desk, stool", "stool, desk, couch" ], "answer": "desk, stool, couch", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: couch, desk, stool? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['desk, stool, couch', 'stool, couch, desk', 'couch, desk, stool', 'stool, desk, couch'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0218_00/01280.jpg", "scene0218_00/01320.jpg", "scene0218_00/01360.jpg", "scene0218_00/01400.jpg" ], "turn_id": 9, "scan_id": "scene0218_00", "origin_question": "By the end of the 3 turn, were you getting closer to or farther away from these two objects: X: \"the desk\", Y: \"the dark, casual, lightweight jacket made of synthetic fiber\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to Y and farther away from X", "user_message": "For the 9 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 3 turn, were you getting closer to or farther away from these two objects: X: \"the desk\", Y: \"the dark, casual, lightweight jacket made of synthetic fiber\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0686_00/00000.jpg", "scene0686_00/00040.jpg", "scene0686_00/00080.jpg", "scene0686_00/00120.jpg" ], "turn_id": 1, "scan_id": "scene0686_00", "origin_question": "Remember, have you seen any sink(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any sink(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0686_00/00160.jpg", "scene0686_00/00200.jpg", "scene0686_00/00240.jpg", "scene0686_00/00280.jpg" ], "turn_id": 2, "scan_id": "scene0686_00", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the moderate-sized white oval plastic soap dispenser\"; \"the white rectangular wall-mounted sink with two faucets\"; \"the dispenser\" ", "option": [ "the moderate-sized white oval plastic soap dispenser", "the white rectangular wall-mounted sink with two faucets", "the dispenser" ], "answer": "the dispenser", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the moderate-sized white oval plastic soap dispenser\"; \"the white rectangular wall-mounted sink with two faucets\"; \"the dispenser\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the moderate-sized white oval plastic soap dispenser', 'the white rectangular wall-mounted sink with two faucets', 'the dispenser'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0686_00/00320.jpg", "scene0686_00/00360.jpg", "scene0686_00/00400.jpg", "scene0686_00/00440.jpg" ], "turn_id": 3, "scan_id": "scene0686_00", "origin_question": "Is the moderate-sized white oval plastic soap dispenser to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the moderate-sized white oval plastic soap dispenser to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0686_00/00480.jpg", "scene0686_00/00520.jpg", "scene0686_00/00560.jpg", "scene0686_00/00600.jpg" ], "turn_id": 4, "scan_id": "scene0686_00", "origin_question": "Please recall \"the moderate-sized white oval plastic soap dispenser\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.72", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the moderate-sized white oval plastic soap dispenser\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0686_00/00640.jpg", "scene0686_00/00680.jpg", "scene0686_00/00720.jpg", "scene0686_00/00760.jpg" ], "turn_id": 5, "scan_id": "scene0686_00", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the white rectangular wall-mounted sink with two faucets\" from your position?", "option": [], "answer": "92.51", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the white rectangular wall-mounted sink with two faucets\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0686_00/00800.jpg", "scene0686_00/00840.jpg", "scene0686_00/00880.jpg", "scene0686_00/00920.jpg" ], "turn_id": 6, "scan_id": "scene0686_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0686_00/00960.jpg", "scene0686_00/01000.jpg", "scene0686_00/01040.jpg", "scene0686_00/01080.jpg" ], "turn_id": 7, "scan_id": "scene0686_00", "origin_question": "Is the white rectangular wall-mounted sink with two faucets to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the white rectangular wall-mounted sink with two faucets to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0686_00/01120.jpg", "scene0686_00/01160.jpg", "scene0686_00/01200.jpg", "scene0686_00/01240.jpg" ], "turn_id": 8, "scan_id": "scene0686_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0686_00/01280.jpg", "scene0686_00/01320.jpg", "scene0686_00/01360.jpg", "scene0686_00/01400.jpg" ], "turn_id": 9, "scan_id": "scene0686_00", "origin_question": "What will be the first-time appearance order of the following categories: bin, mirror, toilet?", "option": [ "mirror, toilet, bin", "mirror, bin, toilet", "toilet, mirror, bin", "bin, mirror, toilet" ], "answer": "mirror, toilet, bin", "user_message": "For the 9 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bin, mirror, toilet? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['mirror, toilet, bin', 'mirror, bin, toilet', 'toilet, mirror, bin', 'bin, mirror, toilet'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0631_02/00000.jpg", "scene0631_02/00040.jpg", "scene0631_02/00080.jpg", "scene0631_02/00120.jpg" ], "turn_id": 1, "scan_id": "scene0631_02", "origin_question": "Remember, have you seen any stand(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any stand(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0631_02/00160.jpg", "scene0631_02/00200.jpg", "scene0631_02/00240.jpg", "scene0631_02/00280.jpg" ], "turn_id": 2, "scan_id": "scene0631_02", "origin_question": "When did you first discover \"the lamp\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the lamp\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0631_02/00320.jpg", "scene0631_02/00360.jpg", "scene0631_02/00400.jpg", "scene0631_02/00440.jpg" ], "turn_id": 3, "scan_id": "scene0631_02", "origin_question": "At the end of which turn were both of \"the lamp\" and \"the box\" on your right side?", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the lamp\" and \"the box\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0631_02/00480.jpg", "scene0631_02/00520.jpg", "scene0631_02/00560.jpg", "scene0631_02/00600.jpg" ], "turn_id": 4, "scan_id": "scene0631_02", "origin_question": "What will be the first-time appearance order of the following categories: book, shelf, tv?", "option": [ "tv, book, shelf", "shelf, tv, book", "shelf, book, tv", "book, shelf, tv" ], "answer": "book, shelf, tv", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: book, shelf, tv? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['tv, book, shelf', 'shelf, tv, book', 'shelf, book, tv', 'book, shelf, tv'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0631_02/00640.jpg", "scene0631_02/00680.jpg", "scene0631_02/00720.jpg", "scene0631_02/00760.jpg" ], "turn_id": 5, "scan_id": "scene0631_02", "origin_question": "Using your orientation at the end of turn 4 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "119.07", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 4 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0631_02/00800.jpg", "scene0631_02/00840.jpg", "scene0631_02/00880.jpg", "scene0631_02/00920.jpg" ], "turn_id": 6, "scan_id": "scene0631_02", "origin_question": "In which turn did you see both \"the smooth rectangular single bed with soft white material\" and \"the clothes\" simultaneously? ", "option": [], "answer": "5", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the smooth rectangular single bed with soft white material\" and \"the clothes\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0631_02/00960.jpg", "scene0631_02/01000.jpg", "scene0631_02/01040.jpg", "scene0631_02/01080.jpg" ], "turn_id": 7, "scan_id": "scene0631_02", "origin_question": "In which turn did you see both \"the window\" and \"the clothes\" simultaneously? ", "option": [], "answer": "4", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the window\" and \"the clothes\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0631_02/01120.jpg", "scene0631_02/01160.jpg", "scene0631_02/01200.jpg", "scene0631_02/01240.jpg" ], "turn_id": 8, "scan_id": "scene0631_02", "origin_question": "Based on your current orientation, in what counterclockwise direction (in degrees) is \"the clothes\" from your position?", "option": [], "answer": "106.93", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what counterclockwise direction (in degrees) is \"the clothes\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0050_00/00000.jpg", "scene0050_00/00040.jpg", "scene0050_00/00080.jpg", "scene0050_00/00120.jpg", "scene0050_00/00160.jpg" ], "turn_id": 1, "scan_id": "scene0050_00", "origin_question": "Remember, how many backpack(s) have you seen so far? ", "option": [], "answer": "2", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many backpack(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0050_00/00200.jpg", "scene0050_00/00240.jpg", "scene0050_00/00280.jpg", "scene0050_00/00320.jpg", "scene0050_00/00360.jpg" ], "turn_id": 2, "scan_id": "scene0050_00", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the rectangular light gray wooden desk\"; \"the rectangular brown leather ottoman with a smooth top\"; \"the medium-sized brown leather couch\" ", "option": [ "the rectangular light gray wooden desk", "the rectangular brown leather ottoman with a smooth top", "the medium-sized brown leather couch" ], "answer": "the rectangular light gray wooden desk", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the rectangular light gray wooden desk\"; \"the rectangular brown leather ottoman with a smooth top\"; \"the medium-sized brown leather couch\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular light gray wooden desk', 'the rectangular brown leather ottoman with a smooth top', 'the medium-sized brown leather couch'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0050_00/00400.jpg", "scene0050_00/00440.jpg", "scene0050_00/00480.jpg", "scene0050_00/00520.jpg", "scene0050_00/00560.jpg" ], "turn_id": 3, "scan_id": "scene0050_00", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0050_00/00600.jpg", "scene0050_00/00640.jpg", "scene0050_00/00680.jpg", "scene0050_00/00720.jpg", "scene0050_00/00760.jpg" ], "turn_id": 4, "scan_id": "scene0050_00", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the bag\" from your position?", "option": [], "answer": "128.56", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the bag\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0050_00/00800.jpg", "scene0050_00/00840.jpg", "scene0050_00/00880.jpg", "scene0050_00/00920.jpg", "scene0050_00/00960.jpg" ], "turn_id": 5, "scan_id": "scene0050_00", "origin_question": "Which two objects are on the same side of you now? \"the rectangular, creamy textured pillow made of cotton\", \"the medium-sized brown leather couch\", \"the plain gray vertical printer made of plastic\". ", "option": [ "the rectangular, creamy textured pillow made of cotton, the medium-sized brown leather couch", "the rectangular, creamy textured pillow made of cotton, the plain gray vertical printer made of plastic", "the medium-sized brown leather couch, the plain gray vertical printer made of plastic" ], "answer": "the rectangular, creamy textured pillow made of cotton, the medium-sized brown leather couch", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the rectangular, creamy textured pillow made of cotton\", \"the medium-sized brown leather couch\", \"the plain gray vertical printer made of plastic\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular, creamy textured pillow made of cotton, the medium-sized brown leather couch', 'the rectangular, creamy textured pillow made of cotton, the plain gray vertical printer made of plastic', 'the medium-sized brown leather couch, the plain gray vertical printer made of plastic'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0050_00/01000.jpg", "scene0050_00/01040.jpg", "scene0050_00/01080.jpg", "scene0050_00/01120.jpg", "scene0050_00/01160.jpg" ], "turn_id": 6, "scan_id": "scene0050_00", "origin_question": "Which direction is the rectangular, creamy textured pillow made of cotton in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the rectangular, creamy textured pillow made of cotton in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0050_00/01200.jpg", "scene0050_00/01240.jpg", "scene0050_00/01280.jpg", "scene0050_00/01320.jpg", "scene0050_00/01360.jpg" ], "turn_id": 7, "scan_id": "scene0050_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0050_00/01400.jpg", "scene0050_00/01440.jpg", "scene0050_00/01480.jpg", "scene0050_00/01520.jpg", "scene0050_00/01560.jpg" ], "turn_id": 8, "scan_id": "scene0050_00", "origin_question": "Is the cylindrical teal-handled cup with a white interior to your left/right now?", "option": [ "left", "right" ], "answer": "left", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Is the cylindrical teal-handled cup with a white interior to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0050_00/01600.jpg", "scene0050_00/01640.jpg", "scene0050_00/01680.jpg", "scene0050_00/01720.jpg", "scene0050_00/01760.jpg" ], "turn_id": 9, "scan_id": "scene0050_00", "origin_question": "In which turn were you closest to \"the white rectangular tissue box with floral patterns\"? ", "option": [], "answer": "5", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the white rectangular tissue box with floral patterns\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0050_00/01800.jpg", "scene0050_00/01840.jpg", "scene0050_00/01880.jpg", "scene0050_00/01920.jpg", "scene0050_00/01960.jpg" ], "turn_id": 10, "scan_id": "scene0050_00", "origin_question": "In which turn did you see both \"the rectangular brown leather ottoman with a smooth top\" and \"the clothes\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the rectangular brown leather ottoman with a smooth top\" and \"the clothes\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "1mp3d_0084_region2/1edaa01e4a684f99b080cffdd684e9aa_i1_3.jpg", "1mp3d_0084_region2/1edaa01e4a684f99b080cffdd684e9aa_i2_3.jpg", "1mp3d_0084_region2/1edaa01e4a684f99b080cffdd684e9aa_i2_4.jpg", "1mp3d_0084_region2/622968ca9a434c64bf723c17f1e6a46c_i1_3.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0084_region2", "origin_question": "Remember, how many stove(s) have you seen so far? ", "option": [], "answer": "0", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many stove(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "1mp3d_0084_region2/622968ca9a434c64bf723c17f1e6a46c_i2_3.jpg", "1mp3d_0084_region2/e7300eb50a4549fbacf3a07c6ca894c6_i1_3.jpg", "1mp3d_0084_region2/e7300eb50a4549fbacf3a07c6ca894c6_i2_3.jpg", "1mp3d_0084_region2/7c24eb9bf0c54e0c89224dffd66ce227_i2_1.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0084_region2", "origin_question": "In which turn did you see both \"the square black writing desk made of dark material\" and \"the large rectangular wooden door in deep brown\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the square black writing desk made of dark material\" and \"the large rectangular wooden door in deep brown\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "1mp3d_0084_region2/e7300eb50a4549fbacf3a07c6ca894c6_i1_2.jpg", "1mp3d_0084_region2/e7300eb50a4549fbacf3a07c6ca894c6_i2_2.jpg", "1mp3d_0084_region2/622968ca9a434c64bf723c17f1e6a46c_i1_2.jpg", "1mp3d_0084_region2/622968ca9a434c64bf723c17f1e6a46c_i2_2.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0084_region2", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the square black writing desk made of dark material\"; \"the large rectangular orange velvety blanket\"; \"the conical brown lampshade with a black cylindrical lamp base\" ", "option": [ "the square black writing desk made of dark material", "the large rectangular orange velvety blanket", "the conical brown lampshade with a black cylindrical lamp base" ], "answer": "the conical brown lampshade with a black cylindrical lamp base", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the square black writing desk made of dark material\"; \"the large rectangular orange velvety blanket\"; \"the conical brown lampshade with a black cylindrical lamp base\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the square black writing desk made of dark material', 'the large rectangular orange velvety blanket', 'the conical brown lampshade with a black cylindrical lamp base'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "1mp3d_0084_region2/1edaa01e4a684f99b080cffdd684e9aa_i1_2.jpg", "1mp3d_0084_region2/1edaa01e4a684f99b080cffdd684e9aa_i2_2.jpg", "1mp3d_0084_region2/622968ca9a434c64bf723c17f1e6a46c_i1_1.jpg", "1mp3d_0084_region2/622968ca9a434c64bf723c17f1e6a46c_i2_1.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0084_region2", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the rectangular single bed with a patterned white bedsheet\"; \"the tall wooden cabinet with a deep reddish-brown hue\"; \"the conical brown lampshade with a black cylindrical lamp base\" ", "option": [ "the rectangular single bed with a patterned white bedsheet", "the tall wooden cabinet with a deep reddish-brown hue", "the conical brown lampshade with a black cylindrical lamp base" ], "answer": "the tall wooden cabinet with a deep reddish-brown hue", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the rectangular single bed with a patterned white bedsheet\"; \"the tall wooden cabinet with a deep reddish-brown hue\"; \"the conical brown lampshade with a black cylindrical lamp base\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular single bed with a patterned white bedsheet', 'the tall wooden cabinet with a deep reddish-brown hue', 'the conical brown lampshade with a black cylindrical lamp base'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Judgement)", "new_observations": [ "1mp3d_0084_region2/e7300eb50a4549fbacf3a07c6ca894c6_i1_0.jpg", "1mp3d_0084_region2/e7300eb50a4549fbacf3a07c6ca894c6_i2_0.jpg", "1mp3d_0084_region2/e7300eb50a4549fbacf3a07c6ca894c6_i1_1.jpg", "1mp3d_0084_region2/e7300eb50a4549fbacf3a07c6ca894c6_i2_1.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0084_region2", "origin_question": "Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance forward or backward from that position?", "option": [ "Forward", "Backward" ], "answer": "Backward", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Assuming the direction you are facing at the end of the turn 1 is forward, did you move a certain distance forward or backward from that position? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Forward', 'Backward'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0039_00/00000.jpg", "scene0039_00/00030.jpg", "scene0039_00/00060.jpg", "scene0039_00/00090.jpg" ], "turn_id": 1, "scan_id": "scene0039_00", "origin_question": "Remember, have you seen any door(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any door(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0039_00/00120.jpg", "scene0039_00/00630.jpg", "scene0039_00/00660.jpg", "scene0039_00/00690.jpg" ], "turn_id": 2, "scan_id": "scene0039_00", "origin_question": "In which turn were you farthest from \"the dark frame standard double or queen-size bed\"? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the dark frame standard double or queen-size bed\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0039_00/00720.jpg", "scene0039_00/00750.jpg", "scene0039_00/00780.jpg", "scene0039_00/00810.jpg" ], "turn_id": 3, "scan_id": "scene0039_00", "origin_question": "Remember, have you seen any paper(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any paper(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0039_00/00840.jpg", "scene0039_00/00870.jpg", "scene0039_00/00900.jpg", "scene0039_00/00930.jpg" ], "turn_id": 4, "scan_id": "scene0039_00", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the tall, dark wooden shelf with open sections\"; \"the rectangular wooden cabinet painted light beige\"; \"the modern lamp with a white lampshade and metal base\" ", "option": [ "the tall, dark wooden shelf with open sections", "the rectangular wooden cabinet painted light beige", "the modern lamp with a white lampshade and metal base" ], "answer": "the modern lamp with a white lampshade and metal base", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the tall, dark wooden shelf with open sections\"; \"the rectangular wooden cabinet painted light beige\"; \"the modern lamp with a white lampshade and metal base\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the tall, dark wooden shelf with open sections', 'the rectangular wooden cabinet painted light beige', 'the modern lamp with a white lampshade and metal base'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0039_00/00960.jpg", "scene0039_00/00990.jpg", "scene0039_00/01020.jpg", "scene0039_00/01050.jpg" ], "turn_id": 5, "scan_id": "scene0039_00", "origin_question": "Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the rectangular glossy black stand made of sturdy materials\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the rectangular glossy black stand made of sturdy materials\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0146_01/00000.jpg", "scene0146_01/00040.jpg", "scene0146_01/00080.jpg", "scene0146_01/00120.jpg" ], "turn_id": 1, "scan_id": "scene0146_01", "origin_question": "Remember, have you seen any computer(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any computer(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0146_01/00160.jpg", "scene0146_01/00200.jpg", "scene0146_01/00240.jpg", "scene0146_01/00280.jpg" ], "turn_id": 2, "scan_id": "scene0146_01", "origin_question": "Please recall \"the rectangular white piece of thin paper\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.99", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the rectangular white piece of thin paper\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0146_01/00320.jpg", "scene0146_01/00360.jpg", "scene0146_01/00400.jpg", "scene0146_01/00440.jpg" ], "turn_id": 3, "scan_id": "scene0146_01", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the medium-sized brown wooden door\"; \"the rectangular black and grey plastic soap dispenser\"; \"the small, circular silver stopcock with two handles\" ", "option": [ "the medium-sized brown wooden door", "the rectangular black and grey plastic soap dispenser", "the small, circular silver stopcock with two handles" ], "answer": "the medium-sized brown wooden door", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the medium-sized brown wooden door\"; \"the rectangular black and grey plastic soap dispenser\"; \"the small, circular silver stopcock with two handles\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the medium-sized brown wooden door', 'the rectangular black and grey plastic soap dispenser', 'the small, circular silver stopcock with two handles'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0146_01/00480.jpg", "scene0146_01/00520.jpg", "scene0146_01/00560.jpg", "scene0146_01/00600.jpg" ], "turn_id": 4, "scan_id": "scene0146_01", "origin_question": "What will be the first-time appearance order of the following categories: backpack, paper, sink?", "option": [ "sink, backpack, paper", "paper, backpack, sink", "paper, sink, backpack", "backpack, paper, sink" ], "answer": "paper, sink, backpack", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: backpack, paper, sink? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['sink, backpack, paper', 'paper, backpack, sink', 'paper, sink, backpack', 'backpack, paper, sink'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0146_01/00640.jpg", "scene0146_01/00680.jpg", "scene0146_01/00720.jpg", "scene0146_01/00770.jpg" ], "turn_id": 5, "scan_id": "scene0146_01", "origin_question": "What will be the first-time appearance order of the following categories: backpack, bin, mirror?", "option": [ "bin, mirror, backpack", "mirror, backpack, bin", "mirror, bin, backpack", "backpack, mirror, bin" ], "answer": "bin, mirror, backpack", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: backpack, bin, mirror? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['bin, mirror, backpack', 'mirror, backpack, bin', 'mirror, bin, backpack', 'backpack, mirror, bin'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0146_01/00810.jpg", "scene0146_01/00850.jpg", "scene0146_01/00890.jpg", "scene0146_01/00930.jpg" ], "turn_id": 6, "scan_id": "scene0146_01", "origin_question": "Compared to your position at the end of the 5 turn, are you now closer or farther away from \"the switch\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 5 turn, are you now closer or farther away from \"the switch\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0146_01/00970.jpg", "scene0146_01/01010.jpg", "scene0146_01/01050.jpg", "scene0146_01/01090.jpg" ], "turn_id": 7, "scan_id": "scene0146_01", "origin_question": "In which turn did you see both \"the small, circular silver stopcock with two handles\" and \"the switch\" simultaneously? ", "option": [], "answer": "3", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the small, circular silver stopcock with two handles\" and \"the switch\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0057_region11/1216ded953194b41a63d635217e6bed7_i1_3.jpg", "1mp3d_0057_region11/1216ded953194b41a63d635217e6bed7_i2_3.jpg", "1mp3d_0057_region11/071418244e994a8284a535fc136de090_i1_3.jpg", "1mp3d_0057_region11/071418244e994a8284a535fc136de090_i2_3.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0057_region11", "origin_question": "Remember, have you seen any bed(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any bed(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "1mp3d_0057_region11/1216ded953194b41a63d635217e6bed7_i1_2.jpg", "1mp3d_0057_region11/1216ded953194b41a63d635217e6bed7_i2_2.jpg", "1mp3d_0057_region11/1216ded953194b41a63d635217e6bed7_i1_1.jpg", "1mp3d_0057_region11/1216ded953194b41a63d635217e6bed7_i2_1.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0057_region11", "origin_question": "When did you first discover \"the tall, narrow wooden cabinet with a white door\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the tall, narrow wooden cabinet with a white door\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "1mp3d_0057_region11/071418244e994a8284a535fc136de090_i1_1.jpg", "1mp3d_0057_region11/071418244e994a8284a535fc136de090_i2_1.jpg", "1mp3d_0057_region11/a9def3677c5d4397a2c12d2876b71254_i1_4.jpg", "1mp3d_0057_region11/a9def3677c5d4397a2c12d2876b71254_i2_4.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0057_region11", "origin_question": "In which turn were you farthest from \"the tall, narrow wooden cabinet with a white door\"? ", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the tall, narrow wooden cabinet with a white door\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "1mp3d_0057_region11/a9def3677c5d4397a2c12d2876b71254_i1_1.jpg", "1mp3d_0057_region11/a9def3677c5d4397a2c12d2876b71254_i2_1.jpg", "1mp3d_0057_region11/a9def3677c5d4397a2c12d2876b71254_i1_2.jpg", "1mp3d_0057_region11/a9def3677c5d4397a2c12d2876b71254_i2_2.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0057_region11", "origin_question": "Which direction is the tall, narrow wooden cabinet with a white door in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the tall, narrow wooden cabinet with a white door in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "1mp3d_0057_region11/071418244e994a8284a535fc136de090_i1_2.jpg", "1mp3d_0057_region11/071418244e994a8284a535fc136de090_i2_2.jpg", "1mp3d_0057_region11/f2ad72c11231453fa9a85ea2cb22be42_i1_5.jpg", "1mp3d_0057_region11/f2ad72c11231453fa9a85ea2cb22be42_i2_5.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0057_region11", "origin_question": "Which direction is the small hardcover book with a white cover in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "front-right", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the small hardcover book with a white cover in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0057_region2/5edf6294b685419f8bd7c76ebd0dac2f_i1_0.jpg", "1mp3d_0057_region2/5edf6294b685419f8bd7c76ebd0dac2f_i2_0.jpg", "1mp3d_0057_region2/5edf6294b685419f8bd7c76ebd0dac2f_i1_1.jpg", "1mp3d_0057_region2/5edf6294b685419f8bd7c76ebd0dac2f_i2_1.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0057_region2", "origin_question": "Remember, have you seen any pillow(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any pillow(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "1mp3d_0057_region2/757807d9031a4c17a4675d7b19e220a5_i1_0.jpg", "1mp3d_0057_region2/757807d9031a4c17a4675d7b19e220a5_i2_0.jpg", "1mp3d_0057_region2/3d8bc9f4323a47ef99f8210711ca1b60_i1_4.jpg", "1mp3d_0057_region2/3d8bc9f4323a47ef99f8210711ca1b60_i2_4.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0057_region2", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the small, rectangular light brown wooden stand\", Y: \"the large black fan with five dark blades\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to Y and farther away from X", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the small, rectangular light brown wooden stand\", Y: \"the large black fan with five dark blades\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "1mp3d_0057_region2/757807d9031a4c17a4675d7b19e220a5_i1_5.jpg", "1mp3d_0057_region2/757807d9031a4c17a4675d7b19e220a5_i2_5.jpg", "1mp3d_0057_region2/3d8bc9f4323a47ef99f8210711ca1b60_i1_5.jpg", "1mp3d_0057_region2/3d8bc9f4323a47ef99f8210711ca1b60_i2_5.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0057_region2", "origin_question": "At the end of which turn were both of \"the deep green plush couch made of cozy microfiber\" and \"the large black fan with five dark blades\" on your right side?", "option": [], "answer": "2", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the deep green plush couch made of cozy microfiber\" and \"the large black fan with five dark blades\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "1mp3d_0057_region2/3d8bc9f4323a47ef99f8210711ca1b60_i1_0.jpg", "1mp3d_0057_region2/3d8bc9f4323a47ef99f8210711ca1b60_i2_0.jpg", "1mp3d_0057_region2/86abf0dfba5c44d383ff3dfc701c6ee6_i1_4.jpg", "1mp3d_0057_region2/86abf0dfba5c44d383ff3dfc701c6ee6_i2_4.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0057_region2", "origin_question": "In which turn did you see both \"the large black fan with five dark blades\" and \"the wooden deck with a chair and railings\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the large black fan with five dark blades\" and \"the wooden deck with a chair and railings\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "1mp3d_0057_region2/98e10164fed14c2baaa107af2560555e_i1_3.jpg", "1mp3d_0057_region2/98e10164fed14c2baaa107af2560555e_i2_3.jpg", "1mp3d_0057_region2/c65a031835d844a1831c35017b790f30_i1_2.jpg", "1mp3d_0057_region2/c65a031835d844a1831c35017b790f30_i2_2.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0057_region2", "origin_question": "In which turn did you see both \"the large black fan with five dark blades\" and \"the wooden deck with a chair and railings\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the large black fan with five dark blades\" and \"the wooden deck with a chair and railings\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "1mp3d_0057_region2/98e10164fed14c2baaa107af2560555e_i1_2.jpg", "1mp3d_0057_region2/98e10164fed14c2baaa107af2560555e_i2_2.jpg", "1mp3d_0057_region2/86abf0dfba5c44d383ff3dfc701c6ee6_i1_2.jpg", "1mp3d_0057_region2/86abf0dfba5c44d383ff3dfc701c6ee6_i2_2.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0057_region2", "origin_question": "Which direction is the deep green plush couch made of cozy microfiber in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the deep green plush couch made of cozy microfiber in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "1mp3d_0057_region2/86abf0dfba5c44d383ff3dfc701c6ee6_i1_3.jpg", "1mp3d_0057_region2/86abf0dfba5c44d383ff3dfc701c6ee6_i2_3.jpg", "1mp3d_0057_region2/73c4028b050e4fc98125415d9cb5d194_i1_5.jpg", "1mp3d_0057_region2/73c4028b050e4fc98125415d9cb5d194_i2_5.jpg" ], "turn_id": 7, "scan_id": "1mp3d_0057_region2", "origin_question": "What will be the first-time appearance order of the following categories: bed, door, fan?", "option": [ "fan, door, bed", "bed, fan, door", "bed, door, fan", "fan, bed, door" ], "answer": "bed, fan, door", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: bed, door, fan? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['fan, door, bed', 'bed, fan, door', 'bed, door, fan', 'fan, bed, door'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "1mp3d_0057_region2/73c4028b050e4fc98125415d9cb5d194_i1_4.jpg", "1mp3d_0057_region2/73c4028b050e4fc98125415d9cb5d194_i2_4.jpg", "1mp3d_0057_region2/73c4028b050e4fc98125415d9cb5d194_i1_3.jpg", "1mp3d_0057_region2/73c4028b050e4fc98125415d9cb5d194_i2_3.jpg" ], "turn_id": 8, "scan_id": "1mp3d_0057_region2", "origin_question": "At the end of which turn were both of \"the small, rectangular light brown wooden stand\" and \"the large black fan with five dark blades\" on your left side?", "option": [], "answer": "4", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the small, rectangular light brown wooden stand\" and \"the large black fan with five dark blades\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0229_00/00000.jpg", "scene0229_00/00030.jpg", "scene0229_00/00060.jpg", "scene0229_00/00090.jpg" ], "turn_id": 1, "scan_id": "scene0229_00", "origin_question": "Remember, have you seen any vase(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any vase(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0229_00/00120.jpg", "scene0229_00/00150.jpg", "scene0229_00/00180.jpg", "scene0229_00/00210.jpg" ], "turn_id": 2, "scan_id": "scene0229_00", "origin_question": "In which turn were you farthest from \"the compact black monochrome laser printer\"? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the compact black monochrome laser printer\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0229_00/00240.jpg", "scene0229_00/00270.jpg", "scene0229_00/00300.jpg", "scene0229_00/00330.jpg" ], "turn_id": 3, "scan_id": "scene0229_00", "origin_question": "Which two objects are on the same side of you now? \"the wooden dining chair with a grid-like backrest\", \"the large rectangular light-colored board made of canvas or wood\", \"the rectangular wooden cabinet coated in light brown paint\". ", "option": [ "the wooden dining chair with a grid-like backrest, the large rectangular light-colored board made of canvas or wood", "the wooden dining chair with a grid-like backrest, the rectangular wooden cabinet coated in light brown paint", "the large rectangular light-colored board made of canvas or wood, the rectangular wooden cabinet coated in light brown paint" ], "answer": "the wooden dining chair with a grid-like backrest, the large rectangular light-colored board made of canvas or wood", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the wooden dining chair with a grid-like backrest\", \"the large rectangular light-colored board made of canvas or wood\", \"the rectangular wooden cabinet coated in light brown paint\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the wooden dining chair with a grid-like backrest, the large rectangular light-colored board made of canvas or wood', 'the wooden dining chair with a grid-like backrest, the rectangular wooden cabinet coated in light brown paint', 'the large rectangular light-colored board made of canvas or wood, the rectangular wooden cabinet coated in light brown paint'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0229_00/00360.jpg", "scene0229_00/00390.jpg", "scene0229_00/00420.jpg", "scene0229_00/00450.jpg" ], "turn_id": 4, "scan_id": "scene0229_00", "origin_question": "Which two objects are on the same side of you now? \"the shelf\", \"the vintage rectangular wooden table with metal handles\", \"the remote control\". ", "option": [ "the shelf, the vintage rectangular wooden table with metal handles", "the shelf, the remote control", "the vintage rectangular wooden table with metal handles, the remote control" ], "answer": "the vintage rectangular wooden table with metal handles, the remote control", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the shelf\", \"the vintage rectangular wooden table with metal handles\", \"the remote control\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the shelf, the vintage rectangular wooden table with metal handles', 'the shelf, the remote control', 'the vintage rectangular wooden table with metal handles, the remote control'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0229_00/00480.jpg", "scene0229_00/00510.jpg", "scene0229_00/00540.jpg", "scene0229_00/00570.jpg" ], "turn_id": 5, "scan_id": "scene0229_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0229_00/00600.jpg", "scene0229_00/00630.jpg", "scene0229_00/00660.jpg", "scene0229_00/00690.jpg" ], "turn_id": 6, "scan_id": "scene0229_00", "origin_question": "In which turn did you see both \"the large rectangular light-colored board made of canvas or wood\" and \"the compact black monochrome laser printer\" simultaneously? ", "option": [], "answer": "2", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the large rectangular light-colored board made of canvas or wood\" and \"the compact black monochrome laser printer\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0229_00/00720.jpg", "scene0229_00/00750.jpg", "scene0229_00/00780.jpg", "scene0229_00/00810.jpg" ], "turn_id": 7, "scan_id": "scene0229_00", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0653_01/00000.jpg", "scene0653_01/00040.jpg", "scene0653_01/00080.jpg", "scene0653_01/00120.jpg", "scene0653_01/00160.jpg" ], "turn_id": 1, "scan_id": "scene0653_01", "origin_question": "Remember, how many chair(s) have you seen so far? ", "option": [], "answer": "2", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many chair(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0653_01/00200.jpg", "scene0653_01/00240.jpg", "scene0653_01/00280.jpg", "scene0653_01/00320.jpg", "scene0653_01/00360.jpg" ], "turn_id": 2, "scan_id": "scene0653_01", "origin_question": "At the end of which turn were both of \"the rectangular black cabinet with two drawers\" and \"the medium-sized paperback book with a red cover\" on your right side?", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the rectangular black cabinet with two drawers\" and \"the medium-sized paperback book with a red cover\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0653_01/00400.jpg", "scene0653_01/00440.jpg", "scene0653_01/00480.jpg", "scene0653_01/00520.jpg", "scene0653_01/00560.jpg" ], "turn_id": 3, "scan_id": "scene0653_01", "origin_question": "Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0653_01/00600.jpg", "scene0653_01/00640.jpg", "scene0653_01/00680.jpg", "scene0653_01/00720.jpg", "scene0653_01/00760.jpg" ], "turn_id": 4, "scan_id": "scene0653_01", "origin_question": "In which turn did you see both \"the rectangular whiteboard with a black plastic frame\" and \"the large rectangular window with a wooden frame\" simultaneously? ", "option": [], "answer": "4", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the rectangular whiteboard with a black plastic frame\" and \"the large rectangular window with a wooden frame\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0653_01/00800.jpg", "scene0653_01/00840.jpg", "scene0653_01/00880.jpg", "scene0653_01/00920.jpg", "scene0653_01/00960.jpg" ], "turn_id": 5, "scan_id": "scene0653_01", "origin_question": "", "option": [], "answer": "", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0653_01/01000.jpg", "scene0653_01/01040.jpg", "scene0653_01/01080.jpg", "scene0653_01/01120.jpg", "scene0653_01/01160.jpg" ], "turn_id": 6, "scan_id": "scene0653_01", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the medium-sized grey backpack with multiple compartments\" from your position?", "option": [], "answer": "50.54", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the medium-sized grey backpack with multiple compartments\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0653_01/01200.jpg", "scene0653_01/01240.jpg", "scene0653_01/01280.jpg", "scene0653_01/01320.jpg", "scene0653_01/01360.jpg" ], "turn_id": 7, "scan_id": "scene0653_01", "origin_question": "Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the rectangular wooden door with a glass panel\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 4 turn, are you now closer or farther away from \"the rectangular wooden door with a glass panel\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "scene0653_01/01400.jpg", "scene0653_01/01440.jpg", "scene0653_01/01480.jpg", "scene0653_01/01520.jpg", "scene0653_01/01560.jpg" ], "turn_id": 8, "scan_id": "scene0653_01", "origin_question": "Among these three objects(\"the clear plastic bottle with a white pump dispenser\",\"the white ceramic cup with a red interior\",\"the rectangular wooden door with a glass panel\" ), which one is the closest to you?", "option": [ "the clear plastic bottle with a white pump dispenser", "the white ceramic cup with a red interior", "the rectangular wooden door with a glass panel" ], "answer": "the white ceramic cup with a red interior", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects(\"the clear plastic bottle with a white pump dispenser\",\"the white ceramic cup with a red interior\",\"the rectangular wooden door with a glass panel\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the clear plastic bottle with a white pump dispenser', 'the white ceramic cup with a red interior', 'the rectangular wooden door with a glass panel'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0653_01/01600.jpg", "scene0653_01/01640.jpg", "scene0653_01/01680.jpg", "scene0653_01/01720.jpg", "scene0653_01/01760.jpg" ], "turn_id": 9, "scan_id": "scene0653_01", "origin_question": "Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0653_01/01800.jpg", "scene0653_01/01840.jpg", "scene0653_01/01880.jpg", "scene0653_01/01920.jpg", "scene0653_01/01960.jpg" ], "turn_id": 10, "scan_id": "scene0653_01", "origin_question": "When did you first discover \"the standard medium-sized black keyboard made of plastic\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the standard medium-sized black keyboard made of plastic\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0545_02/00000.jpg", "scene0545_02/00030.jpg", "scene0545_02/00060.jpg", "scene0545_02/00090.jpg" ], "turn_id": 1, "scan_id": "scene0545_02", "origin_question": "Remember, have you seen any chair(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any chair(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0545_02/00120.jpg", "scene0545_02/00150.jpg", "scene0545_02/00180.jpg", "scene0545_02/00210.jpg" ], "turn_id": 2, "scan_id": "scene0545_02", "origin_question": "At the end of which turn were both of \"the light brown rectangular wooden desk with a drawer\" and \"the window\" on your right side?", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the light brown rectangular wooden desk with a drawer\" and \"the window\" on your right side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0545_02/00240.jpg", "scene0545_02/00270.jpg", "scene0545_02/00300.jpg", "scene0545_02/00330.jpg" ], "turn_id": 3, "scan_id": "scene0545_02", "origin_question": "Is the natural light brown wooden shelf with two levels to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the natural light brown wooden shelf with two levels to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement3)", "new_observations": [ "scene0545_02/00360.jpg", "scene0545_02/00390.jpg", "scene0545_02/00420.jpg", "scene0545_02/00450.jpg" ], "turn_id": 4, "scan_id": "scene0545_02", "origin_question": "Which two objects are on the same side of you now? \"the sleek black chair with a mesh backrest\", \"the light brown rectangular wooden desk with a drawer\", \"the vertical white semi-transparent curtain\". ", "option": [ "the sleek black chair with a mesh backrest, the light brown rectangular wooden desk with a drawer", "the sleek black chair with a mesh backrest, the vertical white semi-transparent curtain", "the light brown rectangular wooden desk with a drawer, the vertical white semi-transparent curtain" ], "answer": "the sleek black chair with a mesh backrest, the light brown rectangular wooden desk with a drawer", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Which two objects are on the same side of you now? \"the sleek black chair with a mesh backrest\", \"the light brown rectangular wooden desk with a drawer\", \"the vertical white semi-transparent curtain\". To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the sleek black chair with a mesh backrest, the light brown rectangular wooden desk with a drawer', 'the sleek black chair with a mesh backrest, the vertical white semi-transparent curtain', 'the light brown rectangular wooden desk with a drawer, the vertical white semi-transparent curtain'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0545_02/00480.jpg", "scene0545_02/00510.jpg", "scene0545_02/00540.jpg", "scene0545_02/00570.jpg" ], "turn_id": 5, "scan_id": "scene0545_02", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the light brown rectangular wooden desk with a drawer\", Y: \"the vertical white semi-transparent curtain\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "closer to both X and Y", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the light brown rectangular wooden desk with a drawer\", Y: \"the vertical white semi-transparent curtain\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0545_02/00600.jpg", "scene0545_02/00630.jpg", "scene0545_02/00660.jpg", "scene0545_02/00690.jpg" ], "turn_id": 6, "scan_id": "scene0545_02", "origin_question": "What will be the first-time appearance order of the following categories: chair, door, radiator?", "option": [ "radiator, chair, door", "radiator, door, chair", "chair, door, radiator", "chair, radiator, door" ], "answer": "chair, radiator, door", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: chair, door, radiator? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['radiator, chair, door', 'radiator, door, chair', 'chair, door, radiator', 'chair, radiator, door'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0545_02/00720.jpg", "scene0545_02/00750.jpg", "scene0545_02/00780.jpg", "scene0545_02/00810.jpg" ], "turn_id": 7, "scan_id": "scene0545_02", "origin_question": "Please recall \"the vertical white semi-transparent curtain\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.92", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the vertical white semi-transparent curtain\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0088_00/00000.jpg", "scene0088_00/00040.jpg", "scene0088_00/00080.jpg", "scene0088_00/00120.jpg" ], "turn_id": 1, "scan_id": "scene0088_00", "origin_question": "Remember, have you seen any kitchenware(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any kitchenware(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0088_00/00160.jpg", "scene0088_00/00200.jpg", "scene0088_00/00240.jpg", "scene0088_00/00280.jpg" ], "turn_id": 2, "scan_id": "scene0088_00", "origin_question": "Remember, have you seen any chair(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any chair(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0088_00/00320.jpg", "scene0088_00/00360.jpg", "scene0088_00/00400.jpg", "scene0088_00/00440.jpg" ], "turn_id": 3, "scan_id": "scene0088_00", "origin_question": "In which turn were you closest to \"the medium-sized light brown wooden table\"? ", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the medium-sized light brown wooden table\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0088_00/00480.jpg", "scene0088_00/00520.jpg", "scene0088_00/00560.jpg", "scene0088_00/00600.jpg" ], "turn_id": 4, "scan_id": "scene0088_00", "origin_question": "How far is your current position from where you were at the end of the turn 3? (in meters)", "option": [], "answer": "2.45", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 3? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0088_00/00640.jpg", "scene0088_00/00680.jpg", "scene0088_00/00720.jpg", "scene0088_00/00760.jpg" ], "turn_id": 5, "scan_id": "scene0088_00", "origin_question": "Which direction is the cylindrical black bin with a closed lid in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the cylindrical black bin with a closed lid in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0088_00/00800.jpg", "scene0088_00/00840.jpg", "scene0088_00/00880.jpg", "scene0088_00/00920.jpg" ], "turn_id": 6, "scan_id": "scene0088_00", "origin_question": "Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "93.9", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0558_01/00000.jpg", "scene0558_01/00040.jpg", "scene0558_01/00080.jpg", "scene0558_01/00120.jpg" ], "turn_id": 1, "scan_id": "scene0558_01", "origin_question": "Remember, have you seen any shelf(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any shelf(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0558_01/00160.jpg", "scene0558_01/00200.jpg", "scene0558_01/00240.jpg", "scene0558_01/00280.jpg" ], "turn_id": 2, "scan_id": "scene0558_01", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0558_01/00320.jpg", "scene0558_01/00360.jpg", "scene0558_01/00400.jpg", "scene0558_01/00440.jpg" ], "turn_id": 3, "scan_id": "scene0558_01", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "107.24", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0558_01/00480.jpg", "scene0558_01/00520.jpg", "scene0558_01/00560.jpg", "scene0558_01/00600.jpg" ], "turn_id": 4, "scan_id": "scene0558_01", "origin_question": "Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "21.75", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0558_01/00640.jpg", "scene0558_01/00680.jpg", "scene0558_01/00720.jpg", "scene0558_01/00760.jpg" ], "turn_id": 5, "scan_id": "scene0558_01", "origin_question": "How far is your current position from where you were at the end of the turn 2? (in meters)", "option": [], "answer": "1.59", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 2? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "scene0558_01/00800.jpg", "scene0558_01/00840.jpg", "scene0558_01/00880.jpg", "scene0558_01/00920.jpg" ], "turn_id": 6, "scan_id": "scene0558_01", "origin_question": "Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation?", "option": [], "answer": "107.59", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 2 as a reference, by how many degrees has your current orientation rotated clockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0558_01/00960.jpg", "scene0558_01/01000.jpg", "scene0558_01/01040.jpg", "scene0558_01/01090.jpg" ], "turn_id": 7, "scan_id": "scene0558_01", "origin_question": "Please recall \"the blue backpack made of sturdy material with straps\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.46", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the blue backpack made of sturdy material with straps\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "scene0558_01/01130.jpg", "scene0558_01/01170.jpg", "scene0558_01/01210.jpg", "scene0558_01/01250.jpg" ], "turn_id": 8, "scan_id": "scene0558_01", "origin_question": "How far is your current position from where you were at the end of the turn 6? (in meters)", "option": [], "answer": "1.29", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 6? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0558_01/01290.jpg", "scene0558_01/01330.jpg", "scene0558_01/01370.jpg", "scene0558_01/01410.jpg" ], "turn_id": 9, "scan_id": "scene0558_01", "origin_question": "Which direction is the blue backpack made of sturdy material with straps in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 9 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the blue backpack made of sturdy material with straps in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0648_00/00000.jpg", "scene0648_00/00040.jpg", "scene0648_00/00080.jpg", "scene0648_00/00120.jpg", "scene0648_00/00160.jpg" ], "turn_id": 1, "scan_id": "scene0648_00", "origin_question": "Remember, how many bed(s) have you seen so far? ", "option": [], "answer": "2", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, how many bed(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0648_00/00200.jpg", "scene0648_00/00240.jpg", "scene0648_00/00280.jpg", "scene0648_00/00320.jpg", "scene0648_00/00360.jpg" ], "turn_id": 2, "scan_id": "scene0648_00", "origin_question": "Please recall \"the modern adjustable lamp with a cylindrical white shade\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.29", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the modern adjustable lamp with a cylindrical white shade\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0648_00/00400.jpg", "scene0648_00/00440.jpg", "scene0648_00/00480.jpg", "scene0648_00/00520.jpg", "scene0648_00/00560.jpg" ], "turn_id": 3, "scan_id": "scene0648_00", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the rectangular white plastic telephone of standard size\" from your position?", "option": [], "answer": "164.8", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the rectangular white plastic telephone of standard size\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0648_00/00600.jpg", "scene0648_00/00640.jpg", "scene0648_00/00680.jpg", "scene0648_00/00720.jpg", "scene0648_00/00760.jpg" ], "turn_id": 4, "scan_id": "scene0648_00", "origin_question": "In which turn were you closest to \"the fan\"? ", "option": [], "answer": "1", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the fan\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0648_00/00800.jpg", "scene0648_00/00840.jpg", "scene0648_00/00880.jpg", "scene0648_00/00920.jpg", "scene0648_00/00960.jpg" ], "turn_id": 5, "scan_id": "scene0648_00", "origin_question": "When did you first discover \"the deep navy blue medium-sized sturdy backpack\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: When did you first discover \"the deep navy blue medium-sized sturdy backpack\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0648_00/01000.jpg", "scene0648_00/01040.jpg", "scene0648_00/01080.jpg", "scene0648_00/01120.jpg", "scene0648_00/01160.jpg" ], "turn_id": 6, "scan_id": "scene0648_00", "origin_question": "At the end of which turn were both of \"the silver rectangular aluminum laptop with a slim profile\" and \"the deep navy blue medium-sized sturdy backpack\" on your left side?", "option": [], "answer": "4", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the silver rectangular aluminum laptop with a slim profile\" and \"the deep navy blue medium-sized sturdy backpack\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0648_00/01200.jpg", "scene0648_00/01240.jpg", "scene0648_00/01280.jpg", "scene0648_00/01320.jpg", "scene0648_00/01360.jpg" ], "turn_id": 7, "scan_id": "scene0648_00", "origin_question": "Which direction is the silver rectangular aluminum laptop with a slim profile in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-right", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the silver rectangular aluminum laptop with a slim profile in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0648_00/01400.jpg", "scene0648_00/01440.jpg", "scene0648_00/01480.jpg", "scene0648_00/01520.jpg", "scene0648_00/01560.jpg" ], "turn_id": 8, "scan_id": "scene0648_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0648_00/01600.jpg", "scene0648_00/01640.jpg", "scene0648_00/01680.jpg", "scene0648_00/01720.jpg", "scene0648_00/01760.jpg" ], "turn_id": 9, "scan_id": "scene0648_00", "origin_question": "In which turn were you closest to \"the deep navy blue medium-sized sturdy backpack\"? ", "option": [], "answer": "9", "user_message": "For the 9 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the deep navy blue medium-sized sturdy backpack\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0648_00/01800.jpg", "scene0648_00/01840.jpg", "scene0648_00/01880.jpg", "scene0648_00/01920.jpg", "scene0648_00/01960.jpg" ], "turn_id": 10, "scan_id": "scene0648_00", "origin_question": "Please recall \"the black pair of shoes with a simple design\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "0.56", "user_message": "For the 10 turn, these are the 5 images you took in chronological order. The question for this turn is: Please recall \"the black pair of shoes with a simple design\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 10 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "1mp3d_0030_region11/2dae060398e8498d8e8f6f319bdf8b40_i1_0.jpg", "1mp3d_0030_region11/2dae060398e8498d8e8f6f319bdf8b40_i2_0.jpg", "1mp3d_0030_region11/4a3c031b6787433fb02da9d40a359c68_i1_1.jpg", "1mp3d_0030_region11/4a3c031b6787433fb02da9d40a359c68_i2_1.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0030_region11", "origin_question": "Remember, how many chair(s) have you seen so far? ", "option": [], "answer": "2", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many chair(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "1mp3d_0030_region11/bd01fe5cdc7f4d7580134b2a3ffd08a8_i1_0.jpg", "1mp3d_0030_region11/bd01fe5cdc7f4d7580134b2a3ffd08a8_i2_0.jpg", "1mp3d_0030_region11/4a3c031b6787433fb02da9d40a359c68_i1_2.jpg", "1mp3d_0030_region11/4a3c031b6787433fb02da9d40a359c68_i2_2.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0030_region11", "origin_question": "When did you first discover \"the clear glass cup shaped like a cylinder\"? (index of the turn)", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the clear glass cup shaped like a cylinder\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "1mp3d_0030_region11/2dae060398e8498d8e8f6f319bdf8b40_i1_1.jpg", "1mp3d_0030_region11/2dae060398e8498d8e8f6f319bdf8b40_i2_1.jpg", "1mp3d_0030_region11/cbeb8dcdebb34b1daf46a8aa1983139e_i1_0.jpg", "1mp3d_0030_region11/cbeb8dcdebb34b1daf46a8aa1983139e_i2_0.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0030_region11", "origin_question": "In which turn were you farthest from \"the medium-sized hardcover book with a black cover\"? ", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the medium-sized hardcover book with a black cover\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-position(Estimation)", "new_observations": [ "1mp3d_0030_region11/c69e398f10eb40938d19c2c736b0c084_i1_1.jpg", "1mp3d_0030_region11/c69e398f10eb40938d19c2c736b0c084_i2_1.jpg", "1mp3d_0030_region11/c69e398f10eb40938d19c2c736b0c084_i1_0.jpg", "1mp3d_0030_region11/c69e398f10eb40938d19c2c736b0c084_i2_0.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0030_region11", "origin_question": "How far is your current position from where you were at the end of the turn 3? (in meters)", "option": [], "answer": "1.26", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: How far is your current position from where you were at the end of the turn 3? (in meters) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Estimation)", "new_observations": [ "1mp3d_0030_region11/cbeb8dcdebb34b1daf46a8aa1983139e_i1_4.jpg", "1mp3d_0030_region11/cbeb8dcdebb34b1daf46a8aa1983139e_i2_4.jpg", "1mp3d_0030_region11/cbeb8dcdebb34b1daf46a8aa1983139e_i1_5.jpg", "1mp3d_0030_region11/cbeb8dcdebb34b1daf46a8aa1983139e_i2_5.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0030_region11", "origin_question": "Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation?", "option": [], "answer": "60.84", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, by how many degrees has your current orientation rotated counterclockwise relative to that previous orientation? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "1mp3d_0030_region11/ebb3e7ba8ccb4ff7b4fecb33cddfdd03_i1_3.jpg", "1mp3d_0030_region11/ebb3e7ba8ccb4ff7b4fecb33cddfdd03_i2_3.jpg", "1mp3d_0030_region11/ebb3e7ba8ccb4ff7b4fecb33cddfdd03_i2_2.jpg", "1mp3d_0030_region11/ebb3e7ba8ccb4ff7b4fecb33cddfdd03_i1_1.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0030_region11", "origin_question": "Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the medium-sized hardcover book with a black cover\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 1 turn, are you now closer or farther away from \"the medium-sized hardcover book with a black cover\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0406_02/00000.jpg", "scene0406_02/00040.jpg", "scene0406_02/00080.jpg", "scene0406_02/00120.jpg" ], "turn_id": 1, "scan_id": "scene0406_02", "origin_question": "Remember, have you seen any curtain(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any curtain(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0406_02/00160.jpg", "scene0406_02/00200.jpg", "scene0406_02/00240.jpg", "scene0406_02/00280.jpg" ], "turn_id": 2, "scan_id": "scene0406_02", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the square-framed picture with a blue-green background\"; \"the standard rectangular white bathtub made of porcelain\"; \"the beige curtain with a wavy pattern\" ", "option": [ "the square-framed picture with a blue-green background", "the standard rectangular white bathtub made of porcelain", "the beige curtain with a wavy pattern" ], "answer": "the square-framed picture with a blue-green background", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the square-framed picture with a blue-green background\"; \"the standard rectangular white bathtub made of porcelain\"; \"the beige curtain with a wavy pattern\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the square-framed picture with a blue-green background', 'the standard rectangular white bathtub made of porcelain', 'the beige curtain with a wavy pattern'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0406_02/00320.jpg", "scene0406_02/00360.jpg", "scene0406_02/00400.jpg", "scene0406_02/00440.jpg" ], "turn_id": 3, "scan_id": "scene0406_02", "origin_question": "At the end of which turn were both of \"the white oval-shaped porcelain toilet\" and \"the medium dark silver cylindrical bin\" on your left side?", "option": [], "answer": "3", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the white oval-shaped porcelain toilet\" and \"the medium dark silver cylindrical bin\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0406_02/00480.jpg", "scene0406_02/00520.jpg", "scene0406_02/00560.jpg", "scene0406_02/00600.jpg" ], "turn_id": 4, "scan_id": "scene0406_02", "origin_question": "At the end of which turn were both of \"the standard rectangular white bathtub made of porcelain\" and \"the white oval-shaped porcelain toilet\" on your left side?", "option": [], "answer": "4", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the standard rectangular white bathtub made of porcelain\" and \"the white oval-shaped porcelain toilet\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0406_02/00640.jpg", "scene0406_02/00680.jpg", "scene0406_02/00720.jpg", "scene0406_02/00760.jpg" ], "turn_id": 5, "scan_id": "scene0406_02", "origin_question": "In which turn were you closest to \"the white oval-shaped porcelain toilet\"? ", "option": [], "answer": "5", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the white oval-shaped porcelain toilet\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0406_02/00800.jpg", "scene0406_02/00840.jpg", "scene0406_02/00880.jpg", "scene0406_02/00920.jpg" ], "turn_id": 6, "scan_id": "scene0406_02", "origin_question": "Is the medium dark silver cylindrical bin to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the medium dark silver cylindrical bin to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "scene0406_02/00960.jpg", "scene0406_02/01000.jpg", "scene0406_02/01040.jpg", "scene0406_02/01080.jpg" ], "turn_id": 7, "scan_id": "scene0406_02", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the standard rectangular white bathtub made of porcelain\" from your position?", "option": [], "answer": "102.63", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the standard rectangular white bathtub made of porcelain\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0193_00/00000.jpg", "scene0193_00/00030.jpg", "scene0193_00/00060.jpg", "scene0193_00/00090.jpg" ], "turn_id": 1, "scan_id": "scene0193_00", "origin_question": "Remember, have you seen any bed(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any bed(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0193_00/00120.jpg", "scene0193_00/00150.jpg", "scene0193_00/00180.jpg", "scene0193_00/00210.jpg" ], "turn_id": 2, "scan_id": "scene0193_00", "origin_question": "In which turn did you see both \"the curtain\" and \"the simple rectangular bed with a purple mattress\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the curtain\" and \"the simple rectangular bed with a purple mattress\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0193_00/00240.jpg", "scene0193_00/00280.jpg", "scene0193_00/00310.jpg", "scene0193_00/00340.jpg" ], "turn_id": 3, "scan_id": "scene0193_00", "origin_question": "Please recall \"the sleek rectangular beige plastic bin\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.43", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the sleek rectangular beige plastic bin\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0193_00/00390.jpg", "scene0193_00/00420.jpg", "scene0193_00/00450.jpg", "scene0193_00/00480.jpg" ], "turn_id": 4, "scan_id": "scene0193_00", "origin_question": "In which turn were you closest to \"the simple rectangular bed with a purple mattress\"? ", "option": [], "answer": "3", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the simple rectangular bed with a purple mattress\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "scene0193_00/00510.jpg", "scene0193_00/00540.jpg", "scene0193_00/00570.jpg", "scene0193_00/00600.jpg" ], "turn_id": 5, "scan_id": "scene0193_00", "origin_question": "At the end of which turn were both of \"the sleek black chair with a mesh backrest\" and \"the simple rectangular bed with a purple mattress\" on your left side?", "option": [], "answer": "3", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the sleek black chair with a mesh backrest\" and \"the simple rectangular bed with a purple mattress\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 5 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0427_00/00000.jpg", "scene0427_00/00040.jpg", "scene0427_00/00080.jpg", "scene0427_00/00120.jpg" ], "turn_id": 1, "scan_id": "scene0427_00", "origin_question": "Remember, have you seen any table(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any table(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "scene0427_00/00160.jpg", "scene0427_00/00200.jpg", "scene0427_00/00240.jpg", "scene0427_00/00280.jpg" ], "turn_id": 2, "scan_id": "scene0427_00", "origin_question": "In which turn were you closest to \"the round black table with wooden edges\"? ", "option": [], "answer": "2", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you closest to \"the round black table with wooden edges\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0427_00/00320.jpg", "scene0427_00/00360.jpg", "scene0427_00/00400.jpg", "scene0427_00/00440.jpg" ], "turn_id": 3, "scan_id": "scene0427_00", "origin_question": "What will be the first-time appearance order of the following categories: blackboard, door, table?", "option": [ "blackboard, table, door", "blackboard, door, table", "table, door, blackboard", "table, blackboard, door" ], "answer": "blackboard, table, door", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: blackboard, door, table? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['blackboard, table, door', 'blackboard, door, table', 'table, door, blackboard', 'table, blackboard, door'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "scene0427_00/00480.jpg", "scene0427_00/00520.jpg", "scene0427_00/00560.jpg", "scene0427_00/00600.jpg" ], "turn_id": 4, "scan_id": "scene0427_00", "origin_question": "When did you first discover \"the rectangular wooden door with a smooth surface\"? (index of the turn)", "option": [], "answer": "3", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the rectangular wooden door with a smooth surface\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Estimation)", "new_observations": [ "scene0427_00/00640.jpg", "scene0427_00/00680.jpg", "scene0427_00/00720.jpg", "scene0427_00/00760.jpg" ], "turn_id": 5, "scan_id": "scene0427_00", "origin_question": "Please recall \"the bin\", what is the horizontal distance between you and this object now (in meter)?", "option": [], "answer": "1.09", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Please recall \"the bin\", what is the horizontal distance between you and this object now (in meter)? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0427_00/00800.jpg", "scene0427_00/00840.jpg", "scene0427_00/00880.jpg", "scene0427_00/00920.jpg" ], "turn_id": 6, "scan_id": "scene0427_00", "origin_question": "Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the round black table with wooden edges\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 2 turn, are you now closer or farther away from \"the round black table with wooden edges\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0427_00/00960.jpg", "scene0427_00/01000.jpg", "scene0427_00/01040.jpg", "scene0427_00/01080.jpg" ], "turn_id": 7, "scan_id": "scene0427_00", "origin_question": "Compared to your position at the end of the 6 turn, are you now closer or farther away from \"the round black table with wooden edges\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 6 turn, are you now closer or farther away from \"the round black table with wooden edges\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0427_00/01120.jpg", "scene0427_00/01160.jpg", "scene0427_00/01200.jpg", "scene0427_00/01240.jpg" ], "turn_id": 8, "scan_id": "scene0427_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0365_01/00000.jpg", "scene0365_01/00040.jpg", "scene0365_01/00080.jpg", "scene0365_01/00120.jpg", "scene0365_01/00160.jpg" ], "turn_id": 1, "scan_id": "scene0365_01", "origin_question": "Remember, have you seen any shovel(s) so far? ", "option": [ "Yes", "No" ], "answer": "No", "user_message": "For the 1 turn, these are the 5 images you took in chronological order. The question for this turn is: Remember, have you seen any shovel(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "scene0365_01/00200.jpg", "scene0365_01/00240.jpg", "scene0365_01/00280.jpg", "scene0365_01/00320.jpg", "scene0365_01/00360.jpg" ], "turn_id": 2, "scan_id": "scene0365_01", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the medium-sized rectangular black bin\"; \"the door\"; \"the vertical black plastic tissue dispenser with a blue base\" ", "option": [ "the medium-sized rectangular black bin", "the door", "the vertical black plastic tissue dispenser with a blue base" ], "answer": "the door", "user_message": "For the 2 turn, these are the 5 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the medium-sized rectangular black bin\"; \"the door\"; \"the vertical black plastic tissue dispenser with a blue base\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the medium-sized rectangular black bin', 'the door', 'the vertical black plastic tissue dispenser with a blue base'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "scene0365_01/00400.jpg", "scene0365_01/00440.jpg", "scene0365_01/00480.jpg", "scene0365_01/00520.jpg", "scene0365_01/00560.jpg" ], "turn_id": 3, "scan_id": "scene0365_01", "origin_question": "In which turn did you see both \"the white rectangular wall-mounted sink with rounded corners\" and \"the vertical black plastic tissue dispenser with a blue base\" simultaneously? ", "option": [], "answer": "1", "user_message": "For the 3 turn, these are the 5 images you took in chronological order. The question for this turn is: In which turn did you see both \"the white rectangular wall-mounted sink with rounded corners\" and \"the vertical black plastic tissue dispenser with a blue base\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0365_01/00600.jpg", "scene0365_01/00640.jpg", "scene0365_01/00680.jpg", "scene0365_01/00720.jpg", "scene0365_01/00760.jpg" ], "turn_id": 4, "scan_id": "scene0365_01", "origin_question": "Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 4 turn, these are the 5 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 3 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "scene0365_01/00800.jpg", "scene0365_01/00840.jpg", "scene0365_01/00880.jpg", "scene0365_01/00920.jpg", "scene0365_01/00960.jpg" ], "turn_id": 5, "scan_id": "scene0365_01", "origin_question": "Which direction is the plant in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 5 turn, these are the 5 images you took in chronological order. The question for this turn is: Which direction is the plant in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0365_01/01000.jpg", "scene0365_01/01040.jpg", "scene0365_01/01080.jpg", "scene0365_01/01120.jpg", "scene0365_01/01160.jpg" ], "turn_id": 6, "scan_id": "scene0365_01", "origin_question": "Compared to your position at the end of the 5 turn, are you now closer or farther away from \"the vertical black plastic tissue dispenser with a blue base\"?", "option": [ "Closer", "Farther" ], "answer": "Closer", "user_message": "For the 6 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 5 turn, are you now closer or farther away from \"the vertical black plastic tissue dispenser with a blue base\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0365_01/01200.jpg", "scene0365_01/01240.jpg", "scene0365_01/01280.jpg", "scene0365_01/01320.jpg", "scene0365_01/01360.jpg" ], "turn_id": 7, "scan_id": "scene0365_01", "origin_question": "What will be the first-time appearance order of the following categories: door, sink, toilet?", "option": [ "sink, door, toilet", "toilet, sink, door", "door, toilet, sink", "door, sink, toilet" ], "answer": "sink, door, toilet", "user_message": "For the 7 turn, these are the 5 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: door, sink, toilet? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['sink, door, toilet', 'toilet, sink, door', 'door, toilet, sink', 'door, sink, toilet'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement1)", "new_observations": [ "scene0365_01/01400.jpg", "scene0365_01/01440.jpg", "scene0365_01/01480.jpg", "scene0365_01/01520.jpg", "scene0365_01/01560.jpg" ], "turn_id": 8, "scan_id": "scene0365_01", "origin_question": "Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the white rectangular wall-mounted sink with rounded corners\"?", "option": [ "Closer", "Farther" ], "answer": "Farther", "user_message": "For the 8 turn, these are the 5 images you took in chronological order. The question for this turn is: Compared to your position at the end of the 3 turn, are you now closer or farther away from \"the white rectangular wall-mounted sink with rounded corners\"? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Closer', 'Farther'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 5 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0074_region6/d512f07e48aa4fe9a559a1a708ab3c02_i1_4.jpg", "1mp3d_0074_region6/d512f07e48aa4fe9a559a1a708ab3c02_i2_4.jpg", "1mp3d_0074_region6/ff43ac7b797541b5a89ce2373de329a4_i1_5.jpg", "1mp3d_0074_region6/ff43ac7b797541b5a89ce2373de329a4_i2_5.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0074_region6", "origin_question": "Remember, have you seen any couch(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any couch(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc1)", "new_observations": [ "1mp3d_0074_region6/d512f07e48aa4fe9a559a1a708ab3c02_i1_5.jpg", "1mp3d_0074_region6/d512f07e48aa4fe9a559a1a708ab3c02_i2_5.jpg", "1mp3d_0074_region6/ff43ac7b797541b5a89ce2373de329a4_i1_0.jpg", "1mp3d_0074_region6/ff43ac7b797541b5a89ce2373de329a4_i2_0.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0074_region6", "origin_question": "When did you first discover \"the rectangular wooden stand with dark wood grain\"? (index of the turn)", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: When did you first discover \"the rectangular wooden stand with dark wood grain\"? (index of the turn) To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "1mp3d_0074_region6/a0e680c2599e473c9d64183b65dac6b3_i1_3.jpg", "1mp3d_0074_region6/a0e680c2599e473c9d64183b65dac6b3_i2_3.jpg", "1mp3d_0074_region6/a0e680c2599e473c9d64183b65dac6b3_i1_2.jpg", "1mp3d_0074_region6/a0e680c2599e473c9d64183b65dac6b3_i2_2.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0074_region6", "origin_question": "What will be the first-time appearance order of the following categories: chair, couch, statue?", "option": [ "chair, statue, couch", "statue, couch, chair", "couch, statue, chair", "chair, couch, statue" ], "answer": "statue, couch, chair", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: chair, couch, statue? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['chair, statue, couch', 'statue, couch, chair', 'couch, statue, chair', 'chair, couch, statue'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-diversity(Judgement)", "new_observations": [ "1mp3d_0074_region6/6be9623980e245ce80ced0bdb70db93c_i1_3.jpg", "1mp3d_0074_region6/6be9623980e245ce80ced0bdb70db93c_i2_3.jpg", "1mp3d_0074_region6/856a21dd20ce41c3ac1c2a160cec9a0f_i1_4.jpg", "1mp3d_0074_region6/856a21dd20ce41c3ac1c2a160cec9a0f_i2_4.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0074_region6", "origin_question": "Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the bowl filled with yellow fruit\"; \"the washbasin\"; \"the rectangular wooden stand with dark wood grain\" ", "option": [ "the bowl filled with yellow fruit", "the washbasin", "the rectangular wooden stand with dark wood grain" ], "answer": "the bowl filled with yellow fruit", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects, which one was newly discovered in this turn(had not appeared before)? \"the bowl filled with yellow fruit\"; \"the washbasin\"; \"the rectangular wooden stand with dark wood grain\" To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the bowl filled with yellow fruit', 'the washbasin', 'the rectangular wooden stand with dark wood grain'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "1mp3d_0074_region6/6be9623980e245ce80ced0bdb70db93c_i1_4.jpg", "1mp3d_0074_region6/6be9623980e245ce80ced0bdb70db93c_i2_4.jpg", "1mp3d_0074_region6/16bcabb86c9546f89e60b1caf0fc0a13_i1_3.jpg", "1mp3d_0074_region6/16bcabb86c9546f89e60b1caf0fc0a13_i2_3.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0074_region6", "origin_question": "What will be the first-time appearance order of the following categories: chair, curtain, fruit?", "option": [ "fruit, curtain, chair", "chair, curtain, fruit", "chair, fruit, curtain", "curtain, chair, fruit" ], "answer": "chair, fruit, curtain", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: chair, curtain, fruit? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['fruit, curtain, chair', 'chair, curtain, fruit', 'chair, fruit, curtain', 'curtain, chair, fruit'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "1mp3d_0074_region6/219e774e6cf942a5af295147d7bf945e_i1_0.jpg", "1mp3d_0074_region6/219e774e6cf942a5af295147d7bf945e_i2_0.jpg", "1mp3d_0074_region6/2dd65c204fd64835b8cd417df5aa4149_i1_0.jpg", "1mp3d_0074_region6/2dd65c204fd64835b8cd417df5aa4149_i2_0.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0074_region6", "origin_question": "", "option": [], "answer": "", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Temporal-loc)", "new_observations": [ "1mp3d_0074_region6/a34ec23dd5284d8fbca553acefd49c03_i1_3.jpg", "1mp3d_0074_region6/a34ec23dd5284d8fbca553acefd49c03_i2_3.jpg", "1mp3d_0074_region6/68f5cd251e4c457894ea04e8184286d0_i1_2.jpg", "1mp3d_0074_region6/68f5cd251e4c457894ea04e8184286d0_i2_2.jpg" ], "turn_id": 7, "scan_id": "1mp3d_0074_region6", "origin_question": "At the end of which turn were both of \"the medium-sized wooden chair with a curved backrest\" and \"the bowl filled with yellow fruit\" on your left side?", "option": [], "answer": "7", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: At the end of which turn were both of \"the medium-sized wooden chair with a curved backrest\" and \"the bowl filled with yellow fruit\" on your left side? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "1mp3d_0074_region6/68f5cd251e4c457894ea04e8184286d0_i1_1.jpg", "1mp3d_0074_region6/68f5cd251e4c457894ea04e8184286d0_i2_1.jpg", "1mp3d_0074_region6/68f5cd251e4c457894ea04e8184286d0_i1_0.jpg", "1mp3d_0074_region6/68f5cd251e4c457894ea04e8184286d0_i2_0.jpg" ], "turn_id": 8, "scan_id": "1mp3d_0074_region6", "origin_question": "Among these three objects(\"the bowl filled with yellow fruit\",\"the long, soft gray curtain made of lightweight material\",\"the small silver traditional telephone made of plastic\" ), which one is the closest to you?", "option": [ "the bowl filled with yellow fruit", "the long, soft gray curtain made of lightweight material", "the small silver traditional telephone made of plastic" ], "answer": "the small silver traditional telephone made of plastic", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects(\"the bowl filled with yellow fruit\",\"the long, soft gray curtain made of lightweight material\",\"the small silver traditional telephone made of plastic\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the bowl filled with yellow fruit', 'the long, soft gray curtain made of lightweight material', 'the small silver traditional telephone made of plastic'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Estimation)", "new_observations": [ "1mp3d_0074_region6/a34ec23dd5284d8fbca553acefd49c03_i1_4.jpg", "1mp3d_0074_region6/a34ec23dd5284d8fbca553acefd49c03_i2_4.jpg", "1mp3d_0074_region6/2dd65c204fd64835b8cd417df5aa4149_i1_5.jpg", "1mp3d_0074_region6/2dd65c204fd64835b8cd417df5aa4149_i2_5.jpg" ], "turn_id": 9, "scan_id": "1mp3d_0074_region6", "origin_question": "Based on your current orientation, in what clockwise direction (in degrees) is \"the long, soft gray curtain made of lightweight material\" from your position?", "option": [], "answer": "33.46", "user_message": "For the 9 turn, these are the 4 images you took in chronological order. The question for this turn is: Based on your current orientation, in what clockwise direction (in degrees) is \"the long, soft gray curtain made of lightweight material\" from your position? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 9 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0614_02/00000.jpg", "scene0614_02/00040.jpg", "scene0614_02/00080.jpg", "scene0614_02/00120.jpg" ], "turn_id": 1, "scan_id": "scene0614_02", "origin_question": "Remember, how many chair(s) have you seen so far? ", "option": [], "answer": "3", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many chair(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement3)", "new_observations": [ "scene0614_02/00160.jpg", "scene0614_02/00200.jpg", "scene0614_02/00240.jpg", "scene0614_02/00280.jpg" ], "turn_id": 2, "scan_id": "scene0614_02", "origin_question": "By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the clothes\", Y: \"the radiator\" ?", "option": [ "closer to both X and Y", "farther away from both X and Y", "closer to X and farther away from Y", "closer to Y and farther away from X" ], "answer": "farther away from both X and Y", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: By the end of the 1 turn, were you getting closer to or farther away from these two objects: X: \"the clothes\", Y: \"the radiator\" ? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['closer to both X and Y', 'farther away from both X and Y', 'closer to X and farther away from Y', 'closer to Y and farther away from X'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "scene0614_02/00320.jpg", "scene0614_02/00360.jpg", "scene0614_02/00400.jpg", "scene0614_02/00440.jpg" ], "turn_id": 3, "scan_id": "scene0614_02", "origin_question": "Is the radiator to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the radiator to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0614_02/00480.jpg", "scene0614_02/00520.jpg", "scene0614_02/00560.jpg", "scene0614_02/00600.jpg" ], "turn_id": 4, "scan_id": "scene0614_02", "origin_question": "What will be the first-time appearance order of the following categories: blackboard, door, keyboard?", "option": [ "door, keyboard, blackboard", "keyboard, blackboard, door", "door, blackboard, keyboard", "blackboard, door, keyboard" ], "answer": "keyboard, blackboard, door", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: blackboard, door, keyboard? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['door, keyboard, blackboard', 'keyboard, blackboard, door', 'door, blackboard, keyboard', 'blackboard, door, keyboard'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0614_02/00640.jpg", "scene0614_02/00680.jpg", "scene0614_02/00720.jpg", "scene0614_02/00760.jpg" ], "turn_id": 5, "scan_id": "scene0614_02", "origin_question": "Remember, how many shelf(s) have you seen so far? ", "option": [], "answer": "2", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many shelf(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0614_02/00800.jpg", "scene0614_02/00840.jpg", "scene0614_02/00880.jpg", "scene0614_02/00920.jpg" ], "turn_id": 6, "scan_id": "scene0614_02", "origin_question": "Using your orientation at the end of turn 5 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Counterclockwise", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 5 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 6 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-quantity(Counting)", "new_observations": [ "scene0690_00/00000.jpg", "scene0690_00/00030.jpg", "scene0690_00/00060.jpg", "scene0690_00/00090.jpg" ], "turn_id": 1, "scan_id": "scene0690_00", "origin_question": "Remember, how many chair(s) have you seen so far? ", "option": [], "answer": "3", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, how many chair(s) have you seen so far? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "None", "new_observations": [ "scene0690_00/00120.jpg", "scene0690_00/00150.jpg", "scene0690_00/00180.jpg", "scene0690_00/00210.jpg" ], "turn_id": 2, "scan_id": "scene0690_00", "origin_question": "", "option": [], "answer": "", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. There is no question for this turn.", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_state-orientation(Judgement)", "new_observations": [ "scene0690_00/00240.jpg", "scene0690_00/00270.jpg", "scene0690_00/00300.jpg", "scene0690_00/00330.jpg" ], "turn_id": 3, "scan_id": "scene0690_00", "origin_question": "Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation?", "option": [ "Clockwise", "Counterclockwise" ], "answer": "Clockwise", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Using your orientation at the end of turn 1 as a reference, has your current orientation rotated clockwise or counterclockwise by a certain angle (<180) relative to that orientation? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Clockwise', 'Counterclockwise'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0690_00/00360.jpg", "scene0690_00/00390.jpg", "scene0690_00/00420.jpg", "scene0690_00/00450.jpg" ], "turn_id": 4, "scan_id": "scene0690_00", "origin_question": "What will be the first-time appearance order of the following categories: couch, paper, table?", "option": [ "table, couch, paper", "table, paper, couch", "paper, table, couch", "couch, table, paper" ], "answer": "couch, table, paper", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: couch, paper, table? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['table, couch, paper', 'table, paper, couch', 'paper, table, couch', 'couch, table, paper'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0690_00/00480.jpg", "scene0690_00/00510.jpg", "scene0690_00/00540.jpg", "scene0690_00/00570.jpg" ], "turn_id": 5, "scan_id": "scene0690_00", "origin_question": "What will be the first-time appearance order of the following categories: chair, light, shelf?", "option": [ "chair, light, shelf", "light, shelf, chair", "shelf, chair, light", "shelf, light, chair" ], "answer": "chair, light, shelf", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: chair, light, shelf? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['chair, light, shelf', 'light, shelf, chair', 'shelf, chair, light', 'shelf, light, chair'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "scene0690_00/00600.jpg", "scene0690_00/00630.jpg", "scene0690_00/00660.jpg", "scene0690_00/00690.jpg" ], "turn_id": 6, "scan_id": "scene0690_00", "origin_question": "Remember, have you seen any couch(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any couch(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "scene0690_00/00720.jpg", "scene0690_00/00750.jpg", "scene0690_00/00780.jpg", "scene0690_00/00810.jpg" ], "turn_id": 7, "scan_id": "scene0690_00", "origin_question": "What will be the first-time appearance order of the following categories: couch, paper, table?", "option": [ "couch, table, paper", "paper, couch, table", "paper, table, couch", "table, couch, paper" ], "answer": "table, couch, paper", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: couch, paper, table? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['couch, table, paper', 'paper, couch, table', 'paper, table, couch', 'table, couch, paper'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 7 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Judgement)", "new_observations": [ "1mp3d_0040_region16/605c759ce3ae47259eea6154501b62ba_i1_3.jpg", "1mp3d_0040_region16/605c759ce3ae47259eea6154501b62ba_i2_3.jpg", "1mp3d_0040_region16/605c759ce3ae47259eea6154501b62ba_i1_4.jpg", "1mp3d_0040_region16/605c759ce3ae47259eea6154501b62ba_i2_4.jpg" ], "turn_id": 1, "scan_id": "1mp3d_0040_region16", "origin_question": "Remember, have you seen any cap(s) so far? ", "option": [ "Yes", "No" ], "answer": "Yes", "user_message": "For the 1 turn, these are the 4 images you took in chronological order. The question for this turn is: Remember, have you seen any cap(s) so far? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['Yes', 'No'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Temporal-loc)", "new_observations": [ "1mp3d_0040_region16/73843243a62a46c4a7451799e909c9f7_i1_3.jpg", "1mp3d_0040_region16/73843243a62a46c4a7451799e909c9f7_i2_3.jpg", "1mp3d_0040_region16/73843243a62a46c4a7451799e909c9f7_i1_4.jpg", "1mp3d_0040_region16/73843243a62a46c4a7451799e909c9f7_i2_4.jpg" ], "turn_id": 2, "scan_id": "1mp3d_0040_region16", "origin_question": "In which turn were you farthest from \"the rectangular black medium-sized speaker with a glossy surface\"? ", "option": [], "answer": "1", "user_message": "For the 2 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn were you farthest from \"the rectangular black medium-sized speaker with a glossy surface\"? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement2)", "new_observations": [ "1mp3d_0040_region16/b62f605b7c14452bb33bc503fef6a9c6_i1_1.jpg", "1mp3d_0040_region16/b62f605b7c14452bb33bc503fef6a9c6_i2_1.jpg", "1mp3d_0040_region16/1e5d5e45bf0844ce8032a9aad112260c_i1_5.jpg", "1mp3d_0040_region16/1e5d5e45bf0844ce8032a9aad112260c_i2_5.jpg" ], "turn_id": 3, "scan_id": "1mp3d_0040_region16", "origin_question": "Which direction is the stylish rectangular black radio made of smooth material in relative to you now: front left, front right, rear left, or rear right?", "option": [ "front-right", "rear-right", "rear-left", "front-left" ], "answer": "rear-left", "user_message": "For the 3 turn, these are the 4 images you took in chronological order. The question for this turn is: Which direction is the stylish rectangular black radio made of smooth material in relative to you now: front left, front right, rear left, or rear right? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['front-right', 'rear-right', 'rear-left', 'front-left'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-direction(Judgement1)", "new_observations": [ "1mp3d_0040_region16/b62f605b7c14452bb33bc503fef6a9c6_i1_2.jpg", "1mp3d_0040_region16/b62f605b7c14452bb33bc503fef6a9c6_i2_2.jpg", "1mp3d_0040_region16/1e5d5e45bf0844ce8032a9aad112260c_i1_1.jpg", "1mp3d_0040_region16/1e5d5e45bf0844ce8032a9aad112260c_i2_1.jpg" ], "turn_id": 4, "scan_id": "1mp3d_0040_region16", "origin_question": "Is the flat-screen TV with a silver casing to your left/right now?", "option": [ "left", "right" ], "answer": "right", "user_message": "For the 4 turn, these are the 4 images you took in chronological order. The question for this turn is: Is the flat-screen TV with a silver casing to your left/right now? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['left', 'right'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "1mp3d_0040_region16/1e5d5e45bf0844ce8032a9aad112260c_i1_0.jpg", "1mp3d_0040_region16/1e5d5e45bf0844ce8032a9aad112260c_i2_0.jpg", "1mp3d_0040_region16/f469638902b44bd78c32856f03e89cb7_i1_0.jpg", "1mp3d_0040_region16/f469638902b44bd78c32856f03e89cb7_i2_0.jpg" ], "turn_id": 5, "scan_id": "1mp3d_0040_region16", "origin_question": "What will be the first-time appearance order of the following categories: hat, light, tv?", "option": [ "tv, light, hat", "hat, tv, light", "hat, light, tv", "tv, hat, light" ], "answer": "tv, hat, light", "user_message": "For the 5 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: hat, light, tv? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['tv, light, hat', 'hat, tv, light', 'hat, light, tv', 'tv, hat, light'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-existence(Temporal-loc2)", "new_observations": [ "1mp3d_0040_region16/f469638902b44bd78c32856f03e89cb7_i1_5.jpg", "1mp3d_0040_region16/f469638902b44bd78c32856f03e89cb7_i2_5.jpg", "1mp3d_0040_region16/f469638902b44bd78c32856f03e89cb7_i1_4.jpg", "1mp3d_0040_region16/f469638902b44bd78c32856f03e89cb7_i2_4.jpg" ], "turn_id": 6, "scan_id": "1mp3d_0040_region16", "origin_question": "In which turn did you see both \"the stylish rectangular black radio made of smooth material\" and \"the rectangular wooden bed with a brown and beige bedspread\" simultaneously? ", "option": [], "answer": "6", "user_message": "For the 6 turn, these are the 4 images you took in chronological order. The question for this turn is: In which turn did you see both \"the stylish rectangular black radio made of smooth material\" and \"the rectangular wooden bed with a brown and beige bedspread\" simultaneously? To answer this question, you need to combine image/coordinate information from past turns. Please provide a numerical value as the result based on history information.The information I provided is sufficient for you to infer the value; do not refuse to answer! Give me your answer and your reason in a JSON format({answer:float/int, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_object_spatial-distance(Judgement2)", "new_observations": [ "1mp3d_0040_region16/f469638902b44bd78c32856f03e89cb7_i1_3.jpg", "1mp3d_0040_region16/f469638902b44bd78c32856f03e89cb7_i2_3.jpg", "1mp3d_0040_region16/93528274a86b47e6a3b0a6f398090741_i1_2.jpg", "1mp3d_0040_region16/93528274a86b47e6a3b0a6f398090741_i2_2.jpg" ], "turn_id": 7, "scan_id": "1mp3d_0040_region16", "origin_question": "Among these three objects(\"the rectangular, warm brown wooden box\",\"the rectangular wooden bed with a brown and beige bedspread\",\"the tall, slender metal lamp with a cream lampshade\" ), which one is the closest to you?", "option": [ "the rectangular, warm brown wooden box", "the rectangular wooden bed with a brown and beige bedspread", "the tall, slender metal lamp with a cream lampshade" ], "answer": "the rectangular wooden bed with a brown and beige bedspread", "user_message": "For the 7 turn, these are the 4 images you took in chronological order. The question for this turn is: Among these three objects(\"the rectangular, warm brown wooden box\",\"the rectangular wooden bed with a brown and beige bedspread\",\"the tall, slender metal lamp with a cream lampshade\" ), which one is the closest to you? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['the rectangular, warm brown wooden box', 'the rectangular wooden bed with a brown and beige bedspread', 'the tall, slender metal lamp with a cream lampshade'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." }, { "type": "Agent_visible_info-order(Judgement)", "new_observations": [ "1mp3d_0040_region16/93528274a86b47e6a3b0a6f398090741_i1_3.jpg", "1mp3d_0040_region16/93528274a86b47e6a3b0a6f398090741_i2_3.jpg", "1mp3d_0040_region16/81250dc8c43340689897b178d434c23b_i1_0.jpg", "1mp3d_0040_region16/81250dc8c43340689897b178d434c23b_i2_0.jpg" ], "turn_id": 8, "scan_id": "1mp3d_0040_region16", "origin_question": "What will be the first-time appearance order of the following categories: picture, stand, tv?", "option": [ "picture, tv, stand", "tv, stand, picture", "tv, picture, stand", "stand, tv, picture" ], "answer": "tv, picture, stand", "user_message": "For the 8 turn, these are the 4 images you took in chronological order. The question for this turn is: What will be the first-time appearance order of the following categories: picture, stand, tv? To answer this question, you need to combine image/coordinate information from past turns. Please choose the answer from ['picture, tv, stand', 'tv, stand, picture', 'tv, picture, stand', 'stand, tv, picture'] based on history information. Give me your answer and your reason in a JSON format({answer:str, reason:str}) ", "system_prompt": "Assume you are currently exploring a room where all objects are stationary. Over time, you change your position and orientation within the room and take images.\nNow, I will engage you in a multi-turn dialogue (a total of 8 turns). In each turn, I will provide you with 4 images taken from the beginning to the end of that turn, \nPlease answer my questions based on your state(position/orientation) at the end (last image) of each turn." } ]